climate change to land surface change

LINKING CLIMATE CHANGE TO LAND SURFACE CHANGE

ADVANCES IN GLOBAL CHANGE RESEARCH

VOLUME 6

Editor-in-ChiefMartin Beniston, Institute of Geography, University of Fribourg, Perolles,

Switzerland

Editorial Advisory BoardB. Allen-Diaz, Department ESPM-Ecosystem Sciences, University of California, Berkeley,

CA, U.S.A.R.S. Bradley, Department of Geosciences, University of Massachusetts, Amherst, MA,

U.S.A.W. Cramer, Department of Global Change and Natural Systems, Potsdam Institute for Cli-

mate Impact Research, Potsdam, Germany.H.F. Diaz, NOAA/ERL/CDC, Boulder, CO, U.S.A.S. Erkman, Institute for Communication and Analysis of Science and Technology – ICAST,

Geneva, Switzerland.M. Lal, Centre for Atmospheric Sciences, Indian Institute of Technology, New Delhi, India.M.M. Verstraete, Space Applications Institute, EC Joint Research Centre, Ispra (VA)‚

Italy.

The titles in this series are listed at the end of this volume.

LINKING CLIMATE CHANGE TOLAND SURFACE CHANGE

Edited by

Sue J. McLaren

and

Dominic R. KnivetonDepartment of Geography,

University of Leicester,Leicester, England, U.K.

KLUWER ACADEMIC PUBLISHERSNEW YORK, BOSTON, DORDRECHT, LONDON, MOSCOW

eBook ISBN: 0-306-48086-7Print ISBN: 0-7923-6638-7

©2003 Kluwer Academic PublishersNew York, Boston, Dordrecht, London, Moscow

Print ©2000 Kluwer Academic Publishers

All rights reserved

No part of this eBook may be reproduced or transmitted in any form or by any means, electronic,mechanical, recording, or otherwise, without written consent from the Publisher

Created in the United States of America

Visit Kluwer Online at: http://kluweronline.comand Kluwer's eBookstore at: http://ebooks.kluweronline.com

Dordrecht

TABLE OF CONTENTS

Table of contents

Preface

Contributing Authors

v

vii

xi

SECTION A: SHORT-TERM CLIMATE VARIABILITY

Chapter 1 Brooks, N. and Legrand, M.Dust variability over Northern Africa and rainfall in the Sahel

Chapter 2 Agnew, C. T. and Chappell, A.Desiccation in the Sahel

Chapter 3 Yair, A. and Bryan, R. B.Hydrological response of desert margins to climate change: The

Effect of Changing Surface Properties

Chapter 4 Viles, H. and Goudie, A. H.Weathering, geomorphology and climaticvariability in the Central Namib Desert

Chapter 5 Adegoke, J. O. and Carleton, A. M.Warm season land surface-climate interactions in the UnitedStates Midwest from mesoscale observations

Chapter 6 Wilby, R. L. and Dettinger, M.D.Streamflow changes in the Sierra Nevada, California,simulated using a statistically downscaled General CirculationModel scenario of climate change

Chapter 7 Schmidt, M. and Dehn, M.Examining links between climate change and landslide activityusing GCMS: Case Studies from Italy and New Zealand

1

27

49

65

83

99

123

vi SECTION B: LONG-TERM CLIMATE VARIABILITY

Chapter 8 Bachhuber, F. W. and Catto, N. R.Geologic evidence of rapid, multiple and high magnitudeclimate change during the last glacial (Wisconsinan) of NorthAmerica

Chapter 9 Catto, N. R. and Bachhuber, F. W.Aeolian geomorphie response to climate change: an examplefrom the Estancia valley, Central New Mexico, U.S.A.

Chapter 10 White, K., McLaren, Black, S. and Parker, A.Evaporite minerals and organic horizons in sedimentarysequences in the Libyan Fezzan: implications forpalaeoenvironmental reconstruction

Chapter 11 Gurney, S. D.Relict cryogenic mounds in the UK as evidence of climateChange

Chapter 12 Burgess, P. E. ,Palutikof, J. P. and Goodess, C. M.Investigations into Long-Term Future Climate Changes

SECTION C: SUMMARY

Chapter 13 Kniveton, D. and McLaren, S.Geomorphological and climatological perspectives on land surface –climate change

Index

143

171

193

209

231

247

261

PREFACE

The relationships that exist between changes in climate and land surface change aretopical issues, but research and collaboration between researchers from the differentdisciplines of climatology, geomorphology and Quaternary Sciences, is often hamperedby the different approaches; the incompatibility of scales of involvement (both spatialand temporal) of the various models used; and by differences of interest in such topicsas mean values for climatic parameters and the probabilities of extreme events. In termsof approaches there are those researchers who have tried to model past, present andfuture climatic changes, and there are people who have used proxy data (such assediments and landforms) to reconstruct past climates. Only relatively recently haveattempts been made to integrate the two distinct approaches.

In order to improve our understanding of the relationships that exist between changingclimates and land surfaces, a number of factors need to be considered including: - thespatial and temporal scales of climate variability and geomorphological change; the impactsof climate change on various landforms; the modification of climate by surface processes;modelling climate change on a global scale as well as downscaling of such model outputsso that they are applicable on regional scales; and prediction and management of landsurface changes as a result of future climate changes. These factors will be discussedfurther in Chapter 13.

To understand how climate is likely to change in the future, it is necessary to have anunderstanding of how climate has changed in the past in order to identify any underlyingtrends in natural climatic change. Many of the studies that use various proxies to makeinterpretations of past environmental conditions from landforms and other land surfacefeatures, as well as the small scale recent process-based research all need to be placed ina larger framework to aid our understanding of global climate change. Palaeo-reconstructions are needed to provide evidence of past changes, to help in thecomprehension of the responses of terrestrial surfaces and to help validate predictivemodels of climate change. Present day studies rely on the processes of observation,measurement (using both field work and analysis of remotely sensed images) as well asmodelling.

This book by no means attempts to be a summary of the main research on looking at therelationship between climate change and land surface change, but rather gives aselection of papers that show some of the different approaches that have been

viii McLAREN AND KNIVETON

undertaken to address the many issues and to highlight the importance ofmultidisciplinary research over different timescales (from 101 to 106 years) and from thescale of local catchment studies to global processes. Recent technical advances intechniques such as absolute dating; geochemical analyses, remote sensing and climatemodelling have aided these studies.

The book stems from a one-day conference held at the Royal Geographical Society withthe Institute of British Geographers (R.G.S with I.B.G.) Annual Conference held inLeicester on January 5th 1999. The symposium was jointly organised by the Associationof British Climatologists (A.B.C.) and the British Geomorphological Research Group(B.G.R.G.), and was organised by Sue McLaren, Dominic Kniveton and JohnMcClatchey.

The selection of peer-reviewed papers included in this book address a wide range ofissues ranging from looking at long-term climate changes through modelling (Burgess etal), palaeoenvironmental reconstructions (e.g. White et al, Catto & Bachhuber,Bachhuber & Catto) through to evidence of short-term climatic variability (e.g.Adegoke & Carleton, Brooks & Legrand, and Yair & Bryan) and attempts to downscalefrom General Circulation Models (GCM’s) to allow modelling of regional-scale patternsof climatic change and the effects on various surface and geomorphological processes(e.g. Wilby & Dettinger and Schmidt & Dehn). The chapters show just a smallselection of the wide-ranging nature of research currently being undertaken in thegeneral area of climate change and terrestrial surface processes.

The main division of papers has been made in terms of the spatial and temporal scales ofthe studies rather than between climatology and geomorphology because the editors ofthe book wish to stress the importance of trying to link these two areas. The finalchapter develops the main themes of the preceding chapters in the context of the widerfield of scientific literature.

The authors hope that this book makes an early attempt to present some recent advancesin understanding the linkages between climates and land surfaces in order to further ourability to predict environmental change.

The success of the conference and the production of the book were as a result of manypeople. We would like to thank the A.B.C., the B.G.R.G. and the R.G.S. (with theI.B.G.) for providing funds for guest speakers to attend the conference. We are gratefulto John McClatchey (Nene), Dave Thomas (Sheffield), Alan Werritty (Dundee) andNorm Catto (Newfoundland), for acting as Chairpersons at the symposium. In terms ofthe preparation of the book we would like to thank all the contributors (especially for

PREFACE ix

meeting all the deadlines); the many reviewers; Susan Draycott, Ruth Pollington andKate Moore for help with printing and cartography; and to Mariette Ph de Jong andAstrid Zandee who approached the authors with the offer of publishing the book withKluwer Academic Publishers.

SUE McLARENDOMINIC KNIVETON

Department of Geography, UniversityOf Leicester, Leicester LE1 7RH

This page intentionally left blank

CONTRIBUTING AUTHORS: -

JIMMY ADEGOKE: Department of Geography and Earth System Science Center,The Pennsylvania State University, University Park PA 16802, U.S.A.

CLIVE AGNEW: Department of Geography, University College London, 26Bedford Way, London WC1H OAP U.K.

FRED BACHHUBER: University of Nevada, Las Vegas, Las Vegas, NV,USA, 89154-4010.

STUART BLACK: Postgraduate Research Institute for Sedimentology, TheUniversity of Reading, Whiteknights, Reading, RG6 6AB, U.K.

NICK BROOKS: Climatic Research Unit, School of Environmental Sciences,University of East Anglia, Norwich, NR4 7TJ, U.K.

RORKE BRYAN: Faculty of Forestry, The University of Toronto, Toronto,Canada.

PAUL BURGESS: Climatic Research Unit, University of East Anglia, NorwichNR4 7TJ, U.K.

ANDREW CARLETON: Department of Geography and Earth System ScienceCenter, The Pennsylvania State University, University Park PA 16802, U.S.A.

NORM CATTO: Memorial University of Newfoundland, St. John’s, Canada,

A1B 3X9

ADRIAN CHAPPELL: Telford Institute of Environmental Systems, Departmentof Geography, University of Salford, Manchester, M5 4WT U.K.

MARTIN DEHN: Dept. of Geography, University of Bonn, Meckenheimer Allee166, D-53115 Bonn, Germany

MICHAEL DETTINGER: U.S. Geological Survey, Water Resources Division,California, Scripps Institution of Oceanography, 9500 Gilman Drive, La Jolla,California, 92093-0224

CLARE GOODESS: Climatic Research Unit, University of East Anglia, NorwichNR4 7TJ, UK

xii

ANDREW GOUDIE: School of Geography, University of Oxford, MansfieldRoad, Oxford OX1 3TB

DOMINIC KNIVETON: Department of Geography, University of Leicester,University Road, Leicester LE1 7RH

MICHEL LEGRAND: Laboratoire d’Optique Atmosphérique Universitéde Sciences et Technologies de Lille-1, F59655 Villeneuve d’Ascqcedex, France.

SUE McLAREN: Department of Geography, University of Leicester, UniversityRoad, Leicester, LE1 7RH, U.K.

JAN PALUTIKOF: Climatic Research Unit, University of East Anglia, NorwichNR4 7TJ, U.K.

ADRIAN PARKER: Geography Department, Oxford Brookes University, GipsyLane Campus, Headington, Oxford, OX3 0BP, U.K.

MICHAEL SCHMIDT: Dept. of Geography, University of Bonn, MeckenheimerAllee 166, D-53115 Bonn, Germany

HEATHER VILES: School of Geography, University of Oxford, Mansfield Road,Oxford OX1 3TB, U.K.

KEVIN WHITE: Landscape and Landform Research Group, Department ofGeography, The University of Reading, Whiteknights, Reading, RG6 6AB, U.K.

ROBERT L. WILBY: Division of Geography, University of Derby, KedlestonRoad, Derby, DE22 1GB, UK. National Center for Atmospheric ResearchBoulder, Colorado, 80307-3000, USA

AARON YAIR: Department of Geography, The Hebrew University, Jerusalem,Israel.

DUST VARIABILITY OVER NORTHERN AFRICAAND RAINFALL IN THE SAHEL

NICK BROOKSClimatic Research UnitSchool of Environmental Sciences,University of East Anglia, NorwichNR4 7TJ, U.K.

MICHEL LEGRANDLaboratoire d’Optique AtmosphériqueUniversité de Sciences et Technologiesde Lille-1, F59655 Villeneuve d’Ascqcedex, France.

Abstract

The Infra-Red Difference Dust Index (IDDI) is a new dataset that uses reductions inatmospheric brightness temperature (derived from METEOSAT IR-channelmeasurements) to map the distribution of mineral aerosols over continental Africa. TheIDDI dataset is described, and the IDDI data are used to identify the major African dustsources, located in the Sahel-Sahara zone. The seasonal variations in these sources arediscussed. Annual, seasonal and monthly dust indices are constructed from the IDDIdata for different latitudinal zones in the Sahel-Sahara zone. The temporal and spatialvariability of dust production in the Sahel and Sahara is inferred from these indices andthe latitudes of maximum dust production are identified. Interannual variability of dustproduction is described in conjunction with a consideration of variations in annualrainfall over the Sahel. Relationships between rainfall and subsequent dust production inthe Sahel are investigated by correlating zonally averaged rainfall and IDDI values atlags of one and two years.

The spatial and temporal patterns of dust production suggest that spring and summerdeflation is associated with the passage of convective disturbances across the Sahel.There is evidence that wet-season rainfall totals have an impact on dust production inthe later part of the following dry season. The results also suggest a cumulative impactof rainfall on December dust production. However, there is no evidence from this studythat dust production is associated with widespread land degradation.

KEY WORDS: dust, rainfall, Sahel, Sahara, variability1

S.J. McLaren and D.R. Kniveton (eds.), Linking Climate Change to Land Surface Change, 1–25.© 2000 Kluwer Academic Publishers. Printed in the Netherlands.

2 BROOKS AND LEGRAND

The Sahel is the semi-arid transition zone between the Sahara desert and humidequatorial Africa. It is characterised by a steep north-south temperature gradient andhigh interannual rainfall variability. The timeseries of spatially aggregated rainfallanomalies for the Sahel (Figure 1) suggests that the region has experienced a desiccationsince the late 1960s. Rainfall has been below the regional twentieth century mean formost years since 1968. Large rainfall deficits in 1972 and 1973 contributed to famine inthe Sahel, and the largest rainfall deficit this century was associated with the Ethiopianfamine of 1984. In both of these cases the impact of drought was exacerbated by otherfactors.

West African visibility data indicate that levels of atmospheric dust over the Sahelthroughout the year have increased dramatically since the 1950s, and it has beensuggested that dust loadings over the Sahel now exceed those over the Sahara (N’Tchayiet al., 1994, 1997). Middleton (1985) found an increase in dust storm activity in certainparts of the Sahel during drought years. Prospero and Nees (1986) reported elevateddust concentrations in the atmosphere over the North Atlantic after the deficient wetseasons of the early 1970s. More recently, Tegen and Fung (1995) and Tegen et al.(1996) have suggested that 30-70% of the global mineral aerosol budget is the result ofdeflation from soils which have been degraded by climate change and/or humanactivity. They invoke human activity in the Sahel, and a climatic shift in the boundary

1. Introduction

DUST VARIABILITY OVER NORTHERN AFRICA 3

between the Sahel and Sahara, as major factors in determining the global atmosphericdust budget. These studies have resulted in the widely held opinion that dust productionin northern Africa has largely shifted from the Sahara to the Sahel as a result of climaticdesiccation and inappropriate land-use practices. Until now, it has been difficult toassess such assumptions using observational data as such data have been somewhatlimited in spatial extent. However, a new proxy dust-loading dataset for continentalAfrica now exists, based on METEOSAT infra-red channel measurements. This datasetis known as the Infra-Red Difference Dust Index (IDDI). While the IDDI detects anyaerosols which reduce the infra-red radiance at the top of the atmosphere, it may beinterpreted in terms of dust concentrations over the arid and semi-arid regions ofnorthern Africa, where mineral dust is the dominant atmospheric aerosol.

The IDDI dataset has been used in a preliminary investigation of spatial and temporaldust variability over the Sahel-Sahara zone of northern Africa (i.e. Africa north of theEquator). This paper presents results detailing the spatial and temporal variability ofatmospheric dust loadings for the period 1984-1993. Spatial variability and seasonalityare addressed via a visual analysis of dust/IDDI fields. A more quantitative presentationof seasonality and meridional variation in dust production is achieved by plotting meanmonthly IDDI values, spatially averaged over different latitudinal zones, against time. Aqualitative interpretation of dust variability in response to rainfall is presented, followedby a discussion of correlations between wet-season rainfall and subsequent dustloadings as represented by zonally averaged IDDI values. The short length of the IDDItime series means that many of the conclusions are speculative. However, aconsideration of the results within the context of existing knowledge enables a plausibleconceptual model of rainfall influences on dust production to be constructed.

This study concentrates on the aerosol signal in the IDDI fields over the Sahel-Saharazone, because of the recent changes in observed dust concentrations and also becausethis region contains the major African dust sources. We may also be confident thatsignals in the IDDI fields over the arid and semi-arid regions of northern Africa are theresult of the episodic transport of dust (see below). However, IDDI signals over otherparts of Africa are also discussed where appropriate. Possible explanations for thepresence of strong signals in the IDDI data where dust is unlikely to be a majoratmospheric constituent are presented.

2. The Infra-Red Difference Dust Index

The IDDI dataset has been developed at the Laboratoire d’Optique Atmosphérique atthe Université des Sciences et Technologies de Lille, France (Legrand et al., 1994).IDDI data represent the reduction in the measured infra-red (IR) brightness temperature(BT) of the atmosphere from that which would result from an aerosol-free atmosphere.Brightness temperature values are derived from METEOSAT IR-channel radiometriccount measurements taken daily at approximately 11:30 UTC. Fields of maximumbrightness temperature over non-overlapping 15-day periods are constructed. Fields ofdifferences between these composite fields and daily brightness temperature fields


within each 15-day period are then calculated. The resulting difference fields aredivided into 10x10-pixel boxes and a statistical algorithm based on the spatial coherencemethod (Legrand et al., 1994) is used to classify pixels as cloudy or non-cloudy. Cloudypixels are assigned a cloud-masking code, and the remaining pixels represent the IDDIvalues, where brightness temperature reductions are due to the presence of aerosolsalone.

The IDDI signal results from the reduction in the temperature of the underlying landsurface by reduced solar insolation (resulting in less emitted IR radiation), and also fromthe attenuation of the outgoing longwave radiation (OLR) by the aerosol layer.Attenuation of OLR will be greatest when the aerosol particles have effective diametersof the same order of magnitude as the wavelength of the radiation, i.e. of the order of 10

Sub-micron particles are transparent in the infra-red (Maley, 1982). Theoreticalconsiderations and recent, as yet unpublished, modelling studies (Legrand, pers. comm.)indicate that, in the case of mineral aerosols, small dust particles cause thegreatest reduction in daytime temperatures, while coarse dust causes the greatestdaytime reduction in IR radiance at the top of the atmosphere (TOA). The strongestsignals in the IDDI will therefore result from dust events with a high proportion of large

particles, although events comprised of small particles in high concentrationswill be detected due to the reduction in emitted IR radiation from the cooler surface.

The IDDI data are converted to a 1° latitude x 1° longitude geographical grid, and existover land regions only. The geographical coverage extends from 35° south to 38° northand 18° west to 45° east, covering all of Africa and parts of the Middle East (see Figure2). The dataset will be updated to the present day in the near future.

The IDDI data have been validated against ground-based visibility and aerosol opticaldepth (AOD) measurements at a number of sites throughout West Africa (Legrand etal., 1994). During these validation studies, it was found that IDDI values correlated wellwith near-surface visibilities. IDDI values of 5 K and above corresponded to dustyconditions, when visibility was reduced below 10 km, and values of 10 K and abovecorresponded to severely dusty conditions, with visibility reduced below 5 km. IDDIimages have also been compared with fields of AOD over the eastern tropical Atlanticin order to verify continuity across the West African coast.

Nonetheless, there are several potential pitfalls to be considered when interpreting theIDDI data. The detection of aerosols depends on the variability in their concentration. Ifconcentrations are generally elevated over the whole of the 15-day reference period,they will be interpreted as part of the “clear-sky” background, reducing the BT values ofthe reference field. Long-term dust haze is therefore unlikely to be detected. A similarproblem may occur over regions which are covered by cloud throughout the referenceperiod. Long-term cloud cover will result in misleading reference values, and may alsoaffect the efficiency of the cloud-detection algorithm, leading to the erroneousidentification of cloud as IDDI (i.e. aerosol) data. Over very cloudy regions such asthose near the Equator, the IDDI data may be unreliable due to this “cloud


contamination”. Problems may also arise where low, relatively warm, clouds arepresent; these may be identified as aerosols, resulting in large IDDI values.

Also to be considered is the presence of aerosols resulting from biomass burning, whichis widespread throughout much of Africa in the dry season. Such aerosols typically havedimensions of less than (Artaxo et al, 1994); they will have some impact on theOLR, but their dominant effect will be one of cooling of the land surface. Theseparticles should therefore have a similar effect on the measured TOA radiance to finedust aerosols. However, because of the extent of burning, they may constitute a constantsmoke haze lasting for periods of days to weeks, resulting in their not being detected inthe IDDI fields, but rather being incorporated into the reference fields.

The above considerations notwithstanding, the IDDI data represent a useful semi-quantitative measure of dust loadings over the arid and semi-arid regions of Africa.Over the Sahara and Sahel, dust events are highly episodic and contain high proportionsof aerosols large enough to strongly attenuate the OLR, resulting in strong IDDI signals.The incidence of cloud over these regions is low enough to present no significantproblems of cloud contamination. The issues of biomass burning aerosols and fine dusthaze are discussed in more detail below, although these features do not appear to inhibitthe detection of episodic dust events over the main regions of interest in this study,which lie north of 10° N.

To date, IDDI fields over the Sahel and Sahara have not been converted to AOD values,and cannot be interpreted in terms of specific volumes of dust or thicknesses of dustlayers. The reduction in brightness temperature due to dust aerosols will depend on thevertical distribution of the dust, the particle density and the particle size distribution, aswell as the reflective properties of the underlying surface. Nonetheless, IDDI fieldsreliably reflect the distribution and abundance of atmospheric mineral aerosols overnorthern Africa, and exhibit a sufficient degree of spatial and temporal invariance to beused in studies of large-scale dust mobilisation and transport (Legrand et al., 1994).

3. Distribution of Saharan and Sahelian dust sources

It may be assumed that dust concentrations and particle sizes will be greatest closest todust source regions. Fields of IDDI data may therefore be employed to identify themajor source regions throughout Africa. Use of different averaging periods enables thetemporal variation in the activity of dust sources to be analysed. The major dust sourcesin northern Africa have been identified in this fashion by Legrand et al. (1994). Thissection elaborates on their description, within the context of other studies of dustsources and climatological considerations of known or likely dust mobilisationprocesses. Discussion of the major dust sources is restricted to northern Africa,focussing on the Sahelian and Saharan zones.

Monthly mean IDDI fields were created by averaging daily IDDI fields for cells wherefewer than eighty per cent of days were classed as cloudy. Over most of the Sahel-


Sahara zone, where cloud is scarce, this approach results in continuous spatial coveragein the monthly fields. Annual mean IDDI fields were created for each year by averagingthe monthly mean fields over twelve-month periods. Seasonal mean fields were createdby averaging the monthly fields over shorter periods for each year. Mean annual,seasonal and monthly fields were created by averaging the yearly fields over the period1984-1993. The mean annual IDDI field for 1984-1993 is shown in Figure 2.


Three broad regions in which IDDI values exceed 5 K are apparent. This threshold isarbitrary, but delineates distinct zones within which mean dust levels are elevated abovethe background. Further detail within these zones is apparent in the form of areas withIDDI values in excess of 5.5 K. These regions are interpreted as coinciding broadly withareas containing dust sources. One such region is the north-central Sahel, between about5° E and 20° E, and 13° N and 18° N. Two maxima are apparent within this region,centred approximately at 16° E., 17° N and 9° E, 15° N. The former maximum extendsover parts of the Erg of Bilma and the alluvial plain northwest of the town of Largeau inChad. This region has been identified as an important dust source by other authors (e.g.McTainsh, 1980; Drees et al., 1993). The latter maximum lies to the south of the AïrMountains in Niger, in the vicinity of a region of enhanced generation of convectivedisturbances (Rowell and Mitford, 1992) which result in spring and summer dustmobilisation (Dubief, 1979; McTainsh, 1996).

A second source region (or collection of sources), which may be labelled the WestSahara region, lies between about 7°-0° W and 20°-25° N. This area corresponds to aregion that includes the Erg Iguidi and Erg Chech of northern Mali, northern Mauritaniaand southwestern Algeria. A nearby maximum in the IDDI field lies over a region ofseasonal watercourses in the Morocco-Western Sahara border region. Dust transportedlarge distances over the Atlantic and to Europe has been identified as originating inthese regions (Reiff et al., 1986; Coudé Gaussen et al., 1987; Chiapello et al., 1997).

The third major source region extends from about 13° N to 25° N, and some 1° to 3°either side of the 30° E meridian, from northern Sudan into southern Egypt. Hereafterthis is referred to as the East Sahel-Sahara region. This region is characterised by theHaboob dust storms of the Nile Valley (McTainsh, 1996), and dust from thenortheastern Sudan has been transported to the eastern Mediterranean (Middleton, 1986,1997).

A minor region of activity is indicated by high IDDI values over a small area centred on14° E, 22.5° N, between the Plateau de Djado in northern Niger and the Idhan Murzuqerg in southwestern Libya. This region is hereafter referred to as the northern Nigerregion.

All the source regions identified above are characterised by fields of sand dunes orseasonal watercourses, or both. This suggests that erodible material is supplied by dunefields or by water erosion, or a combination of the two. The source region near the AïrMountains extends into the zone of degraded soils as suggested by UNEP (1992),suggesting that land use and climatic desiccation of soils may be partly responsible fordeflation in this area. However, the region is dominated by numerous water channelsand few permanent human settlements, suggesting that water erosion is an importantfactor in providing erodible material.


Also present in the annual field are strong IDDI signals over the Horn of Africa, west-central Africa and southeastern Africa. The two former regions exhibit IDDI values ashigh or higher than the highest Sahelo-Saharan values. Dust transport over the Horn ofAfrica is associated with the Asian Monsoon circulation in summer (Husar et al., 1997).The west-central African signal is unlikely to be due to dust aerosols, while the reasonfor the southeastern African signal is open to debate. The high IDDI values over thesethree regions are discussed further in Section 4, within the context of seasonal changesin the regional environment.

3.1. DUST SOURCES AND LAND DEGRADATION

The northern limit of the region characterised by land degradation is placed in the regionof 17° N on soil degradation maps published by UNEP (1992). However, estimates ofthe extent of soil degradation in the Sahel are extremely unreliable and subjective(Warren, 1996; Williams and Balling, 1996). In the absence of reliable soil degradationdata it is impossible to identify new dust sources arising from land-use practices orclimatic desiccation, or to quantify the contribution of disturbed soils to the regionaldust budget. However, soil degradation is likely to be minimal in regions of low rainfalland outside of the zone of rainfed agriculture, the limit of which is placed at the locationof the 300 mm isohyet by WMO (1976). Fields of annual rainfall totals derived from thedataset of New et al., (1999, not shown) indicate that the 300 mm isohyet lies to thesouth of 17° N. These considerations suggest that the 17° N latitude represents areasonable and liberal (if somewhat arbitrary) working limit for the zone containingdegraded soils. This limit will be employed when the role of soil-state in dustproduction is considered in Sections 5-8.

Examination of the mean annual IDDI field suggests that the major dust source regionsin the Sahel and Sahara conform to the accepted, or “classical”, sources of dust, createdby “natural” processes of sediment production and deflation. A possible exception is thesource region in the north-central Sahel in the vicinity of the Aïr Mountains.

It is possible that material from anthropogenically degraded soils does not produce astrong signal in the IDDI data, resulting in an underestimation of the extent of the majorsource regions. Aerosols from degraded soils are likely to be very different in naturefrom those deflated from arid to hyper-arid desert regions. Dust consisting of suchaerosols will contain more organic material and have a higher clay content, resulting ina high proportion of small aerosol particles (McTainsh and Walker, 1982).Organic material has been detected in dust deposited in Niger (Drees et al., 1993) andnorthern Nigeria (McTainsh and Walker, 1982). However, it is not clear whether theorganic input is due to the long-term desiccation of vegetated areas or if it is a long-termfeature of the soil-dust cycle. As previously discussed, the IDDI signal from dust with alow mean particle size will be predominantly the result of surface cooling. McTainshand Walker (1982) report a tendency for lower visibility and reduced solar radiation tobe associated with finer mean particle sizes. The correlation of IDDI values withmeasured visibilities (Legrand et al., 1994) suggests that the IDDI are capable of


detecting such fine material. It is possible that such fine material from degraded soilsexists as a semi-permanent dust haze throughout much of the year, in which case itwould not be detected by the IDDI for the reasons outlined above. However, it isreasonable to suppose that episodic dust events would originate over such degraded landin the same fashion as over other regions, and as a result of the same atmosphericprocesses. This would be particularly true outside of the wet season, when the Saharaand Sahel are both subject to the Harmattan circulation. The lack of a regional signal inthe IDDI data over the hypothesised regions of widespread land degradation (thevicinity of the Aïr Mountains notwithstanding) therefore calls into question theassumption that aerosols from degraded soils contribute significantly to the regionaldust budget, and the budget of material exported from northern Africa (Tegen and Fung,1995).

4. Seasonal variations in dust production and non-dust IDDI signals

Seasonally averaged fields of IDDI are presented in Figure 3. Again a threshold of 5 Kdelineates broad regions of dust activity, with further detail apparent in the form ofIDDI values in excess of 6 K. While this analysis focuses on northern Africa, thestructure of the seasonal IDDI fields in southern, eastern and central Africa is alsodiscussed where appropriate.

In JFM the most active areas are the East Sahel-Sahara, the north-central Sahel and thenorthern Niger regions. A broad shift in dust activity from east of 5° E in JFM to west of15° E in AMJ is apparent. The East Sahel-Sahara sources remain active in AMJ,although the geographical extent of IDDI values greater than 6 K is reduced. AMJ IDDIvalues are high over southern Morocco and western Algeria, and also in the western partof the north-central Sahara.

JAS represents the peak of the Sahelian wet season, when the surface discontinuitybetween the West Africa Monsoon airmass and the dry Saharan airmass lies at itsnorthernmost limit, around 20° N in August (Hastenrath, 1991). IDDI values greaterthan 6 K are confined to the west of 5° E and between 17° N and 25° N. The southernlimit of this zone is very distinct; the 4 K/5 K boundary occurs close to the 5 K/6 Kboundary at approximately the same latitude from the West African coast to 7° E. Thissuggests that large dust loadings are prevented from occurring south of the northernlimit of the monsoon rains, which extend to within several hundred kilometres south ofthe surface discontinuity (Hastenrath, 1991). The southern latitudes of this regioncoincide with an area identified by Rowell and Milford (1992) as a region of enhancedgeneration of convective disturbances or disturbance lines (DLs), encompassing theplains to the north of the Niger Bend.


Another major feature of the JAS field is the region of very high IDDI values over theHorn of Africa in JAS. These values are considerably higher than the maximum valuesover the Sahel and Sahara. This signal over the Horn of Africa coincides with very highequivalent aerosol optical thickness (EAOT) measurements over the Arabian Sea (Husaret al., 1997 – based on data from July 1989 to June 1991). The parts of Arabia visible inthe IDDI fields exhibit low IDDI values, suggesting that dust transport over the ArabianSea is predominantly from the Horn of Africa (Sirocko and Sarnthein, 1991).Mobilisation and transport of dust is aided by the East African (or Somali) low-level jet,


which is active at this time of year as part of the summer monsoonal circulation(Hastenrath, 1991, Husar et al., 1997). Transport of dust over large distances occursabove the monsoon inversion, in a fashion analogous to the transport of Saharan dustabove the trade wind inversion in the Saharan air layer (Kalu, 1979; Sirocko andSarnthein, 1989). The OND field exhibits low IDDI values except in a small regionwithin the north-central Sahel zone and another over northern Niger. Examination of themean monthly fields (not shown) illustrates that dust loadings are lowest in November,and that the high-IDDI regions in the OND field are due to the “switching on” ofsources in these regions in December.

The major IDDI signals outside of the regions discussed above are detailed andinterpreted below.

4.1. THE GUINEA COAST

In JAS a zone of relatively high IDDI values exists over the Guinea Coast region,extending in places to some 12° N and exhibiting a maximum in the east over Nigeria.The period JAS corresponds to the “Little Dry Season” (Barry and Chorley, 1995) inthis region and it might therefore be expected that widespread biomass burning wouldbe prevalent. Monthly maps of fire distribution are available for some years from theWorld Fire Atlas, compiled by the European Space Agency and the European SpaceResearch Institute (ESA/ESRIN) as part of the Ionia programme (Arino and Melinotte,1995; Arino et al., 1997). These maps have been produced from AVHRR and ATSRsatellite data. A visual comparison of the monthly IDDI fields with monthly fire mapsfor 1993 suggests that the JAS high IDDI values over the Guinea Coast are not due tocombustion products, as fires are almost entirely absent from this region in this periodaccording to the fire maps. At this time of year detectable fires are concentrated betweenthe Equator and 20°S, where IDDI values are low. Strong fire signals in theESA/ESRIN data occur over and to the east of the Guinea Coast throughout the winter,with fires being most widespread in January. Again, the regions of high IDDI values donot correspond to those characterised by fires; the January 1993 IDDI field exhibits lowvalues over the Guinea Coast. However, the relationship between the distributions offires as detected by satellite remote sensing methods, and high concentrations ofbiomass burning aerosol products is not necessarily straightforward. Fires will only bedetected if they exist under relatively clear-sky conditions. Both clouds and highconcentrations of airborne combustion products will obscure the ground from satellitedetectors operating in the visible part of the electromagnetic spectrum. Thus, fires thatproduce large quantities of aerosols may not be detected. It is plausible that materialfrom such fires is responsible for some of the high IDDI signals apparent in figures 5.4to 5.6, providing at least a partial explanation for the summer Guinea Coast signal.

Another plausible explanation for the high IDDI values over the Guinea Coast insummer is that dust is transported from the Sahel-Sahara to a zone of relatively stagnantair over this region, where it remains in the atmosphere for some time. Between theGuinea Coast and the Sahel-Sahara transition zone, dust will be removed from theatmosphere by rainfall, resulting in short residence times, low aerosol concentrations


and hence low IDDI values. A further possibility is that of cloud contamination arisingfrom persistent cloudy conditions throughout the periods used to create the referencefields. This is most likely over Nigeria, where the highest regional IDDI values exist inthe vicinity of a region of frequent cloud cover.

4.2. WEST-CENTRAL AFRICA

Large quantities of combustion aerosols also provide a plausible explanation for highIDDI values over regions where dust is unlikely to be a major feature of the atmosphere.Such values are seen over west central Africa (stretching from Gabon to the DemocraticRepublic of Congo and southwards over Angola) in all the fields, and are greatest inOND, JFM. Some biomass burning occurs in this region in these periods, particularly inOctober (based on 1993 data from ESA/ESRIN). However, the frequency and density offires during the periods in question is far greater between 0° and 15° N, where IDDIvalues remain low. Again, these discrepancies between the IDDI and fire data may bedue to the complex relationship between fire and smoke aerosol distributions. Thisregion is adjacent to a region of frequent cloud cover in JFM and OND (i.e. southernhemisphere spring and summer), when the IDDI values are highest. It is possible thatsome cloud contamination occurs in these periods.

4.3. EASTERN AFRICA

High IDDI values also occur over many of the eastern coastal regions of Africa south ofthe Equator, particularly in AMJ and JAS. These regions contain no extensive deserts,but do include semi-arid and dry sub-humid zones. The boreal summer high IDDI signaloccurs during the dry season in East Africa. It is possible that dust mobilisation occursfrom disturbed soils in these regions, although a complex biomass burning aerosolsignal is again highly plausible, as burning is widespread in the dry season. Cloudcontamination is likely in JFM and OND, but during AMJ and JAS the elevated IDDIvalues exist well away from areas of frequent cloud cover.

4.4. SOUTHERN AFRICA

Finally it is worth mentioning the southern hemisphere African deserts in terms of dustsources as defined by the IDDI data. These regions do not stand out in the seasonal orannual fields, although elevated IDDI values are apparent over the Kalahari in JFM. It isstriking that the Namib Desert does not appear to be a significant source of dust. Thecold Benguela Current to the immediate west of the desert results in a highly stableatmosphere that is not conducive to the generation of the type of large convective eventsthat are responsible for dust mobilisation and transport in northern Africa. While duststorms do occur over the sandy desert in the Namibian interior, it appears that the spatialand time scales associated with these events are such that they do not produce a majorsignal in the mean IDDI fields. The coastal atmosphere is very different from that overWest Africa, and it is likely that the atmospheric environment over the Namib desert issuch that dust aerosols are not carried to the elevations necessary for long-rangetransport. Middleton (1997) states that dust mobilisation and transport from the southern


African deserts is poorly understood, but suggests that the scale of such phenomena isnot comparable with that which characterises the northern African regions.

4.5. SUMMARY

The above discussion further illustrates some of the caveats to be considered wheninterpreting the IDDI data. The question of whether the IDDI is a reliable means ofdetecting combustion aerosols remains open, and will only be resolved when therelationship between detected fires and the nature and distribution in the atmosphere oftheir products is better understood. It also appears that the IDDI is less reliable underpersistently cloudy conditions. Further work is required to decouple the effects ofbiomass burning products and cloud contamination from the impacts of dust on theIDDI signal.

However, over the regions of interest in this study, the IDDI appears to perform well,exhibiting cumulative signals from large dust events and identifying the major sourcesof dust aerosols. It may therefore be used with confidence in studies of Saharan andSahelian aerosols and their relationships with the regional climate. Seasonal andgeographical variations in the IDDI data may also be used to infer informationconcerning the behaviour of the major aerosol sources in northern Africa.

5. Meridional variation in dust production

In order to assess the seasonal variation in dust production in northern Africa in a morequantitative fashion, several different zones were defined. These zones are theaggregated Sahel (10° - 20 ° N), the aggregated Sahara (20° - 30° N), the South Sahel(10° - 15° N), the North Sahel (15° - 20° N), the South Sahara (20° - 25° N), the NorthSahara (25° - 30° N), the zone from 15° - 17° N and the zone from 18° - 20° N. The lasttwo zones are used to examine dust seasonality either side of the suggested limit of soildegradation (Section 3.1).

Spatially aggregated, mean monthly IDDI values over the zones described above(Figure 4) illustrate a broad commonality of dust loadings over the Sahel-Sahara region.Values are generally high in the first half of the calendar year, falling to a minimum inOctober or November and rising again in December. However, the evolution of theNorth Sahara zone departs from that of the other zones outside of March-June. This is tobe expected as a result of the influence of mid-latitude weather systems such asMediterranean and Atlantic cyclones.


From June to September, dust levels are higher over the Sahel than over the Sahara.Sahel dust loadings peak in June; Saharan dust loadings are at a maximum in March andApril. The lowest dust levels occur in November over the Sahel, and in October overthe Sahara.

IDDI values are consistently higher over the North Sahel than over the South Sahel, andthe North Sahel exhibits the highest values of all the 5°-latitude zones in December andJanuary and from June to September. The North Sahel contains the transition zonebetween the Sahel and Sahara and the nominal northern geographical limit of soildegradation. The 15°-17° N band lies to the south of this limit, so variations of IDDIwithin this band may be interpreted as reflecting variability of dust production frompotentially disturbed soils, with a component due to advection from zones to the north,particularly during the dry-season. IDDI values in the 18°-20° N band may be assumedto reflect variability of dust production from undisturbed soils. However, theuncertainties in the estimates of the extent of soil degradation (Section 3.1) should berecalled.

The 15°-17° N band yields the larger IDDI signal from December to March and in Mayand June. (The April value is similar to that in the 18°-20° N. band.) This indicates thatthe meridional maximum in dust loadings lies in the 15°-17° N band in December,January and June, when the maximum values in the 5° latitude zones occur over theNorth Sahel. Similarly, dust loadings are highest in the 18°-20° N band from July toSeptember. (These results are unchanged if other 2°-latitude bands within the NorthSahel are considered.) During the summer the 2°-latitude bands exhibiting the highestIDDI values lie to the south of the average position of the surface discontinuity (Tetzlaffand Peters, 1988; Hastenrath, 1991), i.e. within the monsoonal air mass. It is arguable


that these high IDDI values represent advected material from the Sahara overlying themonsoon air. However, if this is the case, still higher IDDI values should be apparentcloser to the northerly source regions. Therefore, these meridional maxima in IDDIvalues may be interpreted as representing meridional maxima in dust levels resultingfrom dust mobilisation in the shallow northern part of the monsoon air layer.

Thus dust mobilisation is at a maximum within the zone containing potentially degradedsoils in the early to mid dry season and in the early phase of the wet season.Mobilisation may remain high in this zone in JAS, but rainfall will remove dust fromthe atmosphere, shifting the maximum in the IDDI signal to the northern fringes of theactive rainfall zone.

It is likely that the June maximum in the 15°-17° N band is due to the intensity of thedeflation processes and the balance between dust mobilisation and removal, rather thanthe sensitivity of the soils to deflation. In June this band corresponds to thenorthernmost extent of the wedge of monsoonal air (Tetzlaff and Peters, 1988), wherethe convective disturbances that generate rainfall and mobilise dust are weak due to thesmall thickness of the monsoon air layer (Hastenrath, 1991). Such weak disturbancesmay be sufficient to cause deflation, but too weak to produce significant amounts ofprecipitation. Thus the June maximum may be simply a manifestation of the regionalclimatology. The same processes are likely to be responsible for deflation in the 18°-20°N band in JAS.

In December and January both the Sahel and Sahara are subject to the regional-scaleHarmattan circulation, characterised by northeasterly winds over most of northernAfrica (McTainsh, 1996). Deflation processes are therefore associated with large-scaleatmospheric circulation patterns, suggesting that dust mobilisation will be greatestwhere soils are most vulnerable. The December-January maximum in dust productionbetween 15° and 17° N is therefore likely to represent a meridional maximum in theavailability of erodible material. This may be due to the fragility of degraded soils inthis region, or a maximum in water erosion arising from the action of rainfall andrainfall-runoff on semi-arid surfaces. Low vegetation cover may also play a role; it islikely that the combination of relatively high rainfall (when compared with the drySahara) and the lack of vegetation protection of the land surface together result in highwater erosion rates. Degraded soils will be more susceptible to water erosion, but it isnot necessary to invoke land degradation in order to explain this maximum in dustproduction.

6. Interannual variability of dust and rainfall

Figure 5 shows rainfall anomalies for the period 1983-1994, standardised with respect tothe 1983-1984 mean. This period represents the period over which IDDI data areavailable, and includes 1983 in order to show all years that may affect dust values at alag of +1 year.


Figure 6 shows yearly, spatially averaged annual IDDI anomalies calculated over fourdifferent periods, for the various latitudinal zones described in the previous section. Theprimary objective of such a representation is to illuminate interannual variability ofatmospheric dust loadings over bands subject to different rainfall regimes. While themain zones of interest are those in the Sahel, values for Saharan zones are included sothat rainfall-dominated regions may be compared with arid regions.

The annual period represents the mean IDDI values over the period November-October,chosen to commence around the beginning of the dry season. The wet-season is liberallydefined as the period May-October, during which deflation mechanisms are most likelyto be associated with the westward travelling disturbance lines (DLs), which bring themajority of rainfall to the Sahel (Rowell and Milford, 1992). The early dry-season isdefined as November-December, the part of the dry season in which the vegetationcover as represented by NDVI values is significantly greater than the dry-seasonminimum (Hess et al., 1996). The late dry-season is defined as January-April, duringwhich vegetation cover is close to the dry-season minimum, and in which dustmobilisation and transport in both the Sahel and Sahara are subject to the Harmattancirculation (Adeyfa and Holmgren, 1996).


In many cases the anomalies over one region reflect those over the other regions,suggesting a common atmospheric influence on the primary deflation mechanisms.Major differences between Sahelian and Saharan regions are likely to be due to theinfluence of rainfall in the Sahel. The impact of the severe 1984 drought is evident inannual, early dry-season and late dry-season anomalies in 1984/85, 1984 and 1985respectively. Over the Sahel the anomalies for these years are large and positive. Overthe Sahara these anomalies are small or negative. Rainfall influences therefore serve todecouple the Saharan and Sahelian dust signals.

6.1. ANNUAL ANOMALIES

The three driest years in the Sahel in the 1984-93 period were 1984, 1987 and 1990(Figure 5). The annual periods following these wet-seasons exhibit the largest positiveIDDI anomalies in the South Sahel series (Figure 6). The largest-magnitude negativeanomalies in the Sahel occur after the wet-seasons of 1985, 1989 and 1991. These IDDIanomalies occur after dry or intermediate-rainfall years. This pattern of large negativeIDDI anomalies is also reflected in the South Sahara, suggesting that atmosphericinfluences (for example a low frequency of strong surface winds) may be partlyresponsible for these periods of low dust loadings.

6.2. WET SEASON ANOMALIES

The largest positive wet-season IDDI anomalies in the South Sahel occur in 1988, 1989and 1991, the wettest years in the 1984-93 period. This further supports the hypothesisthat DLs (which are more frequent and intense in wet years) are largely responsible fordust mobilisation in the wet-season. For these three relatively wet years, IDDI anomalymagnitude decreases with increasing rainfall. While three years do not representsufficient data to constitute a trend, this result suggests the possibility that summer dustloadings may be generally higher in wetter years but that, above a certain rainfallthreshold, dust levels decline as rainfall increases. This is physically plausible: intenseDLs will mobilise greater quantities of dust than weak DLs, but will also produce morerainfall, which will remove dust from the atmosphere. Thus spring/summer Sahelatmospheric dust loadings are likely to be controlled by two processes that act inopposition to each other. The relative strengths of these processes will depend on thefrequency and intensity of the DLs in any given wet-season. This conceptual model hasimportant implications for the identification of the mechanisms behind the observedincreases in Sahelian dust production. Lamb et al. (1998) report a decrease in both thefrequency and intensity of DLs over the Sahel since the onset of dry conditions in thelate 1960s. Enhanced spring and summer dust loadings over the Sahel may therefore bethe result of a change in the balance between processes controlling the mobilisation andremoval of dust particles, rather than, or in addition to, changes in soil properties.

In the North Sahel and South Sahara the dustiest wet-seasons occur in 1987, 1988 and1991. Dust mobilisation in these regions is likely to be related to DL activity within thevicinity of the surface discontinuity, where the monsoon air layer is not thick enough to


allow rainfall generation. Rowell and Milford (1992) have identified August DLsgenerated as far north as 20° N.

Wet-season dust levels are lowest in 1985, 1986 and 1990 (South Sahel) or 1989 (NorthSahel). All of these years are dry except 1989, which follows the wet year of 1988. Lowlevels of dust in dry years may be explained by weak or infrequent DLs. The low levelof dust in 1989, a relatively wet year (Figure 5), suggests that there was not muchmaterial available for deflation in this year. This may be due to the removal of suchmaterial after heavy water erosion in 1988 and/or a recovery in the vegetation cover ofthe Sahel over 1988 and 1989. An alternative explanation is that dust levels are high inwet-seasons dominated by weak DLs (which do not produce much rainfall) and low inwet-seasons dominated by intense DLs. If the rainfall in 1989 was the result of apredominance of the latter, removal of dust by rainfall may have dominated over dustmobilisation.

6.3. EARLY DRY SEASON ANOMALIES

The most striking aspect of the November-December anomalies is the switch frompositive anomalies in the Sahel until 1988 to negative anomalies from 1989 onwards.This pattern is punctuated by small positive anomalies in the South Sahel in 1985 and1986 and a small positive anomaly in the North Sahel in 1992.

A single year of drought may not have a long-term impact on soil or vegetation(Bullard, 1997). However, several consecutive years of drought, as occurred in theearly-mid 1980s, are likely to have a cumulative impact on vegetation and hence on theorganic matrix of the soil, leading to loss of soil cohesion. It is suggested that the wetterconditions prevailing from 1988 onwards led to a recovery in soil cohesion byencouraging vegetation cover, which would result in a greater degree of protection ofSahelian soils from deflation (Bullard, 1997). The dry year of 1990 occurred inisolation, and would not have had a long-term impact on soil properties. The positiveIDDI anomaly following the 1988 wet-season is probably due to water erosion causedby the action of heavy summer rainfall on soils with little vegetation cover (eitherbecause of the distribution of rainfall or due to the dying off of vegetation under theprevious dry conditions). In the short-term this would lead to an increase in the amountof erodible material (Baird, 1997).

The anomaly series for the Saharan regions do not closely reflect those for the Sahelianregions, further reinforcing the interpretation that dust production in this period islargely a function of earlier rainfall. However, the large negative IDDI anomaly in 1989is apparent in all the series except that representing the South Sahel (where the anomalyis negative but not of great magnitude), suggesting that the regional-scale circulationalso modulates dust production in this period. The positive IDDI anomaly of 1988 isalso not confined to the Sahel, suggesting a possible atmospheric influence on dustlevels throughout the region.


6.4. LATE DRY SEASON ANOMALIES

The Sahel exhibits lower interannual variability in dust loadings over the January-Aprilperiod than over the other periods described here. As rainfall exhibits considerableinterannual variability, these results suggest that rainfall generally has a small impact onJanuary-April dust production. However, a very large positive IDDI anomaly is evidentin all Sahelian zones in 1985, after the extremely dry years of 1983 and 1984, providingcompelling evidence for a cumulative impact of multiple years of large rainfall deficits.1987 and 1990 are years characterised by extreme rainfall deficits that follow years thatare dry, but not extreme in terms of rainfall. 1987 and 1990 are not followed by largepositive late dry season IDDI anomalies. It is speculated that the small 1988 IDDIanomalies in the Sahelian regions may be due to the lack of rainfall and the consequentreduction in erodible material produced by water erosion.

7. Rainfall-dust correlations

Rainfall over the May-October period and monthly mean IDDI values were spatiallyaveraged over the zones defined in Section 5. The resulting timeseries, representingaggregated dust and rainfall values over spatially coincident areas, were correlated.Correlations were performed between rainfall and monthly IDDI values representingtwelve months commencing in the November immediately following the wet-season(lag = +1 year), and between rainfall and IDDI values representing twelve monthscommencing in November of the following year, i.e. thirteen months after the end of thewet-season (lag = +2 years). The results were tested for statistical significance using asimple monte-carlo style randomisation procedure. For each correlated pair, one of thetimeseries was randomised 10,000 times and the two series correlated for eachrandomisation. If the original correlation was exceeded fewer than 500 out of 10,000times the result was deemed to be significant at the 5 per cent level. Correlations at the 1per cent level were also noted. Correlations not significant at the 5 per cent level wererejected.

Correlations were calculated for Saharan zones for purposes of comparison: significantrelationships would not be expected over Saharan regions where rainfall is low andinfrequent. The rainfall averaging period is arbitrary in the case of Saharan rainfall,further reducing the likelihood of meaningful statistical dust-rainfall relationships overSaharan regions. If such relationships were found, they would suggest that statisticallysignificant results over the both the Sahara and the Sahel were artefacts of the statisticalprocedure employed.

Comparisons with the Sahara notwithstanding, the short length of the timeseries meansthat the resulting correlations should not be interpreted as demonstrating definitephysical relationships between rainfall and dust loadings. Nevertheless, considerationsof the probable mechanisms of dust production provide a conceptual context withinwhich such correlations may be interpreted. Significant correlations may therefore beused to infer likely impacts of rainfall on dust production, as well as the temporal


distribution of such lagged relationships. Such an approach is useful in reinforcing orrejecting existing hypotheses, and suggesting new hypotheses, of dust variability.

7.1. LAG 1-YEAR RELATIONSHIPS

No significant correlations result from the lag 1-year analysis for the Saharan zones.This is encouraging as it suggests that the Sahel correlations described below are notmerely coincidental results arising from an analysis of short time series.

Significant negative correlations at a lag of one year were found for all the Sahelianzones in March, and for the South Sahel in April (Table 1). The strongest apparentrelationships occurred in March over the aggregated Sahel, the North Sahel and the 15°-17° N band. These results suggest that what variability there is in dust productionthroughout the Sahel in March is significantly influenced by the previous year’s rainfall.March falls within the period characterised by large dust loadings and low dustvariability, so the proportion of the March dust production that results from theinfluence of rainfall on the soil-state is likely to be relatively small.

Significant positive correlations occur in October over the aggregated Sahel and theSouth Sahel. This result is difficult to explain. If it is physically meaningful, it may bedue to rainfall-driven soil erosion in one year sensitising the soil to the particulardeflation mechanisms operating in the following October. These mechanisms are likelyto be related to DL activity at the end of the wet-season. Such DLs may be strongenough to mobilise dust but too weak to produce much rainfall. The same deflationmechanisms will operate throughout the wet-season, but removal of dust from theatmosphere by rainfall will result in a weak IDDI signal, masking the relationshipbetween soil-state and dust mobilisation. This conceptual model is likely to beappropriate only in an arid regime where soils are fragile and susceptible to rainfall


erosion. The higher correlation in the wetter South Sahel therefore suggests thatadvection from the more arid northern zones may be responsible for this relationship.This assumes that rainfall variability is coherent between the South Sahel and the morenortherly regions, as the correlation is the result of consideration of South Sahel rainfallonly. This interpretation is highly speculative.

7.2. LAG 2-YEAR RELATIONSHIPS

Significant negative correlations for the 2-year lagged timeseries are observed inDecember over the North Sahel and over the two narrower bands lying within the NorthSahel (Table 2). The North Sahel signal gives rise to a smaller significant negativecorrelation over the aggregated Sahel. These results suggest that rainfall variability has acumulative impact on December soil properties and hence on dust production in thenorthern latitudes of the Sahel, where rainfall is low and where some regions may becharacterised by soil degradation (UNEP, 1992).

A similar relationship is suggested for the North Sahel in September. This mayrepresent the impact of desiccation-related soil degradation on dust production.However, this result is not reflected in the correlations for the 15°-17° N and 18°-20° Nbands. Also of note is the fact that a significant positive correlation is observed over theNorth Sahara for August. This signal results in a smaller significant correlation for theaggregated Sahara. For such short time series, any physical interpretation of this isolatedsignificant Saharan result would be wildly speculative. It is highly plausible that it is aphysically meaningless artefact of the statistics. Hence the isolated September result forthe North Sahel should also be treated with caution. A positive correlation for May overthe North Sahel may reflect a multi-year sensitising of soils to deflation by rainfallerosion (as suggested for the lag 1-year October results), or may also be spurious.


8. Discussion and Conclusions

The IDDI data enable the major dust sources to be identified, and the seasonal evolutionof these sources to be described. Dust sources are identified with regions of sandy desertand regions of seasonal watercourses. The distribution of airborne dust in summer isclosely associated with the position of the surface discontinuity between the monsoonand Saharan air masses, and suggests a role for convective disturbance lines (DLs) insummer dust production. The role of DLs in dust production is further supported by thenorthward migration of the meridional dust maximum over the Sahel in the summermonths.

The monthly zonally averaged IDDI values indicate that dust production in the Sahel-Sahara zone is at a maximum between 15° and 17° N in December, January and June,and between 18° and 20° N from July to September. Dust levels over the Sahara exceedthose over the Sahel in much of the dry Season, and mean zonally averaged dust valuesbetween 20° and 25° N (Sahara) exceed those between 10° and 15° N (Sahel) in allmonths except May and June.

The zonal maximum in dust production is therefore located in Sahelian latitudes onlyduring part of the year, and in the zone of potential land degradation for only threemonths of the year. In June, this maximum is likely to be the result of the balancebetween dust mobilisation and wet deposition as determined by the prevailingmeteorology. In December and January, the strongest IDDI signals occur over theaccepted natural dust sources located in the north-central Sahel. While material fromdisturbed soils may contribute to the dust budget in these months, the notion that suchprocesses have created major new source regions in areas not previously associated withdust production, and extending throughout much of the Sahel, is not supported by thisstudy. It is possible that the maximum in December/January dust production within the15°-17° N zone may be the result of a meridional maximum in the generation ofdeflatable material by water erosion, arising from the balance between rainfall intensity(greater than in more northerly regions) and vegetation cover (less extensive than inmore southerly regions). It should also be noted that mean dust concentrations arerelatively low in December, and that dust activity in northern Africa as a whole isgreater throughout much of the year than in January. This is particularly so in regionsnear the West African coast. It is therefore unlikely that dust production from disturbedsoils in these two months makes a large contribution to the regional and global annualdust budget.

Rainfall-dust correlations indicate that enhanced dust production in April and May isassociated with reduced rainfall in the previous year. Limited evidence for a cumulativeimpact of drought on December dust production is provided by correlations betweenrainfall and IDDI values at a lag of two years. April falls within the period during whichinterannual variability in Sahelian dust concentrations is low and dust events in northernAfrica are frequent. The component of the April dust budget associated with rainfall in


preceding years is therefore likely to represent relatively small percentage changes inthe quantities of dust mobilised.

The results of this study suggest that rainfall does exert some influence on dustproduction during certain parts of the year, although rainfall does not appear to be thedominant factor in determining the amount of dust mobilised in the Sahel on interannualtimescales. The research described here does not support the notion that dust-eventfrequency in the Sahel has increased as a result of widespread land degradation, nor thatthe Sahel has become a more important source of mineral aerosols than the Sahara. Itappears that the role of the land surface (and particularly of human activity) inmodulating atmospheric dust concentrations has been over-emphasised, while too littleattention has been paid to the role of meterological processes in determining theregional dust budget. In particular, observed changes in the nature of summer rain-bearing disturbances may have played a key role in decadal-scale changes in the amountof dust mobilised within, and exported from, Sahelian regions.

Acknowledgements

This work was completed as part of a PhD project funded by the Climatic Research Unitin the form of the Hubert Lamb Studentship, and supervised by Dr Mike Hulme. IDDIdata were obtained from the Laboratoire d’Optique Atmosphérique, Université deSciences et Technologies de Lille. Rainfall data were provided by Mark New at theClimatic Research Unit. Thanks are also extended to an anonymous reviewer for theircomments.

References

Adeyfa, Z. D. and Holmgren, B. (1996) Spectral solar irradiance before and during a Harmattan dust spell,Solar Energy 57 (3), 195-203.

Arino O., Melinotte, J-M. Rosaz, J. M. and Monjoux, E. (1997) ESA Fire Product, Proceedings of the 7thISPRS conference on Physical Measurement and Signatures in Remote Sensing, 7-11 April, Courchevel.

Arino O. and Melinotte, J-M. (1995) Fire Index Atlas, Earth Observation Quarterly 50, T.D. Guyenne (ed.),ESA Publications Division, ESA/ESTEC, Keplerplaan 12200 AG, Noordwijk, The Netherlands.

Artaxo, P., Gerab, F., Yamasoe, M. A. and Martins, J. V. (1994) Fine mode aerosols compositions at 3 long-term atmospheric monitoring sites in the Amazon Basin, Journal of Geophysical Research - Atmospheres99 D11, 22857-22868.

Baird, A. J. (1997) Overland flow generation and sediment mobilisation by water, in Thomas, D. S. G. (Ed.)Arid zone geomorphology: Process, form and change in drylands, 2nd Edition, John Wiley and Sons Ltd,165-184.

Barrey, R. G. and Chorley, R. J. (1995) Atmosphere, weather and Climate, 6th Edition, Routledge, p. 259.Bullard, J. E. (1997) Vegetation and geomorphology, in Thomas, D. S. G. (Ed.) Arid zone geomorphology:

Process, form and change in drylands, 2nd Edition, John Wiley and Sons Ltd, 109-131.Chiapello, I., Bergametti, G., Gomes, L., Chatenet, B., Dulac, F., Pimenta, J., Santos Suares, E. (1995). An

additional low layer transport of Sahelian and Saharan dust over the North-Eastern Tropical Atlantic,Geophysical Research Letters 22, 3191-3194.

Coudé-Gaussen, G., Rognon, P., Bergametti, G., Gomes, L., Strauss, B, Gros, J. M., Le Coustumer, M. N.(1987) Saharan dust on Fuertaventura Island: Chemical and mineralogical characteristics, air masstrajectories, and probable sources, Journal of Geophysical Research 92, 9753-9771.

25

Drees, L. R., Manu, A. and Wilding, L. P. (1993) Characteristics of aeolian dusts in Niger, West Africa,Geoderma 59, 213-233.

Dubief, J. (1979) Review of the North African climate with particular emphasis on the production of eoliandust in the Sahel zone and in the Sahara, in Morales, C. (Ed.) Saharan Dust, John Wiley and Sons Ltd.

Hastenrath, S. (1991) Climate Dynamics of the Tropics, Kluwer Academic Publishers, Dordrecht.Hess, T., Stephens, W. and Thomas, G. (1996) Modelling NDVI from decadal rainfall data in the north east

arid zone of Nigeria, Journal of Environmental Management 48, 249-261.Husar, R. B., Prospero, J. M. and Stowe, L. L. (1997) Characterisation of tropospheric aerosols over the

oceans with the NOAA advanced very high resolution radiometer optical thickness operational product,Journal of Geophysical Research 102 D14, 16,889-16909.

Kalu, A. E. (1979) The African dust plume: its characteristics and propagation across West Africa in winter, inSaharan Dust: mobilisation, transport, deposition: papers and recommendations from a workshop held inGothenburg, Sweden, 25-28 April 1977, C. Morales (ed.), Wiley.

Lamb, P. J., Bell, M. A. Finch J. D. (1998) Variability of Sahelian disturbance lines and rainfall during 1951-1987, Water Resources Variability in Africa during the XXth Century (Proceedings of the Abidjan ’98Conference held at Abidjan, Cote d’lvoire, November 1998). IAHS Publications No. 252.

Legrand, M., N'Doume, C. and Jankowiak, I. (1994) Satellite-derived climatology of the Saharan aerosol,Passive Infrared remote sensing of clouds and the atmosphere II, D. K. Lynch (ed.), Proc. SPIE 2309,127-135.

McTainsh, G. H. (1980) Harmattan dust deposition in northern Nigeria, Nature 286, 587-588.McTainsh, G. (1996) Dust concentrations and particle-size characteristics of an intense dust haze event: inland

delta region, Mali, West Africa, Atmospheric Environment 30, 1081-1090.Maley, J. (1982) Dust, clouds, rain types, and climatic variations in tropical North Africa, Quaternary

Research 18, 1-16.Middleton, N. J. (1985) Effect of drought on dust production in the Sahel, Nature 316, 431 -434.Middleton, N. J. (1997) Desert dust, in Thomas, D. S. G. (Ed.) Arid Zone Geomorphology, Wiley.New, M. G., Hulme, M. and Jones, P. D. (1999) Representing 20th century space-time climate variability. II:

Development of a 1901-1996 monthly terrestrial climate fields. Journal of Climate, in press.N’Tchayi, G. M., Bertrand, J., Legrand, M. and Baudet, J. (1994) Temporal and spatial variations of the

atmospheric dust loading throughout West Africa over the last thirty years, Annales Geophysicae 12, 265-273.

N’Tchayi Mbourou, G., Bertrand, J. J., Nicholson, S. (1997) The diurnal and seasonal cycles of wind-bornedust over Africa north of the equator, Journal of Applied Meteorology 36, 868-882.

Prospero, J. M. and Nees, R. T. (1986) The Impact of the North African Drought and El-Niño on Mineral Dustin the Barbados Trade Winds, Nature 320, 735-738.

Reiff, J. Forbes, S., Spieksma, T.Th. M., Reynders, J. J. (1986) African dust reaching northwestern Europe: Acase study to verify trajectory calculations, Journal of Climate and Applied Meteorology 25, 1543-1567.

Rowell, D. P. and Milford, J. R. (1993) On the generation of African squall lines, Journal of Climate 6, 1181-1193.

Sirocko, F., Sarnthein. M., Lange, H. and Erlenkeuser, H. (1991) Atmospheric summer circulation and coastalupwelling in the Arabian Sea during the Holocene and the last Glaciation, Quaternary Research 36, 72-93.

Tegen, I. and Fung, I. (1995) Contribution to the atmospheric mineral aerosol load from land surfacemodification, Journal of Geophysical Research, 100, 18,707-18,726.

Tegen, I., Lacis, A. A. and Fung, I. (1996) The influence on climate forcing of mineral aerosols fromdisturbed soils, Nature 380, 419-422.

Tetzlaff, G. and Peters, M. (1988) A composite study of early summer squall lines and their environment overWest Africa, Meteorology and Atmospheric Physics 38, 153-163.

UNEP (1992) World Atlas of Desertification, N. Middleton and D. S. G. Thomas (Eds.), Edward Arnold,London.

Warren, A. (1996) Desertification, in Adams, W. M., Goudie, A. S. and Orme, A. R. (Eds.) The PhysicalGeography of Africa, pp 343-355.

Williams, M. A. J. and Balling, R. C. (1996) Interactions of Desertification and Climate, WMO, UNEP,Arnold, London, pp 25-28.

WMO (1976) Special environmental report No. 9: An evaluation of climate and water resources fordevelopment of agriculture in the Sudano-Sahelian zone of West Africa, WMO – No. 459.

DUST VARIABILITY OVER NORTHERN AFRICA

DESICCATION IN THE SAHEL

C. T. AGNEW

Department of Geography, UniversityCollege London, 26 Bedford Way,London WC1H OAP UK([email protected])

A. CHAPPELL

Telford Institute of Environmental Systems,Department of Geography, Universityof Salford, Manchester, M5 4WT UK([email protected])

Abstract

The Sahel region of West Africa is well known as a region of environmental degradation.The reported incidence of drought and desertification has been challenged but regionaldesiccation is still widely accepted. This paper investigates the evidence for regionaldesiccation and in particular the effect of aggregating rainfall statistics across the area.Regression analysis reveals that the recent regional downward trend in rainfall is notreproduced at all stations at the 1% level of significance but is significant when data isaggregated. Geostatistical methods were used to investigate the spatial variability of rainfall.The results suggest that changes in the raingauge network since 1945 rather than climate maybe influencing regional rainfall statistics. It was found that the distribution of raingaugesbetween 1945 and 1975 was not adequate to sample latitudinal changes in rainfall and thatthe annual rainfall for the region was largely a product of poor sampling east to west untilsufficient stations were reporting data from 1970. These results raise questions over the useof regional statistics such as rainfall anomalies and the fitting of regional trend lines to depictclimate change in the Sahel. Geostatistical methods offer a more complex but more reliableapproach to the estimate of regional rainfall characteristics.

27


28 DESICCATION IN THE SAHEL

1. Introduction

The West African Sahel is well known due to reports on drought, desertification and faminethat span at least three decades (Figure 1). Evidence of a change in the region’s climate isusually portrayed by a standardised rainfall anomaly plot (Figure 2) which displays almostcontinuous negative anomalies since the 1960s where,

AGNEW AND CHAPPELL 29

This has been noted for some time with Lamb (1974), Nicholson (1979) and Winstanley(1973) writing about a downward trend in Sahelian rainfall after international attentionfocused on the drought of the early 1970s. A decade later Copans (1983), Druyan, (1989),Flohn (1987) and Tickell (1986), supported the view of lower than average rainfall and someeven wrote about a persistent drought. The notion of a desiccating environment continuesinto the 1990s with some workers linking this to an advancing Sahara and claims thatdesertification is affecting the region (Hulme and Kelly, 1993; Nicholson and Palao, 1993;

is the standardised rainfall anomaly.is the station rainfall for the ith station and kth year.is the time mean of the ith station.is the standard deviation of ith precipitation station

(after Jones and Hulme, 1996).


UNEP, 1992; Zheng and Eltahir, 1998) with statements such as ‘The desert is advancingpartly because of recurring cycles of drought’ (Pritchard, 1990 p.246); ‘After a 20 year seriesof droughts, the Sudano-Sahelian region remained the most permanently vulnerable area’(Odingo, 1992, p.6); ‘the prolonged drought that has struck the Sahel for 25 years now.’(D’Amato and Lebel, 1998 p.956).

A major problem with the above reports is that they fail to distinguish between desiccation,drought or desertification as defined in Table 1. It is clear that in order to adopt the mostappropriate response it is necessary to distinguish between these different types ofenvironmental degradation. The aim of this paper is then to answer the question; is the Sahelclimate desiccating?

2. Why Investigate Desiccation?

It may, at first, seem a waste of effort to investigate desiccation in the Sahel given theenormous evidence in support of persistent drought and a downward trend in rainfalls. Yetthere are several reasons why the question is pertinent: refutation of the idea of an advancingdesert; challenges to the notion of persistent drought; changes to the base period over whichclimate change is measured; changes to the number of climate stations available for analysis;reported increase in the variability of the data.

The notion of land degradation in the Sahel and an advancing Sahara has been challenged,both in scientific and popular publications (Agnew, 1995; Agnew and Warren, 1996; Binns,1995; Mainguet, 1991; Thomas, 1993). The notion of persistent drought has also beenquestioned for some time (Agnew 1989, 1990; Franke and Chasin, 1980; Garcia, 1981;


Glantz, 1987, 1994; Wijkman and Timberlake, 1985). High rainfalls have been evident inthe Sahel during the period of desiccation e.g., 1988 and 1998, yet there has been littlecritical examination of the notion of widespread desiccation. In addition, the thirty year baseperiod upon which climate change is determined has recently (Hulme, 1992) been changedby the WMO from 1931-1960 to 1961-1990 with uncertain impacts upon the determinationof rainfall trends. Hulme (1992) reported that rainfall is becoming more variable in the Saheland the downward trend more persistent. He also noted that the number of rainfall stationsavailable is declining. Ba et al. (1995) noted in their analysis of satellite-determined rainfallsin the Sahel, that the number of stations available between 1983 and 1988 fell from 271 to147 (a 46% reduction). A trend that UNEP (1992) noted starting in the late 1970s andcontinuing through to 1990. But perhaps of greater importance is the variability in thenetwork of Sahelian rainfall stations ‘....that are unevenly distributed in space, sparse incritical regions, and/or reported irregularly...it is often impossible to obtain a sufficientrainfall dataset over wide areas from a conventional rain gauge network.’ (Ba et al, 1995 p.412).

Hence, there is some concern over the reliability of the rainfall network employed in theSahel to assess climate change while the impact of changing the number (and location) ofrainfall stations used to determine rainfall trends in the Sahel is uncertain. Doubt can also bedirected to the strategy of focusing on spatially aggregated rainfall statistics for all, if notmajor parts, of the Sahel that ignore local variations. This paper then seeks to investigatecritically the evidence for widespread desiccation during changes in the climatic base periodand aggregation of trends from different sets of stations each year. We first look at thetemporal rainfall patterns through an examination of recent trends in the Sahel beforeundertaking a geostatistical analysis of its spatial variation.

3. Results and Discussion

3.1 RAINFALL TRENDS: IS THE CLIMATE IN THE SAHEL DESICCATING?

The analysis is based on data provided by the Climate Research Unit of the University ofEast Anglia. There have been several attempts to define the Sahelian region ranging fromclimate (Sivakumar 1989) to ecological conditions (Davy et al. 1976). As we are primarilyinterested in rainfall we can use the results of previous work that shows Continental Sahel(Niger, Burkina Faso and Mali east of 5°W) can be viewed as a climatic region (Ba et al. ,1995 and Moron, 1994).

A regression line fitted to the standardised rainfall anomalies displayed in figure 2 for theyears 1961 to 1994 produces a strong negative relationship between rainfall and time


0.73) such that rainfall decreases by 0.069 of a standardised anomaly per annum. This isequivalent to a reduction in rainfall of 8 mm each year or 244 mm between 1961 and 1990,averaged across Continental Sahel. Given that the mean rainfall for these stations and for thisperiod is 511mm this is a massive reduction in annual precipitation. Hulme (1996)examined reports of desiccation in the world’s drylands and found little global evidence oflong term drying except for the Sahel. The Sahel showed a significant downward trend of96.8mm per century, equivalent to a decline of around 1mm a year in annual rainfall, (basedon the whole of the Sahel with an annual average rainfall of 451mm). This is much less thanwe have reported, even if the desiccation is assumed to only take place over the last 30 yearshence producing a reduction of 3mm/year. Nicholson and Palao (1993) note the downwardtrend started early in the 1950s. They separated West Africa into nineteen regions andcalculated for each, the standardised departures of rainfall from a long-term mean between1950 and 1985. A regression line was fitted to the standardised departures for each region(Table 2). They found the downward trend was most pronounced in the wetter south, less indrier and western areas. Rainfall for Continental Sahel was also found to decrease by 0.055of a standardised anomaly per annum (Nicholson and Palao, 1993) which is close to the valueof 0.069 we reported above for a more recent time period.

Thus, there is some general agreement over the amount of desiccation that has taken place inrecent decades. However, figure 3 (and table 2) suggest that this general figure of 0.069 of astandardised anomaly each year may be misleading. Higher values are found for Mali andBurkina Faso (0.087) but much lower for Niger (0.037). It is also evident from figure 3 thatin fitting the trend line it is being strongly influenced by the very high rainfalls in the early1960s and very low rainfalls in recent years. In between these periods there is much variation,especially for Niger where the correlation coefficient is only 0.255 but significant at the1% level. The variation raises concern over the analysis of aggregated rainfall anomaliesrather than aggregated trends. The former assumes rainfall in Continental Sahel is spatiallyhomogeneous. It is also uncertain whether extreme values are unduly influencing the fittedregression lines and misleading the desiccation trend. These two points are discussed below.


3.2 RAINFALL TREND LINES

The analysis here focuses upon the rainfall stations in Niger as the data collected havealready been shown to behave in a peculiar fashion. The persistent downward trend evidentin the data aggregated for Continental Sahel (Mean) is not present in the data aggregated forNiger in figure 2. This difference is clearer in figure 4 where low rainfall during the early1970s and mid 1980s should be contrasted with the large rainfalls in the early 1960s, late1970s and 1994. Although the correlation coefficient of the relationship betweenstandardised anomaly and time is small it remains significant at the 1 % level.

To predict the desiccation for Niger between 1961 and 1990 (latest WMO period) regressionlines were fitted to the standardised rainfall anomaly for each station and for groups ofstations over this period. The reduction in rainfall equivalent to the decrease in standardisedanomaly produced by the regression lines is shown in figure 5. The stations are arranged inascending order of annual rainfall (Bilma is lowest with 12 mm, Gaya is highest with796mm). The stations are also grouped based on annual totals of less than 250mm; between250 and 500mm and above 500mm (used by Agnew, 1990 to identify pastoral; agro-pastoraland rainfed agriculture regions). There is a mean desiccation during this period of 150 mm ofannual rainfall but this varies considerably. Surprisingly the variation is not simply based onthe mean annual rainfall; Maine-Soroa (annual rainfall of 342mm) experiences a predicteddrying of 237mm, equivalent to that of Maradi (annual rainfall 493mm). Whereas, Gaya(796mm) has a predicted drying of 124mm which is the same as that predicted for Zinder(annual rainfall 369mm). Notably, when the rainfall stations are grouped according toAgnew’s (1990) classification the predicted rainfall decrease does vary according to meanannual rainfall.

When rainfall stations are aggregated into broad groups the trend lines for all stationsare found to be significant as shown in figure 6. We have plotted the regression F statistic asthe analyses all have the same degrees of freedom so the results are directly comparableirrespective of the annual amount of rainfall or magnitude of the rainfall anomalies. Incontrast the results for individual stations contain a high degree of variability. Some stationse.g., Tahoua and Maine-Sora display a highly significant pattern of desiccation while atothers e.g., Gaya and N’Guigmi rainfall does not fit into a linear drying trend. This isdemonstrated in figures 7 and 8. The key point here is that the aggregation of rainfallanomalies conceals those stations that do not display any rainfall trend. The suspicion is thatthose stations that contain extreme anomalies are masking other stations where there is noclear linear trend. This can be illustrated by plotting the predicted amount of drying from thedesiccation trend line against the difference between the highest and the lowest observedrainfall value (figure 9). The result is significant at the 1% level of significance. Ifstations with a weak trend line, (N’Guigmii, Zinder, and Gaya) are excluded then the


relationship between desiccation trend and rainfall maximum minus rainfall minimum is evenstronger

In a semi-arid region of highly variable rainfall it does seem unwise to accept a generalstatement of regional desiccation when this is based on a few extreme observations.Furthermore, when these extreme values occur at the start (early 1960s) and end (late 1990s)points of the period it can be no surprise that a linear trend line can often be fitted to the data.This raises the question as to whether the period 1961 to 1990 is appropriate for such ananalysis of desiccation in the Sahel.


No significant trend is observed for the rainfall stations with the longest run of data, Niameyand Zinder (from 1905) and Tahoua (from 1922) plotted in figure 10. The abnormally highrainfalls of the 1950s and early 1960s are then biasing perceptions of climate change in theSahel. In addition, the affect on aggregation of changes in the location and number of rainfallevents and changes to the location and number of stations, is largely unknown and will bedealt with next.

3.3 DOES AGGREGATION MISLEAD?

We have demonstrated that the spatial aggregation of rainfall anomalies can lead to theerroneous conclusion that all parts of the Sahel have experienced similar rates of recentdesiccation. There is then a need to understand the spatial variability of the Sahelian rainfalland the impact of changes to the location and number of stations used to aggregate climatestatistics for the Sahel. Ba et al. (1995), using Meteosat data to predict seasonal rainfalls,found a marked difference between the position of the isohyets when comparing (a) onlyMeteosat derived rainfalls based upon the terrestrial raingauge network to (b) Meteosatderived rainfalls based on full regional coverage. The implication was that the number anddistribution of raingauges influenced the estimation of rainfall totals. The suggestion is thatareal estimates of rainfall are unreliable and simply taking the mean of all stations biases the


wetter south where there are more stations. This suggestion is tested by geostatisticalanalyses of total boreal summer (June, July, August and September) rainfall (TBSR) for theWest African Sahel (10-20°N and 20°W to 20°E) between 1931 and 1990. A summary of theresults is presented here, further details of the analysis can be found in can be found inChappell and Agnew (in review).

Intuitively reasonable trend due to latitudinal variation in the annual TBSR data was removedby fitting quadratic polynomials on their spatial co-ordinates using least squares regression(Chappell et al., 1996). Ordinary experimental variograms were computed for the residualsfrom the trend of TBSR every year between 1931 and 1990 in the two principal directions(N-S and E-W) of spatial variation. These variograms were fitted with models, (see Chappelland Agnew, in review for further details), using weighted least squares in Genstat (Genstat 5Committee, 1992), which all included a nugget variance, typical in sparsely sampledcontinuous data (Chappell and Oliver, 1997). Although measurement error and stochasticvariation in data contribute to the nugget, the largest source of variation is commonly due tospatially dependent variation that occurs over distances much smaller than the shortestsampling interval (Webster and Oliver, 1990). The majority of the variograms were boundedand included sill variance and range parameters. The dissimilarity between TBSR foraverage separation distances between rainfall stations increases until it reaches a maximumknown as the sill variance. The lag separation distance at which the variogram reaches its sillis the range; this is the limit of spatial dependence (Webster and Oliver, 1990). Beyond thislimit the variance bears no relation to the separation distance. Some of the variograms areunbounded (only linear models were appropriate and the model parameter includes thegradient) and have a structure, which appears to increase indefinitely at this scale ofinvestigation. This suggests that as the area of interest increases, so more sources of variationare encountered (Chappell et al., 1996, Webster and Oliver, 1990) i.e., TBSR remains spatiallydependent with increasing distance.

The model parameters of the fitted variograms are plotted for each year in figures 11 – 13. Theparameters of the variograms were used for anisotropic punctual kriging to estimate TBSRresiduals on a 1 degree grid across the West African Sahel. The quadratic polynomials for eachyear were added back to the kriging estimates. Finally, isohyets were threaded through thekriging estimates of TBSR (with the same isoline frequency) and plotted for every year between1931 and 1990. Figure 14a, b, c, d, e, f and g are approximately decadal examples of the rainfallmaps produced.


The magnitude and variability of the E-W nugget variance (Figure 13a) decreases over timesuggesting that the spatial dependence of rainfall is better sampled over time. Since the E-Wrange (figure 12) is not decreasing over this period it suggests that the structure of the rainfallduring this time is not varying. Thus, the reduction in spatial dependence (E-W nugget variance)is due to improved sampling of rainfall by the station network. The station locations for selectedyears (figures 14a, b, c, d, e, f, g) show an extension into the easterly end of the region. An E-Wgradient in rainfall exists as shown by the isohyets threaded through the kriging estimates oftotal summer rainfall for several years. The isohyets are more compressed in the west of theregion than in the east. The importance of this eastwards extension of rainfall stations is thatover time more rainfall stations are located in the area of lower rainfall. Thus, the spatialvariation in rainfall throughout the region is increasingly better sampled.


The inter-annual variation in the N-S nugget variance (figure 13b) is considerably larger thanthe E-W nugget variance and the general trend in the former is very different from the latter.The nugget variance before ca. 1945 and after ca. 1970 is generally smaller than that shown inthe period between these dates. This suggests that the spatial dependence of rainfall is bettersampled in the early and late periods of the rainfall record and that between ca. 1945 to 1975 theconfiguration of rainfall stations in the N-S direction has poorly sampled the spatial dependenceof rainfall. The reason is not evident in the selection of maps showing the rainfall stations(Figure 14) because the inter-annual variation of the nugget variance is very large. Moreover,an examination of all maps suggests that there has been very little N-S variation in the locationof rainfall stations. The most likely explanation is that the N-S configuration of the rainfallstations was inadequate to sample the generally large spatial dependence (range; figure 12)during the period between ca. 1945 to 1975. This would have been obvious if more of the N-Svariograms had been fitted with bounded models. That they were not is evidence itself forincreasing sources of variation with increasing separation distance. However, the inter-annualvariation in the N-S nugget variance is larger than that in the E-W direction because thevariation in the range of spatial dependence as a ratio of the total distance is much larger in theN-S direction. This interpretation is complicated by the difficulty of modelling the variograms inthis direction where fewer pairs are available than in the E-W direction. The isohyets threadedthrough the kriging estimates of total summer rainfall for several years (Figure 14a, b, c, d, e, f,g) show the expected anisotropic variation, whereby the rainfall gradient is greater in the N-Sdirection than in the E-W direction. The N-S rainfall gradient is greater within the period 1945to 1970 (Figure 14c, d, e) than outside this period (Figure 14a, b, f, g) as evident from theisohyet compression in the south of the region.

It is no coincidence that the pattern of N-S nugget variance (Fig. 13b) bears a strikingresemblance to the pattern of average annual rainfall aggregated for the region (Fig. 2). Thedifferent sources contributing to the nugget cannot individually be quantified. However, it ishighly likely that measurement error is small and that randomness is not large due to climaticcontrol. Most other geostatistical applications have shown that where measurement error can beruled out and randomness assumed small the main source of the nugget is the scale and intensityof sampling. The distribution of rainfall stations over the period ca. 1945 and 1975 has notadequately sampled the N-S variation in rainfall. The average rainfall during this period iscontrolled by the predominance of rainfall stations in the more southerly locations associatedwith larger rainfall. The average annual rainfall for the region is also a product of the poor E-Wsampling configuration early in the rainfall record and the general increase in the effectivenessof sampling more recently. It appears that the only period when the spatial dependence ofrainfall has been adequately sampled by the rainfall station network (i.e., the nugget variance inthe N-S and E-W directions have been at their smallest) is since 1970.


Ironically, this is the period when many workers have reported a persistent downward trend inthe average annual rainfall. However, the results here suggest that this trend is a return to amore precise estimate of the rainfall in this region. Other patterns are artefacts of the use ofspatial aggregation which depends on the relationship between rainfall station location andrainfall spatial dependence.

4. Conclusions

There is overwhelming evidence in research publications that the Sahel, as a region, isdesiccating. Most reports place the start of this change in climate in the 1960s though somesuggest a decade earlier. Examination of standardised rainfall anomalies for ContinentalSahel supports this view with annual rainfalls declining through the 1970s and 1980s. Therainfall averages in 1961 to 1990 are significantly lower than during 1931 to 1960. A morelocalised analysis of rainfall trends has shown that this regional aggregation can mask localvariations. A downward trend over the last 30 years is not significant at all stations and Nigerin particular displays different patterns compared to Mali and Burkina Faso. A downwardtrend is also not evident over a twentieth century long perspective giving rise to concern thatthe abnormally high rainfalls of the 1950s and 1960s are leading us to believe that the 1970sand 1980s are abnormally low.

The danger with all of these points is that they merely illustrate the variability of the data set.The results of a geostatistical analysis show that patterns in the mean annual rainfall are anartefact of the rainfall station location and rainfall spatial dependence. The recent downturnin mean annual rainfall appears to be a return to a more precise estimate as a consequence ofimproved station sampling of the rainfall distribution. The analysis supports the suggestionthat areal estimates of rainfall are unreliable and the simply taking the mean of all stationsbiases the estimate (Ba et al., 1995). However it is too simplistic to assume that the bias isdue to the location of more stations in the wetter south. The geostatistical analysis has shownthat the temporal variation in location (N-S and E-W) of the rainfall station network has aconsiderable effect on the sampling distribution of rainfall across the region.

Even if we accept that conditions in many parts of the Sahel are correctly represented by theclaims of widespread drought and a persistent downward trend in rainfalls, this fails toexplain the impact upon people living in this region. Despite alarm that the environment isdegrading food production has increased over the last 30 years (Agnew, 1995). Livestocknumbers have been affected by the droughts of the 1970s and 1980s but numbers quickly re-established when rainfalls recovered (IUCN, 1989). Claims that the Sahara is advancing havebeen refuted (Thomas, 1993) and poor food supply has been explained by factors such as


price, distribution and the role of institutions rather than climate (Norse, 1994 and Olsson,1993). Those who state that the Sahel has recently desiccated but then ignore what this meansfor the inhabitants and the rest of the physical environment of the region are guilty ofoversimplifying rainfall patterns and of assuming the landuse systems in the Sahel are basedupon the higher rainfalls experienced in the middle of this Century. Taking only a 30 yearperiod to establish climate change is then unwise for the Sahelian region and any statistic thatclaims to represent Sahelian conditions as a whole should be treated with caution.

Acknowledgements

The monthly rainfall totals were generously provided by the Climate Research Unit at theUniversity of East Anglia.

References

Agnew, C. T. (1989) Sahel drought, meteorological or agricultural? International Journal of Climatology 9, 371 -382.

Agnew, C. T. (1990) Spatial aspects of drought in the Sahel. Journal of Arid Environments 18, 279-293.Agnew, C. T. (1995) Desertification, drought and development in the Sahel, in Binns, A. (ed.) People and

environment in Africa. J. Wiley and Sons, Chichester p137-149.Agnew, C. T. and Warren, A. (1996) A framework for tackling drought and land degradation. Journal of Arid

Environments 33, 309-320.Ba, M. B., Frouin, R. and Nicholson, S. E. (1995) Satellite derived interannual variability of West African Rainfall

during 1983-88. Journal of Applied Meteorology 34, 411 -431.Binns. T. (1990) Is desertification a myth? Geography 75, 106-113.Chappell, A. and Oliver, M. A., (1997) Geostatistical analysis of soil redistribution in SW Niger, West Africa, in E.

Y. Baafi and N. A. Schofield (eds.) Quantitative Geology and Geostatistics, Kluwer, Dordrecht. pp. 961-972Chappell, A. Oliver, M. A. Warren, A. Agnew, C. T. and Chariton, M. (1996) Examining the factors controlling the

spatial scale of variation in soil redistribution processes from south-west Niger. In M. G. Anderson and S. M.Brooks (eds.) Advances in Hillslope Processes. J. Wiley and Sons, Chichester. pp 429-449

Copans, J. (1983) The Sahelian drought, in Hewitt, K. (ed.) Interpretations of calamity. Allen and Unwin, London.pp 83-97

D'Amato, N. and Lebel, T. (1998) On the characteristics of the rainfall events in the Sahel with a view to theanalysis of climatic variability. International Journal of Climatology 18, 955-974

Davy, E. G., Mattei, F and Solomon, S. I. (1976) An evaluation of the climate and water resources for developmentof agriculture in the Sudan-Sahelian zone of West Africa. WMO Special Environmental Report No.9. WMO,Geneva.

Druyan, L. M. (1989) Advances in the study of sub-saharan drought. International Journal of Climatology 9, 77-90.

Flohn, H. (1987) Rainfall teleconnections in northern and eastern Africa. Theoretical & Applied Climatology 38,191-7

Franke, F. and Chasin, B. (1980) Seeds of famine. Allanheld, Osman and Co., New Jersey.Garcia, R. V. (1981) Drought and Man: Volume 1: Nature Pleads Not Guilty. Pergamon, Oxford.Glantz, M. (ed.) (1987) Drought and hunger in Africa. Cambridge University Press. Cambridge.


Glantz, M. (ed) (1994) Drought follows the plow. Cambridge University Press, Cambridge.Hulme, M. (1992) Rainfall changes in Africa. International Journal of Climatology 12, 685-699Hulme, M. (1996) Recent climatic change in the world's drylands. Geophysical Research Letters, 23 (1), 61-64Hulme, M. and Kelly, M. (1993) Desertification and climate change. Environment 35 (6), 39-45, International

Union for Conservation of Nature 1989 Sahel Studies. IUCN, Nairobi.Jones, P. D. and Hulme, M. (1996) Calculating regional climatic time series for temperature and precipitation:

methods and illustrations. International Journal of Climatology 16, 361-377Lamb, H. H. (1974) The Earth's changing climate. Ecologist 4, 10-15Mainguet, M, (1991) Desertification: natural background and human mismanagement. Springer-Verlag, Berlin.Moron, V. (1994) Guinean and Saharan rainfall anomaly indices at annual and monthly time scales (1933-1990).

International Journal of Climatology 14, 325-341.Nicholson, S. E. (1979) Revised rainfall series for the West African subtropics. Monthly Weather Review 107, 620-

23.Nicholson, S. E. and Palao, I. M. (1993) A re-evaluation of rainfall variability in the Sahel. International Journal

Climatology 13, 371-389.Norse, D. (1994) Multiple threats to regional food production, environment, economy, population. Food Policy 19

(2), 113-148Odingo, R. S. (1992) Implementation of the plan of action to combat desertification (PACD) 1987-1991

Desertification Control Bulletin 21, 6-14.Olsson, L. (1993) On the causes of famine - drought, desertification and market failure in the Sudan. Ambio 22 (6),

395-403Pritchard, J. M. (1990) Africa. Longman, Harlow.Sivakumar, M. V. K. (1989) Agroclimatic aspects of rainfed agriculture in the Sudano-Sahelina zone, in

Proceedings of Workshop on Soil, crop and water management systems for rainfed agriculture in the Sudano-Sahelian zone, Niamey January 1987. ICRISAT Sahelian Center, Niamey. ICRISAT, Patancheru, AP 502 324India, pp 17-38

Thomas, D. G. (1993) Sandstorm in a teacup I Understanding desertification. Geographical Journal 159 (3), 318-331.

Tickell, C. (1986) Drought in Africa: impact and response. Overseas Development 102,United Nations Environment Programme (1992) World Atlas of Desertification. Edward Arnold, London.Warren, A. and Khogali, M. (1992) Assessment of desertification and drought in the Sudan-Sahelian region 1985-

1991 UNSO, New York.Webster, R. and Oliver, M. A. (1990). Statistical methods in soil and land resource survey. Oxford Univ. Press.Wijkman, A. and Timberlake, L. (1985) Is the African drought an act of God or of man ? Ecologist 15 (112), 9-18.Winstanley, D. W. (1973) Rainfall patterns and general atmospheric circulation. Nature 245, 190-194Zheng, X and Eltahir, E. A. B. (1998) The role of vegetation in the dynamics of West African monsoons. Journal of

Climate 11, 2078-2096.

HYDROLOGICAL RESPONSE OF DESERTMARGINS TO CLIMATIC CHANGE:THE EFFECT OF CHANGING SURFACEPROPERTIES

A. YAIRDepartment of Geography, The Hebrew University,Jerusalem, Israel.

R. B. BRYANFaculty of Forestry, The University of Toronto,Toronto, Canada.

Abstract

Arid and Semi-arid ecosystems are regarded by ecologists as highly resistant to stressdue to their adaptation to the extreme variability in the climatic conditions over a timescale of decades. Under such conditions a rather extreme change in climate, mainlyrainfall, would be required in order seriously to affect natural semi-arid and aridenvironments. The above approach disregards the fact that one of the forms of land-surface change that may result from climatic change in deserts, and especially at a desertfringe, is not limited to purely climatic variables such as precipitation and temperature.It is always accompanied by quite rapid alteration of surface properties, connected todeposition of loess or sand. In subtropical semi-arid and arid areas loess deposition, at agiven site, is often attributed to wet periods; while sand deposition to dry periods. Thenew surface properties can be expected to exercise strong influence on infiltration,runoff and soil moisture. An aspect not yet answered is how much sand or loessdeposition is required to affect the hydrological regime and related water resources. Inorder to check the effect of thin topsoil sand or fine-grained layers on infiltration andrunoff sprinkling experiments were conducted in the laboratory, at various rainintensities and duration. Data obtained show that a slight change in surface propertieshas a rapid and significant hydrological effect. A sand layer 1-2 cm thick is enough toeliminate runoff generation; whereas a fine-grained layer 1-2 mm thick has an oppositeeffect, significantly increasing runoff generation. One can therefore conclude that aridand semi-arid environments, although highly adapted to extreme variations in rainfall,may be extremely sensitive to slight changes in their surface properties, which alter theirhydrological regime quickly and efficiently.

49


50 HYDROLOGICAL RESPONSE OF DESERT MARGINS

1. Introduction

The term resilience is often used by ecologists (Holling, 1973) to describe the degree towhich an ecosystem can be disturbed and yet return to its previous composition andstructure. A system disturbed beyond its resilience will develop into a new ecosystemcharacterised by an altered composition and structure. Such drastic change can betriggered by human activity. For example, the introduction of grazing into a grasslandarea is often responsible for the replacement of a grass cover by a shrubland. The sameresult has been predicted for semiarid areas due to warmer and drier climatic conditions(Schlesinger et al., 1990). However, semi-arid and especially arid ecosystems are alsoregarded by ecologists as highly resistant to stress (Holling, 1973; Thiery, 1982; Wiens,1985). These ecosystems are adapted to the extreme variability in the climatic conditionsin the rainy season, from year to year and over a time scale of decades.

The high resilience of arid environments is well demonstrated by the fact that somerocky mountainous areas within extreme deserts (such as the Negev and Sinai) includeup to 30% of Mediterranean and Irano-Turanian species (Yair and Danin, 1980). Thevery existence of such species in an area with less than 100 mm average annual rainfallclearly proves that enough water is provided to such plants even during a sequence ofdry years. Furthermore, Shmida (1982) reports the occurrence of endemicMediterranean species in the Negev and Sinai deserts where present day average annualrainfall is 70-100 mm. As the development of endemic species requires a long period ofisolation, such occurrence can be considered as indicative of stable conditions overhundreds or even thousands of years. To summarise, as stated by Thiery (1982) “speciesadapted to highly variable environmental conditions, and a high rate of mortality, aremore likely to tolerate an extreme stress than are species from very constantenvironments”. In this situation a rather extreme change in climate, mainly in rainfall,would be required in order seriously to affect natural semi-arid environments.

The above approach disregards the fact that climatic change in subtropical deserts, andespecially at a desert fringe, is never limited to climatic and environmental variablessuch as precipitation, temperature, vegetation cover and soil properties. It is almostalways accompanied by quite rapid alteration of surface properties. The new surfaceproperties exercise strong influence on infiltration, runoff, soil moisture and thus on thevegetal cover (Yair, 1983; 1994; Yair and Danin, 1980). Scientists working in theSahara (Coude-Gaussen, 1991) and in the Negev desert (Goldberg, 1981; Yaalon andDan, 1974; Goring-Morris and Goldberg, 1990) tend to agree that one type of systemchange is related to aeolian deposition. According to these authors, loess deposition inthe subtropical desert fringe, took place during relatively wet periods, while sanddeposition is characteristic of dry climatic periods. Studying the environmental effectsof loess penetration into the Northern Negev, Yair (1983; 1994), Yair and Shachak(1985) and Kadmon et al., (1995) showed that although loess penetration occurredduring a wet period, it resulted in an increase in salt input (by rainfall and dust) coupledwith a limited leaching depth. This scenario led to soil salinisation and desertificationprocesses. An opposite trend occurred during the following dry period. The negative

YAIR AND BRYAN 51

effect of rainfall decrease was counteracted by sand penetration that allowed deeprainwater infiltration, deep leaching and good water preservation.

The effects described above reflect the response of the environments studied to climaticchanges, during the late Quaternary, at a geological time scale. An aspect not yetconsidered is how much sand or loess deposition is required to affect the hydrologicalprocesses. Is a sand layer of several centimetres, or a thin loess layer of severalmillimetres, sufficient to pass a threshold, which irreversibly affects the water regime.The thinner the layer the shorter the time necessary for impact on the environment. It isobvious that aeolian deposition rates vary tremendously from one geographic area toanother, as well as within a given area, in relation to the availability of the material, thedistance from the source area and the regional and local wind regime.

Sand accumulation by wind can be very rapid. Field monitoring in a sandy area in theNegev desert shows deposition, and / or erosion rates, of 10-100 cm during a singleyear, most of it in one extreme windstorm (Kadmon and Leshner, 1995). On the basis ofair photos, taken in 1968 and 1982 along the Egyptian –Israeli border, Tsoar and Moller,(1986) report sand accretion up to 5 m at the crest of linear dunes over an 18-yearperiod. Sand incursion into the area, based on C14, TL ages and prehistoric implements,is assumed to have begun up to 43,000 years ago (Goldberg, 1981; Magaritz and Enzel,1990; Rendell et al., 1995). Boreholes dug in the area show that sand thickness is some30 m at the crest of the linear dunes. This gives an average net accumulation rate of ~0.7 mm per year. Accumulation rate is lower within the interdune corridors where sandthickness is only 6-10 m. The time required for the deposition of 1-2 cm of sand istherefore very short being of the order of 15-30 years, and probably shorter duringperiods of high deposition rates.

The accumulation rate of loess in the Negev desert, north of the sandy area, is lower.Loess deposition under present day dry climatic conditions is of the order of 0.01mmper year (Yaalon and Ganor, 1975; Bruins and Yaalon, 1979; Goosens and Offer, 1990).According to Yaalon and Dan (1974) the loess deposit in the Beersheva sedimentarybasin is 12-15 m thick and its age is of the order of 100,000 years, which gives a netaverage accumulation rate of 0.12-0.15 mm per year. Accumulation rates are somewhatlower on hillslopes where erosion occurs. Under such conditions the time required todeposit 1-2 mm of loess covers only a few decades. The actual deposition rate musthave been higher as part of the loess has been eroded and transported out of the area intothe Mediterranean Sea.

2. Aim of Present Study

In view of the relatively fast deposition rate of aeolian materials, both sand and loess,we decided to check the effect that thin layers of sand or loess exercise on infiltrationand runoff processes. The hypothesis advanced here is that natural semiaridenvironments, although quite resilient to changes in precipitation and temperature, aremore sensitive to slight changes in surface properties associated with climatic changes.


Such persistent sedimentary changes can be expected to alter quickly the hydrologicalregime, drastically affecting water availability and thus the natural ecosystem over along time scale.

3. Experimental Design

As the hydraulic properties of sand and loess differ greatly, two different sets ofexperiments were planned. The first aimed at checking the effect on runoff of two sandlayers (1-2 cm and 4-5 cm thick) overlying a relatively impervious substratum. By usingtwo sand layers that differ in their thickness we expected to identify the thresholdamount of sand required to have a significant effect on the hydrological regime. In thesecond set the effect of a thin (1-2 mm thick) fine-grained layer, whose composition issimilar to that of loess material, overlying a permeable sand layer was checked. Thestudy is based on sprinkling experiments, conducted in the laboratory, at various rainintensities and duration. Trays of 100 cm x 39.2 cm were set at a slope angle of 4degrees (Figure 1). Rainfall amount and distribution were monitored with rain collectorsplaced at the edge of the boxes (Figure 1). Surface and subsurface flow samples werecollected from the trays every 60-90 seconds and moisture content was determined priorto each run.

YAIR AND BRYAN 53

For the study of the effects of sand cover two trays were first filled up with Kettle Creeksilt loam soil, from Ontario, Canada. The particle size composition of this sediment(Figure 2) is similar to that of loess in the northern Negev desert. Boxes were sprinkledwith rainfall intensity of 31 mm/hr. for two hours, until ponding occurred. The fallheight of the drops was 5.5 m and kinetic energy reproduction was 75-80% of similarnatural rainfall. The material was left to drain overnight. The following day the upperlayer of the wetted soil was removed and replaced with dry sand whose particle sizecomposition is shown on Figure 2. The sand was derived by sieving samples ofPontypool loamy sand, developed on kame deposits in Ontario. The particle sizecomposition of the sand used (Figure 2) is quite similar to that of sand forming thelongitudinal dunes in the north-western Negev. One box was covered with a sand layer1-2 cm thick and the other with a layer 4-5 cm thick. Five sprinkling experiments wereconducted with the sand cover. The protocol of these experiments is given in Table 1.

The two remaining experiments were conducted on the fine-grained layer, spread over adry sand substratum. The first run was conducted on the dry, uncompacted, silt loam.Rainfall was applied at average intensity of 44.3 mm/hr for 17 minutes, representing arain amount of 15.5 mm. The second run was conducted three days later with wettersurface conditions and compacted topsoil. Moisture content of the topsoil was 23.3%.The test lasted 10 minutes with a rain intensity of 41.2 mm/hr, producing 6.9 mmrainfall depth.

YAIR AND BRYAN 55

4. Results

4.1. RUNOFF GENERATION ON THIN SANDY LAYERS

4.1.1. Runs with continuous rainData obtained during the first three experiments, in the two boxes, are presented inFigure 3. Neither surface flow nor ponding occurred on the thick sand layer during anyof the runs. Surface flow did occur on the thin sand layer during the first two runs at 36-37 mm/hr rain intensity (Figure 3A, 3B), but not during the last run with the lowerintensity at 24.5 mm/hr. Local ponding was observed after 14 minutes of rainfall.Ponding on the shallow sand layer started at minute 6 during the first run and runoff atminute 15 with a very sharp increase in discharge. Discharge decreased suddenly fiveminutes later, coincident with the start of subsurface flow. This phenomenon was notobserved during any of the following experiments. During the second run (Figure 3B),under wet surface and subsurface conditions (Table 1), the time to runoff was shorter,total discharge higher and equilibrium conditions were reached within a minute afterrunoff started.

Subsurface flow was observed, on both boxes, during all three runs. During the first runsubsurface flow was delayed compared with surface flow. As could be expected,because of difference in pore volume, subsurface flow started later, and with a lowerdischarge on the thicker sand layer (Figure 3). During the second run (Figure 3B),subsurface flow in both boxes started at the same time as surface flow. Again subsurfaceflow discharge was higher on the thin than on the thick sand layer. On the third run(Figure 3C), conducted at lower rain intensity, trends recorded were similar to those atthe first run, except for a shorter time lag until the beginning of subsurface flow thatresulted in higher discharges at the two boxes.

4.1.2. Runs with intermittent rainData obtained are presented in Figure 4. During the first run (Figure 4A), conducted atthe lower intensities, in the range of 23.6-32 mm/hr, surface runoff did not develop onany of the boxes. However local ponding was observed during the two last rainshowerswhen moisture conditions and rain intensities were the highest. Subsurface flowoccurred only on the thin sand layer and was limited to two rain-showers (Figure 4).

The second run (Figure 4B) was conducted a day later under wet surface conditions.The very high rain intensities applied, coupled with the saturated silty-loam substrate,resulted in a quick response of the thin sand layer. Surface flow developed quickly withthe highest discharges recorded. Despite the extreme conditions surface flow did notdevelop over the thick sand layer.

Subsurface flow occurred simultaneously, at ponding time, on both boxes during eachof the rain-showers applied. Due to the short duration of the rain-showers equilibriumconditions were not reached. Subsurface flow discharge was higher on the thin sandlayer during the first rain-shower, but not during the following one (Figure 4 B).


The hydrological response of the sandy layers to rainfall highlights the following points:despite the extreme rain conditions applied during the experiments surface flow neverdeveloped on the 4-5 cm layer, which was able to absorb and drain all rainwater at allrain intensities applied. This is due to its high pore volume and the rapid drainage at theinterface with the saturated underlying layer. The response of the 1-2 cm sand layer wasdifferent. Despite the high moisture content of the underlying layer (Table 1), runoff didnot develop during the run with 24.5 mm/hr, or during the low intensity intermittentrain-showers. In both cases the amount of rain applied was of the order of 15 mm.However, surface flow developed during all higher intensity rainshowers, even on drysand. Runoff generation over the medium and fine-grained sand used cannot be ascribedto surface sealing processes, but rather to a return flow phenomenon as described byDunne and Black (1970). A perched water table developed above the underlyingsaturated silty loam soil, filling the pore space in the thin sandy layer. Once the porespace was saturated water appeared at the surface, starting first with saturated wedge atthe lower end of the box. During the following stage runoff rate was conditioned by therate of rainfall application and drainage, through subsurface flow, at the interfacebetween the sand and fine-grained soil. The whole process was faster during the runsconducted under wet surface and subsurface conditions.

YAIR AND BRYAN 57


It is important to note that the perched water table developed quickly because theunderlying soil was saturated or nearly saturated during all runs. Had this soil beensignificantly drier, runoff would not have developed on any of the runs because of thehigh absorption capacity of the underlying well-aggregated soil. Two hours ofsprinkling at 31 mm/hr. (representing 61mm rain depth) were needed for ponding toappear over this soil.

Surface and subsurface flow recorded during the experiments are considered to resultfrom the experimental design. They are very unlikely to occur under natural desertconditions, characterised by scarce and intermittent small rainstorms, high temperaturesand high evaporation rates, coupled with dry surface and subsurface soil, that would notallow the development of a perched water table or a saturated subsoil beneath a thinsand layer.

4.2. RUNOFF GENERATION ON A THIN FINE-GRAINED LAYER

Two runs were conducted. The first run was performed over dry sand covered by apowdery, loose fine-grained layer. Within 90 seconds of sprinkling, cracks developed inthe fine-grained material, probably due to hydro-compaction and subsequent sealingprocesses of the unconsolidated fine-grained material. The cracks were 1-2 mm wide.Some of them were already sealed or filled up with water when ponding occurred atminute 7 (Figure 5). Runoff started two minutes later when the accumulated rainamount was 6.5 mm. Discharge increased quickly and stabilised after 2.5 minutes.Runoff rate, at peak flow, was 68 % of the rain applied. Subsurface flow did notdevelop.

The second run was conducted under wet conditions. Runoff started almostimmediately. Runoff coefficient, at peak flow, was 74 %. A small trickle of subsurfaceflow was observed at the last minute.

The effect of the thin fine-grained layer, on top of the highly permeable sand, wasstriking. The fast development of runoff on the first run is a clear indication of veryeffective sealing of the fine-grained layer, supported by the complete lack of subsurfaceflow. Runoff generation during the second run was much faster; due to wet surfaceconditions as well as to the development of a surface crust seal at the end of the firstrun.

5. Discussion

The discussion will focus on the issue of extrapolating results obtained in small-scalelaboratory experiments to field conditions. Such an analysis is important in an attemptto upscale results obtained to a landscape scale, while addressing the question of thelong- term environmental effects of surface changes connected to climatic changes. Anideal situation for this investigation is found in the northern Negev desert, wheredifferent phases of loess and sand deposition have been recorded during the lateQuaternary. In the central part of the northern Negev, the loess deposits cover the flatvalley bottoms and extends over valley hillslopes, where the loess is in direct contactwith older Eocene and Cretaceous bedrock. South of the loess-covered area, thelandscape is rocky with deeply incised valleys. Such a situation allows studying thehydrological and environmental effects of loess deposition over rocky areas. A differentlandscape exists in the western Negev, along the northern part of the Egyptian- Israeliborder. Here, the Nizzana sand field represents the eastern edge of the extensive Sinaierg. It is characterised by linear dunes separated by wide interdune corridors. Severaltrenches dug in an interdune corridor reveal a sequence of loess layers alternating withsandy units. The thickness of the loess layers is in the order of 10-40 cm. (Yair, 1990;

YAIR AND BRYAN 59


Harrison and Yair, 1998). In several places the loess units outcrop at the surface,forming flat surfaces adjacent to dune ridges. The proximity of loess-covered surfaces tosandy covered surfaces allows a comparative study of these two units on the waterregime and related environmental conditions.

Average annual rainfall in the study area varies in the range of 90-200 mm, being higherin the northern loess covered area than in the southern sandy and rocky areas. Thesprinkling experiments conducted lasted 9 days during which the total rain amountapplied was 82.5 mm, very close to the long-term average annual rainfall for thesouthern areas. Rain amounts applied, during each of the runs, occur in this area one tothree times a year. However rain intensities used, and especially their long duration,represent rather extreme to very extreme conditions. Rainfall data collected in this areaduring the last 30 years show that 85% of the rain fell at an intensity below 10 mm/hr(Kutiel, 1978). Rain intensities up to 30-35 mm/hr are recorded almost every year butfor a short duration of 1-6 minutes. Higher intensities are rare and usually last no morethan 1-2 minutes.

5.1. ENVIRONMENTAL EFFECTS OF LOESS DEPOSITION OVER ROCKYSURFACES

As indicated earlier, there is a general agreement, among scientists working in the area,that loess deposition took place during a wet period whereas sand deposition occurredduring dry periods. Several studies were devoted to the environmental effects of loesspenetration into the northern Negev desert. These studies cover hydrological,pedological, botanical and zoological aspects. Long-term hydrological data collected atthe Sede Boqer Instrumented Watershed, located in the rocky Negev where averageannual rainfall is 93 mm, clearly indicate that rocky areas respond quickly to rainfall.(Yair, 1994; 1999). The threshold rain amount necessary to generate runoff is in theorder of 2mm. Runoff occurs with any rainstorm having an intensity exceeding 5mm/hr. Under such conditions runoff frequency and magnitude are high. Runoffgenerated over the rocky areas is absorbed on its way downslope by colluvial mantles,allowing local water concentration, deep water penetration and soil leaching. A differenthydrological response is characteristic of loess covered areas. Stibbe, (1974) and Morinand Jarosh, (1977) show that the threshold rain amount required for runoff generationover the loessial soils of the Beersheva Basin is ~ 8 mm, very close to that obtained inthe laboratory experiment (6.5 mm). Due to rainfall scarcity, the prevalence of lowintensity rainfall, long time intervals between rainstorms, high evaporation rates, higherporosity and higher water absorption capacity of the loess, runoff frequency andmagnitude are much lower than in rocky areas. However, the compacted loess preventsdeep water infiltration. The depth of water penetration seldom exceeds 40 cm (Yair,1994). Under such conditions, leaching is confined to a shallow depth, contributing to agradual soil salinisation process. This process is further enhanced by the increase in saltinput of airborne salts, from rainfall and dust, during a wet climatic period. Thedegradation of the water regime, and soil salinisation process, that followed loesspenetration, had long lasting effects on the vegetation and on the biological activity.Comparative studies were conducted in the Negev desert between the northern, wetter,

YAIR AND BRYAN 61

loess-covered area and the southern rocky climatically drier southern area. These studiesdeal with the density and composition of the vegetal cover, with the abundance andcomposition of burrowing animals and snails and with soil salinity. Data collectedclearly show that soil salinity in the climatically wetter loess area is higher than that ofsoils in the rocky area. The northern loess area is also far more arid and less productivethan the drier southern rocky area (Yair and Shachak, 1987; Kadmon et. al., 1989; Yair,1994). In other words, although loess penetration had occurred during a wet climaticphase it resulted in an overall desertification effect.

5.2 LOESS DEPOSITION OVER A SANDY SUBSTRATUM

Similar environmental studies were conducted in the Nizzana sand field. As could beexpected, runoff generation is faster on the compacted loess covered than on the loosesand covered deposits (Yair, 1990). Depth of water infiltration is limited to 40 cm inthe loess. Infiltrated water is quickly lost by evaporation. The soil is saline (Blume etal., 1995) and devoid of vegetation. The sandy areas represent a far better edaphicenvironment. Deep rainwater infiltration, up to 400 cm in rainy years, (Yair et. al.,1997) combined with low capillary water movement create a water reservoir availablefor plants. Plant cover, on the stabilised section of dune slopes, is 30-40 %, reachingalmost 100 % at the base of the dunes. The pronounced positive effect of a shallow sandcover is evident where small sand mounds develop on the flat loess covered areas,allowing for the formation of local water lenses at the interface loess-sand. A mound 10cm thick, is enough to allow germination of annual plants, whereas a mound 30-40 cmthick can support perennial shrubs.

6. Conclusions

Data presented in this study support the hypothesis that a change in surface propertieshas a rapid environmental effect, in semiarid and arid areas, where climatic changes areaccompanied by the rapid input of aeolian material. A sand layer 1-2 cm thick, even ifdeposited on top of a relatively impervious substratum, is thick enough to eliminaterunoff generation under an arid rainfall regime characterised by infrequent, low intensityand intermittent rainstorms, with limited total rainfall. An opposite effect can beexpected when fine-grained material, such as loess, is deposited above a highlypermeable sandy substratum or over rocky surfaces. The sealing and compactionprocess on the fine-grained layer is so efficient that infiltration through this layer isdrastically reduced, leading to a bad water regime and environmental drier conditions.

Data obtained are in complete agreement with field studies conducted in the northernNegev desert. These studies show that, under the intermittent and infrequent rainstormsprevailing in the area, surface properties play the determinant role in the non-uniformrunoff generation and in the redistribution of water resources in space, greatly affectingthe whole ecosystem (Yair, 1983; 1994).


To summarise, although semiarid ecosystems are highly adapted to extreme variationsin rainfall, over a time scale of decades and centuries, they seem extremely sensitive tochanges in their surface properties, which alter their hydrological regime quickly andefficiently. The time necessary to achieve such a drastic change seems to be very short,at a human rather than a geological time scale.

Finally, it would be quite interesting to study, in a similar way, the impact that thedeposition of loess and sand had on hydrological and ecological processes at the fringeof cold and glaciated deserts in the northern hemisphere (in Europe as well as in theAmerican continent). Climatic conditions in latter areas differ significantly from thoseprevailing in the subtropical belt. Such a complementary study would provide a broaderand deeper understanding of climatic changes on the environment for various climaticconditions.

Acknowledgements

This study was conducted at the Soil Erosion Laboratory of the University of Toronto.The technical help of Mr. Niklaus Kuhn is greatly appreciated. Thanks are due to Mrs.M. Kidron, of the Department of Geography, Hebrew University, for drawing theillustrations.

References

Bruins, H.J. and Yaalon, D.H. (1979). Stratigraphy of the Netivot Section in the Desert Loess of the Negev(Israel). Acta Geologica Academiae Scientarum Hungaricas; Tamus 22:161-169.

Coude- Gaussen, G. (1991). Les Poussieres Sahariennes, Montrouge, Libbey, 485pp.Dunne, T and Black, R.D. (1970). An experimental investigation of runoff production in permeable soils.

Water Resources Research, 6: 478-490.Goldberg, P. (1981). Late Quaternary stratigraphy of Israel: an eclectic view. Colloques internationaux du

CNRS. Prehistoire du Levant, 598: 58-66.Goring, A. M and Goldberg, P. (1990). Late Quaternary dune migration in the southern Levant: Archeology,

Chronology and Paleoenvironments. Quaternary International, 5: 115-137.Goosens, D. and Offer, Z. I. (1990). A wind tunnel simulation and field verification of desert dust deposition

Avdat Experimental Station, Negev desert. Sedimentology, 37: 7-22.Harrison, J.B.J. and Yair, A (1998). Late Pleistocene aeolian and fluvial interactions in the development of the

Nizzana dune field, Negev desert, Israel. Sedimentology, 45: 507-518.Holling, C.S. (1973). Resilience and Stability of Ecological Systems. Annual Review, Ecological Systems, 4:

10-23.Kadmon, R., Yair,A. and Danin, A. (1989). Relationship between soil properties, soil moisture and vegetation

along loess covered hillslopes, Northren Negev, Israel. Catena Supplement 14: 83-92.Kadmon, R. and Leshner, H. (1995). Ecology of linear dunes. Effect of surface stability on the distribution and

abundance of annual plants. Advances in GeoEcology, 28: 125-143.Kutiel, H. (1978). The distribution of rain intensities in Israel. MSc.thesis, the Hebrew University, Jerusalem

(in Hebrew).Morin, J. and Jarosh, H. S. (1977). Rainfall-runoff analysis for bare soils. Pamphlet no 164, Volcani Institute

for Agricultural Research Center, Beit Dagan, Israel. 23 pp.Magaritz M; and Enzel Y. (1990). Standing water deposits as indicators of late Quaternary dune migration in

the northwestern Negev. Climatic Change, 16: 307-318.

YAIR AND BRYAN 63

Rendell, H. M., Yair, A. and Tsoar H. (1993). Thermoluminescence dating of periods of sand movement andlinear dune formation in the northern Negev. In A. C. Millington and K. Pye (eds) The Dynamics andEnvironmental Context of Eolian Sedimentary Systems. Geological Society Special Publication 72: 69-74.

Stibbe, E. (1974). Hydrological balance of Limans in the Negev. Volcani Institute for Agricultural Research.Publication no 304, Beit Dagan, Israel. 35 pp.

Schlesinger, W.H., Reynolds, J.F., Cunningham, G.L., Huenneke, L.F., Jarell, W.M., Virginia, R.A. andShmida, A. (1982). Endemic plants of Israel, Rotem, Bulletin of the Israel Plant Information Centre, 3: 3-47

(in Hebrew).Thiery, R. G. (1982). Environmental instability and community diversity. Biological Reviews, 57: 691-710.Whitford, W.C. (1990). Biological feedbacks in global desertification. Science, 247: 1043-1048.Wiens, A.J. (1985). Vertebrate Responses to Environmental Patchiness in Arid and Semiand Ecosystems, in

Pickett, S.T.A. and White, P.S. (eds), The Ecology of Natural Disturbance and Patch Dynamics.Academic Press, NY, pp.: 169-196.

Yaalon, D.H. and Dan, J.(1974). Accumulation and distribution of loess-derived deposits in the semi-desertand desert fringe area of Israel. Zeitschrift fur Geomorphologie, Supplement Band 20: 91-105.

Yaalon, D.H. and Ganor, E. (1975). Rates of aeolian accretion in the Mediterranean and desert fringeenvironments of Israel. International Congress of Sedimentology, Nice, France : 169-174.

Yair, A. (1983). Hillslope hydrology, water harvesting and areal distribution of some ancient agriculturalsystems in the northern Negev desert. Journal of Arid Environments, 6: 283-301.

Yair, A. (1987). Environmental effects of loess penetration into the northern Negev desert. J. of AridEnvironmnets,13: 9-24.

Yair, A. (1990). Runoff generation in a sandy area; the Nizzana sands, western Negev, Israel. Earth surfaceProcesses and Landforms, 15: 597-609.

Yair, A. (1992). The control of headwater area on channel runoff in a small arid watershed. In Parsons, T andAbrahams, a. (eds), Overland Flow, pp.53-68.

Yair, A. (1994). The ambiguous impact of climate change at a desert fringe, Northern Negev, Israel, inMillington, A. C. and Pye, K. (eds). Environmental Change in Drylands: Biogeographical andGeomorphological Perspectives, Chichester, John Wiley and Sons, pp. 199-227.

Yair, A. (1999). Spatial variability in the runoff generated in small arid watersheds: implications for waterharvesting, in Hoekstra, T. M. and Shachak, M. (eds), Arid Lands Management toward EcologicalSustainability, pp. 212-222.

Yair, A. and Danin, A. (1980). Spatial variations in vegetation as related to the soil moisture regime over anarid limestone hillside, Northern Negev, Israel. Oecologia, 47: :83-88.

Yair, A. and Shachak, M. ( 1985). Studies in watershed ecology of an arid area. in Berkofsky, L.and Wurtele,G. (eds). Progress in Desert Research, Rowman and Littlefield, New Jersey, pp. 45-193.

Yair, A., Lavee, H and Greitser, N. (1997). Spatial and temporal variability of percolation and watermovement in a system of longitudinal dunes, western Negev, Israel. Hydrological Processes, 11: 43-58.

WEATHERING, GEOMORPHOLOGY AND CLIMATICVARIABILITY IN THE CENTRAL NAMIB DESERT

HEATHER VILES and ANDREW GOUDIESchool of Geography, University of Oxford,Mansfield Road, Oxford OX1 3TB

Abstract

Weathering is an important component of geomorphological change in the CentralNamib Desert. Previous studies have reported on the weathering role of salt anddissolution, allied with wind abrasion. However, many surface are covered by luxuriantlichen growths, fed by fog precipitation, whose weathering role has not been clarified.Here we present preliminary investigations of the role of lichens and other rock surfacemicroorganisms in weathering and surface protection, using field observations from arange of sites between 2 and 80 km from the coast, coupled with Scanning ElectronMicroscope (SEM) observations of lichen:substrate interactions. A model of lichenweathering activity is proposed, illustrating the different roles of lichens on various rocktypes. Spatial segregation of lichen and other weathering processes is seen to occur at arange of scales.

1. Introduction

The Central Namib Desert, Namibia, Southern Africa, covers some and isone of the driest deserts in the world. It is characterised by gravel plains often underlainby gypsum crusts (Eckardt and Spiro, 1999), interspersed with weathered rock outcropsand numerous dry riverbeds. The area is separated from the Namib sand sea to the southby the Kuiseb River, and from the coastal dunes in the north by the Huab River.Classified as hyper-arid, the area receives very little rainfall, although fog candramatically increase overall precipitation amounts. Southgate et al., (1996) haveanalysed climate records over the period 1962-1991 from the desert research station atGobabeb (23° 34’S, 15° 03’E) and found mean annual rainfall to be 19 mm (range 0 -107 mm) and mean annual fog precipitation to be 37 mm (range 14 - 68 mm). Gobabebis over 70 km from the coast (see figure 1), and areas nearer the sea will experiencegreater amounts of fog. The geology of the area is dominated by a suite of metamorphic

65

KEY WORDS: Biological weathering; salt weathering; rock-surface microenvironments.


66 VILES AND GOUDIE

and igneous rocks from the Nosib and Swakop groups of the Damara System of lateProterozoic age. The Damaran metamorphics are highly variable and include micaschists, marble, granitic gneiss and quartzite. Into these rocks are intruded granites andblack dolerite dykes (Goudie, 1972).

As in other desert environments with major outcrops of rock and desert pavement,weathering plays a key role in both shaping residual rock outcrops and in producingfine-grained sediment. The nature and rate of weathering within arid environments hasbeen the subject of much debate among geologists and geomorphologists over the years,with early views that deserts represented sterile, and highly restricted weatheringenvironments being gradually replaced by the view that deserts can experience severe (ifsuperficial and selective) weathering (Goudie, 1997; Goudie et al., 1997). However,there is still much debate about how desert weathering processes work, and the controlson them. As Smith (1994, p. 39) puts it: ‘Weathering research is thus not a question ofwhat we know about desert weathering, but what we do not know…’ Much previousdesert weathering research has failed to study different mechanisms in association, in aneffort to improve knowledge on individual processes.

Desert weathering is often evidenced by flaking and disintegrating rock outcrops, andless commonly by distinctive small rills and pits. It is often difficult to ascribe thegenesis of these features to particular weathering processes, or combinations ofprocesses. Several major types of desert weathering processes have been studied interms of their operation and importance. The role of insolation weathering (essentiallycaused by heating and cooling of rock surfaces producing steep temperature gradients

WEATHERING IN THE CENTRAL NAMIB DESERT 67

towards the inner part of the rock) in shaping the desert landscape has been muchdebated (Cooke et al., 1993; Smith, 1994). Micro-environmental conditions (such asaspect) and rock type may be critical in determining how important insolationweathering is. Salt weathering has been proposed by many workers as a major, largelyphysical, weathering process in deserts, probably playing an important role in producinglandforms such as alveole and tafoni (Mustoe, 1982), and as a source of large amountsof fine-grained debris (although chemical weathering processes have also been proposedas important to the formation of tafoni by Young, 1987, and Campbell, 1999). Threeimportant groups of factors control the operation of salt weathering according to Cookeet al., (1993, p. 33), that is environmental and micro-environmental conditions, materialproperties and the characteristics of the salts themselves. Lichens, algae and other lowerplants and microorganisms have been shown to play a significant, if often localised, rolein desert weathering. Studies from the Negev (e.g. Danin and Garty, 1983) and colddeserts (e.g. Friedmann, 1982; Sun and Friedmann, 1991) indicate that microorganismsinteract with rock surfaces in a range of ways. Firstly, some can bore into rock surfaces(the euendolithic niche, following the terminology of Golubic et al., 1981). Othersinhabit cracks (the chasmoendolithic niche) or preformed cavities in the rock(cryptoendolithic niche). Each of these three types can contribute significantly toweathering through the production of small-scale pitting, and aiding the development ofsurface flaking and disintegration. Even those microorganisms which live purely on thesurface (the epilithic niche) or under stones (the hypolithic niche) can alter chemicalconditions at the rock surface, encouraging weathering. Lower plants andmicroorganisms also contribute to surface protection through the formation of biologicalcrusts. Thus, mosses, liverworts and fruticose and foliose lichens growing on soils androcks increase the wind resistance of the surface reducing erosion and trapping dust-blown debris (Danin and Ganor, 1991, 1997). Again, micro-environmental conditionsseem to be a very important factor in the nature of microorganic influences onweathering and surface stability. In the Negev Desert for example, Kappen et al., (1980)found that NE facing slopes have a much richer lichen cover than SW facing ones,because of differences in receipt of solar radiation and moisture retention.

Chemical weathering may be more important in deserts than has been previouslyrecognised (Smith, 1994) and there have been many reports of microscale karrenfeatures on limestones and marbles within deserts which may be produced by smallamounts of precipitation given suitable lithological and micro-environmental conditions(Lowdermilk and Woodruffe, 1932; Sweeting and Lancaster, 1982; Smith, 1987).However, it is often difficult to differentiate rillenkarren from wind-hewn fluting, whichmay also be found on desert rocks. Such wind-eroded forms are likely to beconcentrated on slopes facing into prevailing or dominant wind directions, and shouldoccur across a range of lithologies. The foregoing discussion of current knowledge ondesert weathering processes and controls indicates that much of the evidence iscircumstantial and that although we can make some generalisations about what factorscontrol the operation of weathering processes (often based on laboratory simulationswhich do not replicate the complex nature of real desert environments), we do not have

68 VILES AND GOUDIE

adequate knowledge of how they interact one with another to produce landforms anddebris.

Associations between weathering processes may be synergistic (i.e. one may encouragethe action of another), or one may slow down or prevent the action of others. Such co-associations may operate sequentially, or at the same time. Thus, during previous wetterphases, for example, more intense chemical weathering might have ‘pre-stressed’ therock leaving it more vulnerable to weathering by salts. Figure 2 shows an example ofsequential weathering environments from Swartbankberg, with lichens colonizing apreviously exfoliated boulder. The exfoliation may have provided suitable roughenedsurfaces, thus encouraging lichen growth. On the other hand, lichens, although perhapsproducing micro-scale chemical attack under their thalli, might protect rock surfacesfrom the action of wind and salts. Different weathering processes may also be spatiallysegregated, because of the varying climatic and micro-environmental controls whichdetermine their operation.

Previous work on weathering in the Central Namib Desert has highlighted theimportance of salt weathering (Goudie et al., 1997) and dissolution of carbonateminerals (Sweeting and Lancaster, 1982). This paper focuses on biological weatheringprocesses (which have received little previous attention here), and how they relate toother weathering processes. The Central Namib desert has a rich lichen flora (asdescribed by Schieferstein and Loris, 1992 and shown in figure 3), largely supported byfog, along with a range of cyanobacteria and other microorganisms. Lichens andmicroorganisms are poikilohydric and can take up water from air with relative humidityhigher than 70% and most can survive long periods of desiccation (Walter, 1986).

The aims of this paper are to investigate three aspects of weathering in the CentralNamib Desert, i.e.:

1.

2.

3.

Spatial differences in biological and other weathering processes at a range ofscales.Co-associations between biological and other weathering processes.

The likely impacts of temporal variability in climate on biological and otherweathering processes.

Before considering the detailed evidence of biological weathering it is important toclarify the nature of environmental gradients within the Central Namib Desert.

70 VILES AND GOUDIE

2. Environmental variability in space and time across the central namib desert

The Central Namib Desert stretches over 100 km inland from the coast and ischaracterised by clear E-W gradients in climate, lichen cover, lichen biomass, altitudeand ground type. Rainfall increases markedly inland, whereas fog is highest nearer thecoast with some 120 fog days per year at the coast, tailing off to around 40 days at 40km inland, and 5 fog days at around 100 km from the coast (Olivier, 1995). Altitudeincreases inland towards the foot of the great escarpment around 80 to 140 km from thecoast and peaks around 900m a.s.l. The occurrence of lichens seems to decrease inlandwith most luxuriant growths found within about 30 km of the coast. Detailed studies inthe area around Swakopmund by Schieferstein and Loris (1992) show that maximumlichen coverage occurs around 5 km from the coast and biomass peaks around 1 kmfrom the coast. Gypsum crusts are found predominantly within the coastal zone, andtheir eastern limit is around 50 - 70 km from the shore at an elevation of 400-500m(Eckardt and Spiro, 1999). The low-lying coastal areas are prone to the accumulation ofa wide range of salts at the surface, producing a range of pan forms. Thus, it might beexpected that the nature and intensity of weathering will also vary across these gradientsbecause of the different controlling factors influencing the physical, chemical andbiological processes thought to operate in this area.

There are also much smaller scale variations in the nature and intensity of weathering inthe Central Namib Desert, as aspect, lithology, and microclimate vary across the cm - mscale. For example, East-facing slopes are preferentially affected by seasonal dryeasterly winds, known as Berg winds, whereas West-facing slopes are more protected.North-facing slopes experience higher levels of incoming radiation, greater fluctuationsin temperature, and higher levels of evaporation than south-facing slopes in this southernhemisphere environment. Such microclimatic variability will have ecological andgeomorphological impacts. Schieferstein and Loris (1992), in their study of near-coastallichen fields, found that fruticose and foliose lichens were dominant on SW-facingslopes, whereas crustose (and often euendolithic) species dominated on NE and E-facingslopes. Sweeting and Lancaster (1982) suggested that solutional rillenkarren, andmicroorganic growths including crustose lichens are found predominantly on west-facingmarble surfaces at around 50 km inland, with wind-abraded surfaces on east-facingslopes (see figure 4).


There is also, at one site, some vertical differentiation of micro-environment and alsoprobably therefore of weathering regime. Thus, rock surfaces at ground level in contactwith salt-enriched soils are likely to be prone to salt weathering, whereas rock faceshigher above the ground surface may be colonised by lichens, or affected by wind orsolution.

There are also patterns of temporal variability in climate and environment at a range ofscales which will also have impacts upon weathering regimes. The climate of the Namibis influenced by the cold Benguela current off the west coast and by the El NiñoSouthern Oscillation. Although there are a few meteorological stations within the CentralNamib Desert only the one at Gobabeb has a sufficiently long and continuous record topermit analysis of temporal trends. Southgate et al., (1996) found that, over the period1962-1991, annual rainfall at Gobabeb had a coefficient of variability of 113% whereasthat of fog was only 36%. Analysis of the cumulative deviation from the mean bySouthgate et al., (1996) showed that in terms of rainfall the period from 1962/3 to1974/5 was much drier than average, and between 1975/6 and 1978/9 much wetter thanaverage, with below average conditions since then. Fog variability was found to begenerally in inverse relation to that of rain, with low fog precipitation between 1974/5and 1985/6 and then a marked increase after that.

3. Methods

During two short field seasons in 1994 and 1996 a range of sites was visited across theCentral Namib Desert (as shown in figure 1). At each site observations were made on

72 VILES AND GOUDIE

the geology and geomorphology, and a range of small rock samples was removed forobservation under the Scanning Electron Microscope (SEM). On return to thelaboratory, SEM samples were obtained by fracturing the rock with a cold chisel toobtain a cross-sectional view from the surface into the interior of the rock. At two sites,Swartbankberg and Tomato Pan, field estimates of lichen cover were made using a 25 x25 cm quadrat divided into 5 cm squares. Between 10 and 20 quadrats were randomlylocated and the % cover of lichens estimated by counting the number of 5 cm squarescontaining appreciable amounts of lichens.

4. Results

The geomorphological and ecological characteristics of the sites studied are listed inTable 1. At all sites there was clear evidence of weathering, often in the form of surfaceflaking and small scale pitting. Lichens were found at all sites, although there was verylittle lichen cover at Mirabib (80 km from coast) and Gobabeb (70 km from coast).

At many sites there was clear evidence of small-scale spatial patterning in weatheringmicroenvironments. Thus, at Tomato Pan upstanding dolerite outcrops were coveredwith lichen-covered boulders (with crustose, foliose and fruticose types), as were loweroutcrops of schist. However, just above the level of the pan surface lichens decreased innumber, to be replaced by signs of harsh salt weathering (Goudie et al., 1997). AtSwartbankberg, most rock outcrops were covered by a mosaic of crustose lichens, exceptwhere the surface was exfoliating rapidly. On the boulder strewn slopes at the base ofSwartbankberg there was extensive lichen cover, even on boulders previously subjectedto exfoliation (figure 2). At Gobabeb, on finely sculpted granite outcrops with alveoliand tafoni extensively developed, lichen cover was very rare - limited to a few, smallgreen crustose types growing on dark, fine-grained parts of the rock, and on the visors oftafoni. At Vogelfederberg, where the Hamilton mountains marbles outcrop, and also atthe Karibib marble outcrop there were clear signs of spatial segregation of weatheringmicro-environments into east-facing (with wind-blown flutes) and west facing (withbrown lichen-pocked surfaces and occasional small rills) as shown in figure 4. On areasof lichen fields near Tomato Pan and the airport at Rooikop, a lush growth of foliose,fruticose and crustose lichens was found with lichens covering boulders and gypsumcrusts (figure 3). The only bare patches were along commonly used tracks, aroundbushes, and at the bottom of ephemeral washes. At all sites there was a general trend forcrustose lichens to dominate on outcrop surfaces, with foliose and fruticose lichens onlybecoming common on boulder-strewn and gravel plains.

Some estimates of lichen cover made at Swartbankberg and Tomato Pan on a range ofsurface types are presented in Table 2. The high standard deviations between individualquadrats on each surface type illustrate the patchiness of cover at this scale. AtSwartbankberg the highest mean % cover, and the highest variability, are found on theschist outcrop at the highest point sampled on the profile (c. 350 m a.s.l.), with thelowest mean value at the base of the slope (around 300 m a.s.l.). This would conform to


the hypothesis that lichens are supported by fog here, as the higher parts of the hill willintercept more fog. At Tomato Pan, on the most stable, least salty surface on the top ofthe dolerite dyke, very high mean % cover is found (83.1%), with a similar decliningtrend down to the gravel plain just above the salty pan surface. In this case, salt is likelyto be the controlling factor, as fog will be adequate throughout the site.

Study of rock and lichen samples collected in the field indicates that lichen thalli of up toa few mm in thickness are commonly found (see table 3), although many lichen growthsare extremely thin and in several cases at least partly euendolithic. There arecryptoendolithic growths present on several samples (evidenced by a green line around 1mm below the surface of the rock), especially more porous and lighter colouredoutcrops.

74 VILES AND GOUDIE

WEATHERING IN THE CENTRAL NAMIB 75

76 VILES AND GOUDIE

Three major types of lichen growth are revealed in hand specimen. Firstly, individualepilithic lichen thalli forming circular or near circular patches, sometimes peeling awaytowards the centre. These growths provide only a patchy cover of the underlyingsubstrate. Secondly, there are mosaics of adjoining eplithic and euendolithic lichenthalli which entirely cover the surface of some areas of rock, boulders or gypsum crust.These growths are often very thin (<< 500 microns) but extensive. Finally, there aredominantly euendolithic growths which are revealed on the surface only as narrowgrowths along cracks and grain boundaries, but which form an extensive sub-surfacelayer. Each growth style will have a corresponding impact on weathering.

Scanning electron microscope (SEM) reveals more information about the nature of thesubstrate:lichen interactions as summarised in table 3. Foliose and fruticose lichensgenerally have only sparse attachments (through rhizines or rootlets) to the underlyingsurface and have not been investigated further here for any weathering role. Crustoselichens here generally have thalli of up to 1 mm thick growing on top of the rock, andbeneath the surface fungal hyphae or whole parts of the lichen thallus may extend downby 1 mm or so.

Several lichens show clear borehole production by fungal hyphae in suitable crystalsubstrates (usually calcite within marbles) as shown in figure 5a and 5b. Such boreholeshave been noted to occur in limestones and marbles throughout the world, with stressedor frequently wetted environments (such as coastal and arid environments and streams)having particularly well-developed features. In general, within the 25 samples studiedhere, boreholes extend down about into the calcite crystals (a lot less thancommonly found on limestone coasts, e.g. by Le Campion-Alsumard, 1979, but morethan found on terrestrial limestones on Aldabra Atoll by Viles, 1987). Fungal hyphalboreholes were also rare on other rock types here, notably on gypsum at Swartbankberg,


where calcification of hyphae was also observed. On schist, dolerite, diorite and granitehyphal boreholes were not found, but there was abundant evidence of lichen thallipenetrating into the rock surface and leading to detachment of grains and flakes (in asimilar fashion to the action of Lecidea auriculata on gabbroic boulders on moraines inNorth Norway observed by McCarroll and Viles, 1995). Figure 6a and 6b illustrategrain detachment under lichens on schist from Swartbankberg. However, this activityappeared spatially very patchy at the small scale, with some lichens not creating anyobvious impact on their substrate. On all rock types many of the crustose lichens werepeeling away from the surface. Several samples also showed the impact of acryptoendolithic layer on weathering, as shown in figure 7a and 7b. Here,cryptoendolithic microorganisms within a diorite pebble from Tomato Pan fill porespaces and appear to be aiding grain detachment.

78 VILES AND GOUDIE

From these observations in the field and under the microscope, a model of lichenweathering and surface protecting activity in the area can be put forward (figure 8).Folios and fruticose lichens, because of the nature of their attachment to the rock(limited to point contact) do not create much weathering impact on the underlyingsurface. In contrast, they play a major role in trapping airborne debris, and in bindingtogether any unconsolidated material on which they grow. Their dominant effect is thusto protect the underlying surface from erosion. Crustose lichens, on the other hand maybe dominantly protective or preferentially cause weathering, depending largely on thenature of the underlying rock type and their growth and decay characteristics. Thus, onmarble the biochemical action of lichens (forming fungal hyphal boreholes) tends todominate, whereas on schist, dolerite, diorite and granite they may be important in grainand flake detachment - through what might be called biophysical action, although itprobably involves chemical and physical processes of deterioration and disintegration ofthe underlying rock.

On quartz pebbles, lichen weathering activity is generally minimal (although one sampleshowed some evidence of shallow fungal boreholes). In all three cases, and especiallythat of the quartz pebbles, the net effect of lichens is probably a protective one, at leastwhile they are still alive. In this hyper-arid environment, lichen cover buffers theunderlying rock from wind, changes its thermal response to heating and cooling stressesand will prevent water from making contact with the surface (thus limiting dissolution).However, when lichens die, decay or otherwise are removed from the surface (perhapsby grazing) net erosion will occur, as grains detached from the rock and encased withinthe lichen thallus will be removed. The overall importance of lichen weathering dependsalso on the nature of growths (e.g. predominantly euendolithic forms vs. epilithic types),the contribution of any cryptoendolithic community, and the percentage cover of lichens.

In order to test this model, and apply a temporal framework, it is necessary to collectmore information about the growth rates of the various lichens constituting the lichenmosaics. Furthermore, the impact of lichens on the thermal response of a range ofdifferent rock types needs testing. In some cases, lichen surface coloration mayencourage insolation weathering by reducing or increasing albedo in comparison withthe underlying rock.


5. Discussion and conclusions

The model of lichen weathering and surface protection activity proposed above needs tobe set within the context of other weathering and erosional processes within the CentralNamib Desert, and against the backdrop of environmental variability over space andtime at a range of scales discussed earlier. Figure 9 shows the three main processes seento be sculpting rock surfaces in the Central Namib Desert plotted onto three axes. Thebottom axis shows the major environmental gradient from the coastal zone to the inlandextremes of the desert, as one moves from the salt and fog-affected coastal environmenttowards the inland area where wind (and insolation, and possibly rainfall) effectsdominate. The side axes depict the smaller-scale gradients found at a site. On the lefthand side is what is in reality often a vertical gradient from stable to unstable surfaces,or from sound upstanding rock down onto salty surfaces, as is found at Tomato Pan. Onthe right hand side is a gradient from benign to harsh, which in reality is a gradient ofaspect from west-facing to east-facing (as found on the marble outcrops atVogelfederberg). The dominance of salt, wind (and insolation) and lichen weatheringprocesses can be mapped onto these axes as shown in figure 9.

80 VILES AND GOUDIE

This figure illustrates the generally spatially-segregated nature of the differentdenudation processes, on both small and large scales reflecting the notableenvironmental gradients at these scales. It is clear that each of the three denudationdomains (i.e. lichen-dominated, salt-dominated and wind-dominated) possess individualcharacteristics. Thus, the lichen-dominated domain includes a diverse series (as shownin figure 8) of protective and weathering effects, all of which are highly superficialaffecting the top few mm at most, and showing clear differentiation in impacts ondifferent rock types. Over time, the rate of weathering and removal of material in thelichen-dominated domain is controlled probably by the nature of fog inputs (which, asSouthgate et al., 1996 show, contribute twice the moisture as rainfall here with a third ofthe variability) which control lichen growth. The salt-dominated domain, however,shows similar selectivity in terms of rock types affected (although for the case of saltweathering, it appears that porosity and easy ingress of water is the key factorpredisposing rocks to salt attack, rather than mineralogy in the case of lichen-effects).The depth of weathering caused by salt is in the order of centimetres rather thanmillimetres. Over time, changes in the sources of salt and groundwater levels are likelyto play a key role in determining salt weathering effectiveness. Finally, the wind-dominated domain also shows selectivity according to rock type (most previous reportsof wind flutes have been on marble, e.g. Sweeting and Lancaster, 1982), and greatsuperficiality. Changes in the strength of the Berg wind over time will affect the roleplayed by wind in denudation here.


The preliminary field and laboratory observations presented above have been used tocreate simple conceptual models of the role of lichens in weathering, in comparison withthe role of other weathering processes. Clear evidence has been presented of a range ofimpacts of lichens on weathering and surface protection, backing up findings from otherdesert areas (e.g. the Negev). Lack of long-term climatic records, and the sparsity ofmeteorological sites within the Central Namib Desert, means that definitive statementsabout large-scale environmental gradients over space and time are hard to make. Thealmost complete lack of microenvironmental data (e.g. rock surface temperature andhumidity) means that more concrete pronouncements on micro-scale environmentalvariability are also impossible. Until such data are available it will be difficult to bemore precise about the overall role of lichen weathering in geomorphological change inthe Central Namib Desert. However, lichens are clearly a neglected component ofweathering here which deserve more attention in the future.

Acknowledgements

We thank Amy and Alice Goudie for indefatigable field assistance, and Adrian Parkerfor help with preparing SEM samples. Steve Jones kindly drew the figures.

References

Campbell, S.W. (1999) Chemical weathering associated with tafoni at Papago Park, Central Arizona, EarthSurface Processes and Landforms 24, 271-8.

Cooke, R. U., Warren, A. and Goudie, A. S. (1993) Desert Geomorphology, London: University CollegeLondon Press.

Danin, A. and Ganor, E. (1991) Trapping of airborne dust by mosses in the Negev Desert, Israel, EarthSurface Processes and Landforms 16, 153-62.

Danin, A. and Ganor, E. (1997) Trapping of airborne dust by Eig’s meadowgrass (Poa eigii) in the JudeanDesert, Israel, Journal of Arid Environments 35, 77-86.

Danin, A. and Garty, J. (1983) Distribution of cyanobacteria and lichens on hillsides of the Negev Highlandsand their impact on biogenic weathering, Zeitschrift für Geomorphologie 27, 423-444.

Eckardt, F. D. and Spiro, B. (1999) The origin of sulphur in gypsum and dissolved sulphate in the CentralNamib Desert, Namibia, Sedimentary Geology 123, 255-273.

Friedmann, E. I. (1982) Endolithic microorganisms in the Antarctic cold desert, Science 215, 1045-53.Golubic, S., Friedmann, E. and Schneider, J. (1981) The lithobiontic ecological niche, with special reference

to microorganisms, Journal of Sedimentary Petrology 51, 475-478.Goudie, A. S. (1972) Climate, weathering, crust formation, dunes, and fluvial features of the Central Namib

Desert, near Gobabeb, South West Africa, Madoqua series 2, 1, 15-31.Goudie, A. S. (1997) Weathering processes, in: Thomas, D.S.G. (ed.) Arid zone geomorphology.

edition, Chichester: John Wiley and Sons Ltd., 25-39.Goudie, A. S., Viles, H. A. and Parker, A. G. (1997) Monitoring of rapid salt weathering in the central

Namib Desert using limestone blocks, Journal of Arid Environments 37, 581-98.Kappen, L., Lange, O. L., Schulze, E.-D., Buschboom, U. and Evenari, M. (1980) Ecophysiological

investigations on lichens of the Negev Desert. VII The influence of the habitat exposure on dewimbibition and photosynthetic productivity, Flora 169, 216-229.

Le Campion Alsumard, T. (1979) Les cyanophycées endolithes marines. Systématique, ultrastructure,écologie et biodestruction, Oceanologica Acta 2, 143-56.

82 VILES AND GOUDIE

Lowdermilk, J.D. and Woodruffe, A.O. (1932) Concerning rillensteine, American Journal of Science series5, 23, 135-43.

McCarroll, D. and Viles, H. A. (1995) Rock-weathering by the lichen Lecidea auriculata in an arctic-alpineenvironment, Earth Surface Processes and Landforms 20, 199-206.

Mustoe, G. E. (1982) The origin of honeycomb weathering, Geological Society of America Bulletin 93,108-115.

Olivier, J. (1995) Spatial distribution of fog in the Namib, Journal of Arid Environments 29, 129-138.Schieferstein, B. and Loris, K. (1992) Ecological investigations on lichen fields on the Central Namib. 1.

Distribution patterns and habitat conditions. Vegetatio 98, 113-128.Smith, B. J. (1987) An integrated approach to the weathering of limestone in an arid area and its role in

landscape evolution: a case study in S E Morocco, in: Gardiner, V. (ed.) International Geomorphology1986. Chichester: Wiley, Vol. II, 637-57.

Smith, B. J. (1994) Weathering processes and forms, in: Abrahams, A.D. and Parsons, A.J. (eds)Geomorphology of desert environments. London: Chapman and Hall, 39-63.

Southgate, R. I., Masters, P. and Seely, M. K. (1996) Precipitation and biomass changes in the Namib Desertdune ecosystem, Journal of Arid Environments 33, 267-80.

Sweeting, M. M. and Lancaster, N. (1982) Solutional and wind erosion forms on limestone in the CentralNamib Desert, Zeitschrift für Geomorphologie 26, 197-207.

Sun, H.J. and Friedmann, E.I. (1991) Long-term ( to years) biogenous weathering and dynamics ofmicrobial growth in Ross Desert sandstone. (Abstract). Transactions, American Geophysical Union 72,102.

Viles, H. A. (1987) Blue-green algae and terrestrial limestone weathering on Aldabra Atoll: an SEM andlight microscope study, Earth Surface Processes and Landforms 12, 319-330.

Walter, H. (1986) The Namib Desert. In: Evenari, M., Noy-Meir, I. And Goodall, D.W. Hot deserts andarid shrublands B, Ecosystems of the World 12B. Amsterdam: Elsevier, 245-282.

Young, A. (1987) Salt as an agent in the development of cavernous weathering, Geology 15, 962-966.

WARM SEASON LAND SURFACE – CLIMATEINTERACTIONS IN THE UNITED STATESMIDWEST FROM MESOSCALE OBSERVATIONS

J. O. ADEGOKE and A. M. CARLETONDepartment of Geography and Earth SystemScience Center, The Pennsylvania StateUniversity, University Park PA 16802, U.S.A.

Abstract

The United States Midwest over the last two decades has experienced marked warm seasonclimate anomalies, including droughts and major floods. While the development of theseextreme events can usually be traced to anomalies in atmospheric circulation, and mayinclude teleconnections, studies based on model simulations have shown that land surfaceforcing may be partly responsible for the persistence of these climate anomalies. This studyevaluates the presence and strength of long-term land surface-climate interactions in theU.S. Midwest. We do this via an analysis of the cross-seasonal (spring and summer)associations between temperature and moisture (Palmer Drought Severity Index-PDSI, CropMoisture-Z Index, and precipitation) anomalies. Direct and lag correlations for the 1895-1995 and 1948-1995 periods show that warm and dry summers tend to follow warm springseasons. These results imply that springtime precipitation anomalies may help to determinethe temperature regime of the following summer, possibly via the moisture content of theupper soil. We also show that broad land cover types tend to modulate summer climateanomalies in the U. S. Midwest.

1.

Over the last two decades, there has been great interest in understanding the nature andcauses of interactions among the biosphere, oceans, ice and atmosphere. In particular, thecoupling between the terrestrial biosphere and the atmosphere has been shown to operate ontime and space scales ranging from hours to decades and from plot level to regional andeven global scales (Yeh et al., 1984; Pielke et al., 1993).

83


Introduction

84 ADEGOKE AND CARLETON

Anthropogenic activities that modify terrestrial vegetation, such as large scale deforestationfor logging and agricultural purposes, urban expansion and industrial activities such as stripmining, are now known to affect local to global climate conditions by altering the surfaceenergy and moisture budgets and, thereby, the characteristics of the planetary boundarylayer (PBL). The primary physical processes involved in these vegetation-climateinteractions are changes in surface albedo, soil heat flux, roughness length, and thepartitioning of sensible to latent heat fluxes (or Bowen ratio), as they influence moisture andconvection in the PBL. Early evidence of these effects came from studies of the devastatingSahel droughts of the early 1970s, which increased concerns about desertification inmarginal areas (Hammer, 1970; Charney et al., 1977; Yeh et al., 1984).

Our understanding of the scale interactions of land-atmosphere processes over differentsurfaces has been derived, in particular, from intensive field experiments in drier areas andfor generally restricted time periods (Li and Avissar, 1994; Bonan et al., 1993). Theseinclude the International Land Surface Climatology Project (ISLSCP), HAPEX-MOBILHY(Hydrologic Atmospheric Pilot Experiment and Modelisation du Bilan Hydrique),HAPEX-Sahel and the southwest Australian “Bunny Fence” Experiment, or BUFEX (e.g.,Andre et al., 1986; Sellers et al., 1992; Lyons et al., 1993; Goutorbe et al., 1989).Observational data from these field experiments have proved very useful for calibratingmodel simulations of interactions between the Earth’s land surface and the atmosphere. Inparticular, these calibrations have provided vegetation and soil property information fordifferent land cover types, and have led to improvements in the fidelity with which surfacebiosphere models reproduce vegetation-climate interactions. However, the results of theobservational studies are not directly transferable to more humid areas having differentvegetation and soil moisture, or ambient atmospheric, conditions (e.g., Gibson and VonderHaar, 1990; Raymond et al., 1994; Travis, 1997). Moreover, they are not easily "scaled up"to a region the size of a typical synoptic system (Shuttleworth, 1991; Molders and Raabe,1996). Following the earlier field experiments, similar studies have been conducted or areplanned for other regions, for example, NOPEX (Northern Hemisphere Climate-ProcessesLand-Surface Experiment) and LBA (The Large scale Biosphere-Atmosphere Experiment inAmazonia). These experiments are coordinated under the Biosphere Aspects of theHydrological Cycle (BAHC) research initiative of the International Geosphere BiosphereProgram (IGBP). These research initiatives are designed to generate new knowledge neededto understand the climatological, ecological, biogeochemical and hydrological functioningof the Earth system and the impact of land use change on these functions. They underscorethe need for continued observational studies in humid mid-latitude regions.

In this chapter, we present our observational results to date on the role of land surfaceconditions in the warm season (April through September) climate of the humid lowlandregion of the Midwest U.S. centered on the "Corn Belt". This area encompasses the five core

LAND SURFACE-CLIMATE INTERACTIONS 85

Midwest States of Illinois, Iowa, Indiana, Ohio and Missouri, and the adjoining states ofMichigan, Wisconsin, Minnesota and Kentucky (Figure 1). This economically vitalagricultural region is susceptible to substantial interannual and interdecadal variations insummer climate. The severe drought of 1988 and devastating floods of 1993 are recentexamples of extreme summertime climate anomalies that affected much of the Central U.S.These two events resulted in an estimated $52 billion in farm losses and property damage inthe Midwest (Lott, 1993). Our recent and ongoing research efforts have focused on the inter-relations between Midwest land cover (primarily vegetation, evapotranspiration: ET, surfaceand soil moisture) and climate parameters (temperature, rainfall and convective cloudiness)and their expression across a range of spatial and temporal scales. These investigations arebased on the analysis of long term climate division and digital land cover data, used inconjunction with satellite data (the Advanced Very High Resolution Radiometer-AVHRRderived vegetation index, the Geosynchronous Operational Environmental Satellite-GOES).We used these datasets to determine land surface - climate interactions for sub-areas andtime periods characterized by contrasting vegetation status, surface moisture andatmospheric conditions.

2. Midwest Land Surface-Climate Associations from Historical Data

Over the last two decades, the U.S. Midwest (Fig. 1) has experienced marked warm seasonclimate anomalies, including drought and major floods (Changnon and Kunkel, 1992;Lozano-Garcia et al., 1995). While the development of these extreme events can usually betraced to anomalies in atmospheric circulation, and may include remote forcing viateleconnections, studies have suggested that land surface forcing may be partly responsiblefor the persistence of these climate anomalies (McNab, 1989; Namias, 1991). Specifically,surface moisture availability has been implicated in the prolonged persistence of recentanomalous wet and dry spells in the Midwest (Betts et al., 1994; Wetzel et al., 1996).

In approaching the role of land surface conditions in recent Midwest U.S. climate anomalies,we conducted an analysis of the cross-seasonal associations between temperature andmoisture (precipitation, Palmer Drought Severity Index: PDSI, Crop Moisture or Z-index)anomalies for the 1895-1995 period and also sub-periods. The primary data set used is themonthly state climate division temperature, precipitation and drought indices (PalmerDrought Severity Index-PDSI and Crop Moisture, or Z Index) for the full period, availablefrom the National Climatic Data Center (NCDC). The PDSI is a drought index based on theconcept of the water balance, and was developed to use monthly temperature andprecipitation as input data. The Palmer model uses parameters for evapotranspiration(calculated using the Thornthwaite water balance model), and parameters for soil-moisturerecharge, runoff, and water capacity of the soil. The values given by the index range from -7to +7, with negative values denoting dry spells and positive values indicating wet spells. The


monthly Crop Moisture Index (Z) is used to assess short-term crop needs versus availablewater in the upper 5-feet (1.52 meters) of the soil profile, and is a measure of the departurefrom normal of the moisture climate for a given month (Karl, 1986). Of the two indices, thePDSI is slower to respond to changes in environmental conditions owing to the inclusion ofa lag term from the previous month. Therefore, the PDSI is more indicative ofclimatological drought severity. Other data sets include gridded daily 700mb height data andanomaly charts for the U.S. (1948-1993), and U.S. crop moisture index charts for selectedcase study years. The tropospheric height data are used to identify the surface climateassociations with synoptic scale atmospheric circulation.

The cross-seasonal (spring and summer) associations between the anomalies of temperatureand moisture (precipitation, PDSI and Z-index) were determined via direct and lagcorrelations for the 1895-1995 and 1948-1995 periods. Spring (MAM) and Summer (JJA)


seasonal averages for each climate variable were first calculated and then aggregated forfive core states (Illinois, Iowa, Missouri, Ohio and Indiana) and also for the entire Midwestregion (five core states plus Michigan, Wisconsin, Minnesota and Kentucky). This is tocheck on the stability of the correlations so derived. The results are similar to thoseidentified in studies for other regions, and the U.S. as a whole (Namias, 1960, 1991; Karl,1986; Huang et al., 1996). They show, in particular, that warm and dry summers tend tofollow warm spring seasons (Table 1).

The correlations for the expanded region (Table 2) show very little sensitivity to domainsize. In both cases (i.e., the core and expanded Midwest region) there are fewer significantcorrelations in the recent time period, although they are all of the same (negative) sign.These results imply that springtime precipitation anomalies may help to determine thetemperature regime of the following summer, possibly via the moisture content of the uppersoil.


To address the extent to which the rainfall-temperature associations so revealed are due toland surface forcing (e.g., soil moisture), or to persistence of atmospheric circulationpatterns across seasons (e.g., Simmonds, 1993), we classified surface and mid-tropospheric(500mb) synoptic circulation over the Midwest using manual methods (Arnold, 1994), forthe spring (MAM) and summer (JJAS) seasons of the period 1948-95. We used aclassification scheme that stratified the monthly-averaged 500mb height charts into sixsynoptic categories: strong zonal flow, weak zonal flow, strong meridional flow, weakmeridional flow, cyclonic circulation and anticyclonic circulation. We also noted thepresence or absence of storms in the study area for individual months during the 1948-95period with a storm track index derived by simply tallying the number of storms recordedfor each month. The results of this analysis show that during the wetter spring months, thecirculation regime was mostly cyclonic with a strong zonal mean flow configuration. Thecirculation regime during the warmer summer months, on the other hand, waspredominantly anticyclonic and meridional. This implies that, in the mean, monthly toseasonal large-scale circulation tends to be the major influence on the warm season climateof the Midwest.

Notwithstanding this result, large-scale circulation does not fully account for the observedsummer climate surface conditions in all situations, particularly during anomalously wet ordry years. Composite summer circulation indices for the eight driest (1953, 1957, 1960,1967, 1976, 1983, 1988, 1991) and six wettest (1961, 1972, 1977, 1986, 1992, 1993) yearsshow that the cross-seasonal coherence of surface climate anomalies indicated in Tables 1and 2 above (e.g., dry spring and warm dry summer following) are maintained acrossdifferent circulation patterns. For example, the 1988 severe summer drought persisted forseveral weeks following the flattening of a strong ridge that had stalled over the Midwest inearly to mid summer (Figure 2 a, b, c). Throughout the central Midwest, abnormally dry soilmoisture conditions prevailed throughout July and August as PDSI values stayed between –4 and –7 (Figure 3 a, b, c), despite the switch in circulation around mid-July. The reducedsoil moisture is likely to have helped to maintain or even to amplify the 1988 summerdrought by enhancing surface sensible heating, and reducing the local evaporativecontribution to the atmosphere (Kunkel, 1989).


3. Surface Heterogeneity and Midwest Climate Anomalies

To further define the role of mesoscale land surface conditions in the warm season climatedynamics of the U.S. Midwest, we address the issue of whether the strength of theassociations between land surface conditions and summer climate anomalies differ amongcover types, and what may have happened to these associations as the boundaries betweenmajor types have evolved over the last century with increased human activities. We do thisby stratifying the Midwest climate division temperature and surface moisture data according


to dominant land cover types. Midwest climate divisions were stratified into one of fivebroad land cover categories and the 100-year NCDC climate data were re-analyzed on thebasis of this stratification. The land cover categories are as follows: cropland, mixedcropland/urban, mixed cropland/forest, mixed forest/cropland/urban, and forest. TheMidwest land use/land cover (LU/LC) map used to stratify the climate divisions wasproduced from the online LU/LC 1:250,000 digital data provided by the United StatesGeological Survey (USGS). The USGS compiled the LU/LC maps from aerial photographsacquired by the US National Aeronautic Space Administration (NASA) high-altitudemissions during the 1980s. Secondary sources from earlier land use maps and field surveyswere also incorporated into the LU/LC maps as needed. The maps were subsequentlydigitized to create a national digital LU/LC database. The USGS mapped and coded land usein the 1:250,000 quadrangles using the Anderson classification system (Anderson et. al.,1976), for levels one and two.

The data, originally provided as quadrangles in the geographic information retrieval andanalysis system (GIRAS) format, were converted into ARC/INFO export format andimported into the ARC/INFO software environment. The quads were then edge-matched andcombined to produce a georeferenced digital LU/LC map of the Midwest (Fig. 4). A climatedivision map of the Midwest was overlain onto the LU/LC map and each division assignedinto one of the five land use types, based on the dominant land cover type associated withparticular climate divisions.

Five-year running means of the summer temperature and Z-index time series (standardizedanomalies) for the five broad land cover types (Figures 5 and 6) show significant differencesbetween forest and the other land cover types. The major difference is the damping of theamplitudes of the anomalies of both climate variables (i.e., temperature and Z-index) in theforest regions. Cross correlation of the time series shows that forest correlates very poorlywith all other land cover types (see the highlighted numbers on Table 3). Similar resultswere obtained for seasonal precipitation and PDSI anomalies (not shown). It is plausible thatfactors other than land cover, such as topography, may be partly responsible for this lack ofcorrelation between climate variables for forest and other land cover types, althoughtopography is likely to be a minor factor in the Midwest “flatlands”. This result suggests thatthe conversion of forest area to other land cover types, such as cropland and urban surfaces,may enhance the magnitude of interannual anomalies in summer-time precipitation andtemperature.


4. Conclusion

Land surface-climate interactions in the warm season have been relatively little studied forhumid areas of middle latitudes, in contrast with those areas that are more marginal in termsof rainfall. The humid lowlands of the Midwest U.S. “Corn Belt” comprise a close to ideallaboratory for undertaking such an analysis, given the lack of appreciable topographicchange over wide areas; the variety of natural and human-made land covers, and theirheterogeneity on a range of spatial scales; the summertime peak in precipitation occurrence;the marked interannual variability of climate; and the economic significance of thisdominantly agricultural region. Our preliminary studies, documented in this chapter, suggestthat human modifications to the land surface of this region are detectable in the climaterecord of almost the last 100 years. These impacts are most evident when the prevailingsynoptic situation is characterized by weak wind flow near the surface and aloft on dailytime scales, and seasonally, during times of more extreme anomalies of precipitation(especially droughts).

The availability of high resolution satellite data, GIS, and methodologies for multi-scaleanalysis such as fractal techniques, offer promise of quantifying the relationships betweenMidwest land surface conditions (land cover, soil moisture) and a range of climate andatmospheric indices including convective clouds. A key area of application of thesetechniques is in assessing the scale invariance of surface-atmosphere relationships across


given scale ranges. These analyses, which we are currently undertaking, go beyond thesimpler, yet still necessary, linear correlation studies that show a land cover-climateassociation over the U.S. Midwest region. Important questions, which we will soon be ableto answer, include the following: Over what scale range(s) does the land surface convectivecloud relationship hold? What is the relative importance of surface heterogeneity andatmospheric conditions in determining when and where convective clouds develop in theMidwest? The answers to these more immediate questions will enable the role of landsurface–atmosphere interactions in the Midwest climate variations to be assessed, andshould better permit the role of land surface–atmosphere interactions to be considered inshort- to medium-range weather predictions for the Midwest. They should also help toinform modelers seeking to improve the parameterization of important surface processes inGCMs (e.g., the areal integration of energy flux point measurements), and also regionalmeteorological models for more humid locations.

Acknowledgements

The authors wish to thank two anonymous reviewers for their valuable comments andsuggestions on the manuscript. The support of Penn State’s Earth System Science Center(now Environment Institute) is also gratefully acknowledged. Partial funding for thisresearch came from NSF grant ATM 98-76753. Jason Allard assisted in preparing some ofthe graphics included in this paper.

References

Anderson, J.R., Hardy, E.E., Roach, J.T. and Witmer, R. E., (1976). ‘A land use and land cover classificationsystem for use with remote sensing data’ U.S. Geological Survey Professional Paper, 964, 28p.

Andre, J.-C., Goutorbe, J.-P. and Perrier, A., (1986). ‘HAPEX-MOBILHY: a hydrological atmospheric experimentfor the study of water budget and evaporation flux at the climatic scale’ Bull. Amer. Meteor. Soc., 67, 138-144.

Arnold, D.L., (1994). Synoptic and Mesoscale Climatologies of Severe Local Storms for the American MidwestPh.D. dissertation, Dept. of Geography, Indiana University, Bloomington. 257p.

Betts, A. K., Hall, J.H., Beljaars, A.C.M., Miller, M.J., and Viterbo, P.A., (1994). ‘Coupling between land surfaceboundary layer parameterizations and rainfall on local and regional scales: Lessons from the wet summer of1993’ Preprints, Fifth Symp. On Global Change Studies, Nashville, Amer. Meteor. Soc., 174-181

Bonan, G.B., Pollard, D. and Thompson, S.L., (1993). ‘Influence of subgrid-scale heterogeneity in leaf-area index,stomatal resistance, and soil moisture on grid-scale land-atmosphere interactions’ J. Climate, 6. 1882-1897.

Changnon, S.A. and Kunkel, K.E., (1992). ‘Assessing impacts of a climatologically unique year (1990) in theMidwest’ Phys. Geog., 13, 180-190.

Charney, J.G., Quirk, W.J., Chow, S.H., and Kornfield, J., (1977). ‘A comparative study of the effects of albedochange on drought in semi-arid region’ J. Atmos. Sci., 34, 1336-1385.

Gibson, H.M. and Vonder Haar, T.H., (1990). ‘Cloud and convection frequencies over the southeast United Statesas related to small-scale geographic features’ Mon. Weath. Rev., 118, 2215-2227.


Goutorbe, J.-P., Noilhan, J., Valancogne, C. and Cuenca, R.H., (1989). ‘Soil moisture variations during HAPEX-MOBILHY’ Annals. Geophysics., 7, 415-425.

Hammer, R.M., 1970. ‘Cloud development and distribution around Khartoum’ Weather, 28, 411-414.Huang, J., van den Dool, H. and Georgakakos, K.P., (1996). ‘Analysis of model-calculated soil moisture over the

United States (1931-1993) and applications to long-range temperature forecasts’ Journal of. Climate, 9, 1350-1362.

Karl, T.B., (1986). ‘The relationship of soil moisture parameterizations to subsequent seasonal and monthly meantemperature in the United States’ Mon. Weath. Rev., 114, 675-686.

Kunkel, K.E., (1989). ‘A surface energy budget view of the 1988 Midwestern United States drought’ Bound.-LayerMeteorology, 48, 217-225.

Li, B. and Avissar, R., (1994). ‘The impact of spatial variability of land surface characteristics on land-surface heatfluxes’ Journal of. Climate, 7, 527-537.

Lott, N., (1993). ‘The summer of 1993: Flooding in the Midwest and drought in the Southeast’ National ClimaticData Center Research Customer Service Group Technical Report, 93-04. 21 p.

Lozano-Garcia, D.F., Fernandez, R.N., Gallo, K.P. and Johannsen, C.J., (1995). ‘Monitoring the 1988 severedrought in Indiana, U.S.A. using AVHRR’ International. Journal of Remote Sensing, 16, 1327-1340.

Lyons, T.J., Schwerdtfeger, P., Hacker, J.M., Foster, I.J., Smith, R.C.G. and Xinmei, H., (1993). ‘Land-atmosphereinteraction in a semiarid region: the bunny fence experiment’ Bulletin of the American Meteorological Society,74, 1327-1334.

McNab, A. L., (1989). ‘Climate and drought’ EOS, Transactions of the American Geophysics Union, 70, 882-883.Molders, N. and Raabe, A., (1996). ‘Numerical investigations on the influence of subgrid-scale surface

heterogeneity on evapotranspiration and cloud processes’ Journal of Applied Meteorology, 35, 782-795.Namias, J., (1960). ‘Factors in the initiation, perpetuation and termination of drought’ Extract of Publ. No. 51,

IASH Commission of Surface Waters, 81-94. [Avail. from IASH, UNESCO, Paris].Namias, J., (1991). ‘Spring and summer 1988 drought over the contiguous United States-causes and prediction’

Journal of Climate, 4, 54-65.Pielke, R.A., Dalu, G.A., Lee, T.J., Rodriguez, H., Eastman, J. and Kittel, T.G.F., (1993). ‘Mesoscale

parameterization of heat fluxes due to landscape variability for use in general circulation models, In: ExchangeProcesses at the Land Surface for a Range of Space and Time Scales (Proc. of the Yokohama Symp., July1993). IAHS Publ. No. 212. 331-342.

Raymond, W.H., Rabin, R.M. and Wade, G.S., (1994) Evidence of an agricultural heat island in the lowerMississippi river floodplain’ Bulletin American Meteorological Society, 75, 1019-1025.

Sellers, P.J., Hall, F.G., Asrar, G., Strebel, D.E., and Murphy, R.E., (1992) ‘An overview of the First InternationalSatellite Land Surface Climatology Project (ISLSCP) Field Experiment (FIFE)’ Journal of GeophysicalResearch, 97, 18,345-18,371.

Shuttleworth, W.J., (1991). ‘Insight from large-scale observational studies of land/atmosphere interactions’ In:Wood, E.F. (ed.) Land Surface-Atmosphere Interactions for Climate Modeling: Observations, Models andAnalysis. Kluwer, Dordrecht. p. 3-20.

Travis, D.J., (1997). ‘An investigation of Wisconsin’s anthropogenically-generated convergence boundary andpossible influences on climate’ Wisconsin Geogr., 12, 34-46.

Wetzel, P.J., Argentini, S., and Boone, A., (1996) ‘Role of land surface in controlling daytime cloud amount: twocase studies in the GCIP-SW area’ J. Geophys. Res., 101, 7359-7370.

Yeh, T-C., Wetherald, R.T. and Manabe, S., (1984) ‘The effect of soil moisture on the short-term climate andhydrology change- a numerical experiment’ Mon. Weath. Rev, 112, 474-490.

STREAMFLOW CHANGES IN THE SIERRA

NEVADA, CALIFORNIA, SIMULATED USING

A STATISTICALLY DOWNSCALED GENERAL

CIRCULATION MODEL SCENARIO OF CLIMATE

CHANGE

ROBERT L. WILBYDivision of GeographyUniversity of Derby, Kedleston Road,Derby, DE22 1GB, UK@National Center for Atmospheric ResearchBoulder, Colorado, 80307-3000, USA

MICHAEL D. DETTINGERU.S. Geological SurveyWater Resources Division, CaliforniaScripps Institution of Oceanography9500 Gilman Drive, La Jolla, California,92093-0224

Abstract

Simulations of future climate using general circulation models (GCMs) suggest that risingconcentrations of greenhouse gases may have significant consequences for the globalclimate. Of less certainty is the extent to which regional scale (i.e., sub-GCM grid)environmental processes will be affected. In this chapter, a range of downscaling techniquesare critiqued. Then a relatively simple (yet robust) statistical downscaling technique and itsuse in the modelling of future runoff scenarios for three river basins in the Sierra Nevada,California, is described. This region was selected because GCM experiments driven bycombined greenhouse-gas and sulphate-aerosol forcings consistently show major changes inthe hydro-climate of the southwest United States by the end of the 21st century.

99


100 STREAMFLOW CHANGES IN CALIFORNIA

The regression-based downscaling method was used to simulate daily rainfall andtemperature series for streamflow modelling in three Californian river basins under current-and future-climate conditions. The downscaling involved just three predictor variables(specific humidity, zonal velocity component of airflow, and 500 hPa geopotential heights)supplied by the U.K. Meteorological Office couple ocean-atmosphere model (HadCM2) forthe grid point nearest the target basins. When evaluated using independent data, the modelshowed reasonable skill at reproducing observed area-average precipitation, temperature,and concomitant streamflow variations. Overall, the downscaled data resulted in slightunderestimates of mean annual streamflow due to underestimates of precipitation in springand positive temperature biases in winter.

Differences in the skill of simulated streamflows amongst the three basins were attributed tothe smoothing effects of snowpack on streamflow responses to climate forcing. The Mercedand American River basins drain the western, windward slope of the Sierra Nevada and aresnowmelt dominated, whereas the Carson River drains the eastern, leeward slope and is amix of rainfall runoff and snowmelt runoff. Simulated streamflow in the American Riverresponds rapidly and sensitively to daily-scale temperature and precipitation fluctuations anderrors; in the Merced and Carson Rivers, the response to the same short-term influences ismuch less. Consequently, the skill of simulated flows was significantly lower in theAmerican River model than in the Carson and Merced.

The physiography of the three basins also accounts for differences in their sensitivities tofuture climate change. Increases in winter precipitation exceeding +100% coupled withmean temperature rises greater than +2°C result in increased winter streamflows in all threebasins. In the Merced and Carson basins, these streamflow increases reflect large changes inwinter snowpack, whereas the streamflow changes in the lower elevation American basinare driven primarily by rainfall runoff. Furthermore, reductions in winter snowpack in theAmerican River basin, owing to less precipitation falling as snow and earlier melting ofsnow at middle elevations, lead to less spring and summer streamflow.

Taken collectively, the downscaling results suggest significant changes to both the timingand magnitude of streamflows in the Sierra Nevada by the end of the 21st Century. In thehigher elevation basins, the HadCM2 scenario implies more annual streamflow and morestreamflow during the spring and summer months that are critical for water-resourcesmanagement in California. Depending on the relative significance of rainfall runoff andsnowmelt, each basin responds in its own way to regional climate forcing. Generally, then,climate scenarios need to be specified – by whatever means – with sufficient temporal andspatial resolution to capture subtle orographic influences if projections of climate-changeresponses are to be useful and reproducible.

WILBY AND DETTINGER 101

1. Introduction

Simulations of future climate using general circulation models (GCMs) suggest that risingconcentrations of greenhouse gases may have significant consequences for the globalclimate. Of less certainty is the extent to which regional scale (i.e., sub-GCM grid)environmental processes will be affected. This is because the length scales of GCMs (whichare typically about 200 kilometres) are too coarse to resolve complex orography andimportant sub-grid scale processes such as convective precipitation. Furthermore, GCMoutput representing the surface climate under current conditions is commonly unreliable atthe scale of individual grid points (see Table 1 below). Ironically, these are the scales thatare likely to be of greatest interest to resource managers who have functional responsibilitiesthat cover relatively small geographical areas. In other words, there is a scale mismatchbetween the scale of global change scenarios and the data requirements of the impactsanalyst (Hostetler, 1994). “Downscaling” techniques have subsequently emerged as a meansof bridging the gap between what climatologists currently are able to supply and whatregional climate-change impact studies require.

In this chapter, a range of downscaling techniques are critiqued. Then a relatively simple(yet robust) statistical downscaling technique and its use in the modelling of future runoffscenarios for three river basins in the Sierra Nevada, California, is described. This regionwas selected because GCM experiments driven by combined greenhouse-gas and sulphate-aerosol forcings consistently show major changes in the hydroclimate of the southwestUnited States by the end of the 21st century. Shown in Figure 1, for example, are projectedchanges in winter (DJF) precipitation from the Canadian Centre for Climate Modelling andAnalysis (CGCM1) and by the U.K. Meteorological Office's Hadley Centre for ClimatePrediction and Research (HadCM2) transient climate-change simulations (Flato et al.,1999; Boer et al., 1999a,b; Johns et al., 1997; Mitchell and Johns, 1997). Both experimentshave been central to the U.S. National Assessment of the Potential Consequences of ClimateVariability and Change (see http://www.nacc.usgcrp.gov/), and both predict increases inwinter precipitation over California by 2090-99.

Although the scenarios in Figure 1 (along with accompanying temperature increases) implymajor changes in regional snowpack, snowmelt, and runoff, confidence in HadCM2scenarios at the basin scale is low. For example, Table 1 shows significant differencesbetween observed and HadCM2-derived precipitation statistics for the heterogeneouslandscape of the San Juan basin, Colorado. This deficiency is addressed herein by exploitingobserved correlations between climate variables at the GCM grid-scale (such as geopotentialheight fields) and daily weather at the station scale (such as single-site precipitation). Theseempirical relationships are used to project future changes in atmospheric circulation and


humidity in the HadCM2 climate-change scenarios to the station scale. A hydrologicalmodel is then used to simulate streamflows in each basin under the downscaled current- andfuture-climate conditions.


2. Downscaling Techniques

The theory and practice of downscaling has been reviewed elsewhere (see Giorgi andMearns, 1991; Wilby and Wigley, 1997; Wilby et al., 1998b). Therefore, we provide only abrief overview of the main downscaling approaches, namely (a) dynamical, (b) weathertyping, (c) stochastic, and (d) regression-based methods.

2.1 DYNAMICAL

Dynamical downscaling includes the nesting of a high-resolution regional climate model(RCM) within a GCM (Christensen et al., 1997; McGregor, 1997). The RCM uses the GCMto define time-varying atmospheric boundary conditions around a finite domain, withinwhich the physical dynamics of the atmosphere are modelled using horizontal grid spacingsof 20-50 km. The main limitation of RCMs is that they are as computationally demanding asGCMs (placing constraints on the domain size, number of experiments, and duration ofsimulations). However, RCMs can better resolve smaller scale atmospheric features, such asorographic precipitation, than the host GCM (Jones et al., 1995) and are able to respond inphysically consistent ways to different external forcings such as land-surface oratmospheric-chemistry changes (Giorgi and Mearns, 1999).

2.2 WEATHER TYPING

Weather-typing approaches involve the stratification of local meteorological variations byconcomitant, synoptic-scale (1000 km) atmospheric circulation patterns (e.g., Hay et al.,1992; Matyasovsky et al., 1994). Future regional climate scenarios are then constructed byresampling observed variables from probability distributions conditioned on synthetic seriesof circulation patterns (e.g., Bardossy and Plate, 1992; Dettinger and Cayan, 1992; Goodessand Palutikof, 1998). The main appeal of circulation-based downscaling is that it is foundedon sensible linkages between climate on the large scale, which GCMs are best suited to


project, and weather at the local scale. The technique is also readily applicable to a widevariety of environmental variables and can preserve some of the spatial auto-correlationbetween multiple sites and multiple variables (e.g., precipitation and temperature). However,weather-typing schemes commonly are parochial, have difficulty simulating extreme events,and must assume stationary circulation-to-surface climate conditioning (Wilby, 1997).Precipitation scenarios produced by circulation changes alone are also relatively insensitiveto future climate forcing (see Wilby et al., 1998b).

2.3 STOCHASTIC

The most popular stochastic downscaling approach involves modifying parameters inconventional weather generators such as Richardson’s (1981) Weather-GENeration program(WGEN). The standard WGEN program simulates precipitation occurrence using a two-state, first-order Markov chain; precipitation amounts on wet days using a gammadistribution; and temperature and radiation components using first-order trivariateautoregression that is conditional on precipitation occurrence. Future-climate scenarios aregenerated stochastically using revised parameter sets that have been scaled in directproportion to the corresponding variable changes in a GCM (Wilks, 1992). The mainadvantage of the technique is that it can exactly reproduce key climate statistics and hasbeen widely used for climate-impact assessment (e.g., Mearns et al., 1996). The keydisadvantages relate to the arbitrary manner in which model parameters are changed forfuture-climate conditions, to the unanticipated effects that these changes can have onconditional variables (Katz, 1996), and to the poor representation of interannual variabilityin stochastic models (Gregory et al., 1993).

2.4 REGRESSION

Regression-based downscaling methods employ empirical relationships between localscale/single-site predictand(s) and synoptic-scale predictor(s). Techniques differ accordingto the choice of mathematical transfer function, predictor variable suite, or statistical-fittingprocedure. Methods include linear and non-linear regression, artificial neural networks,canonical correlation, and principal components analyses (e.g., Conway et al., 1996; Craneand Hewitson, 1998; von Storch et al., 1993). The main strengths of regression downscalingare the relative ease of application and the parsimony of the models. However, regressionmodels typically explain only a fraction of the observed climate variability. In common withweather-typing methods, stationarity of the empirical relationships is also assumed, anddownscaled scenarios can be sensitive to the choice of predictor variables and regressionmethod (Winkler et al., 1997). Still, regression models provide an efficient compromise


between simpler, purely stochastic weather generators and computationally expensive,dynamical models. Regressions are inexpensive to apply but are able to reproducephysically realistic intervariable, temporal, and spatial relationships as well as sequences inpredicted fields that are present in historical records.

3. Data and Modelling Methods

In the remainder of this chapter, we describe the application of a regression-baseddownscaling model to streamflow simulation under current- and future-climate scenarios.Two sets of GCM output were used: the first to calibrate and then verify the coupleddownscaling-hydrological model, and the second to downscale GCM output in order tosimulate future streamflow in the Sierra Nevada.

3.1 PREDICTOR VARIABLES

Table 2 lists 15 candidate variables that were originally selected by Wilby et al. (1999) forpossible use as downscaling predictors. All variables were derived from combinations ofdaily grid-point estimates of mean sea level pressure (mslp); 500 hPa geopotential height(H); 2-metre (near-surface) temperature (T2m); and 0.995-sigma-level (near-surface)relative humidity (RH), obtained from the National Center for Environmental Prediction /National Center for Atmospheric Research (NCEP/NCAR) Reanalysis (Kalnay et al., 1996)of atmospheric observations for the period 1979 to 1995. The Reanalysis estimates were re-gridded from the NCEP grid (1.875° of latitude by 1.875° of longitude) to the 2.5° of latitudeby 3.75° of longitude grid on which climate variations are represented in the HadCM2simulations. The pressure data were used to calculate five daily airflow indices (U, V, F, Z,D) for both the surface (s) and the 500-mb level in the upper (u) atmosphere, according to amethodology described by Jones et al. (1993). Daily mean temperatures and relativehumidities were used to estimate daily mean specific humidities (SH) by Richards' (1971)non-linear approximation.


The GCM used to drive the downscaling model in climate-change experiments was the U.K.Meteorological Office Hadley Centre's coupled ocean/atmosphere model (HadCM2) forcedby combined and albedo (as a proxy for sulphate aerosol) changes (Johns et al., 1997;Mitchell and Johns, 1997). In this sulphate-plus-greenhouse gas experiment (their "SUL"experiment), the model run begins in 1861 and is forced with an estimate of historicalradiative conditions to 1993 followed by a projected future-forcing scenario with 1%increases in and sulphate per year from 1994 to 2100. HadCM2 output for the period1980-99 was used as a proxy for the current climate (as in previous downscaling studies,such as Conway et al., 1996; Pilling et al., 1998; Wilby et al., 1998a,b). Output for 2080 to2099 was used to downscale climate conditions arising from future anthropogenic emissionsof greenhouse gases and aerosols.

3.2 STATISTICAL DOWNSCALING MODEL

The statistical downscaling model (Wilby et al., 1999) was calibrated by regressions linkingselected Reanalysis grid-point values as independent predictor variables with daily weatherdata for seven stations in or near the North Fork American, East Fork Carson, and MercedRiver basins (Figure 2) in the Sierra Nevada as dependent variables. The specificpredictands for which regression models were fitted are the daily series of wet-dayoccurrence (O), wet-day amounts (R), and maximum (TMAX) and minimum (TMIN)temperatures. Regression relations were fitted on daily variables for the 10 years from 1979to 1988 and were evaluated using the 7 years from 1989 to 1995. Separate regressions were


undertaken for each station and each or the climatological seasons, winter (DJF), spring(MAM), summer (JJA), and autumn (SON).

All daily predictor variables were normalised using period means and standard deviations(as advocated by Karl et al., 1990) to increase transferability to GCM simulations, whichmay have different means and standard deviations from observed fields. The three mostpowerful predictor variables were selected following a step-wise multiple linear regressionanalysis of the 15 candidate variables listed in Table 2. The chosen predictors were griddedvalues of daily specific humidity (SH), the zonal velocity component of the surfacegeostrophic wind (Us, hereafter referred to as U), and 500 hPa geopotential heights (H).

3.2.1 Daily Precipitation OccurrenceDaily values of a precipitation occurrence parameter (represented by a series of “1”s and“0”s) were regressed against three grid-box predictor variables SH, U, H, and a lag-1autocorrelation function using the following regression equation:


The parameters are fitted by using linear least squares regression. In simulation mode, auniformly distributed random number is used to determine whether precipitationoccurs. For a given site and day, a wet day is synthesized if

3.2.2 Daily Precipitation AmountsWet-day precipitation amounts for a given day are downscaled using the three grid-box predictor variables SH, U, and H. Since is always non-zero, it is appropriate toformulate the following regression model (following Kilsby et al., 1998):

where the are parameters fitted by linear least squares regression and is a randommodelling error. The expected value is given by:

where is an empirically derived correction ratio that allows for the bias resulting from there-transformation of ln(R) to R and the fact that comes from a skewed distribution. Thevalue of is defined such that observed and downscaled precipitation totals are equal forthe calibration period. Additionally, a random scaling factor Ø (with a mean of 1) is used toincrease the variance of R to obtain better agreement with observations (as used by Hay etal., 1991). Note that a lag-1 autoregressive component is not used to model because itsinclusion did not significantly improve the explained variance in wet-day amounts.However, it is acknowledged that this parameter may be appropriate at other locations.

3.2.3 Daily Temperatures andDaily maximum and minimum temperatures for a given day weredownscaled using the three grid-box predictor variables SH, U, and H, and the precedingday’s maximum and minimum temperatures, respectively. The dailytemperature series were modelled using the following regression equations:

and

where and are parameters fitted by linear least squares regression, and and aremodel errors. Both and are assumed to be normally distributed with mean zero andstandard deviation equal to the standard error of the regression equation. Both sets ofresiduals were modelled stochastically using conventional Monte Carlo methods (Wilby etal., 1998a).


3.3 HYDROLOGICAL MODEL

The three river basins are simulated using parameterizations of daily heat and water budgetsin the Precipitation-Runoff Modelling System (PRMS: Leavesley et al., 1983), a physicallybased, distributed parameter model of precipitation forms, snowpack evolution, and runoffgeneration. The spatial variability of land characteristics that affect snowpack and runoff isrepresented by hydrologic response units (HRUs), within which runoff responses to uniformprecipitation or snowmelt inputs are assumed to be homogeneous. HRUs are characterizedand delineated in terms of those physiographic properties that determine hydrologicresponses: elevation, slope, aspect, vegetation, soils, geology, and climate (e.g., Smith andReece, 1995). In the three models used here, HRUs were designed to incorporate all gridcells, on 100-m grids, that share nearly identical combinations of these seven physiographicproperties, regardless of whether the grid cells in an HRU form a contiguous polygon (Jetonand Smith, 1993). The resulting "pixelated" model delineations represent the basins in termsof 50 HRUs in the American River model, 50 HRUs in the Carson River, and 64 HRUs inthe Merced River.

Within each HRU, the heat- and water-budget responses to daily inputs of precipitation anddaily fluctuations of air temperature are simulated. The daily mixes of rain and snow areestimated from each day’s temperatures by interpolations between the temperatures at whichprecipitation historically has been either all snow or all rain (Willen et al., 1971).Interception losses, sublimation, and evapotranspiration are also parameterized andsimulated in terms of precipitation and daily maximum and minimum temperatures. Runoffis partitioned between surface runoff, shallow-subsurface runoff, deep-subsurface runoff,and deep ground-water recharge on the basis of the simulated accumulations of soil moistureat each HRU and of water in deeper subsurface reservoirs that underlie multiple HRUs. Thevarious processes acting on runoff generation from the basins are represented in sufficientdetail that heat- and moisture-fluxes vary realistically with short- and long-term climaticvariations. However, the particular model parameters (such as temperature thresholds forrain to fall) and various land-surface descriptors (such as plant-canopy densities) were notmodified in the future-climate simulations. Thus, the details of the model’s temperature-based parameterizations are assumed, in the present study, to be unchanged under the future-climate scenarios. This simplification amounts to assumptions that precipitation wouldderive from the same heights in the atmosphere as at present and that land-surfaceproperties, such as vegetation type, would not change under the future scenarios.

Snowpack accumulation, evolution, and ultimately the heat and water balances of thesnowmelt periods are critical components in the simulations and are driven by the dailyinputs of precipitation and daily air temperatures (using the parameterizations of Obled andRosse, 1977). Snowmelt in the Sierra Nevada is driven mostly by solar radiation rather thanby direct inputs of heat from the surface-air temperatures (Aguado, 1985). Thus, the


temperature-based parameterizations of daily solar-radiation inputs are an importantcomponent of the models. The method used is a simple correction of clear-sky insolationestimates – from latitude, HRU slope and aspect, and day of the year – using the occurrenceof precipitation and daily maximum air temperatures as crude indicators of the presence orabsence of cloud cover (see Leavesley et al., 1983). Heat deposited in the snowpack by eachday's sunshine is either lost to the overlying atmosphere the following evening if airtemperatures drop below freezing, or is stored to contribute to eventual snowmelt if eveningtemperatures remain warmer than freezing.

These temperature-based snowpack and insolation parameterizations are assumed to beunchanged in the future-climate conditions. Because of the physical detail of the processrepresentations in the models, this assumption is a reasonable simplification. In the models,however, the solar-radiation estimates are functionally tied to daily air temperatures. Thismeans that simulated solar-radiation inputs increase along with air temperatures, and thus, inaddition to being warmer and wetter, the future-climate condition is represented in thehydrological models – almost inadvertently – as also being less cloudy (on dry days) thanthe present condition. As more information describing future relations between dailytemperatures, cloudiness, and solar radiation at the surface becomes available, theparameterization of solar-radiation inputs used in simulations of future climate conditionsmay need to be modified accordingly. In this study, the parameterization was the same in allsimulations.

The Carson and American River models are described in detail by Jeton et al. (1996). TheCarson River model simulates historical streamflows from 1969 to 1998, and the AmericanRiver model simulates streamflow from 1949 to 1998. These simulations are driven byprecipitation and temperature records from two nearby weather stations in the Carson Rivermodel and by weather observations from four nearby stations in the American River model.Indications of the goodness-of-fit of these models are presented by Jeton et al. (1996), andoverall the fits are satisfactory. For example, 97 percent of the observed fluctuations ofannual flow totals in the American River during a validation (non-calibration) period from1949 to 1968 are present in the simulations, and 80 percent of the annual flow fluctuationsof the Carson River are present in simulations during its validation period from 1969 to1979. The Merced River model was designed to simulate daily flows for the period from1916 to present (Dettinger et al., 1999), and the model is driven by precipitation andtemperature observations from two long-term weather stations in the Sierra Nevada for mostof that time (prior to the mid 1930s, only one of the two stations had weather records andthe model was driven with just one input station). From 1937 to 1996, the model captures 77percent of the observed daily flow variability; during the same period, simulated annual flowtotals capture 83 percent of the observed variations.


4. Modelled Streamflow Under Current-Climate Conditions

The calibrated downscaling model was forced using normalized Reanalysis SH, U. and Hpredictor variables for the verification period 1989-95. Statistically downscaled series ofdaily PRCP, TMAX and TMIN at the seven stations then were used to drive the watershedmodels.


The downscaling model, as shown in Figure 3, captures the timing of the July precipitationminimum but underestimates the magnitude of the March maximum. Overall, the model hasa slight dry bias (<3% error), yielding an annual precipitation total of 1117 mm instead of


the observed 1143 mm. The downscaling model has a warm bias for most of the wintermonths, November to January, but a cold bias in February (Figures 3b and 3c). On average,both the downscaled maximum and minimum temperatures are +0.2°C warmer thanobserved data (16.6 and 3.0°C instead of 16.4 and 2.8°C, respectively).

The success of the combination of hydrological models with precipitation and temperaturedownscaling can be measured in terms of biases in the simulated streamflow. Thehydrological simulations driven by downscaled meteorology (Figure 4) generallyunderestimated annual streamflow, whereas simulations driven by station observations ofmeteorology tended to overestimate observed flows. Flows simulated using downscaledmeteorology averaged 90% of observed flows in the Merced, 93% in the Carson, and 92%in the American, in comparison with 106% for all station data.

Despite differences in simulated gross yields, the downscaled data provide a good


approximation of the seasonal streamflow regimes, most notably for the Merced. However,the month of maximum mean streamflow is too early and too low in the case of the Carson,and well timed but underestimated in the American.

A more severe test of the combined hydrological-downscaling model performance isprovided by analysis of simulated daily flows for the downscaling-validation period from1989 to 1995. As shown in Figure 5, for example, simulations forced by the station datahave greater skill on daily time scales than do simulations with the downscaled meteorologyfor the Carson River model. For both simulations, though, the fits are satisfactory, and asignificant component of the overall bias can be attributed to the hydrological model andchoice of stations used for model calibration. Similarly, the correlation skills of flowssimulated using statistically downscaled (station) data in the Merced and American Riverswere r = +0.84 (0.89) and r = +0.67 (0.81), respectively.

5. Modelled Streamflow Under Future-Climate Conditions

Having demonstrated the ability of the combination of the downscaled historical climateconditions with the hydrological models to reproduce realistic historical streamflowvariations, we next simulated streamflow using downscaled GCM simulations. Thedownscaling model – as calibrated with the historical Reanalysis fields – was forced usingdaily SH, U, and H predictors simulated by HadCM2 for current (1980-99) and future(2080-99) climate conditions. The seasonal regimes of the surface variables downscaledfrom the two HadCM2 scenarios are shown in Figure 6, and the corresponding streamflowand snowpack changes simulated by PRMS are shown in Figure 7.

The downscaled scenarios yield more than a 50% increase in the annual precipitation and a+3°C warming of maximum and minimum daily temperatures. However, indicated in Figure6a, the bulk of the precipitation increase occurs in just three months; precipitation inDecember, January, and February increases by +104%. The increase in maximumtemperature (Figure 6b) ranges from +5.6°C (September) to +1.6°C (February), incomparison with +4.5°C (September) and +1.7°C (May) for the minimum temperatures(Figure 6c).

STREAMFLOW CHANGES IN CALIFORNIA116

These changes in future precipitation and temperature regimes are reflected in simulatedsnowpack and streamflow changes (Figure 7). In response to increased precipitation totals,all three rivers show large increases in annual runoff, ranging from +107% in the Merced,through +103% in the Carson, to +82% in the American. The corresponding annual meansnowpack changes were +41% (Merced), +27% (Carson), and –6% (American). TheMerced has the largest increase in winter snowpack with an earlier peak snow water content,and an increase that persists into the summer months. Increased snowpack is accompaniedby a marked increase in Merced spring streamflows. Similar, but smaller, increases aresimulated for the Carson River. Both of these rivers are mostly at high elevations and havecold winters and springs. Evidently, the warmer temperatures in the simulated future-climateconditions are not sufficiently warm to prevent significant increases in overall snowpack andspringtime streamflow. Indeed, in spite of projected warmer conditions, the percentage ofthe Merced and Carson basins that is simulated as being snowcovered – on average – duringDecember through March decreases by only about 5% of current snow-covered areas. Thus,although warmer temperatures would tend to reduce the snow-covered areas, much wetter


winter conditions roughly compensate for the reduced snow-covered areas in thesesimulations.

In contrast, simulations of the American River indicate only a slight increase in wintersnowpack and a decrease in spring snowpack volumes. Snow-covered areas in the AmericanRiver basin decrease by about 10% of current averages under the warmer and wetter future-climate scenario. The American River basin is lower and warmer than the others, and ityields a mix of rainfall runoff and snowmelt runoff each year, under current conditions.Increased winter streamflows projected under future-climate conditions are due to increasesin winter rainfall runoff rather than increased snowmelt runoff. This results in earlier peakstreamflows for the American and Carson Rivers under future climate conditions. All threebasins yield more cool-season flash flooding under the future-climate conditions, and allhave less autumn snowpack owing to higher maximum and minimum temperatures inSeptember and a slight decrease in autumn precipitation.

6. Discussion

A recent synthesis of seasonal precipitation scenarios for North America predicted by 15GCMs reported changes ranging from –4% to +8% per 1°C global-mean warming, withwintertime precipitation over California being the most sensitive season and region (Wigley,1999). Particular models such as CGCM1 and HadCM2 (Figure 1) simulate even greatersensitivities. For example, changes in the winter precipitation of HadCM2 range from –10%/°C over southern Texas to +20%/°C over California. Although such results may beuseful for identifying regions that are potentially most vulnerable to climate variability andchange, GCMs are unable to capture local climatic effects arising from topographic, coastal,and land-surface processes. Statistical downscaling offers a computationally efficient androbust method of generating the basin-scale climate-change estimates necessary forhydrological impact assessments.

Accordingly, a regression-based downscaling method was used to simulate daily rainfall andtemperature series for streamflow modelling in three Californian river basins under current-and future-climate conditions. The downscaling model employed just three predictorvariables (specific humidity, zonal velocity component of airflow, and 500 hPa geopotentialheights) supplied by HadCM2 for the grid point nearest the target basins. When evaluatedusing independent data, the model showed reasonable skill at reproducing observed area-average precipitation, temperature, and concomitant streamflow variations. Overall, thedownscaled data resulted in slight underestimates of mean annual streamflow that can beattributed to underestimates of precipitation in spring and positive temperature biases inwinter.


Simulated streamflows provide a useful indication of the combined skill of the coupleddownscaling-hydrological model for the current climate because streamflow is an effectiveintegrator (in time and space) of the cumulative effects of multiple climate variables.Streamflow at a single gauging site can also be measured more reliably than areal-averageprecipitation or temperature in complex terrain. However, because rivers act as basin-scaleand season-long integrators of climatic forcings, simple downscaling-skill measures appliedto streamflow simulations can appear more skillful than would similar skill measuresapplied to the driving precipitation and temperature inputs. For example, in basins withsignificant snowpack, the gross accumulations of precipitation in the seasonal snowpack andthe timing of spring melt are critical to the simulations of hydrological responses.Conversely, in rainfall-dominated basins, the timing and magnitude of individual stormevents are of greater concern, and these individual storms are the most difficult for adownscaling model to reproduce given only synoptic-scale atmospheric-circulation inputs.Thus, if a downscaling method reproduces only seasonal totals of precipitation and realistictemperature fluctuations during the critical snowmelt periods, snowmelt-dominated riverswill be much better represented than will nearby rivers dominated by rainfall runoff. Even ina rainfall-driven system, though, soil-moisture and ground-water reservoirs will tend tosmooth the hydrological response and contribute apparent skill to the downscaling

The smoothing effects of snowpack on streamflow responses to climate forcings help toexplain differences between the skill of simulated streamflows in the three basins. TheMerced and American River basins drain the western, windward slope of the Sierra Nevada,whereas the Carson River drains the eastern, leeward slope. Hence, the Carson River basinis in the rain shadow of the Sierra Nevada and is drier than the others. The Merced andCarson River basins are high-elevation basins and cool overall, with the part of the Merceddrainage simulated here ranging from 1,200 to over 4,000 m above sea level and the CarsonRiver drainage ranging from 1,600 to 3,400 m. The American River basin is a lowerelevation basin and warmer overall, ranging from 200 m to 2,500 m above sea level.Consequently, the Merced and Carson Rivers are snowmelt dominated whereas theAmerican River is a mix of rainfall runoff and snowmelt runoff. Simulated streamflow inthe American River responds rapidly and sensitively to daily-scale temperature andprecipitation fluctuations and errors; in the Merced and Carson Rivers, the response to thesame short-term influences is much less. Consequently, the skill of simulated flows wassignificantly lower in the American River model than in the Carson and Merced.

The physiography of the three basins also accounts for differences in their sensitivities tofuture climate change. Increases in winter precipitation exceeding +100% coupled withmean temperature rises greater than +2°C result in increased winter streamflows in all threebasins. In the Merced and Carson basins, these streamflow increases reflect large changes inwinter snowpack, whereas the streamflow changes in the lower elevation American basinare driven primarily by rainfall runoff. Furthermore, reductions in winter snowpack in the


American River basin, owing to less precipitation falling as snow and earlier melting ofsnow at middle elevations, lead to less spring and summer streamflow.

Taken collectively, the downscaling results imply significant changes to both the timing andmagnitude of streamflows in the Sierra Nevada by the end of the 21st Century. In the higherelevation basins, the HadCM2 scenario implies more annual streamflow and morestreamflow during the spring and summer months that are critical for water-resourcesmanagement in California. Not shown here, but of comparable concern, the future winterand spring flow simulations also include more sudden flood events in response to winterand spring storms that yield more rainy mixes of rain and snow than at present, and for thatrain to fall on snopwpacks that are warmer than at present. Nearly all of the additional flowin the lower elevation, warmer American River basin occurs in winter and constitutesincreased flood hazards. Thus, depending on the relative significance of rainfall runoff andsnowmelt, each basin responds in its own way to regional climate forcing. Generally, then,climate scenarios need to be specified – by whatever means – with sufficient temporal andspatial resolution to capture subtle orographic influences if projections of climate-changeresponses are to be useful and reproducible.

Acknowledgments

This research was supported by ACACIA (A Consortium for the Application of ClimateImpact Assessments, National Center for Atmospheric Research) and by the U.S. GeologicalSurvey Global Change Hydrology Program. We also thank David Viner of the ClimateImpacts LINK Project (UK Department of the Environment Contract EPG1/1/16) forsupplying the HadCM2 simulations on behalf of the Hadley Centre and U.K. MeteorologicalOffice. NCAR is sponsored by the National Science Foundation.

References

Aguado, E. (1985). Radiation balances of melting snow covers at an open site in the central Sierra Nevada,California. Water Resources Research, 21, 1649-1654.

Bardossy, A. and Plate, E.J. (1992). Space-time model for daily rainfall using atmospheric circulation patterns.Water Resources Research, 28, 1247-1259.

Boer, G.J., Flato, G.M. and Ramsden, D. (1999a). A transient climate change simulation with historical andprojected greenhouse gas and aerosol forcing: projected climate for the 21st century. Climate Dynamics,submitted.

Boer, G.J., Flato, G.M., Reader, M.C. and Ramsden, D. (1999b). A transient climate change simulation withhistorical and projected greenhouse gas and aerosol forcing: experimental design and comparison with theinstrumental record for the 20th century. Climate Dynamics, submitted.

Christensen, J.H., Machenhauer, B., Jones, R.G., Schär, C., Ruti, P.M., Castro, M., Visconti, G. (1997). Validationof present-day regional climate simulations over Europe: LAM simulations with observed boundaryconditions. Climate Dynamics, 13, 489-506.

STREAMFLOW CHANGES IN CALIFORNIA

Conway, D., Wilby, R.L. and Jones, P.D. (1996). Precipitation and air flow indices over the British Isles. ClimateResearch, 7, 169-183.

Crane, R.G. and Hewitson, B.C. (1998). Doubled CO2 precipitation changes for the Susquehanna basin:downscaling from the GENESIS General Circulation Model. International Journal of Climatology, 18, 65-76.

Dettinger, M.D., and Cayan, D.R., (1992), Climate-change scenarios for the Sierra Nevada, California, based onwinter atmospheric-circulation patterns: Proceedings, 1992 American Water Resources AssociationSymposium and Conference, “Managing Water Resources under Global Change,” Reno, Nevada, 681-690.

Dettinger, M.D., Mo, K., Cayan, D.R., and Jeton, A.E. (1999). Global to local scale simulations of streamflow inthe Merced, American, and Carson Rivers, Sierra Nevada, California: Preprints, American MeteorologicalSociety's 14th Conference on Hydrology, Dallas, January 1999, 80-82.

Felzer, B. (1999). Hydrological implications of GCM results for the U.S. National Assessment. Proceedings of theSpeciality Conference on Potential Consequences of Climate Variability and Change to Water Resources ofthe United States, American Water Resources Association, Atlanta, USA, 69-72.

Flato, G.M., Boer, G.J., Lee, W.G., McFarlane, N.A., Ramsden, D., Reader, M.C. and Weaver, A.J. (1999). TheCanadian Centre for Climate Modelling and Analysis Global Coupled Model and its climate. ClimateDynamics, submitted.

Giorgi, F. and Mearns, L.O. (1991). Approaches to the simulation of regional climate change. A review. Rev.Geophys., 29, 191-216.

Giorgi, F. and Mearns, L.O. (1999). Introduction to special section: Regional climate modelling revisited. Journalof Geophysical Research, 104, 6335-6352.

Goodess, C.M. and Palutikof, J.P. (1998). Development of daily rainfall scenarios for southeast Spain using acirculation-type approach to downscaling. International Journal of Climatology, 10, 1051-1083.

Gregory, J.M., Wigley, T.M.L. and Jones, P.D. (1993). Application of Markov models to area-average dailyprecipitation series and interannual variability in seasonal totals. Climate Dynamics, 8, 299-310.

Hay, L.E., McCabe, G.J., Wolock, D.M. and Ayers, M.A. (1991). Simulation of precipitation by weather typeanalysis. Water Resources Research, 27, 493-501.

Hay, L.E., McCabe, G.J., Wolock, D.M. and Ayers, M.A. (1992). Use of weather types to disaggregate GeneralCirculation Model predictions. Journal of Geophysical Research, 97, 2781 -2790.

Hostetler, S.W. (1994). Hydrologic and atmospheric models: the (continuing) problem of discordant scales.Climatic Change, 27, 345-350.

Jeton, A.E. and Smith, J.L. (1993). Development of watershed models for two Sierra Nevada basins using ageographic information system. Water Resources Bulletin, 29, 923-932.

Jeton, A.E. Dettinger, M.D., and Smith, J.L. (1996). Potential effects of climate change on streamflow, eastern andwestern slopes of the Sierra Nevada, California and Nevada. U.S. Geological Survey Water-ResourcesInvestigations Report 95-4260, 44 p.

Johns, T.C., Carnell, R.E., Crossley, J.F., Gregory, J.M., Mitchell, J.F.B., Senior, C.A., Tett, S.F.B. and Wood,R.A. (1997). The Second Hadley Centre coupled ocean-atmosphere GCM: Model description, spinup andvalidation. Climate Dynamics, 13, 103-134.

Jones, P.D., Hulme, M. and Briffa, K.R. (1993). A comparison of Lamb Circulation Types with an objectiveclassification scheme. International Journal of Climatology, 13, 655-663.

Jones, R.G., Murphy, J.M. and Noguer, M. (1995). Simulation of climate change over Europe using a nestedregional-climate model. I: Assessment of control climate, including sensitivity to location of lateralboundaries. J. Roy. Met. Soc., 121, 1413-1449.

Kalnay, E., Kanamitsu, M., Kistler, R., Collins, W., Deaven, D., Gandin, L., Iredell, M., Saha, S., White, G.,Woollen, J., Zhu, Y., Chelliah, M., Ebisuzaki, W., Higgins, W., Janowiak, J., Mo, K.C., Ropelewski, C.,Wang, J., Leetmaa, A., Reynolds, R., Jenne, R. and Joseph, D. (1996). The NCEP/NCAR 40-year reanalysisproject. Bulletin of the American Meteorological Society, 77, 437-471.

Karl, T.R., Wang, W.C., Schlesinger, M.E., Knight, R.W. and Portman, D. (1990). A method of relating GeneralCirculation Model simulated climate to the observed local climate. Part I: Seasonal statistics. Journal ofClimate, 3, 1053-1079.

Katz, R.W. (1996). Use of conditional stochastic models to generate climate change scenarios. Climatic Change,32, 237-255.

120

WILBY AND DETTINGER

Kilsby, C.G., Cowpertwait, P.S.P., O'Connell, P.E. and Jones, P.D. (1998). Predicting rainfall statistics in Englandand Wales using atmospheric circulation variables. International Journal of Climatology, 18, 523-539.

Leavesley, G.H., Lichty, R.W., Troutman, B.M. and Saindon, L.G. (1983). Precipitation-runoff modelling system:User’s manual. U.S. Geological Survey Water-Resources Investigations Report 83-4238, 207 p.

Matyasovszky, I., Bogardi, I. And Duckstein, L. (1994). Comparison of two GCMs to downscale local precipitationand temperature. Water Resources Research, 30, 3437-3448.

McGregor, J.J. (1997). Regional climate modelling. Meteorol. Atmos. Phys., 63, 105-117.Mearns, L.O., Rosenzweig, C. and Goldberg, R. (1996). The effect of changes in daily and interannual climatic

variability on CERES-wheat yields. A sensitivity study. Climatic Change, 32, 257-292.Mitchell, J.F.B. and Johns, T.C. (1997). On modification of global warming by sulphate aerosols. Journal of

Climate, 10, 245-267.Obled, C. and Rosse, B.B. (1977). Mathematical models of a melting snowpack at an index plot. J. Hydrology, 32,

139-163.Pilling, C., Wilby, R.L. and Jones, J.A.A. (1998). Downscaling of catchment hydrometeorology from GCM output

using airflow indices in upland Wales. In: Wheater, H. and Kirby, C. Hydrology in a Changing Environment,Vol.1, Wiley, Chichester, 191-208.

Richards, J.M. (1971). Simple expression for the saturation vapour pressure of water in the range -50° to 140°. Brit.J., Appl. Phys., 4, Ll5-L18.

Richardson, C.W. (1981). Stochastic simulation of daily precipitation, temperature and solar radiation. WaterResources Research, 17, 182-190.

Smith, J.L. and Reece, B.D. (1995). Watershed characterization for precipitation runoff modelling system, NorthFork American River and East Fork Carson River watersheds, California. U.S. Geological Survey HydrologicInvestigations Atlas HA-734, 1 sheet.

von Storch, H., Zorita, E. and Cubasch, U. (1993). Downscaling of global climate change estimates to regionalscales: an application to Iberian rainfall in wintertime. J. Climate, 6, 1161-1171.

Wigley, T.M.L. (1999). The science of climate change: global and U.S. perspectives. Pew Center on GlobalClimate Change, Arlington, Virginia, 48 pp.

Wilby, R.L. (1997). Nonstationarity in daily precipitation series: implications for GCM downscaling usingatmospheric circulation indices. International Journal of Climatology, 17, 439-454.

Wilby, R.L. and Wigley, T.M.L. (1997). Downscaling General Circulation Model output: a review of methods andlimitations. Progress in Physical Geography, 21, 530-548.

Wilby, R.L., Hassan, H. and Hanaki, K. (1998a). Statistical downscaling of hydrometeorological variables usinggeneral circulation model output. Journal of Hydrology, 205, 1-19.

Wilby, R.L., Wigley, T.M.L., Conway, D., Jones, P.D., Hewitson, B.C., Main, J., and Wilks, D.S. (1998b).Statistical downscaling of General Circulation Model output: a comparison of methods. Water ResourcesResearch, 34, 2995-3008.

Wilby, R.L., Hay, L.E. and Leavesley, G.H. (1999). A comparison of downscaled and raw GCM output:implications for climate change scenarios in the San Juan River basin, Colorado. Journal of Hydrology,submitted.

Wilks, D.S. (1992). Adapting stochastic weather generation algorithms for climate change studies. ClimaticChange, 22, 67-84.

Willen, D.W., Shumway, C.A., and Reid, J.E. (1971). Simulation of daily snow water equivalent and melt.Proceedings of the 1971 Western Snow Conference, Billings, Montana, v. 39, 1-8.

Winkler, J.A., Palutikof, J.P., Andresen, J.A. and Goodess, C.M. (1997). The simulation of daily temperature seriesfrom GCM output. Part II: Sensitivity analysis of an empirical transfer function methodology. Journal ofClimate, 10, 2514-2532.

121

EXAMINING LINKS BETWEEN CLIMATECHANGE AND LANDSLIDE ACTIVITYUSING GCMS:case studies from Italy and New Zealand

MICHAEL SCHMIDT and MARTIN DEHNDept. of Geography, University of BonnMeckenheimer Allee 166D-53115 BonnFax: +49-228-73 9099e-mail: [email protected]

Abstract

Climate is an important forcing parameter for landslide activity and hence of stronggeomorphological interest, especially precipitation and temperature as inputs into thelandslide system. Since climate change is widely accepted, its regional impacts are ofmajor research interest. In fact it is not obviously clear which regions will be affected, inwhich manner and how the local environments will react to these changes. Climatescenarios can display probable outcomes and boundary conditions for linkedenvironmental systems. In this study GCM (General Circulation Model) outputs aredownscaled, with an empirical-statistical method applied to two locations. We presentan activity scenario of a mudslide in the Dolomites and scenarios of landslide eventscalculated with different landslide models for the region of Wellington, New Zealand.Scenario results of the Alvera mudslide (Italy) show a significant activity decrease dueto increasing winter temperature and reduced snow storage while precipitation changesdo not show a clear trend. The projections for Wellington indicate fewer events of highlandslide probability in the hemispherical winter due to decreased precipitation.

Key words: Climate change impact, GCMs, downscaling, landslide scenarios

123


124 SCHMIDT AND DEHN

1. Introduction

Water in the soil profile and slope hydrology generally are of major concern for thelandslide system and therefore of great research interest in terms of modellingoccurrence and activity of landslides. The other important climatic parameter forlandslide studies is air temperature. The estimations of both parameters are outlinedprincipally.

Climate change is defined here as due to increasing emissions of atmospheric tracegases with climatic relevance, as simulated by the IPCC (Intergovernmental Panel onClimate Change) in the IS92 emission scenarios (Houghton et al., 1992). Coupled oceanand atmosphere GCMs are at the moment the best tools to describe future climaticconditions, even if the typical grid resolution of modern GCM is still too coarse for localscenarios (Trenberth, 1996). Several downscaling techniques have been developed tobridge this scale gap (Hewitson and Crane, 1996, Zorita and von Storch, 1998). In thisstudy an analog downscaling technique (Cubasch et al., 1996, Zorita and von Storch,1998) is used to derive local precipitation scenarios, which are then taken as input forlandslide models.

Changes of climatic variables like temperature, precipitation, sea level or soil moistureare likely to occur on a global and regional scale. The current average rate of changemight be greater than any seen in the past 10,000 years. Although global averages arewithin historical rates of changes, regional changes could differ widely (Watson et al.,1998). These human-induced large scale phenomena will interact with sources of naturalvariability like the El Niño-Southern Oscillation (ENSO) and thus influence social andeconomic well-being. This is especially serious if the recently projected increases infrequency and magnitude of extreme events are considered (Cubasch et al., 1995).

2. General methodology

The overall methodology is outlined in Fig. 1. Two site specific landslide models wereused to establish local landslide scenarios. These are a threshold model of occurrence oflandslide events for the region of Wellington, New Zealand, and a hydrological tankmodel coupled to a slope-stability model to describe the activity of a single mudslide inthe Dolomites. Both approaches use regional/local precipitation and temperature asinput parameters. These climatological scenarios are the result of the first three model-steps in Fig. 1.

CLIMATE CHANGE & LANDSLIDE ACTIVITY LINKS 125

The GCMs ECHAM4/OPYC3 (European Centre Hamburg Model, 4. Generation)without consideration of sulphate aerosols (Roeckner et al., 1996) and HadCM2SUL(Hadley Centre Coupled Model, 2. Generation) with sulphate aerosols (Johns et al.,1997) are forced with the emission scenarios IS92a, which results in a doubling ofconcentrations within the next century. It has to be taken into account, that GCMs doquite well in projecting general circulation and large scale variables, but due to thecoarse operating scales (ca. 250x250 km) local variables such as precipitation orcloudiness are only roughly parameterised and of low confidence for site specificstudies. Interrelations between Sea Level Pressure (SLP) as a large scale variable andprecipitation as a local variable provide the possibility for a statistical downscaling (seebelow). The presented results and scenarios must be seen in terms of trends andabsolutely not as event probabilities of single events for a given day.

Precipitation, as a variable with high local variability, requires a downscaling to producereliable scenarios based on GCMs. Temperature variability is also high temporally, butnot spatially, so that it is a common technique to interpolate the local temperaturedirectly from GCMs by considering and adjusting the elevation factor, which is notadequately captured by the GCM.


2.1. THE ANALOG DOWNSCALING TECHNIQUE

We applied an analog downscaling proposed by Cubasch et al. (1996) and Zorita andvon Storch (1998). Two independent data sets of SLP and local precipitation arenecessary for training and validating the analog model. Daily SLP of the training periodis stored together with the local precipitation data in a catalog. Then SLP data of avalidation period are compared with the training period by searching for similarcirculation patterns (operated by SLP fields) and associated with this analog situationthe amount of precipitation from the catalog. This is first done by using EmpiricalOrthogonal Functions (EOF), by which the data are projected in a space with orthogonalspan vectors. These independent vectors describe the internal variability of the data.

Usually the 5 leading EOFs describe more than 90% of the data variance. The analogsituation is then simply found by the next neighbour in the 5 dimensional EOF space


(Fig. 2). The same procedure can be applied to observations as training period and SLPof various target periods of a GCM run. Results are precipitation scenarios of respectiveGCM periods.

Before going to the specific case studies one major assumption should be pointed out.The empirically derived relationships between SLP and local precipitation must beconstant over time and also during changing climate. If this is not the case the scenariosare not valid.

3. Case study of the Alvera mudslide

3.1. DOWNSCALING MODEL

In Fig. 3 the location of Cortina d’Ampezzo is pointed out in the land–sea mask of theECHAM4/OPYC3 climate model with T42 resolution. It is obvious that a downscalingof precipitation values is necessary, especially because the whole alpine region isrepresented by a few grid-boxes only. The analog downscaling was carried outseparately for the 4 meteorological seasons spring (MAM), summer (JJA), autumn(SON) and winter (DJF). The validation results in the form of time series correlationsbetween observed and estimated seasonal precipitation are shown in Table 1.

Autumn and winter represent reliable results. In contrary summer precipitation is poorlyestimated, which might be due to the high quantity of summer precipitation withconvective origin. Convective showers are not so closely related to large-scalecirculation than advective precipitation, which is the major component of winterprecipitation. Summarising, it can be said that the analog technique based on SLP is ableto successfully reconstruct the historical development of precipitation in Cortina for theseasons winter, spring and autumn, while precipitation amounts are generallyunderestimated as well as other precipitation parameters (for details refer to Dehn,1999). Despite some deficiencies, the validation shows the skill of the analog techniqueand therefore it can be used for downscaling large-scale circulation features of GCMexperiments.


3.1 FEATURES OF ALVERA MUDSLIDE

The Alvera landslide is a slow moving mudflow (few centimetres per year), situated NEof Cortina d’Ampezzo. The geological structure of the region is characterized by arepeated succession of dolomitic and pellitic rocks (Angeli et al., 1992). The total lengthof the mudflow is about 1700 m and 100 m of width, with an average inclination of 7.3°(Fig. 3).


The mudslide consists of clayey and silty material which is a weathering product of theSan Cassiano formation. Groundwater records suggest a subdivision into a macroporousroot zone underlain by a less permeable clayey layer (Angeli et al., 1998). The depth inwhich 75% of the movement takes place is at about 5 m below the surface which isconnected by deep cracks with the upper system. The second sliding area is at 20-25 mdepth (Gasparetto et al., 1996). A linear tank model is used for the simulation of this 2component system (Angeli et al., 1998). The onset of displacement is triggered ifgroundwater exceeds a critical level of 0.5 m below ground. Concerning landslideactivity the number of days with supercritical groundwater levels, that is > -0.5 m, wasassessed in the scenarios.

3.2 FUTURE SCENARIOS

In the following activity scenarios for this mudslide are presented, based on downscaledprecipitation and interpolated temperature for Cortina d‘Ampezzo. Precipitationscenarios are shown in Fig. 4 as seasonal changes between GCM control period (1950-79 for HadCM2 and 1960-89 for ECHAM4/OPYC3) and future period 2070-2099 forboth GCMs. In the ECHAM4 scenario seasonal precipitation is generally decreasing butnot significantly on the 95% confidence level. In downscaling results from the HadCM2model there is no significant trend displayed.


In contrary to precipitation, local temperature was derived by simply interpolatingbetween several GCM grid points next to Cortina. The absolute values had to be fitted tolocal observations on a monthly basis. The temperature rise in HadCM2 withconsideration of sulphate aerosols is smaller than in ECHAM4/OPYC3 without sulphateaerosols (Fig. 5).

Changes in landslide activity are depicted in Fig. 6. For both GCMs a significantdecrease of future landslide activity in spring (MAM) is visible. Changes in otherseasons are less dramatic and not significant. This strong change in MAM is not visiblein the precipitation scenarios. This finding will be discussed later.


4. Case study Wellington region

4.1. DOWNSCALING MODEL

In Fig. 7 the land-sea mask of the HadCM2SUL model with the grid resolution of2.5°x3.75°, which is similar to the ECHAM4/OYPC3 T42 resolution, is shown. For abetter estimation of local precipitation values, an analog downscaling was carried outwith SLP as the large scale variable, similar to the Cortina example. The validationresults for the different seasons are shown in Table 2. Only the winter season (JJA)shows acceptable validation results. Therefore all further conclusions will be drawn onlyfor the winter season.


4.1 REGIONAL SETTINGS

The study area covers with steep and strongly dissected slopes, often orientedalong fault lines, which produce aligned drainage. The relief is about 460 m with anaverage tectonic uplift of 1mm/a. In the area extensively faulted, tilted and folded darkgrey argillite and greywacke sandstones can be found. On this bedrock colluvium soilsand solifluction deposits are developed. The native forest was converted to pasture und


scrub vegetation as a result of increased urbanisation since the 1840s. This area is proneto landslides if the high amount of 1250 mm yearly precipitation is taken into account.

4.2 The landslide models

4.2.1 THRESHOLD MODEL

The probabilities for landslide events are derived by dealing with daily regionalmaximum precipitation values of all the weather stations in the region. Thisprecipitation amount is, in the case of an event, defined as the triggering value. The aimis to define triggering thresholds or the probability of a given amount of rainfall totrigger landslides.

A first modelling approach is to divide the measured maximum precipitation intoprecipitation classes of 20 mm intervals. Minimum thresholds, below which a landslidehad never occurred can be found as well as thresholds above which landslides alwaysoccurred. The classes below this daily minimum threshold of 20 mm precipitation havea probability of 0% and the precipitation classes above the maximum have 100% oflandslide probability (Fig. 8).


The probability of landslide occurrence is calculated as the ratio of triggeringprecipitation number by total number of occurrence for each precipitation class. Thesevalues were calculated by Glade (1998) for the period of 1903-1995 and were thereforetaken as values on a larger statistical basis. The bars in Fig. 8 represent time periods of30 years. The bars present observed data and data of GCM origin, which are downscaledand used to build precipitation scenarios for the control period (1950-1979) and forfuture conditions (2070-99). The datasets can be compared reasonably either by thesame time period in terms of model quality or the same origin in terms of changingconditions. The model represents similar results for the time period 1950-1979,especially for the lower classes, where the number of elements in each class is higher.This result implies that the whole approach should be suitable. For the future run, aslight tendency towards less extreme events is visible.

One major limitation of this model is that it works without preparatory factors. Only theactual daily amount of precipitation is included as triggering factor.


4.2.2 THE ANTECEDENT DAILY RAINFALL MODEL

In this model the antecedent rainfall of 10 days is considered as preparatory factor in theform of the Antecedent Daily Rainfall Index (ADRI) (Crozier and Eyles, 1980). Theprincipal idea that stands behind this landslide model is that preparatory factors are thereason why a stable slope system becomes marginally stable.

In this modelling approach precipitation is the only preparatory factor for a slope tobecome marginally stable. Following the concept of critical water content (Crozier,1998), the pre-existing water within the slope together with the daily precipitation astriggering factor, causes a landslide to occur. All other triggering factors are excludedfor this model. Controlling factors are site specific and describe the behaviour of asliding mass. These factors are not considered here, because in this study only thenumber of events is dealt with. Glade et al. (1999) model the antecedent rainfallinfluence by the following equations

where is the antecedent rainfall for day 0, the rainfall on the i-th day before day0 and the number of considered days before day 0. is a site specific empiricallyderived factor of value –1.52 (Glade et al., 1999). To include the antecedent conditionsof 10 days seems suitable because the use of longer antecedent conditions show nobetter results for this study region (Glade, 1997). Probabilities of landslide occurrenceare thus calculated with the antecedent rainfall conditions as preparatory factors and theactual rainfall as triggering event, as it is shown in the following equation:

represents the daily precipitation, the probability of landslide occurrence of a givenday and is the antecedent rainfall that is held in the soil as described in equation 1.The probabilities derived with this approach are outlined in the following figures.

CLIMATE CHANGE AND LANDSLIDE ACTIVITY LINKS 137

In Fig. 9a-c fixed probability lines for landslide events are drawn. Any point above aprobability line represents the combination of the daily rainfall and ADRI withthe same (or higher) event probability as the value of this line (Fig. 9a). The curvedprobability lines exhibit a negative relationship between antecedent conditions and thedaily rainfall, which means that at high amounts of ADRI less precipitation is needed totrigger a landslide and vice versa.


The bars in Fig. 10 show the number of days exceeding certain probability values. Thecontrol run (1950-79) and the observed values (same period) show similar results. Thefirst two classes seem to be well represented, while the differences in the higher classesare more important.

Comparing control run and the future landslide scenario (2070-99), a trend fordecreasing winter landslide event probabilities, especially for the lower classes, isdisplayed. The numbers of the very high probabilities have to be handled with care,because these are single storm events with intense precipitation amounts for which isnot yet clear, if GCMs can represent them well.

5. Discussion

The main result of the Alvera climate change impact study is the significant decrease oflandslide activity in spring without accompanying decrease of seasonal precipitation.Therefore it has to be explained how this dramatic decrease of landslide activity inspring is caused. The reason has to be searched for in the significant temperatureincreases simulated by both GCMs for all seasons. An explanation is suggested which isfocusing on the role of snow storage of precipitation in winter (DJF). This effectcontributes usually to high meltwater inputs into the slope system in spring and hencecauses high landslide activity rates of the Alvera mudslide in this season. Globalwarming due to the greenhouse effect forces the DJF temperatures in Cortina ofcurrently –1.1°C to exceed 0°C in the mid next century. This happens according to theGCM-experiments with or without consideration of the cooling effect of sulphateaerosols. In a warmer climate, therefore, less water stored as snow is available for theslope in spring with the consequence of lower groundwater levels and lowered rates oflandslide activity. Treatment of snow storage, snow melt and potentialevapotranspiration in the tank model is described in detail in Angeli et al. (1998).Problems of these parameterisations in the context of climate change applications arediscussed in Dehn (1999).

The results for the study area of Wellington, due to the low sample size of highlandslide probabilities, can not be described with statistical parameters. It would be afuture task to find a method that is suitable to construct scenarios for the remainingseasons as well as winter. In JJA the westerlies are the most important feature foradvective precipitation, which is connected with synoptic scale features and are thususeful for this downscaling method The conditions for the other seasons are morevariable. In the warmer months convective processes due to the surplus of energy in theatmosphere, reduces the performance of the analog downscaling.

CLIMATE CHANGE AND LANDSLIDE ACTIVITY LINKS 139

Even if in winter (JJA) almost 40% of the recorded landslides in the study area occur,the most severe storms with huge precipitation amount and direct triggering did nothappen in JJA. Nevertheless landslides occurred, even with a calculated probability ofless than 10%, which leads to the speculation of the coincidence of more than onetriggering factor.

In summary, it was shown that depending on landslide type and setting different climateelements develop as key parameters due to global warming. This problem is moregenerally discussed in Dehn and Buma (1999). A consequence of this finding is thatsuccessful assessments of future landslide activity require landslide modelling conceptswhich are able to represent process variability due to changes in precipitation andtemperature.

6. Conclusions

The following conclusions can be drawn from the two case studies:With the analog downscaling technique large-scale atmospheric conditions canbe exploited for constructing local climate scenarios on a daily time-scale.The analog technique is performing best in the respective winter of Italy andNew Zealand.Temperature increase, especially in winter, is an important causative factor forthe decreasing landslide activity in spring of Alvera mudslide.The treatment of snow processes and evapotranspiration in the applied modelrequires improvements.The number of landslides for Wellington must be seen as minimum, becausefor every event only one landslide is recognised, even if there are several singleslides. The tendency towards a decrease of winter landslide event probability isthe major outcome of this regional study.The boundary conditions for landslides and thresholds are held constant overtime, which is by no means guaranteed, as they could change by humaninfluence like forest removal or slope cuts for urbanisation or infrastructurelines or change after a landslide.Temperature and evapotranspiration are not adequately considered by theempirical decay formula for the study area of Wellington. This shortcoming isimproved in the Antecedent Soil Water Status model (Crozier, 1998, Glade,1999), which will be used in a next step.


Acknowledgements

This paper is part of the CEC Environment Research Programme on ,,New technologiesfor landslide hazard assessment and management in Europe (NEWTECH)“ (ENV4-CT96-0248).

We would like to thank Eduardo Zorita and Hans von Storch, GKSS Research Centre,Geesthacht for assistance and cooperation with the analog technique, Jelle Buma,University of Utrecht, for the use of the hydrological model, Thomas Glade and MikeCrozier for the use of the New Zealand landslide models, and finally Erich Roecknerand Monika Esch from the Max-Planck-Institut für Meteorologie, Hamburg and DavidViner from Climate Impacts LINK project, Norwich, for providing GCM data.

References

Angeli, M.-G., Buma, J., Gasparetto, P., Pasuto, A. and Silvano, S. (1998) A combined hillslope hydrology /

stability model for low-gradient clay slopes in the Italian Dolomites, Engineering Geology, 49, 1-13.Angeli, M.-G., Menotti, R.M., Pasuto, A. and Silvano, S. (1992) Landslide studies in the Eastern Dolomites

Mountains, Italy. In: D.H. Bell (Editor), Proc. 6th International Symposium on Landslides. Balkema,Christchurch (New Zealand), pp. 275-282.

Crozier, M.J. (1998) The climate landslide couple: a Southern Hemisphere perspective, Paleoclimate

Research, 2, 329-350.Crozier, M.J. and Eyles, R.J. (1980) Assessing the probability of rapid mass movement. In: The New Zealand

Institution of Engineers - Proceedings of Technical Groups (Editor), Proc. Third Australia - New ZealandConference on Geomechanics, Wellington, pp. 2.47-2.51.

Cubasch, U., von Storch, H., Waszkewitz, J. and Zorita, E. (1996) Estimates of climate change in SouthernEurope derived from dynamical climate model output, Climate Research, 7, 129-149.

Cubasch, U., Waszkewitz, J., Hegerl, G.C. and Perlwitz, J. (1995) Regional climate changes as simulated intime-slice experiments, Climatic Change, 31, 273-304.

Dehn, M. (1999) Application of an analog downscaling technique to the assessment of future landslide activity- a case study from the Italian Alps, Climate Research, in press.

Dehn, M. and Buma, J. (1999) Modelling future landslide activity based on general circulation models,Geomorphology, 30, in press.

Gasparetto, P., Mosselman, M. and van Asch, T.W.J. (1996) The mobility of the Alvera landslide (Cortinad'Ampezzo, Italy), Geomorphology, 15, 327-335.

Glade, T. (1997) The temporal and spatial occurrence of rainstorm-triggered landslide events in New Zealand.PhD, Victoria University, Wellington.

Glade, T. (1998) Establishing the frequency and magnitude of landslide-triggering rainstorm events in NewZealand, Environmental Geology, 35, 160-174.

Glade, T. (1999) Models of antecedent rainfall and soil water status applied to different regions in NewZealand, in preparation.

CLIMATE CHANGE AND LANDSLIDE ACTIVITY LINKS

Glade, T., Smith, P. and Crozier, M. (1999) Applying Probability Determination to refine Landslide-

Triggering Rainfall Thresholds using an Empirical ,Antecedent Daily Rainfall Model‘, Pure and AppliedGeophysics, In press.

Hewitson, B.C. and Crane, R.G. (1996) Climate downscaling: techniques and application, Climate Research,7, 85-95.

Houghton, J.T., Callander, B.A. and Varney, S.K. (eds.) (1992) Climate change 1992. The supplementaryreport to the IPCC scientific assessment. Cambridge University Press, Cambridge.

Johns, T.C., Carnell, R.E., Crossley, J.F., Gregory, J.M., Mitchell, J.F.B., Senior, C.A., Tett, S.F.B. and Wood,R.A. (1997) The Second Hadley Centre coupled ocean-atmosphere GCM: Model description, spinup andvalidation, Climate Dynamics, 13, 103-134.

Roeckner, E., Oberhuber, J.M., Bacher, A., Christoph, M. and Kirchner, I. (1996) ENSO variability andatmospheric response in a global coupled atmosphere-ocean GCM, Climate Dynamics, 12, 737-754.

Trenberth, K.E. (1996) Coupled Climate System Modelling. In: T.W. Giambelluca and A. Henderson-Sellers(eds.), Climate Change. Developing Southern Hemisphere Perspectives, Wiley, Chichester, 63-88.

Watson, R.T., Zinyowera, M.C., Moss, R.H. and Dokken, D.J. (eds.) (1998) The regional impacts of climatechange. Cambridge University Press, Cambridge.

Zorita, E. and von Storch, H. (1998) A survey of statistical downscaling techniques, GKSS Report 97/E/20,Geesthacht.

141

GEOLOGIC EVIDENCE OF RAPID,MULTIPLE, AND HIGH-MAGNITUDECLIMATE CHANGE DURING THE LASTGLACIAL (WISCONSINAN) OF NORTHAMERICA

F. W. BACHHUBERUniversity of Nevada, Las Vegas,Las Vegas, NV, USA, 89154-4010

N. R. CATTOMemorial University of Newfoundland,St. John’s, Canada, A1B 3X9

Abstract

Over the last few decades data derived from ice cores and dendrochronologic sequencesindicate that former climate change during the Quaternary has been rapid and of highmagnitude. The Estancia Valley, central New Mexico, U.S.A. contains a high-resolutionsedimentologic and palaeontologic pluvial-lake record that supports these data. The knownEstancia Valley record, in outcrop and subcrop, spans the Wisconsinan (last glacial episodeof North America) and represents relatively continuous deposition through that time. Theperiod from early to middle Wisconsinan (ca. 60,000 BP to 30,000 BP) is marked by arelatively stable climate, represented by dry playa and saline lake sediment and fossils. Atthe beginning of the late Wisconsinan, however, climatic instability becomes the norm. Thefirst fresh-water pluvial lake, as evidenced by fossils, appears at ca. 24,000 BP, followed byadditional discrete fresh-water lakes at ca. 21,000 BP, ca. 20,500, ca. 20,000 BP, ca. 17,000BP, and ca. 14,000 BP. Each of these seemingly deeper-water stands was separated bydesiccation events (the last dating after ca. 12,400 BP) indicating that each lake phaseresulted from high magnitude climate change (full pluvial to full inter-pluvial climate). It

143


144 BACHHUBER AND CATTO

also appears that each lake phase developed rapidly, stabilized for only a short period, anddesiccated rapidly. Late Lake Estancia, the highest lake stand of about 52 m deep and datedat ca. 20,000 BP (coinciding with the late Wisconsinan glacial maximum) offers the mostconclusive evidence of rapid lake development from dry-playa climatic conditions. TheLate Lake Estancia sedimentary sequence consists of deposition on an unconformablesurface, an interval of saline-lake adapted organisms, brackish-water-adapted organisms,and finally infrahaline-adapted organisms, all within a 10 cm stratigraphic interval.Radiocarbon ages and sedimentation rates suggest that change from dry-playa conditions tofull pluvial-lake conditions (maximum lake depth) back to a dry playa occurred within a fewdecades to a few centuries. It is likely that the other fresh-water stands developed anddisappeared in a similar time span. The rise and fall of major pluvial systems in thesouthwest U.S.A. demonstrate that the geologic response to rapid climate change can also berapid.

1. Introduction

Detailed palaeontologic analysis of late-Quaternary lacustrine sediments from the EstanciaValley, central New Mexico suggest that pluvial-lake systems respond quickly anddramatically to climate change. The geologic record of the Estancia Valley delineates notonly multiple, high-level, fresh-water lakes, but also depicts waxing and waning phases oflake evolution, including ephemeral, low-level, lake phases of variable salinity, anddesiccation events—all reflecting regional climate change. The climatic record inferredfrom palaeontologic and sedimentologic change in the Estancia Valley is of greatercontinuity and of significantly higher resolution than that of glacial, fluvial, and aeoliansystems recorded in other areas of North America and compares to that of ice core records.It is believed that the degree of resolution in the Estancia stratigraphic section is on the orderof a few centuries, and in some cases lake level appears to fluctuate dramatically within afew decades. Significantly, the Estancia record clearly demonstrates that geologic andbiologic processes respond to rapid climate change, and in restricted environments thischange is recorded in great detail.

1.1 REGIONAL SETTING

The Estancia Valley is a broad, elliptically shaped, physiographic and structural basin nearthe geographic center of the state of New Mexico (Figure 1). Basin closure is afforded bythe Manzano Mountains along the west margin (highest elevations above 3,050 m), the

GEOLOGIC EVIDENCE OF CLIMATE CHANGE 145

Pedernal Hills along the east margin (up to 2,312 m), Juames Mesa to the south (2,106 m),and a board, sloping saddle (~1,983 m) to the north. A topographic sill is located in thesoutheast portion of the basin at an elevation of 1,932 m. The elevation of the central-valleyfloor is approximately 1,853 m where a complex of over 60 playa-floored deflation basins isincised, up to 10 m deep, along the axis of the valley. The Holocene-age deflation basinsare ringed by parabolic dunes, and other aeolian features (see Catto and Bachhuber, thisvolume).

At present, temperature and precipitation varies significantly within the EstanciaValley watershed (see Catto and Bachhuber, this volume). The central portion of the valley,however, is arid with a large annual water-budget deficit of ~1.5 m.


The water-budget deficit precludes the existence of perennial standing water other than atthe highest elevations in the Manzano Mountains. The deflation basins distributed along thevalley axis contain standing water only following heavy rainfall events. Within days or afew weeks, the deflation-basin floors return to dry playa conditions.

The high-resolution geologic and palaeontologic record contained in the Estancia Valley is,in part, related to its geographic and climatic setting. Occurring between 35°- 36°N latitude,the valley is near the southern margin of the area characterized by optimal pluvial lakedevelopment. At comparable latitudes in Nevada, similar closed basins apparently did notcontain large pluvial lakes (Mifflin and Wheat, 1979) during the Wisconsinan, the last NorthAmerican glacial stage (compares to Weichselian in Europe). The relatively high elevationof the Estancia Valley floor (1,853 m), however, supports a climatic threshold that, withmoderate precipitation or temperature change, produces a significant hydrologic response.Depending upon the magnitude of climate change the hydrologic thresholds crossed are dryplaya to wet playa to saline lake to fresh-water lake, or this succession in reverse. Thesedimentologic and biostratigraphic section of the Estancia Valley record these thresholds,and transitions between thresholds. At lower elevations, however, the response to similarclimate change is not as rapid or as sensitive, and only extreme climate events will berecorded in the geologic record.

In addition to the high valley elevation, which creates the potential for rapidly crossingcritical hydrologic thresholds, the numerous late Holocene deflation basins provide verticaland lateral accessibility to the late Pleistocene sediment record. Up to 10 m of Pleistocenelacustrine sediment is exposed along the slopes of the interior deflation basins, withprogressively reduced exposure thickness occurring in the more exterior (mainly eastern)deflation basins. Consequently, the great number of deflation basins and their juxtapositionto near-shore and deep-water lacustrine facies provides unique access to the geologic record.In outcrop, this record is relatively continuous from latest middle Wisconsinan through lateWisconsinan.

1.2 GEOLOGIC SETTING

The Estancia Valley lies within a structural basin of the easternmost Basin and Rangephysiographic province. The east and west margins of the valley are bordered by a systemof north-trending normal faults with uplifted areas exposing Precambrian- and Palaeozoic-age rocks. Mesozoic-age sediments occur in subcrop.


1.2.1. Pre-Wisconsinan

From early Miocene to early Pleistocene time, the Estancia Valley had an eastward-directed,through-flowing, fluvial system into the ancestral Pecos/Brazos river network (Kelley, 1972;Hawley, 1984). Basin closure, through faulting, probably had developed by middlePleistocene time. The first sedimentologic record of closure consists of a 7 m-thicklacustrine section (Early Lake Estancia) (Figure 2) recorded in water wells (Titus, 1969).Little is known of Early Lake Estancia, but geomorphic evidence and possible reworkedostracods in early to middle Wisconsinan-age sediment (Bachhuber, 1992) suggest that thelake was a fresh-water body and of a size probably greater than any high-lake stand of thelate Wisconsinan pluvial period. Early Lake Estancia is believed to be no younger thanIllinoian in age, and probably of Illinoian age (Bachhuber, 1989).

Early Lake Estancia sediment is overlain by 17 m of alluvium and dune material. Alsodescribed from water-well data (Titus, 1969), this succession, termed the Medial Sand(Figure 2), has not been tested for the presence of fossils. However, it is likely that theMedial Sand is no younger than Sangamonian in age (Bachhuber, 1992).

1.2.2. Early and Middle Wisconsinan

The Medial Sand is overlain by 10.5 m of intercalated clay and gypsarenite, only theuppermost portion of which crops out along the slopes of the deepest-incised deflationbasins. The unit, named the La Salina Complex (Figure 2), has a palaeontologic assemblageand sedimentology representing saline lakes of variable salt content, wet playas and dryplayas (Bachhuber, 1992). For clarification, the wet playa is defined as a system supportedby precipitation and ground-water flow. The floor of a wet playa contains water for much ofthe year, but desiccates periodically. The dry playa is exclusively supported by precipitationevents, and, although flooded following rainstorms, the floor is dry for most of the year.

The outcrop portion of the La Salina Complex is clearly middle Wisconsinan with agesof 30,440 ± 520 yr BP (Beta-25542) to 35,650 +3,000/-2,130 yr BP (A-4903). A age of>48,800 yr BP (AA-6330), along with amino acid racemization data (Rutter et al. 1992),from the middle of the La Salina Complex suggest that the lower portion of the unit is noyounger than early Wisconsinan in age. As such, the La Salina Complex is a relativelycontinuous record of cold/dry climate from early through middle Wisconsinan time.


Significantly, there is no record of high-level, fresh-water lakes in the La Salina Complex(Bachhuber, 1992).


1.2.3. Late WisconsinanThe La Salina Complex is overlain by ~ 10 m of thick-bedded flint-gray clay and inter-

bedded silty clay and gypsarenite. The complete late Wisconsinan section is exposed inoutcrop in the more deeply incised deflation basins such as E-28 (Figure 1). Only the upperportion is exposed in the marginally located basins that elevationally step-up away from thevalley axis, such as E-22 (Figure 1). Previously, the late Wisconsinan sediment had beensubdivided into two main pluvial lake stands, Late Lake Estancia and Lake Willard, with anintervening inter-pluvial section, the Estancia Playa Complex (Bachhuber, 1989). Inaddition, based on a rich and varied palaeontologic assemblage, six distinct phases wererecognized in Late Lake Estancia sediment (Bachhuber, 1989). The palaeontologic evidencedepicts fresh-water conditions as compared to more saline conditions. The model developedby Bachhuber (1989) had Late Lake Estancia rise from an early late Wisconsinan saline-lakephase, waxing and waning into three recognized high-water stands and two partial drawdown phases, and finally desiccating at the beginning of Estancia Playa Complex time.While this model is basically correct in recognizing high water and low water conditions, asinferred from lake salinity, we propose here that water level changes were more dramatic.Instead of partial draw down phases, desiccation occurred with these events separating thedistinct high-water phases into spatially and temporally discrete pluvial lakes.

2. Palaeolimnology of the Late Wisconsinan Pluvial System

The palaeolimnology of the late Wisconsinan pluvial and inter-pluvial sections isdetermined by the fossils, sedimentology, and stratigraphy exposed along the flanks ofdeflation basins. Owing to the distribution of the deflation basins inside the loweststrandlines, near-shore and deep-water facies are readily available for examination andsampling. The abundance and diversity of fossils, all extant species, permit palaeoecologicevaluation. From these data inferences can be drawn as to palaeolimnologic conditionsespecially those related to total dissolved solid (TDS) of the various lake stages. TDS, inturn, is believed to reflect water depth. Saline bodies are considered to have lesser depthswith the surface extent of the lake being relatively small. Fresher water bodies areconsidered to be of greater depth, with the lake occupying a larger surface area. Thesecontentions are partially supported by the occurrence of a series of strandlines that vary inelevation from approximately 1860 m to 1897 m along the Estancia Valley margins, and byvertical facies changes indicating transgressive and regressive sequences. While these


relationships have been long recognized, recent detailed palaeontologic analyses andadditional ages have required a re-evaluation of previous work (Bachhuber, 1989).

2.1. PALAEONTOLOGY

The diversity and density of fossils in the late Wisconsinan sediment of the Estancia Valleyare remarkable. Microfossils and macrofossils occur in varying abundance in all previouslydescribed Late Lake Estancia phases and in Lake Willard, from near-shore and deeper-waterfacies. In general, two basic assemblages are recognized: one characteristic of a fresh-wateror near fresh-water environment; and a second that is indicative of more saline to evenhypersaline conditions. The fresh-water indicators or inhabitants are cutthroat trout,gastropods, pelecypods, Pediastrum (an alga), and some ostracod and diatom species.Saline indicators are certain species of ostracods, diatoms and charophytes, foraminifers(Bachhuber and McClellan, 1977), and Ruppia (ditchgrass). In addition, the sedimentcontains head capsules, wings, and mandibles of insects, seeds from various plants, plantfragments and impressions, pollen, salamander bones, ephippia, and innumerable spheresand disks of unknown affinities. Even though some of the Estancia fossil material has notbeen identified or evaluated, based on work previously completed (Bachhuber, 1989; 1992),there is a good understanding of the various late Wisconsinan lake stands, and the conditionsthat prevailed during their existence. The problem with previous work, however, was thatsediment sampling strategies were too coarse. Normally samples were collected at 5 – 10cm intervals, with the vertical thickness of the sample encompassing up to 5 cm. It has nowbeen realized that this sampling strategy resulted in completely missing a number of majorsalinity reversals, and it blurred the boundaries between distinct saline-lake and fresh-waterevents. Single processed samples often contained palaeontologic indicators of both salineand fresh-water environments, suggesting that the organisms lived together. In certainsituations, these occurrences are real, reflecting mixing and rebedding of previouslydeposited material. In most cases, however, the mixing of what should be incompatiblepalaeoecologic assemblages was an artefact of the sampling technique. Consequently,stratigraphic sections were recollected and reanalyzed with a focus on stratigraphic detail atthe 1 – 2 cm interval. It is this highly detailed analysis along with the discovery of a criticaloutcrop section (LP-IS) that provides the additional palaeolimnologic information for thispaper. It is not the purpose of this paper however to detail the wealth of palaeoecologicaldata that presently exists. Instead, the focus is on the rapidity and magnitude ofpalaeolimnological change, and subsequent inferred palaeoclimatological change.


(Figure 3). Obviously L. staplini is extremely euryhaline, but high abundance of theostracod in the Estancia sediment is invariably associated with mid-salinity associationsaround 35,000 ppm TDS. At higher salinity and in fresher-water bodies (as indicated byother organisms), L. staplini occurs in low abundance. Marine foraminifers also occur inEstancia Valley sediment, and where present a salinity of approximately 35,000 ppm isindicated (Figure 3).

Fresh-water IndicatorsThe most typical and definitive fresh-water indicator in the Estancia Valley sedimentappears to be cutthroat trout (Oncorhynchus clarki). However, while cutthroat trout bones,even skin fragments, are common in the fresh-water phases during the late Wisconsinan, itsusefulness as a fresh-water indicator is, in fact, somewhat limited. Cutthroat trout, a nativeNorth American species, is most commonly found in modern infrahaline (<500 ppm) lakes


and rivers, but it is also reported from lakes with TDS varying from 5,000 ppm to 12,000ppm (Bachhuber, 1989). Its occurrence, therefore, in Estancia Valley sediment seeminglyconstrains salinity to <12,000 ppm TDS. Trout fossils, however, are a good initial indicatorof pluvial conditions in general, even though water quality could have been brackish insteadof fresh. Fortunately trout fossils are most often associated with other fresh-water indicatorsduring early lake evolution. As these other indicators disappear from the record and troutcontinues, it is reasonable to assume that water levels dropped with resultant evolution fromfresh water to brackish water.

Certain ostracods in particular constrain water quality to truly fresh-water conditions, andthese taxa are always found in association with trout bones. Although restricted in range,Cytherissa lacustris is the most diagnostic fresh-water indicator of all the organisms foundin Estancia Valley sediment (Figure 3). C. lacustris is found only in modern, deep, cold-water, infrahaline (<500 ppm) water bodies of holarctic distribution (Delorme, 1969). Itsappearance in two discrete stratigraphic intervals, therefore is unquestionable evidence offresh-water conditions, and likely deep-water conditions. Candona caudata is another keyindicator of fresh-water conditions. This ostracod is reported in modern water of salinityrange between 20 and 2054 ppm TDS (Figure 3). Somewhat less useful is Limnocythereceriotuberosa , which is found in modern aquatic bodies ranging in salinity from 500 to25,000 ppm (Figure 3). In isolation, the more euryhaline nature of L. ceriotuberosa canreflect highly brackish conditions as well as fresh-water bodies. However, as L.ceriotuberosa always occurs with trout fossils, its presence in the Estancia sediment reflectssalinity of <12,000 ppm.

Two other ostracods occur in interpreted fresh-water conditions in the Estancia Valley.Candona rawsoni and L. staplini are always found in association with other fresh-waterindicators. Conversely, they are also almost always found in saline environments, and thusthe simple occurrence of either or both of these ubiquitous species poorly constrains waterquality. Nonetheless, an important stratigraphic relationship is seen in much of the Estanciarecord, although not yet substantiated in the ecologic record. The ostracods Candonarawsoni and L. staplini dominate throughout the Estancia Valley stratigraphic sequence andoccur in virtually all stratigraphic horizons. L. staplini is found in modern bodies of watervarying in salinity from fresh water to almost 200,000 ppm TDS. C. rawsoni occurs inmodern environments that range in salinity from fresh water to about 43,000 ppm TDS.Although both are euryhaline in modern environments, there appears to be a strongcorrelation between the relative abundance of these ostracods and lake salinity. When L.


staplini occurs in numbers much greater than C. rawsoni , other organisms (such asforaminifers) indicate that salinity is around 35,000 ppm or higher. Conversely, when the C.rawsoni /L. staplini ratio (Cr/Ls) is approximately 1:3 to greater than 1:1, other indicators(such as cutthroat trout) demonstrate that fresh-water or slightly brackish-water conditionsprevailed. This relationship still needs to be verified in modern environments, but it is arelatively constant relationship in Estancia Valley sediments. In an Estancia Valley pluviallake, a relatively large Cr / Ls ratio is therefore interpreted to indicate fresh-water or nearfresh-water conditions.

The final fresh-water indicator discussed here is Pediastrum cf. boryanum, a commoncosmopolitan green algae found in modern fresh-water environments in Europe and NorthAmerica. Even though the species present within the Estancia sedimentary record has yet tobe definitively identified, the genus as a whole is a good indicator of fresh-waterenvironments (Figure 3). Unlike the other Estancia Valley fossil organisms that areextracted directly from dried sediment (Bachhuber, 1989), Pediastrum is a component ofsamples processed for pollen content.

2.3. STRATIGRAPHY AND BIOSTRATIGRAPHY

The sediment samples used for this study were collected from deflation basin E-28 andsection LP-IS (Figure 1). As palaeontology and sedimentology are consistent throughoutthe Estancia Valley, however, these exposures are representative. Importantly, E-28contains the thickest outcrop section of late Wisconsinan sediment, and appears to representthe deepest-water sedimentation throughout this time interval. The location of E-28coincides with [what is believed to have been] the topographic low point of the depositionalfloors of the various late Wisconsinan pluvial lakes. Even during desiccation events thesedimentary section at E-28 was least likely to be truncated or deflated. This geologicrecord is the most continuous and has the greatest stratigraphic resolution.

The LP-IS section is a small outlier located on the east margin of Laguna del Perro (Figure1). Here, erosion has removed any sediment overlying the upper part of the Late LakeEstancia section, with the exception of a thin cap of late Holocene aeolian material.Fortuitously, erosion of the upper portion of the section and its isolation from the deflationbasin margin has produced a stratigraphic section that is thoroughly dry, and nearly verticalin development. It is here that the lateral aspect of specific stratigraphic intervals can bebest demonstrated. Section LP-IS is approximately 1 km west of the E-28 section. As such,


it contains a stratigraphic section that is slightly thinner and depositionally upslope from thesection outcropping at E-28.

The Estancia Valley stratigraphic section is characterized by intercalated flint-gray silty clayand white, thin-bedded gypsarenites. Palaeontologic data demonstrates that the alternatingnature of clay units and gypsarenite units is reflective of oscillating water depth, withgypsarenite deposition regressing into the interior of the various lacustrine basins as waterdepth decreases. Likewise, an increase in clay deposition mirrors lake expansion.

2.3.1. Basal Unit

At E-28, the late Wisconsinan section (Figure 4) rests on an unconformable surfacedeveloped on clay and gypsarenite of the early to middle Wisconsinan-age La Salinacomplex ages from >48,800 to ca. 30,000 yr BP). The basal unit of the lateWisconsinan consists of a poorly-sorted pebbly, sandy silt that contains charophyte casts(Bachhuber, 1989), highly abraded Ruppia seeds (Rutter et al. 1992), horse-teethfragments, and saline-indicator ostracods. The basal unit, highly variable in thickness to amaximum of 90 cm, also contains an abundance of Palaeozoic-age flint, chert, and limestoneclasts weathered from the Manzano Mountains on the western flank of the Estancia Valley.Sedimentologic aspects of the unit indicate that the palaeontologic assemblage, for the mostpart, represents reworked and rebedded La Salina Complex fossils deposited by flash floodswhen the valley floor was subaerially exposed. Within, and at the top of the section, cut-and-fill structures are common. At one location, the overlying pluvial section is depositedas slumped infill in a 1 m deep trough cut into and through the basal unit.

2.3.2. Fresh-water Lake 1

A 1 m - thick flint-gray clay rests unconformably on the Basal Unit. This stratigraphicsequence, referred to as the “initial freshwater phase” of Late Lake Estancia (Bachhuber,1989) clearly marks the first fresh-water and deeper-water event of the late Wisconsinan.Bachhuber (1989) originally envisioned this fresh-water phase arising and expanding froman established saline lake. The recognition, however, of a lower and upper unconformablesurface of the Basal Unit indicates that the lake expanded from a subaerially exposed valleyfloor and rapidly evolved into a fresh-water body. Palaeontologic evidence traces early lakeevolution from brackish-water to fresh-water conditions. The fresh-water phase ischaracterized by the occurrence of cutthroat trout, Pediastrum, tiger salamander(Ambystoma tigrinum), and the ostracod Candona caudata, a fairly good indicator of salinityless than about 2000 ppm TDS. Salamander bones have been dated at 24,300 ± 560 yr BP


(AA-1868), and ostracods at the same stratigraphic level have been dated at 23,510 ± 240 yrBP(AA-14068).


Owing to the uncertainties associated with bone dates, the latter age appears to be morereliable. The ages date the time of maximum development of fresh-water and inferreddeepest water conditions. Stratigraphically, the fresh-water interval is about 10 cm thickwith the total Lake 1 sequence (including brackish-water conditions) being 40 cm thick,suggesting that the high lake stand had a short life. Ostracod data indicate that following thefresh-water peak, the lake rapidly collapsed into a brackish-water phase, followed by asaline-lake phase, and finally desiccation.

2.3.3. Ruppia Zone 1

Following the collapse of Lake 1, the floor of the Estancia Valley was subaerially exposedon a number of occasions and covered intermittently by shallow, saline lakes at other times.The subaerial and saline lake events encompass approximately 1 m of stratigraphic sectionat E-28, consisting of silty clay deposited during saline lake phases and poorly- to well-sorted gypsarenites. At LP-IS the subaerial event is characterized by a 1 m thickunconformably-bound depositional sequence containing large numbers of abraded Ruppiaseeds, ostracods M. ingens and H. salinas , and intermittent zones with very highabundances of L. staplini . The seeds and some of the ostracods were reworked andredeposited as epiclastic assemblages from one of the many ephemeral saline-lakes thatexisted following Lake 1 desiccation. The Ruppia Zone 1 sedimentologic sequence at LP-ISis highly variable in thickness, locally pinching out completely, juxtaposing the Lake 1sequence directly in contact with younger Lake 2 sediment. Ostracod valves from the upperpart of the interval are dated at 20,520 ± 200 yr BP (AA-14067). Although this valueapproximates the correct time frame, the age is slightly younger than that determined for theoverlying Lake 2 sequence. The Ruppia Zone 1 age determination, even though out ofstratigraphic context, suggests that the Ruppia seeds themselves have been reworked fromlate Wisconsinan sediment, rather than from the La Salina complex that dates older than ca.30,000 yr BP.


As indicated by the LP-IS section, Lake 2 sediment rests unconformably on sediment from afinal subaerial event of Ruppia Zone 1. At E-28, the Lake 2 section is only about 10 cm inthickness. The section contains cutthroat trout, and ostracods C. caudata and L.ceriotuberosa, neither of which occur in abundance. The occurrence of C. caudataconstrains the upper limit of salinity to about 2,000 ppm TDS. The upper and lowerbounding unconformable surfaces of the sedimentary sequence along with the occurrence offish, L. ceriotuberosa, and C. caudata indicate the existence of a discrete fresher-water lake,


albeit one of short-term duration. While water depth is unknown, Lake 2 is assumed to berelatively deep and of large surface extent. Ostracod valves from this horizon have beendated at 21,060 ± 180 yr BP (AA-14066).


Lake 2 is unconformably overlain by Ruppia Zone 2. The sedimentologic aspects here aresimilar to those of Ruppia Zone 1, but the stratigraphic interval is much thinner. Ruppiaseeds are not abundant, but epiclastic deposition is indicated. At LP-IS the zone is less than10 cm thick, bounded by unconformable surfaces, and has a very low abundance ofostracods. In places, Lake 3 sediment seemingly lies conformably on Lake 2 sediment eventhough unconformable relationships are laterally well exposed. At E-28, the section is about15 cm thick, contains saline-indicator ostracods such as M. ingens and H. salinas, and at onestratigraphic level, abundant L. staplini. The sedimentologic and palaeontologic datasuggest that the section at LP-IS was subaerially exposed, but this may not have been true atE-28. Here the ostracod assemblage appears to be depositionally intact, indicating theexistence of a saline lake throughout the time interval. If a continuous saline lake existed,based on basin morphometry, it would have covered no more than a few There is no

chronology for Ruppia Zone 2, but based on stratigraphic thickness and the apparentlack of unconformable surfaces at E-28, the interval must have been of short duration.


At E-28 the flint-gray clay of Lake 3 is approximately 20 cm thick and contains cutthroattrout, Pediastrum, and the ostracod L. ceriotuberosa (Figure 4). The alga Pediastrumespecially constrains salinity to infrahaline conditions. Trout bones were not found at theLP-IS section and the sediment was not processed for Pediastrum, although L. ceriotuberosaoccurs. C. rawsoni /L. staplini ratios vary from 1:2 to greater than 1:1, a relationship thatwe interpret as a reflection of fresh-water to slightly brackish-water conditions. No ageexists for Lake 3, but this phase must be younger than ca. 21,000 yr BP (the age of Lake 2)and older than ca. 20,000 yr BP (the age of the upper portion of the overlying Ruppia Zone3). This is also the age bracket for the underlying Ruppia Zone 2. Obviously, Lake 3 isyounger than the underlying saline lake and subaerially deposited sediment of that zone. Itis estimated here that the age of Lake 3 is approximately 20,500 yr BP.


Ruppia Zone 3 is similar in all aspects to the older Ruppia Zone 1. It unconformablyoverlies Lake 3 sediments, reflects multiple saline-lake sequences and subaerial processes,


albeit one of short-term duration. While water depth is unknown, Lake 2 is assumed to berelatively deep and of large surface extent. Ostracod valves from this horizon have beendated at 21,060 ± 180 yr BP (AA-14066).

2.3.5. Ruppia Zone 2Lake 2 is unconformably overlain by Ruppia Zone 2. The sedimentologic aspects here aresimilar to those of Ruppia Zone 1, but the stratigraphic interval is much thinner. Ruppiaseeds are not abundant, but epiclastic deposition is indicated. At LP-IS the zone is less than10 cm thick, bounded by unconformable surfaces, and has a very low abundance ofostracods. In places, Lake 3 sediment seemingly lies conformably on Lake 2 sediment eventhough unconformable relationships are laterally well exposed. At E-28, the section is about15 cm thick, contains saline-indicator ostracods such as M. ingens and H. salinas, and at onestratigraphic level, abundant L. staplini. The sedimentologic and palaeontologic datasuggest that the section at LP-IS was subaerially exposed, but this may not have been true atE-28. Here the ostracod assemblage appears to be depositionally intact, indicating theexistence of a saline lake throughout the time interval. If a continuous saline lake existed,based on basin morphometry, it would have covered no more than a few There is no

chronology for Ruppia Zone 2, but based on stratigraphic thickness and the apparentlack of unconformable surfaces at E-28, the interval must have been of short duration.


At E-28 the flint-gray clay of Lake 3 is approximately 20 cm thick and contains cutthroattrout, Pediastrum, and the ostracod L. ceriotuberosa (Figure 4). The alga Pediastrumespecially constrains salinity to infrahaline conditions. Trout bones were not found at theLP-IS section and the sediment was not processed for Pediastrum, although L ceriotuberosaoccurs. C. rawsoni /L. staplini ratios vary from 1:2 to greater than 1:1, a relationship thatwe interpret as a reflection of fresh-water to slightly brackish-water conditions. No ageexists for Lake 3, but this phase must be younger than ca. 21,000 yr BP (the age of Lake 2)and older than ca. 20,000 yr BP (the age of the upper portion of the overlying Ruppia Zone3). This is also the age bracket for the underlying Ruppia Zone 2. Obviously, Lake 3 isyounger than the underlying saline lake and subaerially deposited sediment of that zone. Itis estimated here that the age of Lake 3 is approximately 20,500 yr BP.


Ruppia Zone 3 is similar in all aspects to the older Ruppia Zone 1. It unconformablyoverlies Lake 3 sediments, reflects multiple saline-lake sequences and subaerial processes,


and has an upper unconformable surface. At LP-IS the zone is highly variable in thicknessand laterally pinches in and out across the face of the LP-IS outcrop. Numerousunconformable surfaces and cut-and-fill structures are in evidence throughout the interval.These aspects are clearly observable at LP-IS, and are substantiated to a lesser degree at E-28. At E-28, although an upper unconformable surface is difficult to demonstrate, multipleunconformable surfaces appear within the section. Ruppia seeds occur in large numbers inan epiclastic assemblage, and in certain intervals there is a paucity of other fossils, includingvery low numbers of ostracods. These low abundance intervals represent dry playadeposition.

Ostracod fragments associated with abraded Ruppia seeds in a cut-and-fill structure in thelower portion of the zone have a age of 18,300 ± 170 yr BP (AA-14065). From themiddle of the zone, in a similar association, Ruppia seeds have an age of 20,040 ± 240 yrBP (AA-1867). The younger age is significantly out of stratigraphic context, andcontamination from modern roots is suspected. The ca. 20,000 yr age, however, is instratigraphic context.

2.3.8. Fresh-water Lake 4 (Late Lake Estancia)

Lake 4 unconformably overlies Ruppia Zone 3 at E-28 and LP-IS. The sedimentologicinterval contains a wealth of fresh-water indicators that led to the initial recognition of fresh-water conditions in the late Wisconsinan section of the Estancia Valley, and the designationof Late Lake Estancia (Bachhuber, 1971). Early work, however, did not recognize thecomplexity of the Wisconsinan-age section nor the palaeoclimatic implications. Thedesignation Late Lake Estancia (Bachhuber, 1989) included all the previously describedLate Wisconsinan sediment of this paper up to the Estancia Playa Complex (discussionfollows later), a stratigraphic interval of about 6 m in thickness. Here, the term Late LakeEstancia refers only to the sediment section of Lake 4, an interval of approximately 40 cm inthickness.

At E-28, the Late Lake Estancia section consists of flint-gray clay overlying the Ruppia -bearing, iron stained, silty clay of Ruppia Zone 3. Analysis of sediment sampled at 2-cmthick intervals demonstrates a rapid transition into fresh-water conditions with a basalsediment unit comprised of both fresh- and saline-water indicators. A few centimetershigher in the section, Late Lake Estancia sediment consists of a low-abundance ostracodassemblage where the C. candona /L. staplini ratio is approximately 1:1, indicatingdeveloping fresh-water conditions. The assemblage 4 cm vertically above this level


contains trout fossils and Pediastrum, and is dominated by C. caudata and Cytherissalacustris. The latter ostracod is found only in modern, deep, cold-water, infrahaline (<500ppm) water bodies. The occurrence of C. lacustris especially marks the glacio-pluvialmaximum of the late Wisconsinan. High abundance of C. lacustris occurs only through avertical stratigraphic interval of 6 cm, with very low abundance within the overlying 6 cm.Therefore, the freshest-water and deepest-water conditions seemingly prevailed over a veryshort time. It is at this time that the highest strandline ringing the basin may have formed,indicating maximum lake depth of about 52 m and surface extent in excess of

At LP-IS, where the section represents an intermediate position between littoral zonedeposition and the deep-water deposition of E-28 detailed palaeontologic analysis (Figure 5)reflects the same basic sequence as seen at E-28 but with higher resolution. The upper


portion of Ruppia Zone 3 is characterized by a distinct saline-lake interval unconformablyoverlain by the Ruppia -bearing reworked sediment deposited subaerially on a dry playasurface. This sediment, in turn, was reworked as Late Lake Estancia began to develop. Theinitial depositional phase of Late Lake Estancia consists of a mixture of the saline-indicatorsfrom the uppermost portion of Ruppia Zone 3, and the initial fresh-water inhabitants of thegrowing and deepening lake. Ostracod values from this earliest lake-developmental phaseare dated at 19,950± 190 yr BP (AA-14064). The overlying C. lacustris interval is dated at19,740 ± 150 yr BP (AA-14063), and 19,760 ± 160 yr BP (AA-6329) with both ages derivedfrom C. lacustris values from E-28.

Following ca. 19,750 yr BP, C. lacustris disappears rapidly from the E-28 and LP-ISstratigraphic sections. This disappearance signals a decrease in depth of lake and volumewith a resultant progressive increase in salinity to brackish-water conditions to 12,000ppm TDS over the next 35-cm stratigraphic interval. Initially C. caudata remains inrelatively high abundance, and after it disappears, C. candona abundance approximates thatof L. staplini. Cutthroat trout bones are common throughout the highest lake stand andmuch of the stratigraphic interval represents decreasing water depth.

The tight grouping of ages and the contrasting palaeoenvironmental conditionsillustrated by Figure 6 is the clearest evidence of rapid and high magnitude palaeolimnologicchange in the Estancia stratigraphic record.

2.3.9. Zone 4

At LP-IS, following the disappearance of C. lacustris, the Late Lake Estancia sequence isabruptly terminated (Figure 5). The longer stratigraphic fresh-water record exhibited at E-28 apparently has been truncated by subaerial erosion at LP-IS indicating that Late LakeEstancia either desiccated or water level dropped to only a few metres in depth in thedeepest portion of the basin, at E-28. A 15-cm thick gypsarenite capping the Late LakeEstancia sediment at E-28, however, suggests that desiccation occurred.

2.3.10. Lake 5

Following the collapse of Late Lake Estancia, Lake 5 developed rapidly in what appearsto be 2 phases. An initial fresh-water phase is characterized by a Cr/Ls ratio ofapproximately 1:1 and a low-abundance of C. caudata, that indicates infrahaline or nearinfrahaline conditions. This initial phase was followed by a short-lived partial draw down


phase, but water level recovered quickly as indicated by the appearance of ostracod L.ceriotuberosa. The section also exhibits a Cr/Ls ratio of greater than 1:1, and abundant fishfossils. Although L. ceriotuberosa has an extant salinity range of 500 – 25,000 ppm(Forester, 1986), its occurrence in the Estancia sediment, and its association with otherfresher water organisms, seems to reflect salinity closer to fresh- or slightly brackish-waterconditions. Therefore, Lake 5, even though it does not have infrahaline indicators other thanin the initial phase, does appear to have been relatively fresh to slightly brackish throughmost of its existence. L. ceriotuberosa occurs over a stratigraphic interval of 30 cm and thendisappears. In association with the disappearance of L. ceriotuberosa, total ostracodabundance significantly decreases, but the Cr/Ls ratio remains greater than 1:1, and cutthroattrout fossils are present. We believe that at this point the lake has evolved into a brackish-water system with TDS approximating 12,000 ppm, the upper limit tolerated by cutthroattrout.

Although Lake 5 seemingly never approached the infrahaline, deep-water conditions of LateLake Estancia, it remained as a persistent lake body over a stratigraphic interval of 1 m.Therefore, as compared to the other lake systems in the Estancia Valley, Lake 5 may havehad the longest period of quasi-stability during the late Wisconsinan. Ostracod valves fromthe interval characterized by L. ceriotuberosa have a age of 16,890 ± 150 yr BP (AA-14062).

2.3.11. Estancia Playa ComplexThe eventual desiccation of Lake 5 marked the beginning of the Estancia Playa Complex, astratigraphic sequence characterized by subaerial erosion and deposition, and palaeontologicevidence supporting the occurrence of numerous discrete saline lakes. Gypsarenites andgypsiferous silts are dominant, and cut-and-fill structures and unconformable surfaces arecommon. Intervening clay-rich sediments have a great abundance of saline-indicatorostracods M. ingens and L. staplini . In addition, one saline lake from the upper portion ofthe Estancia Playa Complex has an abundance of 2 species of marine foraminifers(Bachhuber and McClellan, 1977). Foraminiferal tests from this interval have an age of14,345 ± 105 yr BP (AA-6328). This age approximates the close of the Estancia PlayaComplex sequence.

2.3.12. Fresh-water Lake 6 (Lake Willard)

Lake Willard was originally identified by Bachhuber (1971) as consisting of a single fresh-water phase. Detailed palaeontologic analyses of the flint-gray clay to gypsiferous silt


section, however, demonstrates that the lake consisted of 2 distinct fresher-water phasesseparated by an intervening more saline phase, and possible desiccation. The first fresh-water phase, here termed Lake Willard A, is characterized by a high-density ostracodassemblage of C. candona (Cr / Ls ratio greater than 1:1), L. ceriotuberosa, and lownumbers of C. lacustris. Cutthroat trout bones and Pediastrum are common.Unquestionably, Lake Willard A was a fresh-water, infrahaline lake of approximately 500ppm TDS. At E-28, Lake Willard A is represented by a total stratigraphic thickness of lessthan 10 cm, and a age derived from ostracod valves of 13,700 ± 105 yr BP (AA-6327).

Following Lake Willard A’s freshest, deepest phase, water level dropped rapidly with asignificant increase in salinity. A series of shallow saline lakes may have occupied thevalley floor until Lake Willard B developed. Although it is suspected that desiccation didoccur, sedimentologic evidence of this is not well demonstrated. Lake Willard B marks thefinal nearly fresh-water lake in the Estancia Valley late Wisconsinan history. Cutthroat troutreappears and C. rawsoni abundance is much greater than L. staplini abundance. Butabsence of other fresh-water indicators suggests that Lake Willard B was not as fresh or asdeep as Lake Willard A. A age on ostracod valves from the interval is 12,460 ± 135 yrBP (Beta-25819). The stratigraphic thickness of Lake Willard B is approximately 10 cm.

Salinity increased and water level decreased rapidly in Lake Willard B, with apparentevolution directly into a saline lake dated at 12,375 ± 95 yr BP (AA-6326). The absence ofpalaeontologic content suggests that this saline lake desiccated, followed by thedevelopment of a final discrete saline lake. Desiccation of this final saline lake marked thebeginning of the development of the Willard Soil (Bachhuber, 1982), and the close of thelate Wisconsinan in the Estancia Valley.

3. Chronology and Lake Duration

The Estancia Valley sedimentologic and biostratigraphic record demonstrates that the lateWisconsinan was a time period of rapid and high magnitude palaeolimnologic change.Many of the palaeolimnologic events are well constrained by ages, with all fresh-waterlake stages with the exception of Lake 3 having at least one definitive radiocarbon age.With the exception of two dates used in this paper, ages have been determined byAccelerator Mass Spectrometer (AMS) techniques on ostracod valves of single species ofostracods. The ostracod valves were meticulously hand picked from sediment samples and


each AMS sample represents up to 2500 valves. Without the advent of AMS technology,few ages would be available from the Estancia Valley stratigraphic sequence.

Based on chronologies, Lake 4 (Late Lake Estancia) and Lake Willard A and B reflectclimate events that are well recorded in other areas of North America (Mickelson et al.1983). Late Lake Estancia, likely the deepest and largest pluvial lake in the lateWisconsinan, should be time synchronous with the late Wisconsinan glacial maximum of ca.18,000 yr BP. The age of Late Lake Estancia, however, is well constrained at ca. 19,800yr BP. We believe the ~1800 year difference in chronology is due to an hard-watereffect in the Estancia samples. A chronological difference of similar magnitude has affectedthe Lake Willard ages. Lake Willard appears to be correlative with the global YoungerDryas event, between ca. 10,000 to 11,000 yrs BP (Peteet, 1995). Lake Willard A dates atca. 13,700 yr BP and Lake Willard B dates at ca. 12,500 yr BP. If Lake Willard A is thetime equivalent to the Younger Dryas, the hard-water effect would represent a difference ofabout 2,700 years. On the other hand, if Lake Willard B represents the Younger Dryas, thehard-water effect is a difference of about 1500 years, an hard-water effect more comparableto that of Late Lake Estancia. The hard water effect and the timing of the Younger Dryas inthe Estancia Valley have yet to be resolved, but it is proposed here that Lake Willard B isequivalent to the Younger Dryas. Lake Willard A is then a major event not yet recognizedin the North American record.

Each one of the 7 pluvial events recorded in the Estancia Valley late Wisconsinanstratigraphic section is believed to be a discrete fresh-, deep-water lake (Figure 6) separatedby either a dry playa environment, or by a small, shallow saline lake. As such, the durationor life history of each of the individual lake stages can be calculated using sedimentation


rates. Although prevalent bioturbation means that laminated clay is not a commondepositional mode during the Estancia Valley lake histories, a number of laminated sectionsdo occur. The laminated sediment usually consists of a couplet with an upper thin, blacklayer and a thicker lighter colored layer. Even though the laminations appear to be varves,this has yet to be verified. If the laminations do represent annual deposition, the averagesedimentation rate derived from a number of laminated sequences from the variousinfrahaline and brackish-water lake stages is 3 mm/yr. This value is somewhat large forhigher latitude North American lakes.

occurring in well vegetated regions. However, 3 mm/yr of accumulation represents a lowrate in environments where aeolian input would be high regardless of a xeric or mesicclimate.At E-28, where the deepest water sedimentation has occurred through time, the various lakephases are well differentiated by lower or upper unconformable surfaces orpalaeontologically distinct highly saline phases. Therefore, the stratigraphic thickness ofeach lake stage is easily and accurately calculated. Using an average annual sedimentaccumulation of 3 mm, the life-duration of each of the 7 lake stages as infrahaline (about500 ppm) or brackish-water (<12,000 ppm) bodies is illustrated in Table 1.


These calculations indicate that almost all of the lakes developed, stabilized, and desiccatedin a time period near that of human perception. The total duration of the existence of fresh-water or near fresh-water conditions during the late Wisconsinan is less than 900 years. Asthe upper part of the middle Wisconsinan La Salina complex (Bachhuber, 1992) is dated atca. 30,000 yr BP, the 1000 years that the Estancia Valley was occupied by pluvial conditionsrepresents only about 4 % of late Wisconsinan time. Conversely, 17,000 years or 96 % oflate Wisconsinan time, was characterized by either dry playa or saline lake environments.The paradigm that the late Wisconsinan in the Southwest USA was a pluvial time period isgrossly exaggerated. Pluvial conditions did exist, and the magnitude of pluvial events wasgreat, but the climates, albeit repetitive, that give rise to these systems were of shortduration.

4. Palaeoclimate

The late Wisconsinan sedimentological and biostratigraphical record of the Estancia Valleyis clear evidence of rapid and high magnitude palaeolimnological change. While the casehas been made for fluctuating lake levels, this is inferred from the biostratigraphic evidencethat actually reflects change in water quality (TDS), which is a function of water volume. Inthe environment of central New Mexico, a lake becomes alkaline enriched with decreasedtotal salinity only by diluting existing high-salinity water with a large volume ofenriched fresh water or by flooding a dry playa with fresh-water flow. This implies a climatechange with resultant increased surface runoff and ground-water recharge that, in turn,produces high lake levels. Conversely, sulphate enrichment and high total salinity is afunction of increased evaporation that produces lower lake levels, and eventuallydesiccation. Therefore, indirectly, dramatically fluctuating lake levels are documented inthe Estancia Valley record. Most, if not all, of these fluctuations in lake level areclimatically controlled.

The volume of water in the valley at any given time is predominantly a function of surfacerunoff and ground-water recharge, and evaporation. As compared with present climaticconditions, the existence of a lake in the geologic record requires that the interrelationshipbetween precipitation and temperature results in a positive water budget, with a large lakerequiring significantly higher available water than that of a saline lake. Albeit an oversimplification, the 7 high lake stages of the late Wisconsinan register times of increasedprecipitation and/or decreased mean annual temperature. Likewise, saline lakes and dryplayas represent climatic conditions more similar to those of the present. In moving fromdry playa to saline lake to a fresher- water lake important hydrologic thresholds must be


crossed. As mentioned previously, the high elevation of the Estancia Valley floor is one ofthe important reasons contributing to the high-resolution biostratigraphic record. It couldthen be argued that valley-floor elevation places the region in a micro-climatic situationwhere crossing of hydrologic thresholds may involve only a slight change in precipitationand/or temperature. This is thought to be the case for crossing the dry playa/saline lakethreshold, but evolution into a deep, fresh-water pluvial system requires a more significantclimatic change. It is beyond the scope and purpose of this paper to dwell upon the variousclimatic scenarios that could give rise to alternating pluvial and inter-pluvial episodes. Thedata presented here simply supports such climatic change. However, the evidencesupporting rapid, multiple, and high magnitude change during the Wisconsinan, along withthe evidence that supports the view that individual pluvial lakes existed for only a shorttime, decades in some cases, brings into question the present disposition of the NorthAmerican and European glacial climatic record. By extension, we question the inferencethat North American and European glacial climate during the late Wisconsinan / lateWeichselian was marked by relatively uniform conditions until the commencement ofglacial retreat, ca. 14,000 BP. Instead, during this time period, it is likely that numerous andsignificant climatic oscillations occurred that were not recorded in the low-resolution glacialrecord.

5. Conclusions

An eight-metre-long high-resolution stratigraphic section from the Estancia Valley spans a12,000 year period of late Wisconsinan time from ca. 24,000 yr BP to ca. 12,000 yr BP. Thesection has been analyzed in detail for its palaeontologic content. Fossils indicate that thetime interval was characterized by a dynamic palaeoenvironments. Wide fluctuations in therange and abundance of various species/genera are believed to be a response to dramaticallychanging palaeolimnological conditions, particularly in terms of total salinity. Variations insalinity must reflect changes in water depth and volume. The record is interpreted asrepresenting 7 discrete fresh-water lake stages and intervening periods of desiccation.Crossing hydrologic thresholds from dry playa to saline lake to fresh-water lake or thissuccession in reverse can only be a function of climate change. Based on the EstanciaValley palaeontologic record, the following conclusions are drawn.

The Estancia Valley has a high-resolution record of rapid, multiple, and high-magnitude palaeolimnological change. During the glacial maximum, Late LakeEstancia attained a depth of about 52 m and had a surface area in excess of 1,200

The other pluvial lakes were likely to have been comparable in depth and size.


Wisconsinan palaeolimnological change, only in part due to the high elevation ofthe valley floor, is mainly a product of rapid, multiple, and high-magnitudepalaeoclimatic change.

Individual fresh-water lakes developed quickly and survived only for shortperiods of time, between a few centuries and a few decades.

The total time duration of the 7 pluvial lakes was less than 900 years. Thisrepresents only 4 % of late Wisconsinan time, inferring that glacio-pluvial climate,while dramatic in outcome, was not the norm. Instead, the late Wisconsinan of theSouthwest USA is better characterized by the long total duration of inter-pluvialclimatic conditions, probably similar to that of the present.

The high-resolution preservation of the Estancia Valley’s palaeolimnologicalrecord and the inferred palaeoclimatological interpretations throughout the lateWisconsinan (late Weichselian) brings into question the inference that NorthAmerican and European glacial climate was marked by fairly uniform conditionsuntil the commencement of glacial retreat ca. 14,000 yr BP.

The rapidity, variability, magnitude, and duration of climate change evidenced inthe Estancia Valley should present a challenge to palaeoclimate modellers in theirefforts to improve our understanding of climate change.

Acknowledgments

Funding for many of the ages used in this paper was provided by the Office ofResearch, University of Nevada, Las Vegas and the NSF – Arizona AMS Facility at theUniversity of Arizona. Their contributions and those of the anonymous reviewers aregreatly appreciated.

References

Bachhuber, F. W. (1971). Palaeolimnology of Lake Estancia and the Quaternary history of the Estancia Valley,

central New Mexico. Unpublished Ph.D. thesis, University of New Mexico, Albuquerque, NM, USA, 238.

Bachhuber, F. W. (1982). ‘Quaternary history of the Estancia Valley, central New Mexico’, in Grambling, J. A.,

and Wells, S. G. (eds.), Albuquerque Country II. New Mexico Geological Society, 343-346.


Bachhuber, F. W. (1989). ‘The occurrence and palaeolimnologic significance of cutthroat trout (Oncorhynchus

clarki) in pluvial lakes of the Estancia Valley, central New Mexico.’ Geological Society of America Bulletin,

101, 1543-1551.

Bachhuber, F. W. (1992). ‘A pre-late Wisconsin palaeolimnologic record from the Estancia Valley, central New

Mexico’, in Clark, P. U., and Lea, P. D. (eds.), The last interglaciation/glaciation transition in North America:

Geological Society of America Special Paper 270, 289-307.

Bachhuber, F. W., and McClellan, W. A. (1977). ‘Palaeoecology of marine Foraminifera in the pluvial Estancia

Valley, central New Mexico.’ Quaternary Research 7, 254-267.

Delorme, L. D. (1969). ‘Ostracodes as Quaternary palaeoecological indicators.’ Canadian Journal of Earth

Sciences, 6, 1471-1476.

Delorme, L. D. (1982). ‘Lake Erie oxygen; the prehistoric record.’ Canadian Journal of Fisheries and Aquatic

Sciences, 39, 1021-1029.

Delorme, L. D. (1990). ‘Freshwater Ostracodes’, in Warner, B.G. (ed.), Methods in Quaternary Ecology:

Geological Association of Canada, Geoscience Canada, reprint series 5, 93-100.

Delorme, L. D., Zoltai, S. C., and Kalas L. L. (1977). ‘Freshwater shelled invertebrate indicators of palaeoclimate

in northwestern Canada during late glacial times.’ Canadian Journal of Earth Sciences, 14, 2029-2046.

Forester, R. M. (1983). ‘Relationship of two lacustrine ostracode species to solute composition and salinity;

Implications for palaeohydro-chemistry.’ Geology, 11, 435-438.

Forester, R. M. (1986). ‘Determination of the dissolved anion composition of ancient lakes from fossil ostracodes.’

Geology 14, 796-798.

Forester, R. M., Delorme, L. D., and Bradbury, J. P. (1987). ‘Mid-Holocene Climate in northern Minnesota.’

Quaternary Research 28, 263-273.

Hawley, J. W. (1984). ‘The Ogallala Formation in eastern New Mexico’, in Proceedings, Ogallala Aquifer

Symposium II, Texas Tech University Water Resources Research Center, 157-176.

Kelley, V. C. (1972). ‘Geology of the Fort Sumner sheet, New Mexico.’ New Mexico Bureau of Mines and

Mineral Resources Bulletin 98, 51.

Mickelson, D. M., Clayton, L., Fullerton, D. S., and Borns, Jr., H. W. (1983).‘The Late Wisconsin glacial record of

the Laurentide Ice Sheet in the United States’, in Porter, S. C. (ed.), Late-Quaternary Environments of the

United States, Volume 1, The Late Pleistocene. University of Minnesota Press, 3-37.

Mifflin, M. D., and Wheat, M. M. (1979). ‘Pluvial lakes and estimated pluvial climates of Nevada.’ Nevada Bureau

of Mines and Geology Bulletin 94, 57.

Peteet, D. (1995). ‘Global Younger Dryas?’ Quaternary International 28, 93-104.

Rutter, N., Bachhuber, F. W., and Lyons, G. (1992). ‘The use of seeds in aminostratigraphy of a Wisconsin

palaeolimnological record from central New Mexico, U.S.A.’ Sveriges Geologiska Undersokning 81, 307-312.

Titus, F. B., (1969). Late Tertiary and Quaternary hydrogeology of the Estancia basin, central New Mexico.

Unpublished Ph.D. thesis, University of New Mexico, Albuquerque, NM, USA, 179.

AEOLIAN GEOMORPHIC RESPONSE TOCLIMATE CHANGE: AN EXAMPLE FROMTHE ESTANCIA VALLEY, CENTRAL NEWMEXICO, USA

N. R. CATTOMemorial University of Newfoundland,St. John's, Canada, A1B 3X9

F. W. BACHHUBERUniversity of Nevada, Las Vegas,Las Vegas, NV, USA, 89154-4010

Abstract

Three generations of Holocene aeolian activity are preserved along the axis of theEstancia Valley, central New Mexico, USA. The transition from large dome dunes toparabolic dunes to loess reflects differences in sedimentation and environments,controlled by climate change. The aeolian landscape developed on the final lacustrineplain of a complex series of freshwater pluvial lakes that occupied the valley from ca.24,000 BP to ca. 10,500 BP, during the maximum development and subsequentdeglacial phases of the Laurentide glacier. Climate change associated with thedisappearance of the final pluvial lake initiated dome dune formation. The dome dunesare characterised by low-angle tabular cross-laminations and tabular ungraded fine-medium beds. The internal structures indicate that limited sediment availabilitycontrolled dome dune formation. Development was initiated and sustained byprevailing westerly winds, consistent in direction, but marked by gustiness.Accumulation was episodic, punctuated by periods of stability with weak mollisoldevelopment. Charcoal from one mollisol dates to ca. 8,500 BP. After this time, thedome dunes attained their greatest relief and were capped by a thick, cementedgypsiferous soil. A series of at least five stacked, weak mollisols lapped on to the flanksof the dome dunes, ca. 4,660 BP.

Overlying the stacked soils and dome dunes is a second generation of parabolic duneswith associated deflation basins. These dunes developed under westerly winds varyingin intensity. Spatial limitation of dune migration contributed to sediment retention,forcing dunes to increase in height and allowing stacking of successive generations oflandforms. Parabolic dunes originating along the eastern margin of Laguna del Perroclimbed from the playa floor and overrode the dome dunes. Sediment movement by

171


172 CATTO AND BACHHUBER

traction and saltation was hindered, slowing transportation rates, enhancing retention,and resulting in rapid accumulation. Sediment supply was limited by the presence ofephemeral standing water in the playa. Deflation of the marginal areas was furtherlimited by periodic rainfall events, allowing diagenetic overgrowths that stabilised theoriginal gypsiferous lacustrine plain. Depositional episodes were transitory in the dunefield, not associated with marked climate change, and were separated by periods ofstability.

Modern aeolian activity involves the deposition of loess. Loess is also reworked bysheetwash, as well as in playa marginal areas of the larger parabolic dunes. Undermodern climate conditions, aeolian sedimentation in the Estancia Valley is marked byvery limited sand supply, almost exclusively locally derived, and high rates of retention.The discontinuous vegetation cover provides localised zones for trapping of finesuspended gypsite, allowing loess accumulation. Loess accumulates in sites preferentialfor vegetation development, such as inactive trough blowouts and hollows betweenadjacent dunes.

1. Introduction

Aeolian deposits and landforms illustrate the impact of previously existing climates onthe environment. The growing interest in the assessment of climate change bygeomorphologists, coupled with increased understanding of the processes and dynamicsof aeolian sedimentology, highlight the necessity for detailed studies of aeoliansuccessions in regions which no longer support active dunes.

Aeolian dunes are common landforms in many basins throughout the southwesternUnited States. Although active dunes categorise many areas, such as the Mojave Desertof California and White Sands, New Mexico, many examples of inactive dunes are alsopresent. The inactive dunes support sparse assemblages of vegetation, and locally havebeen subject to grazing by ungulates and domestic herd animals. Their relative stabilityand inactivity, therefore, must predominantly be attributed to climate change. Thesedunes serve as indicators of previous climate, as well as potential precursors oflandforms that could develop if climate change proceeds in the 21st century. Study ofthe internal structures and morphology of these inactive dunes thus provides proxy dataconcerning past climate fluctuations.

The Estancia Valley of central New Mexico is located 80 km southeast of Albuquerque(Figure 1; Plates 1 and 2). The basin floor, with a minimum elevation of 1,850 m a.s.l.,is surrounded by topographic highs reaching over 3,000 m a.s.l.. At present, theEstancia Valley is thus completely isolated from surface drainage to the adjacent RioGrande and Pecos River drainage basins. The western margin of the Estancia Valley isdefined by Laguna del Perro, a 19 km long deflation basin oriented north-south, whichsupports shallow ephemeral lakes at sporadic intervals.

AEOLIAN GEOMORPHIC RESPONSE TO CLIMATE CHANGE 173

During several phases of the Quaternary, playa and pluvial lakes developed in theEstancia Valley, depositing silt and clay-textured gypsite deposits in the centre of thebasin, and shoreline deposits of gypsarenite and gravel along the basin margins. Themost recent of the playa (lake) events involved the formation of a complex series ofrelatively short-lived but spatially dominant freshwater lakes, during the LateWisconsinan from ca. 24,000 BP to 10,500 BP (see Bachhuber and Catto, this volume).Development of these freshwater (playa) lakes, and the fluctuations in water level andlacustrine chemistry recorded in their sediments and fossil assemblages, responded toclimate changes linked to the Laurentide glacier of western Canada and the northernUnited States. When glacial recession occurred in Montana and Alberta, the changingclimate resulted in the rapid desiccation of the Estancia Valley, exposing the playagypsite and gypsarenite to aeolian reworking and transportation.

Throughout the lattermost Late Wisconsinan and the Holocene, changing climateresulted in the development of a succession of aeolian landforms, dominated by "smalldunes" and "great dunes" (Titus, 1969). That the "small dunes" and "great dunes" are ofdifferent ages was recognized by Titus, and later by Bachhuber (1971), but littlesubsequent aeolian research was undertaken. The small dunes, however, presentedresearchers with a serious problem -- why were the small dunes of such a rounded andlenticular configuration? The explanation advanced by Titus (1969) and later acceptedby Bachhuber (1971) is that the dunes were modified by wave activity of a shallowHolocene lake following dune deposition. Bachhuber (1971) called this lake phase


Lake Meinzer, but indicated that there was no palaeontologic data to support itsexistence. This paper recognizes that the small dunes are domal dunes and their form isnot the result of wave modification.


The dome dunes are developed on the floor of the Willard Soil (Bachhuber, 1971), agypsite unit that caps the Late Wisconsinan Lake Willard plain and sediments. Theinitial dome dune sediments, in turn, are overlain by weakly developed mollisols.Subsequently, a second generation of aeolian activity, represented by parabolic dunes,shaped the Estancia Valley landscape. After a period of stability, the surfaces of theparabolic dunes were reworked by aeolian action in the late Holocene. The aeolianassemblages preserved in the Estancia Valley thus provide a proxy record of climatefluctuations throughout the last 10,500 years.

2. Climate

Today the central portion of the Estancia Valley is arid (10( C mean annual temperature,120 mm mean annual precipitation). The slopes flanking the valley are semiarid. Thewarmest month is July (21° C) and the coldest month is January (0° C). Temperatureand precipitation conditions in the valley produce an annual water-budget deficit inexcess of 1500 mm (Harbour, 1958). As a result, perennial water bodies do not exist atthe lower elevations, and the floors of the deflation basins are dry playas throughout theyear.

3. Distribution and geomorphology of dunes

The Estancia dune field encompasses an area of 240 km2, extending 9 km east ofLaguna del Perro (Plates 1 and 2). More than 60 individual dunes and associateddeflation basins can be recognised. The dunes are confined to the central valley floorthat has a general elevation of 1,835 m, and all overlie playa lacustrine sediments.

The initial generation of aeolian activity was marked by the construction of dome dunes,which are present throughout the Estancia area. In the western part of the field, thedome dunes are overlain by younger parabolic dunes. Dome dunes from the initialphase of aeolian activity are exposed on the surface in the eastern part of the dune field,suggesting that sediment supply from Laguna del Perro during subsequent phases wasinsufficient to inundate these landforms.

The dome dunes exposed on the surface in the eastern area are ovate, with long axesaligned east-west to ENE-WSW. Heights range from 1.5 m to 2.5 m, generallydecreasing eastward. Principal axial lengths range from 15-20 m, and intermediate axialwidths from 10-15 m, with typical length:width ratios of 1.2 : 1. Slope angles rangefrom 0° to 10°.

In the area directly east of Laguna del Perro, the dome dunes have been buried bysubsequent parabolic dune development. Cross-sections exposed along playa lakemargins indicate that the dome dunes had heights of 3 m -6 m, with slope angles ranging


from 0°-12°. These dome dunes are ovate, with long axes oriented between 080°-090°.The maximum width exposed in cross-section is 20 m.

The dominant features on the landscape in the western part of the Estancia Valley arethe second generation parabolic dunes. The dunes are laterally coalescent adjacent toLaguna del Perro, forming a continuous transverse dune 17 km long and up to 44 m inheight above the floor of Laguna del Perro, to an elevation of 1,898 m a.s.l.. Parabolicdunes extend 9 km to the east of Laguna del Perro, systematically decreasing in height,areal extent, and spatial density from west to east. Individual parabolic dunes are up to25 m in height, 1.0 km in length parallel to the net transport direction, and 0.5 km inwidth. Stoss and lee slope angles reach maximum values of 8° and 32°, respectively.On the lee slopes, the maximum angle does not occur directly below the crestline, butgenerally is located in the basal half of the slipface. Crests are flat-topped, althoughsubsequent deflation has greatly modified the surfaces. The ratios of arm length :distance between arms generally approximate 1:1. The northern and southern arms areequal in length, and the dunes are symmetrical about their long axes. Heights, lengths,and widths all decrease eastward. Concomitantly, the isolated playa-floored deflationbasins with associated parabolic dunes increase in elevation towards the east. Theelevation of the playa floors slope downward towards the axis of Laguna del Perro, as aresult of the sloping water table that controlled the maximum deflation elevation.

The easternmost parabolic dunes, with heights of 6 m, modal lengths of 0.5 km andwidths of 250 m, are gradational to shield dunes, with minor slip faces and poor armdevelopment. The shield dunes represent transitional forms to dome dunes. Maximumlee slope angles decrease eastward, varying from 29° on parabolic dunes in the centralpart of the Estancia valley to 20° on the shield dunes.

The azimuth orientations of the modal long axes vary from 060°-095° throughout thedune field, with no systematic spatial variation. Local variations in axial orientationappear to be related to interference from adjacent upwind dunes. The orientationsindicate formation by southwesterly to westerly winds.

Both the initial and second generation dunes have undergone modification by morerecent aeolian activity. Trough blowouts, with depths to 1.5 m and widths to 3 m, arepresent on the surfaces of the transverse dune and the larger parabolic dunes in thewestern part of the dune field. The trough blowouts have locally been reactivated as aresult of cattle and human traffic. Azimuth orientations are variable, and many troughsare irregularly curved due to animal traffic, but the modal orientations are ENE-WSWSaucer blowouts, with maximum diameters of 2 m and maximum depths of 0.3 m, arealso present on crestal surfaces of parabolic dunes, and on the highest flat summits ofthe transverse complex paralleling Laguna del Perro. The saucers vary in shape fromoval, with long axes aligned WSW-ENE, to circular.

In the eastern part of the Estancia Valley, a third generation of small dome dunes hasdeveloped perched on the larger underlying dome dunes of the first generation andsecond-generation parabolic and shield dunes. Typically, these dome dunes are oval in


plan, with lengths up to 8 m, widths up to 5 m, and heights less than 1.5 m. Themaximum lee slope angles are 6°-10°. Slip faces are very poorly defined, and arelocally absent where dome dunes impinge on preexisting topographic rises. Theazimuth orientations of the long axes of these ovate forms are aligned between 060° -100°. Small saucer blowouts are present on the surfaces of these dunes.

Interspersed in the saddles between all the dune types are discontinuous, poorly sortedstructureless infills dominated by gypsite with minor amounts of gypsarenite.Gypsarenite and gypsite slope failure deposits are also present along the margins of thelarger, older dunes, and in areas where the bases have been undercut by playa-inducederosion. This material represents remobilised loess (brickearth). Unstratifiedgypsarenite/gypsite discontinuously infills trough blowouts on the larger parabolicdunes, and on the transverse dune complex. Gypsite and fine gypsarenite loess caps,tapering westward and up to 70 cm thick, occur on the crests of more parabolic dunes inthe western part of the complex.

4. Dune stratigraphy

The three generations of dunes present at Estancia differ in their internal stratigraphy,reflecting differences in the processes of sedimentation and in the depositionalenvironments. A representative succession preserved adjacent to playa E 28, an isolatedplaya-floored deflation basin directly to the east of Laguna del Perro (see Bachhuberand Catto, this volume) allows direct comparison of the stratigraphy of successive dunegenerations, with younger parabolic dune deposits superimposed on older dome dunesediments. The dome dune sediments adjacent to E 28 in turn overlie lacustrinedeposits of Late Wisconsinan Lake Willard B, which desiccated ca. 10,500 BP (seeBachhuber and Catto, this volume).

4.1 DOME DUNE SEQUENCE

The dome dune sediments are illustrated in Plates 3 and 4. The basal 80 cm of thesuccession consists of intercalated pedogenically altered gypsarenite and gypsite, withrecrystallised selenite, interbedded with rippled sediments and discontinuous horizontallaminae. Recrystallised horizontal laminae of selenite, with maximum thicknesses of 2mm, are present along the bases of some ripples. Wedge-form pendants of selenite withmaximum widths of 3 mm and extending to 8 mm depth are also developed beneathsome contacts marked by textural changes or displaying ripple forms.

The ripples are laterally discontinuous, with individual ripple forms and pairs of ripplesseparated by horizontally laminated or structureless gypsite, suggesting that ripplemigration exceeded sediment supply. The ripples have amplitudes of 0.8 to 1.1 cm,wavelengths of 20 to 32 cm, and ripple indices (c.f. Tanner 1967) of 25 to 40, indicatingaeolian origin. Lee faces are commonly preserved in entirety, and are asymptotic toboth the lower and upper surfaces, with a maximum angle of 11°-16°. The direction ofripple migration, varying from azimuth 080° to 107°, indicates formation by westerlywinds.


Sediment within the ripples consists of both gypsarenite and sand-sized aggregatedspheres of gypsite and other clay minerals. Aggregated clasts of clay-sized materialresult from aeolian deflation of adjacent exposed desiccated lacustrine or salinasediments, locally forming distinctive clay-dominated lunette dunes (Hills 1940, Priceand Kornicker 1961) or mixed textural assemblages of less defined form (e.g. Teller1972). At Estancia, the aggregated gypsite clasts do not form morphologically distinctunits, and are interbedded within the gypsarenite. Deposition of the gypsite aggregatesproceeded penecontemporaneously with formation of the gypsarenite ripples, rather thanas texturally distinctive lunette dunes.

The surfaces of the horizontally laminated and structureless strata all are marked by rainpits, 0.5 to 1.5 mm in depth and 5-10 mm in diameter, characteristically with weaklyconvex bases. The surface contacts of structureless sediments display greaterproportions of pits than do the rippled and horizontally laminated strata, withoverlapping pits common. In form and dimensions, the rain pits resemble thosedeveloped on modern gypsum sediments at White Sands NM. Isolated adhesion wartstructures (c.f. Kocurek and Fielder 1982), with axial lengths of 1 cm and heights of 2mm, are present on the upper contacts of isolated horizontal laminae.

The overlying sediments, which form the majority of the dome dune, are largelycomposed of weakly developed horizontal to gently dipping gypsarenite laminae, each5-8 mm thick. The laminae are horizontal in the central part of the dome, and increasein dip to 7° along the western flank and 3° along the eastern flank. Contact surfacesalong the laminae are marked by adhesion wart structures and fossilised rain pits.Selenite crystals, to 1 cm length, are present along some contacts and invariably arealigned along the dip of the laminae, with axes parallel to the inferred wind direction.Isolated ripples within the laminated sequences have heights of 7 -11 mm, wavelengthsof 29-36 cm, and ripple indices of 31 - 46. Ripples on the eastern flank of the domedune indicate transport directions varying between 107° and 166°, with lee side slopesof 5°-7°.

Ripples preserved on the western side of the dune, and in the crestal area, show variabletransport directions ranging from 065° to 233°, with lee side slopes varying from 2° -7°. Rippled horizons are commonly developed overlying gently concave-downwardserosional contacts, and are encased in structureless gypsarenite. Depressions aretypically excavated 3-5 cm below the surface of the uppermost truncated lamination,and have axial dimensions of 90-150 cm. These depressions represent small saucerblowouts excavated on the stoss and crestal areas of the dune surface, and infilled withrippled and structureless gypsarenite. Individual ripple orientations indicate thattransport was locally towards the centre of the depressions along their upwind andlateral margins, with sediment filling the depressions from three sides.

An episode of stability is indicated by the presence of a disturbed, poorly sortedgypsarenite and gypsite stratum 25-30 cm thick, 2.8 m above the base of the dome dunesuccession. The unit contains dikaka structures, rippled strata which have been partiallytruncated, disturbed horizontal laminations, and adhesion warts. This bed is directly


overlain along an erosional contact by rippled and horizontally laminated gypsarenite,similar to the underlying aeolian sediments. The maximum thickness of sediment in thedome dune is 5.3 m.

The uppermost surface of the dome dune is capped by a laterally continuous stratum ofselenite granules, conforming to the underlying domal morphology. Fragments oforganic detritus up to 3 mm in length are also present. Isolated selenite crystals up to 1.5cm in length, aligned 180°-220°, are present along the basal contact, and pendantselenite crystals are also present. This stratum developed during a period of stability,and represents the conclusion of dome dune development. The selenite crystals weremoved by traction, rolled normal to the prevailing wind direction.

4.2 PARABOLIC DUNE SUCCESSION

The dome dunes are overlain by parabolic dune sediments in the western part of theEstancia Valley. Along the eastern margin of Laguna del Perro, laterally coalescentparabolic dunes have formed the transverse dune complex. The succession preservedadjacent to E 28 is typical of lee-side preservation in the interior of parabolic dunesthroughout the Estancia area (Plate 5).


The basal 3.1 m of the parabolic dune sequence at E 28 is dominated by moderatelydefined tabular sets of low-angle cross-laminated sand-sized clay pellet aggregates,some of which contain lacustrine ostracods, and gypsarenite. The laminae are arrangedin sets 2-15 cm thick, marked by moderately defined contacts with rare adhesion wartstructures dipping at 5°-15° towards 075°-095°. Individual laminae are 1-5 mm thick,and dip at angles between 8°- 22° towards 070°-115°. These beds are interpreted torepresent lee-side deposition, directly downwind of the slipface (McKee 1966, 1979;Halsey et al. 1990).

Wedge-form sets of cross-laminated sands, composed of clay pellet aggregates andgypsarenite, are also present, and are more prevalent in the lowermost 1 m of the dunesequence. Some sets of tabular laminae grade laterally into wedge-form sets away fromthe former dune crest line, in the inferred direction of transport and dune migration. Thewedge-form sets have thicknesses of 1-6 cm, thinning towards the west (upwind).Within the sets, individual laminae have thicknesses of 2-5 mm, dip at 5°-20°, andindicate transport directions varying between 060° and 190°. These wedge-formlaminae were produced by preferential downslope deposition by ripples migratingacross the slipface, driven by winds moving parallel or oblique to the slope (Frybergerand Schenk 1981, Halsey et al. 1990).

Also present in the uppermost 1.5 m of the parabolic dune succession are isolatedtabular gypsarenite/sand-sized clay pellet and fine selenite beds, coarser than theunderlying deposits (Plate 6). These beds are internally structureless, and range in


thickness from 1-3 cm. They are bounded by sharp contacts, frequently rippled, withoccasional adhesion wart and rain pit structures. Dips range from horizontal to 2°easterly. Ripples range in height from 5-11 mm, and in wavelength from 28-42 cm,with ripple indices between 28 and 60. In all instances, the trough width of the rippleswas in excess of the crest width. Transport directions vary between 060° and 140°.

The tabular beds were deposited by a combination of saltation and grainfall, wheresand-sized clay pellets, gypsarenite, and fine selenite adhered to temporarily saturatedsurfaces following precipitation events. Ripples with trough width in excess of crestalwidth form over resistant surfaces, under conditions of low to moderate sediment flux(Sharp 1963). Moistened gypsarenite/selenite surfaces would provide suitableconditions for ripple development.

The uppermost 20 cm of the parabolic dune sequence contains numerous dikakastructures, commonly flanked by plano-convex lenses of moderately sorted sand-sizedparticles. The prevalence of these features in the uppermost part of the dune indicatesthat vascular plants became established on the dune surface, leading to interference withsaltation and eventual stabilisation (Goldsmith 1973, Halsey et al. 1990). Theuppermost 10 cm of this stratum has been pedogenically modified.

The sequence of deposition in the parabolic dune at E 28 indicates that the dune becameprogressively more stabilised over time. Migration was spatially limited, and the duneresponded to continuing sedimentation by increasing in height. Similar sequences arepreserved in exposures in parabolic dunes throughout the Estancia dune field.

4.3 YOUNGEST AEOLIAN SEDIMENTATION

The most recent aeolian event is represented by the development of trough blowoutstructures in the parabolic dunes, indicating that these features are not in equilibriumwith the modern climate and sediment flux regime. In addition, small dome dunes andloess deposits have formed overlying the older dunes.

The cores of the youngest dome dunes are dominated by structureless beds ofgypsarenite and sand-sized clay pellet aggregates, with thicknesses of 5-20 cm. Gentlydipping (2°-6°) tabular sets of gypsarenite laminae, with set thicknesses of 2-5 cm andlaminae thicknesses of 1-3 mm, are preserved on the eastern flanks of some domedunes. Pedogenetic disturbance and dikaka structures are common throughout thesedimentary successions. Rippled beds are not preserved. Desiccation cracks, formingpolygonal networks and extending to 6-8 cm depth are common on the exposed surfacesof dome dunes

Caps of loess, with maximum thicknesses of 70 cm, are present over most of theparabolic dunes. The maximum thicknesses are preserved in saddles between the dunesand as cliff-top successions. Differential thicknesses in successions developed at dunecrests, analogous to cliff-top successions in other environments (c.f. Catto 1983)indicate deposition by westerly winds. The fine gypsarenite and gypsite loess is


structureless, except where disturbed by dikaka and pedogenesis. Episodicaccumulation of loess punctuated by stabilisation events is indicated by pendant selenitecrystals, diagenetic gypsite crusts, and discontinuous thin horizontal layers of fineorganic detritus.

Remobilised loess (brickearth) is also present in lowlying areas between the parabolicdunes, as infills in trough blowouts, and along the margins of the playas. The brickearthis structureless and poorly sorted, containing selenite granules, gypsarenite and gypsiteclasts, and fine organic detritus. Discontinuous lag deposits and monolayers of roundedselenite clasts are common, and fine interbedded strata of primary loess are also present.Local thicknesses exceed 2 m.

5. Genesis and stabilization of dome dunes

The geomorphology and sedimentary structures of the dome dunes of the EstanciaValley indicate formation by westerly winds. Gypsite and gypsarenite that hadaccumulated in the nearshore areas of Lake Willard B, particularly in the basin occupiedby the modern Laguna del Perro, were subject to deflation when the water level droppedca. 10,500 BP (see Bachhuber and Catto, this volume). Repeated deflation events,interspersed with periods of high playa levels accompanied by high water tables alongthe littorals, have resulted in the construction of dome dunes and lunettes along themargins of gypsiferous playas and salinas in other arid and semi-arid regions (e.g. Chenand Barton 1991, Chen et al. 1993, Carignano 1996). The source of sediments for domedune construction was primarily confined to the Lake Willard sediments exposed in thebasinal and littoral areas of Laguna del Perro, although minor proportions of quartz andother non-evaporitic minerals indicate that distally transported clasts are present.

Dome dunes are developed in three geomorphic settings: in arid interior areas understrong winds (McKee 1966, 1979); along marine and lacustrine coasts (Bigarella 1972);and in boreal interior regions (Halsey and Catto 1994). In all these settings, genesis ofdome dunes depends upon the inhibition of slip face development (Halsey and Catto1994). In areas where wind strength has increased while sediment supply has remainedconstant or increased, such as White Sands NM, dome dunes are produced by thebevelling of preexisting large parabolic or transverse dunes (McKee 1966, 1979). Theinternal structures in the cores of these dome dunes thus represent the originalsedimentary processes of parabolic and transverse dune construction. High- tomoderate-angle cross-strata aligned parallel to the modal transport direction gradeupwards into horizontal laminations and trough blowout structures infilled withungraded sediment. Development requires a high rate of sedimentation. In dune fieldswith several dune types, these dome dunes thus are commonly found in the mostproximal locations. At White Sands, dome dunes develop directly adjacent to thegypsarenite source of Lake Lucero (McKee 1966).

In contrast, dome dunes developed along marine and lacustrine coastal margins arecharacterised by horizontally laminated beds and low-angle cross-strata, concave-downward bounding surfaces, and thin accumulations of structureless fine sand and


loess in saucer-shaped blowout depressions. High-angle cross-strata are not present,and trough blowout features are rare. These dunes developed under winds of varyingvelocity and orientation, with sediment availability limited by moisture in the nearshoreenvironment (e.g. Bigarella 1972, McKee and Bigarella 1972, Catto 1994). The dunesmay be distal or proximal to the source of sediment.

A third style of dome dune forms in boreal interior regions, where restricted sandsupply, seasonally influenced by snowfall and hail accumulation and surface freezing, incombination with gusting but persistently oriented winds precludes slipfacedevelopment (David 1977, Halsey and Catto 1994). The cores of interior boreal domedunes are dominated by low-angle tabular and wedge-shaped cross-strata, tabularungraded sand beds, and corrugated fine laminae, with high-angle cross-strata confinedto the basal successions of relatively few examples. Although trough blowoutstructures are rarely observed, saucer blowouts are generally present. A lowsedimentation rate is the primary factor promoting dome dune development in borealinterior regions. In dune fields with several morphological types, the dome dunes arethe most distal members of the assemblage (Halsey and Catto 1994).

The dome dunes along the margin of Laguna del Perro are characterised by low-angletabular cross-laminations and tabular ungraded fine-medium gypsarenite/clay pelletbeds. Thin loess and ungraded fine sandy deposits are confined to the flanks of thedunes, and overlie erosional saucer blowout surfaces. The internal structures of thesedunes thus resemble those formed in boreal interior and modern marine coastalenvironments, under conditions of limited or episodic sand supply. The sedimentarystructures indicate that limited availability of sediment was the most important factorresulting in dome dune formation in the Laguna del Perro area, in contrast to the highsedimentation rates associated with the dunes developed in the drier White Sands area tothe south (McKee 1979). Development was initiated and sustained by prevailingwesterly winds, consistent in direction but marked by gustiness. The dunes developeddirectly adjacent to the playa source.

Accumulation was episodic, punctuated by periods of stability indicated by corrugationof bed contacts and weakly developed pedogenetic modification of the gypsarenite. Thepresence of the charcoal horizon 14C dated at 8,585 ( 75 BP (AA-6325), mid-way in thedune stratigraphic sequence, also suggests that episodes of stability suitable for junipergrowth, and consequent termination of aeolian sand influx, occurred during dome duneconstruction.

Stabilisation of gypsarenite dunes involves development of menisci of mobilised andreprecipitated gypsum cement, reinforced by pendants of cement (Schenk and Fryberger1988). In modern, essentially unvegetated environments, the rate of cementation (andhence stabilisation) depends largely on capillary movement of meteoric water. Periodicprecipitation results in alternating dissolution and recementation, producing multiplediagenetic overgrowths on the near-surface gypsarenite and binding the upper surface(Warren 1982, Schenk and Fryberger 1988).


In modern gypsarenite dunes, stabilisation proceeds most rapidly in the capillary zoneand is slowest in the vadose zone. Evidence for stabilisation and cementation featuresthroughout the cores of the dome dunes indicates that the sediments were periodicallyexposed to capillary zone conditions. On the developing gently sloping flanks andcrests of the domes, conditions for diagenesis and stabilisation more resembled those ofmodern interdune and salina margin environments (e.g. Schenk and Fryberger 1988,Carignano 1996) than those found on the surfaces of more elevated parabolic andtransverse dunes.

Where vascular vegetation is not present, small amounts of fine gypsarenite and coarsegypsite may be trapped by mosses and non-vascular plants (Danin and Ganor 1991).Direct frictional adhesion also occurs on dry surfaces, but rates of retention arecharacteristically less than 20 % of sediment influx, involving accumulation of lessthan 50 g/m2/a (e.g. Goosens 1995). The 1.5 m thickness of gypsarenite accumulatedbetween ca. 10,500 BP and ca. 8,500 BP demonstrates that retention rates weresignificantly higher, in excess of 1000 g/m2/a. Enhanced retention of gypsarenite wouldbe facilitated by accumulation on moist or periodically wetted surfaces, as indicated bythe presence of corrugated bounding surfaces in the cores of the dome dunes.

In the absence of thermoluminescence or optical stimulation luminescence analysis,techniques which cannot at present be applied to gypsiferous deposits (e.g. Berger1995), chronological assessment of dune successions involves determination of theperiods of stabilisation, rather than the times of active dune construction. In theEstancia Valley, dome dune formation was initiated at some time following ca. 10,500BP, when Lake Willard B disappeared. The charcoal layer at the stratigraphic mid-pointof the dune indicates that dune accumulation temporarily halted ca. 8500 BP. Furtherepisodic dune development began at some time after 8,500 BP, and continued untilsome time before pedogenic activity formed the multiple soils capping and off-lappingthe dome dunes. Charcoal recovered from the second weakly developed mollisolhorizon was 14C dated at 4,660 (170 BP (GX-13321).

The initial Holocene episodes of gypsarenite deflation and dune construction thus areapproximately correlative to the multiple, relatively poorly dated early to mid-Holocenedune reactivation events noted on the Colorado High Plains by Forman et al. (1992) andMadole (1995), and in the adjacent Nebraska Sand Hills by Loope et al. (1995). InArizona, conditions in the Tusayan dunefield allowed remobilisation of surfacesediment ca. 4,700 BP, while plinth deposits remained undisturbed (Stokes and Breed1993). Enhanced aeolian transport by westerly winds and distal deposition is recordedin Early and Mid-Holocene lacustrine sediments throughout the High Plains andMidwest, particularly between ca. 8,000 and ca. 4,000 BP (Dean 1997). The delayedonset of desiccation and aeolian deflation in the Great Basin following thedisappearance of the Laurentide glacier to the north has been related to polewarddisplacement of summer monsoonal winds, resulting in a period of enhancedprecipitation between ca. 12,500 and ca. 8,000 BP (Spaulding 1991).


6. Genesis of parabolic dunes

The parabolic dunes are composed of gypsarenite and sand-sized clay pellet aggregatesderived from Laguna del Perro and the other playa-floored deflation basins duringperiods when the lacustrine sediment was subaerially exposed. The morphology of thedunes indicates that the modal transport direction was from west to east, although cross-stratification developed on lee slopes around the convex dune margins were depositedby local sediment movements ranging from north to south-southwest. These variationsin sediment transport directions are the products of dune surface morphology, whichinfluences the directions of grain flow, and of localised wind patterns on the lee slopesof individual dunes. The orientations of individual cross-strata thus reflect the localtransport direction, not the modal wind direction, as has been observed in numerousmodern dune successions (e.g. McKee 1966, Howard 1978, Whitney 1978, Sweet1992). In the lee of large parabolic dunes, winds responsible for sediment deposition onthe lower third of the slopes are commonly oriented oblique to the axis of the dune.

Parabolic dune development involves a combination of limited sand supply, efficientsediment retention and surface adhesion, and consistently oriented winds (David 1977,Halsey et al. 1990). At White Sands, parabolic dunes represent the most distal membersof the dune assemblage, and limited sediment supply is the most critical factor in theirdevelopment (McKee 1966, 1979). In contrast, parabolic dunes in boreal and marinecoastal areas frequently occupy the most proximal positions, and sediment retention iscritical in their establishment and maintenance (Halsey et al. 1990, Catto 1994). Winddirection is consistent in most interior boreal environments, but the effects of variablewinds in coastal environments are frequently negated by the limitations in sand supplyfrom exposed littoral sediments and the geomorphically limited areas inland that areavailable for dune expansion.

Parabolic dunes developed proximal to the marginal area east of Laguna del Perro underthe influence of winds varying in intensity, but the internal structures of the dunesindicate consistent westerly wind activity. Sand supply and / or rates of retention werethus critical for dune development, coupled with the spatially limited area to the east ofLaguna del Perro that was available for eastward expansion and migration. Sedimentsupply was limited by the presence of ephemeral standing water in the playa and bymaximum deflation reaching the local water table. Much of the potential gypsarenitesource material not inundated directly was periodically within the capillary zone, furtherlimiting sediment availability. Deflation of the marginal areas of Laguna del Perro wasalso limited by periodic rainfall events, allowing diagenetic overgrowths to stabilise thegypsum surface. Conditions for aeolian deflation were similar to those currentlyprevailing in interdune areas adjacent to Laguna del Perro.

Retention of gypsarenite is enhanced on dune surfaces periodically moistened byprecipitation. The crenulated surfaces promote adhesion of gypsarenite when moist, andact to trap traction sediment during dry intervals. Chemical weathering and diagenesisfurther enhances sediment retention (Tchakerian 1991).


Spatial limitation of dune migration also contributes to sediment retention, forcingdunes to increase in height and allowing stacking of successive generations oflandforms. In the Gran Desierto of Mexico, Lancaster (1992) observed that stacking ofdune generations occurs where sand supply is high (unlike the Estancia Valley), orwhere expansion of the sand sea is topographically restricted. Stacking also is aconsequence of the preexisting dune and playa topography. Parabolic dunes originatingalong the eastern margin of Laguna del Perro would be forced to climb from the playafloor and override the preexisting dome dunes, surmounting slopes of 10° - 15° andheights in excess of 5 m. Sediment movement by traction and saltation up theserelatively steep slopes would be hindered, slowing transportation rates, enhancingretention, and resulting in rapid accumulation. Similar climbing dunes developed onup-wind facing slopes have been noted by Seppälä (1993) and Cros and Serra (1993).

Climbing results in the dominance of traction and saltation processes in thedevelopment of parabolic stoss slopes and crests, and largely precludes grainfall. As thedune grows vertically and sediment is deposited and winnowed during traction,progressive deflation results in coarsening upward sequences and coarsening upwindfrom the crest along the stoss slope (e.g. Vincent 1996). Under these circumstances,deposition is influenced by sporadic wind gusts, and local deceleration events result inaccumulations of sediment (Arens et al. 1995). The accumulation of sediment in thestoss and crestal areas at a greater rate than the spatially and gravitationally restrictedclimbing dune can migrate results in a growth in height and a tendency for lateralcoalescence of adjacent parabolic dunes normal to the modal wind direction. In theEstancia Valley, the highest and most laterally extensive parabolic dunes, locallycoalesced into transverse forms, are located along the eastern margin of Laguna delPerro.

Depositional episodes were transitory in the dune field, and were separated by periodsof stability. Similar episodic dune development, not associated with marked climatechange, has been recorded in several areas (e.g. Lancaster 1992, Bullard et al. 1997). Inaddition, relatively minor climate changes may also result in a substantial change ofstability regime (e.g. Stokes and Breed 1993, Madole 1995, Liu Jian et al. 1997),suggesting that the climatic fluctuations recorded in other palaeoenvironmental recordsthroughout New Mexico would be sufficient to permit alternating dune reactivation andstabilisation.

The parabolic dunes in the Estancia Valley were reactivated shortly after ca. 4660 BP.This reactivation is co-incident with episodes of dune construction and migration notedin the Mojave Desert (Tchakerian 1991), in northern Arizona (Stokes and Breed 1993),and in Colorado (Forman et al. 1992, Madole 1995). Stabilisation may be related to agradual decrease in the evapotranspiration ratio, rather than an absolute increase inprecipitation (c.f. Carignano 1996).

Lateral expansion of the parabolic dunes impedes any surface drainage between playasand salinas, resulting in temporary impoundment and deposition of playa sediments. Inthe Estancia dune field, the tendency for vertical accumulation and restricted lateral


migration resulted in the development of small playas in relatively fixed locations,rather than a series of ephemeral overridden interdune successions. Interdune areas aresites of deposition rather than periodic deflation and erosion resulting from dunemigration. Thus, in contrast to the situation in actively migrating dune fields (e.g.Lancaster and Teller 1988), interdune and plinth deposits are commonly preserved. Themarginal areas of the playa and plinth successions are marked by interbeddedgypsarenite slope failure and surface wash deposits, along with poorly sortedstructureless gypsite brickearth.

The parabolic dunes of the Estancia dune field developed in combination with playa andephemeral channel sedimentation. Impoundment of playas by parabolic dunes mayresult in increased accumulation of gypsite and gypsarenite in basins during periodsfavourable to dune development. Thus, subaqueous precipitation and aeoliansedimentation may be penecontemporaneous, and high playa levels may not necessarilyrepresent periods of enhanced precipitation and dune stabilisation. Similardune/interdune relationships were documented in the driest areas of the Nebraska SandHills by Loope et al. (1995).

7. Most recent reactivation event

Modern aeolian activity involves the deposition of loess, and the reworking of loess bysheetwash and by playa activity along dune flanks to produce brickearth. Small domedunes are ephemerally active throughout the eastern part of the dune field, and erosionaltrough blowouts are developing on the upper stoss and crestal areas of the largerparabolic dunes.

Under modern climate conditions, aeolian sedimentation in the Estancia Valley ismarked by very limited sand supply, almost exclusively locally derived, and high ratesof retention. The irregular cover of shrubs limits surface traction and saltation of sand(Lee 1991), and hence precludes both rippling and slipface development. Thediscontinuous vegetation cover provides localised zones for trapping of fine suspendedgypsite, allowing loess accumulation. Loess thus accumulates in sites preferential forvegetation development, such as inactive trough blowouts and hollows between adjacentdunes. Prolonged accumulation has resulted in smoothing of the topography, withdepressions between adjacent dune hills partially infilled with accumulations of loessand remobilised brickearth (c.f. Goosens 1997).


Much of the sand supply to the modern dome dunes is derived from erosion of troughblowouts on the larger parabolic dunes to the west. In the blowouts, erosion alternateswith collapse of oversteepened slopes and moisture-induced colluviation, and the infillof the blowouts is dominated by brickearth rather than by primary aeoliansedimentation. The resulting brickearth resembles reworked loess deposits associatedwith ephemeral streams (Sneh 1983), and may be confused with reworked depositsresulting from fluvial activity (Jones and Blakey 1997) or climate change (e.g. Liu Jianet al. 1997).

Local reactivation is attributed to a variety of causes. Local overgrazing and cattletracks are responsible for the initiation of specific saucer and trough blowouts,respectively, but ranching is not intensive in the Estancia Valley. The timing ofreactivation in Estancia also does not coincide with the late nineteenth century eventsnoted in overgrazed areas of Colorado and Arizona. The distribution and configurationof the blowouts suggests that most were created by localised deflation from strongwesterly winds, associated with the most exposed, highest, and driest dune surfaces(rather than with those preferentially grazed or lacking vegetation). The small moderndome dunes developed under conditions of extremely limited sediment supply, incontrast to the larger first and second generation features. Under the moister climate


which currently prevails, deflation would be confined to the driest areas at the dunesummits, and sediment flux from the periodically moistened areas surrounding Lagunadel Perro would be severely limited. Under the current climatic regime, aeolian activityis limited to sporadic deflation, eastward migration of small quantities of gypsarenite,and construction of minor dome dunes.

8. Summary

The sequence of latest Wisconsinan and Holocene events related to aeolian activitypreserved in the Estancia Basin involved:(1) Desiccation of the latest Wisconsinan pluvial lake, Lake Willard B, ca. 10,500 BP;(2) development of the Willard Soil, a gypsite capping the youngest pluvial lakesediments;(3) initial development of the dome dunes, prior to ca. 8,500 BP;(4) a period of stability ca. 8,500 BP, indicated by the charcoal horizon;(5) renewed development of dome dunes;(6) intermittent episodes of stability, with formation of stacked weakly developedmollisols off lapping and capping the dome dunes, ca. 4,660 BP;(7) initiation of deflation, forming the parabolic dunes and the playa floored deflationbasin sequence, beginning after ca. 4,660 BP and ceasing prior to the latest Holocene;and(8) ongoing development of the youngest dome dunes and loess sequences, along withpresent stabilization of most aeolian features.

Acknowledgements

Production assistance was provided by Charles Conway, MUNCL, and byPhotographic Services, Memorial University. Technical and editorial assistance wasprovided by Gail Catto.

References

Arens, S. M., Van Kaam-Peters, H. M. E., and Van Boxel, J. H. (1995). 'Air flow over foredunes andimplications for sand transport.' Earth Surface Processes and Landforms 20, 315-322.

Bachhuber, F. W. (1971). Paleolimnology of Lake Estancia and the Quaternary history of the Estancia valley,central New Mexico . Unpublished Ph.D. thesis, University of New Mexico, Albuquerque, NM, USA,238.

Berger, G. W. (1995). 'Progress in Luminescence Dating Methods for Quaternary Sediments', in Rutter,N.W., and Catto, N. R. (eds.), Dating Methods for Quaternary Deposits. Geological Association ofCanada, 81-104.

Bigarella, J. J. (1972). 'Eolian environments - Their characteristics, recognition, and importance', in Rigby, J.K., and Hamblin, W.K. (eds.), Recognition of Ancient Sedimentary Environments. Society of EconomicPalaeontologists and Mineralogists, Special Publication 16, 12-62.

Bullard, J. E., Thomas, D. S. G., Livingstone, I., and Wiggs, G. F. S. (1997). 'Dunefield activity andinteractions with climatic variability in the southwest Kalahari Desert.' Earth Surface Processes andLandforms , 22, 165-174.


Carignano, C. A. (1996). 'Evolucion geomorfologica de las planicies en la Provincia de Cordoba durante elPleistocene superior.' Revista del Instituto de Geologica y Mineria, 11, 7-26.

Catto, N. R. (1983). 'Loess in the Cypress Hills, Alberta, Canada.' Canadian Journal of Earth Sciences, 20,1159-1167.

Catto, N. R. (1994). 'Anthropogenic Pressures and the Dunal Coasts of Newfoundland.' Coastal Zone 1994Conference, Halifax, Nova Scotia, September 1994, Bedford Institute of Oceanography, 2266-2286.

Chen, X. Y., and Barton, C. E. (1991). 'Onset of aridity and dune-building in central Australia:sedimentological and magnetostratigraphic evidence from Lake Amadeus.' Palaeogeography,palaeoclimatology, palaeoecology, 84, 55-73.

Chen, X. Y., Bowler, J. M., and Magee, J. W. (1993). 'Late Cenozoic stratigraphy and hydrologic history ofLake Amadeus, a central Australian playa.' Australian Journal of Earth Sciences, 40, 1-14.

Cros, L., and Serra, J. (1993). 'A complex dune system in Baix Empord... (Catalonia, Spain)', in Pye, K. (ed.)The Dynamics and Environmental Context of Aeolian Sedimentary Systems. Geological Society ofLondon Special Publication 72, 191-199.

Danin, A., and Ganor, E. (1991). 'Trapping of airborne dust by mosses in the Negev Desert, Israel.' EarthSurface Processes and Landforms, 16, 153-162.

David, P.P. (1977). Sand Dune occurrences in Canada: A theme and resource inventory study of eolianlandforms in Canada. National Parks Branch, Department of Indian and Northern Affairs, Governmentof Canada, Contract 74-230, Report 183. Ottawa, Ontario, 183.

Dean, W. E. (1997). 'Rates, timing, and cyclicity of Holocene eolian activity in north-central United States:Evidence from varved lake sediments.' Geology , 25, 331-334.

Forman, S. L., Goetz, A., and Yuhas, R. H. (1992). 'Large-scale stabilized dunes on the High Plains ofColorado: Understanding the landscape response to Holocene climates with the aid of images fromspace.' Geology , 20, 145-148.

Fryberger, S., and Schenk, C. (1981). 'Wind tunnel sedimentation experiments on the origins of aeolianstrata.' Sedimentology, 28, 805-821.

Goldsmith, V. (1973). 'Internal geometry and origin of vegetated coastal dunes.' Journal of SedimentaryPetrology, 43, 1128-1143.

Goosens, D. (1995). 'Field experiments of aeolian dust accumulation on rock fragment substrata.'Sedimentology, 42, 391-402.

Goosens, D. (1997). 'Long-term aeolian loess accumulation modelled in the wind tunnel: the Molenberg case(central loess belt, Belgium).' Zeitscrift für Geomorphologie NF 41, 115-129.

Halsey, L.A., Catto, N. R., and Rutter, N.W. (1990). 'Sedimentology and development of parabolic dunes,Grande Prairie dune field, Alberta.' Canadian Journal of Earth Sciences, 27, 1762-1772.

Halsey, L.A., and Catto, N. R. (1994). 'Geomorphology of Dome Dunes.' Géographie physique etQuaternaire, 48, 97-105.

Harbour, J. (1958). Microstratigraphy and sedimentatinal studies of an early man site near Lucy, NewMexico. Unpublished M.S. thesis, University of New Mexico, Albuquerque, NM, USA, 111.

Hills, E. S. (1940). 'The lunette, a new land form of aeolian origin.' Australian Geographer, 3, #7, 15-21.Howard, A. D. (1978. 'Sand transport model of Barchan dune equilibrium.' Sedimentology, 25, 307-338.Jian, L., Xihao, W., Shaoquan, L., and Mingshu, Z. (1997). 'The Last glacial stratigraphic sequence,

depositional environment and climatic fluctuations from the Aeolian sand dune in Hongguang, Pengze,Jiangxi (China).' Quaternary Science Reviews 16, 535-546.

Jones, L. S., and Blakey, R. C. (1997). 'Eolian-fluvial interaction in the Page Sandstone (Middle Jurassic) insouth-central Utah, USA -- a case study of erg-margin processes.' Sedimentary Geology , 109, 181-198.

Kocurek, G., and Fielder, G. (1982). 'Adhesion structures.' Journal of Sedimentary Petrology, 54, 1229-1241.Lancaster, N. (1992). 'Relations between dune generations in the Gran Desierto of Mexico.' Sedimentology

39, 631-644.Lancaster, N., and Teller, J. T. (1988). 'Interdune deposits of the Namib Sand Sea.' Sedimentary Geology ,

55, 91-107.Lee, J. A. (1991). 'The Role of Desert Shrub Size and Spacing on Wind Profile Parameters.' Physical

Geography, 12, 72-89.Loope, D. B., Swinehart, J. B., and Mason, J.P., (1995). 'Dune-dammed paleovalleys of the Nebraska Sand

Hills: Intrinsic versus climatic controls on the accumulation of lake and marsh sediments.' GeologicalSociety of America Bulletin , 107, 396-406.

Madole, R. F. (1995). 'Spatial and temporal patterns of Late Quaternary eolian deposition, eastern Colorado,USA.' Quaternary Science Reviews , 14, 155-177.


McKee, E. D. (1966). 'Structures of dunes at White Sands National Monument, New Mexico (and acomparison with structures of dunes from other selected areas).' Sedimentology, 7, 1-69.

McKee, E. D. (1979). 'Sedimentary Structures in Dunes, in E. D. McKee, ed., A Study of Global Sand Seas'.United States Geological Survey Professional Paper 1052, Washington, 83-136.

McKee, E. D., and Bigarella, J. J. (1972). 'Deformational Structures in Brazilian Coastal Dunes.' Journal ofSedimentary Petrology, 42, 670-681.

Price, W. A., and Kornicker, S. (1961). 'Marine and lagoonal deposits in clay dunes. Gulf Coast, Texas.'Journal of Sedimentary Petrology, 31, 245-255.

Schenk, C. J., and Fryberger, S. G. (1988). 'Early diagenesis of Eolian dune and Interdune sands at WhiteSands New Mexico.' Sedimentary Geology, 55, 109-120.

Seppälä, M. (1993). 'Climbing and falling sand dunes in Finnish Lapland', in Pye, K. (ed.), The Dynamicsand Environmental Context of Aeolian Sedimentary Systems. Geological Society of London SpecialPublication 72, 269-274.

Sharp, R. P. (1963). 'Wind ripples.' Journal of Geology, 71, 617-636.Spaulding, W. G. (1991). 'Pluvial climatic episodes in North America and North Africa: types and correlation

with global climate.' Palaeogeography, Palaeoclimatology, Palaeoecology, 84, 217-227.Stokes, S., nd Breed, C. S. (1993). 'A chronostratigraphic re-evaluation of the Tusayan Dunes, Moenkopi

Plateau and southern Ward Terrace, Northeastern Arizona', in Pye, K. (ed.), The Dynamics andEnvironmental Context of Aeolian Sedimentary Systems. Geological Society of London SpecialPublication 72, 75-90.

Sneh, A. (1983). 'Redeposited Loess from the Quaternary Besor Basin, Israel.' Israeli Journal of EarthSciences, 32, 63-69.

Sweet, M. L. (1992). 'Lee-face airflow, surface processes, and stratification types: their significance forrefining the use of eolian cross-strata as paleocurrent indicators.' Geological Society of AmericaBulletin, 104, 1528-1538.

Tanner, W. F. (1967). 'Ripple mark indices and their uses.' Sedimentology, 9, 89-94.Tchakerian, V.P. (1991). 'Late Quaternary Aeolian Geomorphology of the Dale Lake Sand Sheet, southern

Mojave Desert, California.' Physical Geography, 12, 347-369.Teller, J. T. (1972). 'Aeolian deposits of clay sand ' Journal of Sedimentary Petrology, 42, 684-686.Titus, F. B. (1969). Late Tertiary and Quaternary hydrogeology of the Estancia basin, central New Mexico .

Unpublished Ph.D. thesis, University of New Mexico, Albuquerque, NM, USA, 179.Vincent, P. (1996). 'Variation in particle size distribution on the beach and windward side of a large coastal

dune, southwest France.' Sedimentary Geology , 103, 273-280.Warren, J. K. (1982). 'The hydrological setting, occurrence and significance of gypsum in late Quaternary salt

lakes in South Australia.' Sedimentology, 29, 609-637.Whitney, M. (1978). The Role of Vorticity in Developing Lineation by Wind Erosion. Geological Society of

America Bulletin, 89, 1-18.

EVAPORITE MINERALS AND ORGANIC HORIZONSIN SEDIMENTARY SEQUENCES IN THE LIBYAN FEZZAN:IMPLICATIONS FOR PALAEOENVIRONMENTALRECONSTRUCTION

KEVIN WHITELandscape and Landform Research Group,Department of Geography, The Universityof Reading, Whiteknights, Reading, RG6 6AB,U.K.

SUE McLARENDepartment of Geography, University of Leicester,University Road, Leicester, LE1 7RH, U.K.

STUART BLACKPostgraduate Research Institute for Sedimentology,The University of Reading, Whiteknights, Reading,RG6 6AB, U.K.

ADRIAN PARKERGeography Department, Oxford Brookes University,Gipsy Lane Campus, Headington, Oxford, OX3 0BP, U.K.

Abstract

Compared to other parts of the Sahara, there is a paucity of reliable palaeoenvironmentalinformation for the Libyan Fezzan. Ongoing archaeological work in the area provides anopportunity to link palaeoenvironmental and archaeological data to improve ourunderstanding of human adaptations to dryland environmental change. A combination ofstable isotope analysis, mineral magnetic analysis, an investigation of the particle sizedistributions and uranium-thorium dating, together with supporting image processing ofremotely sensed data, indicates a trend of falling groundwater level since the latePleistocene, manifested by former lakes, swamps and springlines. Archaeologicalevidence shows how human activity has adapted to these environmental changes.

193


194 WHITE, McLAREN, BLACK AND PARKER

KEY WORDS: Sahara, environmental change, remote sensing, stable isotope analysis

1. Introduction

There has been considerable research into Late Quaternary palaeoenvironments of theSahara, especially connected with changing lake and groundwater levels (e.g. Brookes,1993; Fontes and Gasse, 1991, Petit-Maire et al., 1994). However, there are stillsignificant gaps in the spatial coverage of these investigations, notably in the LibyanFezzan, despite the wealth of geomorphological and archaeological evidence ofenvironmental change thereabouts (Gaven et al., 1981; Cremaschi and Trombino, 1998).

The study area is centred on the town of Germa (ancient Garama) in the Wadi el Agial(Figure 1). The region lies in the hyperarid central core of Libya, receiving an averageannual rainfall of less than 15 mm p.a. (Dubief, 1963), and relies on groundwaterpumped from the Continental Intercalaire (Nubian Sandstone) aquifer. The extensivegroundwater reserves in central and southern Libya have been important throughouthistorical times and are of great strategic significance today, as they are beingincreasingly exploited by the Great Man-Made River Project for use in the coastal zone(McKenzie and El Saleh, 1994; Pim and Binsariti, 1994). These aquifers have beenrecharged at various stages throughout the Quaternary (Edmunds and Wright, 1979).Geomorphological evidence of lake highstands in several parts of North Africa havebeen dated and these periods have been linked to Quaternary aquifer recharge episodes(Gasse et al., 1987). However, the regional pattern of Quaternary environmental changeis still unclear, due to the paucity of information over very large areas, including theLibyan Fezzan. To clarify the palaeoclimate history of North Africa, evidence for pastchanges in groundwater hydrology in the Fezzan needs to be collected and interpreted.Preliminary results from an ongoing project on palaeoenvironmental reconstruction ofthe Wadi el Agial study site are presented here, along with some tentative conclusionsregarding the relationship between palaeoenvironment and human adaptations.

2. Methodology

Due to the absence of palaeoenvironmental data for the Wadi el Agial, the first task hasinvolved a detailed reconnaissance to identify and map geomorphological evidence ofenvironmental change. To assist in this, a Landsat Thematic Mapper image (Path 187,row 42 - Worldwide Reference System, acquired 12th October 1987) has been used toaid field survey. The main purpose of the remote sensing work is to map gypsum. Thereare high levels of Ca and S in contemporary groundwater, although salinity is variable

PALAEOENVIRONMENTAL RECONSTRUCTION, FEZZAN, LIBYAN 195

Image processing involved applying a spectral mixture model to the multispectral image(Settle and Drake, 1993). This approach has been demonstrated to be very effective atmapping surficial gypsum deposits (White and Drake 1993, Eckardt and White 1997).The technique yields a proportion estimate of gypsum content within each pixel of theimage. The image products were georeferenced with a very low R.M.S. error (meanR.M.S. = 0.66 pixel) to a set of 20 ground control points collected during the initial fieldseason (in March 1998) using a 12 channel hand-held GPS receiver. The ground controlpoints were well distributed over the whole study area. The accuracy of the

(Table 1). If we assume that groundwater geochemistry has not changed significantly,groundwater at or near the surface in the past would have precipitated gypsum, whichcan be remotely sensed (Drake, 1995).


georeferencing was checked in the field the following year (January 1999) and wasfound to be excellent, with the location of all the points visited being well within thenominal accuracy of the GPS instrument (c. 80m). No attempt was made to calibrate theremote sensing output with field data, as the success of this technique of mappinggypsum has been demonstrated elsewhere (Bryant, 1996; White and Drake 1993), andhere the aim is to map the distribution of gypsum deposits (which was checked in thefield), not to quantify actual concentrations.

Other fieldwork focused on finding evidence of palaeolake deposits, denoted byfossiliferous dark organic horizons. These deposits are generally only exposed insection, not at the surface, so remote sensing was not employed here. Instead, detailedsurveys were carried out along sections of the base of the escarpment (assumed to be thehighest shoreline) and sections exposed by well excavation at numerous locationsaround the oasis. Two significant exposures of dark organic horizons were found atescarpment proximal locations, one above the town of Germa (26.5058°N 13.0784°E)and one above the town of el-Greifa (26.4965°N, 13.0030°E). Both sites exhibit thindark organic horizons and contain a fossil assemblage dominated by Melanoidestuberculata (Figure 2).

Other exposures of dark organic horizons were found in well sections in the oasis aroundthe town of el-Greifa (Figure 1), though these were not noticed elsewhere in the studyarea. Wells with dark organic horizons are located at 26.5288°N, 12.9898°E;26.5332°N, 12.9903°E; 26.5448°N, 12.9960°E; 26.5403°N, 13.0050°E; 26.5247°N,12.9862°E. Samples of these dark organic horizons were collected from each of thesesections for comparison with each other and with the deposits found along theescarpment, to determine whether the dark organic horizons exposed during wellconstruction can be correlated with the fossiliferous palaeolake deposits found at thebase of the escarpment.


One of the major problems with identifying organic horizons in dryland situations wherehuman activity has been significant in the past is the possibility of confusion with burnedmaterial and fire hearths. To avoid this possibility, we also sampled a known fire hearthlocated among archaeological remains at 26.5633°N, 12.9512°E.

Stable isotope analyses of C and N were carried out on samples from all the dark organichorizons found in the field; both to correlate the samples and to see if specific isotopesignatures could be identified and used to infer conditions of formation (Koch, 1998). Inorder to be sure that like was being compared with like in all the dark organic horizons,isotopic analyses were run on the organic carbon fraction only, inorganic carbonateswere removed by treatment with 1.0 molar HCl (the percentage weight loss on HCltreatment was recorded). Analyses were performed using an automated carbon andnitrogen analyser and a continuous-flow isotope-ratio-monitoring mass-spectrometer(Europa ANCA Roboprep automated Dumas preparation system coupled to a Europa20/20 mass spectrometer). Typical replicate measurement errors are of the order of 0.3for


generate magnetite and maghemite (Maher, 1986). The effect is normally so great thatmagnetic susceptibility alone can usually discriminate burned samples (Rummery et al.,1979). However, is a concentration-dependent magnetic parameter, and the burnedsignal can be lost in a sedimentary environment if it becomes diluted during diagenesis.To avoid this problem, we have used the ratio of the IRM at 20mT divided by theessentially a measure of the concentration of all ferrimagnetic minerals divided by ameasure of the concentration of ferrimagnetic grains in the (stable singledomain) size range (Oldfield, 1991). Full details of mineral magnetic techniques,parameters and their interpretation are given elsewhere (Walden et al., 1999). This studyis only concerned with identifying any samples which may result from burning so thatthey can be excluded from further palaeoenvironmental interpretation.

Particle size analysis of the dark organic sediments was undertaken to establish thetextural characteristics of the sediments. After soaking the samples in sodiumhexametaphosphate overnight to disaggregate the clasts, the sands and coarse silts wereseparated from the fine silts and clays by wet sieving. The sediments were then driedand the coarse fraction was dry sieved and separated into the following phi sizes:

and and then each fraction was weighed. The finer fraction wascalculated using the pipette sedimentation method and separated intoand size fractions and their weights were determined.

U-Th dating was applied to a Melanoides shell sample. The shell was cleaned inultrapure (Milli-Q) water to remove as much detrital contamination from the exterior aspossible. The sample was then crushed in an agate mortar to a fine powder and baked inan oven at 105°C for 24 hours to remove surface moisture. U-series radionuclides weremeasured by high resolution alpha spectrometry and ICP-MS. Uranium and thoriumisotopes for alpha spectrometry were separated by ion-exchange resins andelectrodeposited on stainless steel planchets (Black et al., 1997; Kuzucuoglu et al.,1998) after aliquots were taken for ICP-MS. yield monitor was used foralpha spectrometry with a decay and ingrowth correction applied for the daughternuclide. Yields ranged from 88-100 % for uranium and 91-98 % for thorium. Blank andbackground determinations were carried out, averaging 20 counts in 10,000. EG&G®"Ultra" alpha detectors were used with very low background counts in the yield monitorenergy range. ratios were determined on unspiked aliquots to ensure norecent alteration had taken place. and were also detected on a VarianUltramass ICP-MS. One aliquot in 5 for ICP-MS was spiked with a yield monitorto assess U recoveries. In addition, internal standards were routinely run and checkedagainst the NIST SRM3159 standard and agreed to better than 0.25 %. Detection limitsat masses 232 and 238 were better than 8 ng after correcting for dead time, backgroundand mass bias. Replication of multiple digestions was better than 0.6 % for both U and


Th whereas replicate aliquot analyses (n = 36) was better than 0.3 % for U and 0.2 % forTh.

Corrections were made for decay of excess and detrital on the assumptionthat these were present at formation of the shell. The correction for the detritalcomponent was made from isochron plots after successive total dissolutions wereperformed following the preparation, firing and digestion outlined in Luo and Ku (1991)and Bischoff and Fitzpatrick (1991). In all cases the slopes of the isochrons are bestdeterminations by a method of least squares fitting which takes account of the errors inboth variables (York, 1969).

3. Results

3.1. EVAPORITE MINERALS

The gypsum proportions map produced by applying a spectral mixture model to theThematic Mapper imagery (Figure 3) indicated a widespread distribution of gypsumthroughout the study area. Field observations and XRD analysis of samples back in thelaboratory indicate that this surficial gypsum is found in four different sedimentaryenvironments.

1.2.

3.

4.

As evaporitic crusts on small playas along the base of Wadi el Agial.As crusts formed on fields, often adjacent to known foggara systems(Figure 1). Foggara, or qanats as they are called in Persia, are undergroundcanals which tap higher water tables near mountain fronts and transport thewater down to irrigated fields beyond the piedmont zone (Goblot, 1979).However, the formation of the field crusts may result from more recentirrigation from pumping rather than from ancient irrigation via foggara.As outcrops along the base of the escarpment, indicating a palaeo-springlineat the junction between the Murzuk Sandstone Formation and theunderlying Germa Bed marls (Klitzsch and Baird, 1969). The fibrous habitof these gypsum crystals indicates that they have grown in-situ within thebedrock and surficial materials (Cody and Cody, 1988). In many places, thegypsum has been altered to anhydrite, presumably due to the high summertemperatures encountered in the surface environment.As a constituent of the sediments deposited within relict fluvial channels onthe south-facing dip-slope of the Murzuk plateau.


These results provide strong evidence of the high Ca and S content of the groundwaterwhich fed these fluvial networks when the water table was higher. The chemistry of thepalaeodischarge would appear to have been similar to that of the contemporarygroundwater (Table 1).


3.2 DARK ORGANIC HORIZONS

The mineral magnetic data discriminate the known fire-hearth site at ElHatia (Figure 4), but one of the well-section dark organic horizons (known as foggarawell, as a foggara channel was encountered during the sinking of the well shaft) has asimilar high value. This may mean that the dark colour of these samples could be due toburning rather than enrichment with organic compounds, so they are excluded fromfurther analyses. However, the lower value of the foggara well sample relative to theEl Hatia sample (Table 2), though still significantly higher than the other dark horizons,indicates a lower concentration of ferrimagnetic minerals, possibly through transport andredeposition, or by dilution during diagenesis and weathering (Rummery et al., 1979).

The burned sediments (El Hatia 8 and Foggara well) also have a much higher HClsoluble content (30-37%) than the other dark organic horizons (Table 2). The sameparameter can be used to discriminate between the el-Greifa palaeolake sediments (16-


21%) and the oasis well section dark organic horizons (6-12%). Again this may resultfrom primary sediment composition or from diagenesis and precipitation of evaporiteminerals.

The stable C isotope results (Table 2) show a strong depletion of relative to PDBstandard (Peedee Formation, belemnite). All the samples cluster around the -23threshold between C3 and C4 plant metabolic processes, suggesting either the organicmatter in the dark organic horizons was derived from a mixed C3/C4 community (Bondet al., 1994; Kelly et al., 1998), or that some other fractionation process has operatedduring or after deposition of the sediment; the initial carbon isotope contents can bemodified by various diagenetic effects such as precipitation/dissolution of secondarycalcite and oxidation of reduced carbon (Fontes, 1994).

The stable N isotope results (Table 2) distinguish between the shoreline lake depositsidentified along the base of the escarpment, and the dark organic horizons identified inthe oasis well sections (Figure 4). This may be attributable to differentpalaeoenvironments (the values of the well section dark organic horizons are closeto atmospheric N (Table 2), suggesting direct take-up by cyanobacteria, possibly insmall swamps which have formed in the oasis), but they could equally well represent thedifference between shoreline and deeper-water vegetation communities. Furthermore,post-depositional processes again have to be considered; ion exchange in soils can alsoresult in fractionation effects affecting Unlike nitrite or nitrate ions, readilyexchanges with clays or the so-called “humic-clay complex” of soils. Isotopefractionations between exchange resins and a liquid phase are known to exist, with an


enrichment between 5 and 25 for the solid phase (Létolle, 1980). Two main factorscan control the N isotopic characteristics of the organic matter.

1)

2)

Varying degrees of diagenesis. Decay produces isotopically light nitrogencompounds and the remaining organic matter becomes progressivelyenriched inThe migration and accumulation of decay products, such as humin, fulvin,humic and fulvic acids, in different horizons (Létolle, 1980).

Therefore, the limited amount of stable isotope data available so far are unable toprovide a basis for correlation of the different dark organic horizons. However, thelimited extent of the dark organic horizons exposed in the oasis well sections, coupledwith the lack of fossils and lower HCl soluble content, suggests that these may be verylocalised marsh or swamp deposits, unrelated to the fossiliferous palaeolake sedimentsfound at the base of the escarpment.

Table 3 shows the results of the particle size analyses that have been conducted. Samplesfrom el-Greifa 1.2 and 1.3 were collected adjacent to one another close to the palaeolakeshoreline. Their sediment distributions are similar with mean grain sizes of and theycomprise mostly sand with low amounts of clay (Table 3). el-Greifa 3 is from a locationslightly further away from the shoreline, in deeper water, which may account for thehigher amounts of clay and silt in this sample. The dark organic horizons in the wellsvary in terms of the mean grain sizes between 4.4 and with on average 45.9% sand,37.6% silt and 16.6% clay. The two samples with the coarsest mean size fractions (ElHatia 8 and Foggara well) were highlighted by mineral magnetic analysis as containingburned materials. All of the samples are platykurtic, positively skewed and poorly orvery poorly sorted.

The Melanoides shell gave a U/Th age of 84.5 ± 4.6 ka BP. This falls within the widely-postulated 90-65ka BP Saharan lake highstand (Causse et al. 1989; Fontes and Gasse,1989; Petit-Maire et al. 1980; Szabo et al., 1995), which can be correlated with oxygenisotope interglacial substage 5a (or 5.1). However, as only one age is so far available forthe Wadi el Agial palaeolake deposits, this must be treated with caution. A large numberof other Melaniodes shells are currently being dated, and a fuller analysis of these, andother radiometric ages, will be provided in a later publication.


4. Discussion

Despite the hyper-aridity of the present environment, the Fezzan has a long history ofhuman occupation and a detailed archaeological record which dates back to thePalaeolithic (Mattingly et al., 1998) Extensive rock art, particularly in the Wadi Berjujregion, some 50 km to the south of Wadi el Agial, shows considerable evolution oftechnique and subject matter, indicating the long duration of human activity in thisregion. The earliest identified remains in the vicinity of Wadi el Agial are scatters ofquartzite lithics and debitage along the edge of the Murzuk escarpment. These includehand axes, a cleaver, projectile points, scrapers and denticulates and a burin. The mostprobable date range for these materials is 105-40ka BP (Mattingly et al., 1998), and mayindicate presence of a lake below the escarpment (where Palaeolithic material is veryscarce) during some of this time. From this we can infer a higher regional water table,and greater activity of the relict fluvial systems along the south-facing dip slope of theMurzuk Plateau.

Below the Murzuk plateau, Neolithic finds are concentrated around patches of duricrustsurfaces around and within the Ubari sand sea. The duricrusts contain abundantmineralised plant roots and have been interpreted as palaeoswamp deposits. From thiswe can infer that, during the Holocene the area of surface water had diminished to anextent that large areas below the Murzuk Plateau were exposed and inhabited byhumans.

Numerous burials and settlements on the escarpment piedmont and adjacent oases aremostly attributed to the Garamantean occupation (c. 500 BC to 500 AD). During thisperiod of occupation small lakes may have still been present in the low points of theWadi El Agial; results from ongoing dating studies are awaited to confirm this; butextensive use of foggara technology appears to have been made at this time (Mattinglyet al., 1998), in order to grow irrigated crops (van der Veen, 1992). In the vicinity of the


village of Twesh (Figure 1), field surveys have indicated that motherwells of foggarachannels (the collecting shafts dug at the top of each foggara to tap the groundwater) areassociated with palaeo-springline gypsum deposits. From this we can infer that, by thetime of foggara construction, these springs had ceased to be perennial features, but theymay have still been zones of localised high groundwater levels. The foggaras weresubsequently abandoned, possibly due to a continuing fall in the water table, althoughother factors, such as the introduction of new technologies or a collapse in population,may also be responsible. The time of foggara abandonment is unknown at present.

The pattern of environmental change, and the close relationship with human activity, hasmany similarities with the record from the Western Desert, Egypt and Sudan, which hasreceived more attention (see e.g. Peel, 1966; Wendorf et al., 1976; Wickens, 1975).Both areas demonstrate a close association between significant Palaeolithic andNeolithic populations and palaeolacustrine deposits (Wendorf and Schild, 1980), bothexhibit complex evidence of fluctuating water table, with aeolian sands underlyinglacustrine clays in the base of depressions (Haynes, 1982). In this respect, the Wadi elAgial seems typical of Late Quaternary environments from eastern North Africa.

5. Conclusion

A preliminary geomorphological survey of the Libyan Fezzan has highlighted evidenceof significant environmental changes in the region during the Late Quaternary. A lakehighstand, tentatively dated at 85ka BP, deposited dark organic sediments and a fossilassemblage dominated in this region by Melanoides tuberculata. Mineral magneticanalysis enables discrimination of these dark organic sediments from those produced byburning. Stable isotope analysis of C suggest that either the organic matter in the darkorganic horizons was derived from a mixed C3/C4 community or that some otherfractionation process has operated during or after deposition of the sediment. Thin darkorganic horizons exposed in well sections at the base of the Wadi el Agial differ instable N composition from the fossiliferous palaeolake sediments exposed higher up onthe escarpment slopes. This may indicate a difference in palaeoenvironments, althoughdiagenetic modifications may also be responsible. HCl soluble content also enablesdiscrimination of burned materials, el-Greifa palaeolake sediments and oasis wellsection dark organic horizons. Palaeo-springlines and minor upwelling zones in smallplaya basins have deposited evaporitic gypsum, which have been mapped fromremotely-sensed data. The ongoing Fezzan archaeological mission is documentingchanges in the distribution of human activity; Palaeolithic industry is restricted to thehamada above the Murzuk escarpment, this may be due to the presence of a large


palaeolake covering the Wadi el Agial at certain times during the late Pleistocene,including at c. 85ka BP. Neolithic industry is found below the escarpment, but may alsobe restricted to the shorelines of much smaller Holocene lakes and swamps. Foggaratechnology was exploited during the period of Garamantean occupation of the area, butthe foggaras were subsequently abandoned, possibly due to a continuing fall in the watertable, although other factors, such as the introduction of new technologies or a collapsein population, may also be responsible.

References

Bischoff J. L. and Fitzpatrick, J. A. (1991) U-series dating of impure carbonates: An isochron technique

using total-sample dissolution. Geochim. Cosmochim. Acta, 55, pp. 543-554.Black, S., Macdonald, R. and Kelly, M. R. (1997) Crustal origin for peralkaline rhyolites from Kenya:

Evidence from U-series disequilibria and Th-isotopes. Journal of Petrology, 38, pp. 277-297.

Brookes, I.A. (1993) Geomorphology and Quaternary Geology of the Dakhla Oasis region, Egypt.Quaternary Science Reviews, 12, pp. 529-552.

Bryant, R.G. (1996) Validated linear mixture modelling of Landsat TM data for mapping evaporite minerals

on a playa surface: methods and applications. International Journal of Remote Sensing, 17, pp. 315-330.Bond, W.J., Stock, W.D. and Hoffman, M.T. (1994) Has the Karoo spread? A test for desertification using

carbon isotopes from soils. South African Journal of Science, 90, pp. 391-397.Causse, C., Coque, R., Fontes, J.C., Gasse, F., Gibert, E., Ben Ouezdou, H. and Zouari, K. (1989) Two high

levels of continental waters in the southern Tunisian Chotts at about 90 and 150 ka. Geology, 17, pp. 922-925.

Cody, R.D. and Cody A.M. (1988) Gypsum nucleation and crystal morphology in analogue saline terrestrialenvironments. Journal of Sedimentary Petrology, 58, pp. 247-255.

Cremaschi, M. and Trombino, L. (1998) The palaeoclimatic significance of paleosols in southern Fezzan(Libyan Sahara): morphological and micromorphological aspects. Catena, 34, pp. 131-156.

Drake, N.A. (1995) Reflectance spectra of evaporite minerals (400-2500 nm): applications for remotesensing, International Journal of Remote Sensing, 16, pp. 2255-2572.

Dubief. J. (1963) Le climat du Sahara. Mem. Inst. Rech. Sahariennes Alger. 2, 270 pp.Eckardt, F. and White, K. (1997) Human induced disruption of stone pavement surfaces in the central

Namib desert, Namibia: observations from Landsat Thematic Mapper. International Journal of RemoteSensing, 18, pp. 3305-3310.

Edmunds, W.M. and Wright, F.P. (1979) Groundwater recharge and palaeoclimate in the Sirte and Kufrabasins, Libya. Journal of Hydrology, 40, pp. 215-241.

Fontes, J.C. (1994) Isotope palaeohydrology and the prediction of long-term repository behaviour. TerraNova, 6, pp. 20-36.

Fontes, J.C. and Gasse, F. (1989) On the ages of humid Holocene and late Pleistocene phases in north Africa- remarks on "late Quaternary climatic reconstruction for the Maghreb (north Africa)" by P. Rognon.Palaeogeography. Palaeoclimatology, Palaeoecology, 70, pp. 393-398.

Fontes, J.C. and Gasse, F. (1991) PALHYDAF (Palaeohydrology in Africa) program: objectives, methods,major results. Palaeogeography, Palaeoclimatology, Palaeoecology, 84, pp. 191-215.

Gasse, F., Fontes, J.C., Plaziat, J.C., Carbonel, P., Kaezmarska, I., De Dekker, P., Soulié-Märsche, I., Callot,Y. and Dupeuble, P.A. (1987) Biological remains, geochemistry and stable isotopes for the


reconstruction of environmental and hydrological changes in the Holocene lakes from north Sahara.Palaeogeography. Palaeoclimatology, Palaeoecology, 60, pp. 1-46.

Gaven, C., Hillaire-Marcel, C. and Petit-Maire, N. (1981) A Pleistocene lacustrine episode in southeastern

Libya. Nature, 290, pp. 131-133.Goblot, H. (1979) Les Qanats, une technique d’acquisition de l’eau. Paris: Mouton Éditeur, 236pp.

Haynes, C.V. (1982) The Darb el-Arba’in Desert: a product of Quaternary climatic change. In El Baz, F. andMaxwell, T.A. (Eds.) Desert Landforms of Southwest Egypt: A Basis for Comparison with Mars,

Washington DC: NASA, pp. 91-117.

Kelly, E.F., Blecker, S.W., Yonker, C.M., Olson, C.G., Wohl, E.E. and Todd, L.C. (1998) Stable isotope

composition of soil organic matter and phytoliths as paleoenvironmental indicators. Geoderma, 82, pp.59-81.

Klitsch, E. and Baird, D.W. (1969) Stratigraphy and palaeohydrology of the Germa (Jarma) area, southwestLibya. In Kanes, W.H. (Ed.) Geology, Archaeology and Prehistory of the Southwestern Fezzan, Libya,

Petroleum Exploration Society of Libya, Eleventh Annual Field Conference, Castelfranco-Veneto, pp. 67-80.

Koch, P.L. (1998) Isotopic reconstruction of past continental environments. Annual Review of Earth and

Planetary Sciences, 26, pp.573-613.Kuzucuoglu, C., Pastre, J-F., Black, S., Ercan, T., Fontugne, M., Guillou, H., Hatté, C., Karabiyikoglu, M.,

Orth, P. and Türkecan, A. (1998) Identification and dating of tephras from Quaternary sedimentary

sequences of inner Anatolia. Journal of Volcanology and Geothermal Research, 85, pp. 153-172.Létolle, R. (1980) N-15 in the natural environment. In Fritz, P. and Fontes, J.C. (Eds.) Handbook of

Environmental Isotope Geochemistry Volume 1, Amsterdam: Elsevier, pp. 407-433.Luo, S. and Ku, T. L. (1991) U-series isochron method: A generalised method employing total-sample

dissolution. Geochim. Cosmochim. Acta 55, pp. 555-564.

Maher, B.A. (1986) Characterisation of soils by mineral magnetic measurements. Physics of the Earth and

Planetary Interiors, 42, pp. 76-92.Mattingly, D.J., al-Mashai, M., Aburgheba, H., Balcombe, P., Eastaugh, E., Gillings, M., Leone, A.,

McLaren, S.J., Owen, P., Pelling, R., Reynolds, T., Stirling, L., Thomas, D., Watson, D., Wilson, A.I. andWhite, K. (1998) The Fezzan Project 1998, preliminary report on the second season of work. Libyan

Studies, 29, pp. 115-144.McKenzie, H.S. and El Saleh, B.O. (1994) The Libyan Great Man-Made River Project 1; Project Overview.

Proceedings of the Institution of Civil Engineers; Water Maritime and Energy, 106, pp. 103-122.

Oldfield, F. (1991) Environmental magnetism, a personal perspective. Quaternary Science Reviews, 10, pp.

73-85.Peel, R.F. (1966) The landscape in aridity. Transactions of the Institute of British Geographers, 38, pp. 1-

23.Petit-Maire, N., Delibrias, G. and Gaven, C. (1980) Pleistocene lakes in the Shati area, Fezzan.

Palaeoecology of Africa, 12, pp. 289-295.Petit-Maire, N., Reyss, J.L. and Fabre, J. (1994) A last interglacial palaeolake in the hyperarid Sahara of

Mali. Comptes Rendus de l’Academie des Sciences Serie II, 319, pp. 805-809.Pim, R.H. and Binsariti, A. (1994) The Libyan Great Man-Made River Project 2; The Water Resource.

Proceedings of the Institution of Civil Engineers; Water Maritime and Energy, 106, pp. 123-145.

Rummery, T.A., Bloemendal, J., Dearing, J., Oldfield, F. and Thompson, R. (1979) The persistence of fire-induced magnetic oxides in soils and lake sediments. Annales de Géophysique, 35, pp. 103-107.

Szabo, B.J., Haynes, C.V.Jr. and Maxwell, T.A. (1995) Ages of Quaternary pluvial episodes determined byuranium series and radiocarbon dating of lacustrine deposits of eastern Sahara. Palaeogeography,

Palaeoclimatology, Palaeoecology, 113, pp. 227-242.


Settle, J. J. and Drake, N.A. (1993) Li near mixing and the estimation of ground cover proportions.

International Journal of Remote Sensing, 14, pp. 1159-1177.

Tite, M.S. and Mullins, C.E. (1971) Enhancement of the magnetic susceptibility of soil on archaeologicalsites. Archaeometry, 13, pp. 209-219.

van der Veen, M. (1992) Botanical evidence for Garamantean Agriculture in Fezzan, Southern Libya.Review of Palaeobotany and Palynology, 73, pp. 315-327.

Walden, J., Oldfield, F. and Smith, J.P. (Eds.) (1999) Environmental Magnetism, a Practical Guide.

Technical Guide No. 6, London: Quaternary Research Association, 243pp.Wendorf, F. and Schild, R. (1980) Prehistory of the Eastern Sahara. New York: Academic Press, 414pp.Wendorf, F., Schild, R., Said, R., Haynes, C.V., Gautier, A., and Kobusienwicz, M. (1976) The Prehistory of

the Egyptian Sahara. Science, 193, pp. 103-114.White, K. and Drake, N.A. (1993) Mapping the distribution and abundance of gypsum in south-central

Tunisia from Landsat Thematic Mapper data. Zeitschrift für Geomorphologie N.F., 37, pp. 309-325.Wickens, G.F. (1975) Changes in the climate and vegetation of the Sudan since 20,000 B.P. Boissiera, 24,

pp. 43-65.

York D. (1969) Least squares fitting of a straight line with correlated errors. Earth and Planetary Science

Letters, 5, pp. 320-324.

RELICT CRYOGENIC MOUNDSIN THE UK AS EVIDENCE OFCLIMATE CHANGE

STEPHEN D. GURNEYLandscape and Landform Research Group, Department of Geography,The University of Reading, PO Box 227, Whiteknights, Reading RG6 6AB,U.K.

Abstract

Relict perennial cryogenic mounds, (the remains of pingos or palsas) have long beenidentified in the UK and have been attributed to the Last Glacial or Younger Dryascold periods. These features take the form of circular to oval depressions surroundedcompletely or partially by a raised rim or rampart. The depressions very oftencontain a wetland with an organic soil or peat. At depth within the depressions therecan be up to several metres of soft sediments (very often a silt-clay). The ramparteddepressions are usually found in clusters where individual features may overlap suchthat they share ramparts or even where the depressions merge to produce a ‘figure-of-eight’ shape in plan. Such relict periglacial features are extremely useful forpalaeoclimatic reconstruction, since they can indicate the average thermal conditionof the ground during the period in which they were actively forming. Many relictcryogenic mounds in the UK have been interpreted as the remains of hydraulicpingos, features which in contemporary cold climate areas indicate eitherdiscontinuous permafrost (mean annual air temperature of orcontinuous permafrost of Some of these features, however, havenow been re-interpreted as the remains of ‘mineral-cored’ palsas, features that areindicative of only sporadic permafrost 0f or of or discontinuouspermafrost of . Thus the varying interpretation and classification ofthese geomorphological features can lead to a wide disparity in their palaeoclimaticsignificance. Clearly, to obtain the correct palaeoclimatic information, such features

209


210 GURNEY

must be accurately identified and appropriate modern analogues utilised in theirinterpretation.

KEY WORDS: Relict cryogenic mounds, geomorphology, palaeoclimatic significance.

1. Introduction

Relict perennial cryogenic mounds are very important geomorphological featuresbecause they indicate the former presence of permafrost. Unfortunately, their correctidentification and interpretation is not straightforward and the use of many specificfeatures for palaeoclimatic reconstruction has been questioned, either on the basis oftheir actual interpretation or the use of inappropriate modern analogues.

Contemporary perennial cryogenic mounds are generally classified as either pingos orpalsas, although the reality may be that these two types represent the extremes of a‘landform continuum’. Pingos are the largest of the perennial cryogenic mounds and canbe over 50 m high, or have diameters of over 500 m (see Gurney, 1998). Pingos areclassified as either hydrostatic (formerly known as closed-system) or hydraulic (open-system). Hydrostatic pingos are usually solitary and their type-site is the MackenzieDelta/Tuktoyaktuk Peninsula area of the Northwest Territories of Canada (Mackay,1998). Hydraulic pingos occur in groups or clusters, and their type-site is eastGreenland (Müller, 1959). Both pingo types rely on a pressure system to deliver waterto the core where it is frozen to form the ice-core responsible for local uplift of theoverburden. Palsas are generally smaller than pingos ( or more high andor more diameter) and take two basic forms: those purely composed of peat and thosewhich have a surficial layer of peat but whose core is developed in mineral sediments(mineral-cored palsas, e.g. Åhman, 1976). It must be noted, however, that an overlapexists with features such as ‘peat plateaux’, which are less discrete and can be muchmore areally extensive, although here the most ‘palsa-like’ features may be formedthrough degradation as opposed to aggradation (e.g. Hinkel, 1988). Palsas grow throughthe accumulation of a segregated ice core whose growth is sustained by cryosuction,which draws water to the freezing front (see Washburn, 1983a; 1983b).

One of the main problems with the reconstruction of former permafrost areas fromgeomorphological features is that these features are discrete and are only locallypreserved. Furthermore, to infer only former temperature conditions from the existenceof these features is far too simplistic. Many periglacial features are to some extentpolygenetic and therefore rely on a number of parameters for growth and maintenance,temperature being only one of these. Of considerable importance are the freeze-thawregime, for example, is it a seasonal regime like that of the high latitudes, where solarinsolation is concentrated in the summer and is absent in the winter, or is it more diurnallike that of the mid and low latitudes, where daily freeze-thaw cycles may operateincreasing the total number of cycles per year. The thickness, timing, redistribution (by

RELICT CRYOGENIC MOUNDS IN THE UK 211

wind) and duration of snow cover are also extremely important since snow can reducethe penetration of frost into the ground surface (a process which may be vital to thecreation of palsa type cryogenic mounds, see Seppälä, 1994). The presence/absence ofvegetation and vegetation types and their albedo are also very important in theunderstanding of the ground thermal regime (e.g. Railton and Sparling, 1973). Clearlyall of these factors must be taken into account if relict cryogenic mounds are to be usedas indicators of some quantifiable climate change. This chapter will attempt a review ofthe literature of relict cryogenic mounds and will discuss their possible use aspalaeoclimatic indicators. Although the focus is on relict cryogenic mounds in the UK(for the distribution of which see Figure 1), to ignore evidence from the rest of northwest Europe and further afield would clearly be misleading and therefore an attempt willbe made to synthesise and incorporate this material, where relevant.

212 GURNEY

Previous reviews have been conducted by Wiegand (1965) who thoroughly cataloguedthe major relict ‘pingo’ sites of middle Europe favouring a periglacial ground-icehypothesis for their formation (see also Maarleveld, 1965). Flemal (1976) attempted toprovide a detailed discussion on relict ‘pingos’ (their characteristics and distribution) aswell as their use in the reconstruction of former permafrost environments. AlthoughFlemal does consider that palsas or another type of cryogenic mound might be the causeof some of the features described as relict ‘pingos’. Bryant and Carpenter (1987)reviewed all the ramparted ground-ice depressions in Britain and Ireland and provided agood distribution map and summary table of these features. The overlap between pingosand palsas was not, however, made clear. De Gans (1988) produced a comprehensivediscussion of Pleistocene ‘pingo’ remnants, reviewing evidence from Europe and NorthAmerica.

2. The basic geomorphology and sedimentology of relict cryogenic mounds

The criteria used to identify relict cryogenic mounds should be clear. Generally thesefeatures take the form of circular to oval depressions that are surrounded completely orpartially by a raised rim or rampart (see Table 1). They are generally found in clustersas opposed to single isolated features (see Figures 2 and 3). When the spatial density ofthe features is high, it is common for the ramparts and even depressions to becomemerged at their edges creating a mutually interfering pattern. The clusters of ramparteddepressions can be found in a range of settings from valley bottoms to lower valley sidesand occasionally on plateau sites. The basic sedimentology of the features is rather moreproblematic since many of the sites are understood and characterised in terms of themorphology alone. When considering the basic sedimentology, it is expected that thesediments comprising the rampart should be partially derived from material that hassloughed off the mound and partially derived from material that has been pushedoutwards from the mound core during growth of the ice body (whether this wascomposed of massive-injection ice or segregated ice or a combination of the two). Thematerial within the depression (often referred to as the ‘depression fill’ or ‘pingo fill’)should be related to the material within which the ice core formed and to the overburdenmaterial (since some of this may have collapsed into the depression on the decay of themound, particularly if the decay was top-down). Above the pingo-fill there are veryoften pond or marsh deposits (e.g. peat), which formed after the collapse of the feature,and these are therefore not diagnostic of mound genesis. When relict cryogenic moundshave been investigated, however, clear relationships such as those expressed above haverarely been determined.


3. Previous investigations of relict cryogenic mounds

3.1. EARLIEST WORK

Relict cryogenic mounds (the remains of perennial cryogenic mounds which formed in aperiglacial environment) were first identified in the UK by Pissart (1963). This workfollowed the investigation by Pissart (1956; 1958) of the “viviers” (literally ‘fish-ponds’)of the Hautes Fagnes plateau of north-east Belgium. They take the form of more or lesscircular depressions surrounded by a raised rim or rampart. Pissart ascribed the featuresto a periglacial genesis and termed them pingos, however, he stated that they wereprobably mostly composed of segregated ice (as opposed to massive or injection ice,more typical of contemporary pingo cores) when they were active. The ramparts wereseen to be the key and these appeared to have been built by material soliflucting off thesides of the mound and perhaps also, by a lateral thrust from the interior of the mound.

214 GURNEY

This paper was pioneering, not only for the discovery of relict cryogenic mounds innorth west Europe indicating a colder, permafrost, environment of the late Pleistocene,but also in recognising that the genesis of these forms may not have been exactly likethat of the contemporary arctic pingos that had been studied up to that time.

Pissart’s work on relict cryogenic mounds in Belgium continued over many years andextensive investigations were undertaken which refined the hypotheses of theirformation and ultimately led to their reclassification. As early as 1974, Pissart suggestedthat, in terms of genesis these features when active were more similar to palsas, thanpingos. An important facet of the growth mechanism appeared to be lateral growth ofthe segregated ice core, which displaced sediment outwards (Bastin et al., 1974). This isfar more closely related to palsa growth than to pingo growth where the diameter of themounds is formed early on in the growth cycle and subsequent growth takes placeupwards (Mackay, 1979). Work continued (e.g. Pissart et al., 1975) and the use oftrenches through the features allowed detailed studies of the structure at depth and of thesedimentology (e.g. Pissart and Juvigné, 1980; Pissart, 1983a; Pissart et al., 1975;Juvigné and Pissart, 1979). The discovery of a peat layer that would have existed atground level during the growth phase of the mounds (Pissart and Juvigné, 1980) andsedimentary structures in the base of the ramparts which were thought to be connected tothe lateral growth of segregation ice lenses led to a reclassification of the features fromthe remains of hydraulic pingos to the remains of ‘mineralogic palsas’ (Pissart, 1983a).


Initially the palaeoclimatic inference of these features was believed to be similar to thatof contemporary hydraulic pingos (a MAAT of 7°C), however, this was subsequentlymodified to between 0°C and 5°C (Pissart, 1965). In terms of using these features toreconstruct permafrost type, this was important since it meant that these features wereindicators of discontinuous rather than continuous permafrost.

216 GURNEY


3.2. FEATURES IN THE UK

The first study of relict cryogenic mounds in the UK compared the features of Llangurigin Wales with those of the Hautes Fagnes Plateau of Belgium (Pissart, 1963). TheWelsh examples were formed in a valley bottom, which posed the possibility of anhydraulic pingo genesis, whereas the Belgian pingos were exclusively on the plateau ofthe Ardennes with a generally northerly aspect. The question must be posed here as towhether the genesis of the two groups of features can possibly be the same. The keypoint in this paper and the significant similarity between the two types of relict forms isthe fact that the ramparts in both Belgium and Wales are thought to provide evidence oflateral thrust exerted by a growing ice lens. This may divide these types of mounds fromarctic pingos where massive ice cores are most common. The importance of segregatedice in arctic pingos, as found in Scandinavian palsas, however, is now also known toplay a role.

Other early work on the relict cryogenic mounds was conducted by Trotman (1963) whoused palynological data to report a vegetation history for the peat, which had developedinside the depression of one of the relict cryogenic mounds near Llangurig. The earliestpeat was said to belong to an early pre-boreal time although this peat formed after themound had collapsed and possibly a long time after and therefore is of no use in datingthe active phase of the feature.

The study of relict cryogenic mounds in the UK during the 1970s and 1980s was largelyconducted by the Watsons. Watson (1971) reported a group of relict cryogenic moundsfrom Wales and the Isle of Man. Watson summarised the evidence at the Llangurig siteand described another site in west Wales (Cledlyn) as well as the site on the Isle of Manat Ballaugh where the depressions are without ramparts. All of the features wereclassified as the remains of hydraulic pingos. The Cledlyn site was further documentedby Watson (1972) and by Watson and Watson (1972). The latter paper detailed all ofthe cores taken from the depressions with respect to both the grain size of the includedsediments and the profiles produced from the depth data. The major problem with thesefeatures is the large quantity of clay-silts, which compose the depression fill. Thismaterial was believed to have been deposited in the pingo pond during and after thecollapse of the pingo, however, since the thickness of this deposit reaches over 7 m,questions must be raised whether this deposit, may possibly have existed at depth beforethe formation of the cryogenic mounds. Another site in west Wales at Cletwr wasdocumented in Watson (1972) and in Watson and Watson (1974). Examples of theforms were mapped and levelled and extensive maps and sections were produced fromthese data and from that gained by coring. Linear relict cryogenic mounds at this sitewere likened to those described by Mückenhausen (1960) of the Hohen Venn/Eifelregion, Germany that had also been classified as the remains of hydraulic pingos.

The paper by Sparks et al. (1972), which describes “presumed ground-ice depressions”at East Walton Common in East Anglia, is often cited as evidence for relict cryogenic

218 GURNEY

mounds in the UK. It was concluded that these features were most likely to have anorigin in ground-ice accumulation, but whether this was of the pingo or palsa type wasnever speculated upon (except that springs were believed to have delivered the water tothe core).

Other sites in the UK with relict cryogenic mounds were documented during the 1980s.Hutchinson (1980) speculated upon possible Late Quaternary mound remnants in centralLondon. The features take the form of drift-filled depressions and had been previouslyinterpreted as Late Quaternary scour hollows (Berry, 1979). Hutchinson reinterpretedthem as the remains of hydraulic pingos. Carpenter and Woodcock (1981) reporteddetailed investigations of a ramparted depression at Elstead in Surrey (still regarded asthe most southerly relict cryogenic mound in the UK). The feature was interpreted asthe remains of a Late Devensian hydraulic pingo. Bryant et al. (1985) presented a casestudy of relict cryogenic mounds from the Whicham Valley of Cumbria. These featureswere defined as the remains of hydraulic pingos and taken to indicate the formerexistence of discontinuous permafrost. The features were believed to be Younger Dryasin age. Miller (1990) documented relict cryogenic mounds near Brent Tor, westernDartmoor. It was concluded that they were remnants of hydraulic pingos whose watersupply was derived from the passage of groundwater along a sub-surface thrust plane.Taylor (1987) reported the existence of features believed to be Younger Dryas age, relicthydraulic pingos at Abermad in west Wales.

The coverage of this topic in Ballantyne and Harris (1994) is extremely useful, althoughthe suggestion that the relict cryogenic mounds at Brent Tor are the most likelycandidates for mineral palsa origin is questionable. It is likely indeed, that due to thegeological conditions at the site, for example the existence of faults in the bedrockbeneath the site, that these features are more closely related to hydraulic pingos.

In Gurney (1995) the relict cryogenic mounds of mid and west Wales were re-evaluatedon the basis of new investigations. Despite previous studies classifying these features asthe remains of hydraulic pingos, a tentative reinterpretation was made that these featuresactually represented the remains of mineral-cored palsas on the basis of thesedimentological composition of the depression fills, which, it was felt did not concurwith the previously hypothesised genesis of these features. Further work was thought tobe warranted at this site.

Gurney and Worsley (1996) documented the remains at Owlbury on theShropshire/Powys border, which take the form of classic relict cryogenic mounds.These features were interpreted as the remains of former mineral-cored palsas asopposed to hydraulic pingos, on the basis of morphology and sedimentary setting.


3.3 FEATURES IN THE REST OF NORTH WEST EUROPE

North west Europe has produced sites of relict cryogenic mounds other than those ofBelgium and the UK. These include: Ireland (Mitchell, 1971; Mitchell, 1973; Coxon,1986; Coxon and O’Callaghan, 1987), Denmark (Cailleux, 1957), Germany(Mückenhausen, 1960; Picard, 1961), Luxembourg (Slotboom, 1963), Poland (Dylik,1965), France (Boyé, 1957; Rousset, 1965; Bout, 1986) and The Netherlands (Ploegerand Waateringe, 1964; Paris et al., 1979; De Gans and Sohl, 1981; De Gans, 1982;Bijlsma, 1983; Van der Meulen, 1988). Where an age has been assigned to the activephase of these features it was generally Last Glacial, although occasionally YoungerDryas. Where a genetic system was suggested it was generally of the hydraulic pingotype (e.g. De Gans and Sohl, 1981) although one these sites did appear to suggest anhydrostatic pingo genesis (Van der Meulen, 1988).

The general characteristics of the relict cryogenic mounds in the rest of north-westEurope, takes a similar form to that of the UK (i.e. circular to oval ramparteddepressions), however, there are notable instances where they are located on slopes orplateau sites (as opposed to lower valley sides or valley bottom locations).

3.4 FEATURES OUTSIDE NORTH WEST EUROPE

Outside north-west Europe the documentation of relict cryogenic mounds is ratherscarce. Flemal et al. (1969 and 1973) summarised the characteristics of a group ofraised, oval to circular, nearly flat-topped mounds, which cover at least ofBloomington ground moraine in north-central Illinois, USA. These forms are termed the‘DeKalb mounds’ and were interpreted as the remains of pingos, and their existence wasused to infer at least discontinuous permafrost in Illinois in Woodfordian time (LateWisconsinan). These features are quite unlike any that have been described fromEurope where relict cryogenic mounds are seen to be ramparted depressions not flat-topped mounds. The hypothesis of origin states, however, that an ice-walled lake wascreated when the original cryogenic mounds collapsed and it was into this lake that thesediments comprising the mounds were deposited. What became of the rampart that canbe seen associated with many collapsed contemporary pingos is not clear. Trimble(1978) described what he termed ‘pingo scars’ from southern North Dakota, USA. Thesefeatures were circular and consisted of an outer ridge, which had a moat on the innerside. In the centre there was a dome and the whole of the feature could have relief of upto 17 m.

Stone and Ashley (1992) described relict cryogenic mounds, which were some of thefirst features documented from North America (in the states of Connecticut andMassachsetts, USA) which were similar geomorphologically to the relict cryogenicmounds of north west Europe (see also Stone and Ashley, 1989 and Stone et al., 1991).These circular and sub-circular shallow depressions with subtly raised rims occur on a

220 GURNEY

drained glacial-lake bottom. Marsh (1987) outlined the evidence for relict cryogenicmounds in Pennsylvania, USA that here take the form of elliptical basins surrounded bylow ramparts. At one site (Halfway Run) 55 well-preserved remnants occur in a cluster.These features were interpreted as the remains of hydraulic pingos. A field ofdepressions similar to those at Halfway Run in Snyder County, however, were thought tobe reminiscent of “mineralogical palsas” as described by Pissart (1983b) rather thanrelict pingos as described by Watson (1971).

4 . The use of modern analogues

The morphology and sedimentology of specific relict cryogenic mound sites in the UKhas generally determined the choice of a modern analogue, except in some cases whereparticular analogues were chosen more because they had been previously employed forother sites. Once a modern analogue has been chosen it is often used as a source ofinformation about possible geomorphological processes that would have operated duringthe life-cycle of the, now relict, features. The modern analogues are also used, ofcourse, as a source of information concerning the palaeoclimatic significance of therelict features. Two large assumptions are made by the use of modern analogues: firstlythat the growth and decay processes of the modern analogues are relevant to the relictfeatures in question and secondly that the modern analogues are actually active in theirpresent climatic setting. It is clear that in the past inappropriate modern analogues forrelict cryogenic mounds have been employed on the basis of one or even both of theseassumptions. This has led to misleading palaeoclimatic inferences being drawn.

4.1. PINGO ANALOGUES

Most commonly, hydraulic (open system) pingos have been taken to be the mostappropriate modern analogues for relict cryogenic mounds in the UK. This is notsurprising since classically hydraulic pingos occur in lower valley side and valley bottomlocations, in clusters which are mutually interfering (see Figure 4), characteristics whichare shared with many of the relict cryogenic mounds sites. The term ‘open system’ wascoined by Müller (1959) and the type site for these features is in east Greenland,although good examples are also found in central Alaska (e.g. Holmes et al., 1968).Importantly the key to understanding the growth of these features is often a knowledgeof the local groundwater movement (see Worsley and Gurney, 1996; Gurney, 1998).The groundwater upwelling, which is sometimes called the ‘pingo spring’, is the sourceof the water, which forms the pingo ice-cores, and it is the varying location of upwelling,which causes the clustering of individual forms. Occasionally a pingo will developwithin the depression created by the collapse of a previous pingo, which essentiallycreates ‘a pingo within a pingo’. There are some contemporary pingos whose specificgenesis does not appear to fit neatly into the hydrostatic/hydraulic classification scheme(e.g. Pissart, 1967; Gurney and Worsley, 1997). Such features may be more suitablemodern analogues than the classic or ‘type’ features that are more often employed.


4.2. PALSA ANALOGUES

Modern analogues of the palsa variety have also been cited for the relict cryogenicmounds of the UK. It should be made clear, however. that this is far more problematicthan employing a pingo analogue because palsa terminology is arcane, even misleading,and the definitions employed are generally morphological as opposed to genetic innature (see Nelson et al., 1992). It is not the intention here to attempt to clarify theterminological debate, but rather to indicate what the use of a palsa analogue means forinterpretation and palaeoclimatic reconstruction.

222 GURNEY

When a palsa analogue is invoked for relict cryogenic mounds in the UK, or indeedelsewhere, this is usually because it is believed that a segregated ice core fed bycryosuction was responsible for the growth and maintenance of the original mounds asopposed to a more massive ice core fed primarily by injection ice from water underpressure (whether hydrostatic or hydraulic). It is clearly not employed because it isthought that the mounds were composed of peat or were developed in a mire, or indeedhad even the merest covering of peat. Therefore, for some, the use of a palsa typeanalogue is an inappropriate use of the terminology. A more pragmatic view, however,is that many relict cryogenic mounds, when active, had more in common genetically withpalsas than with pingos and therefore a palsa analogue is more appropriate.

Very often work conducted on relict cryogenic mounds in north west Europe hasinspired the search for suitable modern analogues in present cold climate environments(e.g. Worsley et al., 1995). In the case of the re-interpretation of the relict cryogenicmounds in Belgium, a need was created for an examination of sites, which might formappropriate modern analogues, and these were forthcoming (e.g. Gangloff and Pissart,1983; Pissart and Gangloff, 1984). More recently, investigations in cold climate areashave realised the requirement for more detailed morphological and sedimentologicalcomparison of contemporary and relict cryogenic mounds. For example, Matthews et al.(1997) specifically relate the evidence of cyclic palsa growth and decay at altitude insouth central Norway to relict (Pleistocene) features in the UK. Given the nature ofanalogue use for palaeoclimatic reconstruction thus far, such studies conducted in coldclimate areas of a lower latitude (62°N) are, perhaps, more valuable than the arctic andhigh-arctic sites often cited in the past.

5. Alternative interpretations of depressions

The alternative origins of any features that might be used to infer former climaticconditions should always be considered to ensure that the interpretation is as sound aspossible. All too often features which are used as palaeoclimatic indicators will be takenup by other workers who are not familiar with their geomorphology and this use ofsecondary level information can perpetuate the use of features for palaeoclimaticreconstruction long after they have been re-interpreted or even falsified. Prince (1964)produced perhaps the most extensive piece of work on the depressions and pits ofNorfolk, in which a periglacial origin was considered (see also Prince, 1961). The mainbody of the paper, however, was concerned with the possible anthropogenic origin ofthese depressions. The possibility of mineral workings and marl pits being the two mostobvious modes of origin. The close juxtaposition of the pits with features with anobvious periglacial origin (e.g. patterned ground) is always a problem since it oftenleads one to believe that their existence might be linked. The comprehensive discussionof the alternative origins of the pits, however, forces one to consider the otherpossibilities more fully. This is just as it should be and this paper should remind us all


that although in the literature of north west Europe there are numerous descriptions ofdepressions where a periglacial interpretation is enticing, this is not necessarily thecorrect interpretation. In conclusion Prince states that in Norfolk seemingly identicalhollows in terms of basic morphology cannot possibly be all attributed to the sameorigin, anthropogenic, periglacial or otherwise. The hollows are simply too numerousand are much more complex than is immediately apparent.

6. Palaeoclimatic inferences

The palaeoclimatic inferences of relict cryogenic mounds are based purely on the currentclimatic conditions observed that affect the contemporary cryogenic mounds that arebelieved to represent their modern analogues (see Pissart, 1987). Thus if a particular setof relict cryogenic mounds are interpreted as relict hydraulic pingos, then the climaticregime typical of hydraulic pingos (a MAAT of -3 to -7°C which can sustaindiscontinuous or continuous permafrost) will be applied. In this way the palaeoclimaticsignificance of relict cryogenic mounds is wholly reliant on correct interpretation and theappropriate use of modern analogues. If the features are incorrectly interpreted, or ifindeed there are no suitable modern analogues, then the palaeoclimatic inferences willbe poor or even misleading. Features originally interpreted as the remains of hydraulicpingos and now believed to be more closely related to the former existence of mineral-cored palsas have suffered a major change in their palaeoclimatic significance (from aninferred MAAT of -7°C or less to perhaps only -1 °C).

Compounding this problem is the fact that many if not most periglacial features rely on anumber of climatic and environmental parameters for their growth and/or maintenance.These other climatic parameters are generally not referred to and only the formertemperature regimes are considered. The number of freeze-thaw cycles or snow depth,for example, is not included in palaeoenvironmental reconstructions.

An additional source of error is that it is usually the mean temperatures, which areinferred, as opposed to the lows or indeed the seasonal temperature profile (which wouldclearly have been markedly different in mid-latitude periglacial areas compared to thatof high latitude periglacial areas). With cryogenic mounds the ice core is fully withinthe permafrost, well below the active layer, and therefore, mean temperatures areprobably not such a poor way of characterising the thermal regime. Even here, however,these mean temperatures are still those of the air as opposed to the ground (mean annualair temperatures as opposed to mean annual ground temperatures, MAGT), which is,perhaps, the more significant problem with all works of palaeoclimatic reconstruction.For this reason, it has been suggested that the palaeoclimatic inferences are too low sincethe MAGT is usually higher than of the MAAT by 1 to 5°C because in winter snowprotects the ground from the extreme cold and in summer the ground is warmed more bysolar radiation than the air (Williams, 1975). Yet again the interpretation of the MAGT

224 GURNEY

is a generalisation, which ignores other factors as described earlier (albedo, vegetationcover, snow etc).

7. Problems with the correct identification of relict cryogenic mounds

At most of the sites it has been the morphology of the relict cryogenic mounds, whichhas primarily been used to interpret the genesis of these features. The sedimentologicalcomposition of the features has been used less in their interpretation either because itssignificance has been unclear or because the sediments have largely been inaccessibledue to the lack of sections (natural or artificial). Whilst coring has been employed it haslargely been inconclusive. The obvious exception to this, of course, is the work ofPissart on the relict cryogenic mounds of Belgium, where good sections of the sedimentshave been studied.

In general, the sediments of relict cryogenic mounds in the UK can be divided into fourbasic groups:

1)2)3)

4)5)

6)7)

the rampart sediments,the intra-rampart sediments,which should reflect the general surficial deposits of the

surrounding area and therefore, are very often composed of slopedeposits, which typically mantle the lower valley sides and valleybottom.

the upper basin sediments (where present)where present are generally organic in nature and are commonly

composed of peat. At various sites where this peat has been studiedit has provided a younger age limit for the features (very often avery much younger age limit than that postulated for the date of themound collapse), and

the lower basin sedimentswhich should compose the depression or ‘pingo fill’ and are very

often composed of fine-grained material.

Any interpretation of relict cryogenic mounds should be based upon the morphology andsedimentology of the features in question. The genetic hypotheses proposed shouldaccount for the location and sediment composition and should include the processes ofcollapse and also the evidence that might survive the collapse. It is important toremember that contemporary cryogenic mounds have a growth and decay cycle whichoperates independently of climatic change and therefore some of the relict cryogenicmounds may be the result of decay during the colder climatic phase rather then decaybecause of the end of the colder climatic phase (the climatic amelioration). Even incollapse, the variety of cryogenic mounds operate differently, for example, pingos tendto collapse from the summit down, whereas palsas generally suffer rupture of thematerial insulating the core at the side low down or even at the base of the mound.


Relatively little is known about the growth mechanisms of relict cryogenic mounds,except perhaps, that once established they appeared to grow laterally as well as,presumably, vertically. This is far more similar to that of palsas than pingos as indicatedpreviously.

The growth mechanism is extremely important to understand. In pingos it allowsdevelopment in a wide range of sediments from bedrock through coarse gravels to fine-grained material. In palsas, however, there is a universal requirement for a frostsusceptible, and thus fine-grained substrate, which cans either be of peat or mineralsediments, and it is only in such a medium that cryosuction and the growth of thesegregated ice core can develop. Thus any study of the composition of relict cryogenicmounds should evaluate which sediments sustained the ice core (segregated orotherwise). If a fine-grained sediments body is identified then there is the possibilitythat cryosuction is involved, if there is not then other processes may have contributed tothe growth of the ice core.

The relict cryogenic mounds of the Welsh borders and of west Wales (Gurney, 1995;Gurney and Worsley, 1996) have a large volume of fine-grained sediments within theircentral depressions. It has been postulated that this sediment existed prior to thecryogenic mound growth and, in fact, facilitated that growth acting as a frost susceptiblesubstrate within which cryosuction could operate. Previous interpretations have placedthis sediment as accumulating from the deposition of material into the ‘pingo pond’, thatis, the pond that often develops in the initial crater that develops in the summit of thepingo when it begins to collapse. This material, would, in part, be derived from therampart deposits, however, very often there is no evidence for the coarse-grained slopedeposits, which form the ramparts, in the basin fill.

8. Conclusions

Clearly, the existence of relict cryogenic mounds in the UK has often been used to inferpast climate (specifically for the Late Quaternary). Their almost unique status inproviding information about the thermal state of the ground (i.e. type of permafrostnecessary to create and /or maintain them) is responsible for this. It remains, however,that these features are not as straightforward to identify as was once thought and itshould now be apparent that there are a range of analogues, which may be applied totheir interpretation. Depending upon the choice of analogue, a very wide range ofpalaeoclimatic/palaeoenvironmental inferences may be drawn. It is important that werealise the limitations of using geomorphological features for such palaeoclimaticreconstructions. Such features are only preserved in favourable, discrete locations andthe temperature inferences are not clear cut given the confusion over current airtemperatures and the actual importance of ground temperatures. Clearly more work isneeded to investigate the internal structure and sedimentology of these features in order

226 GURNEY

so that the most appropriate analogue can be applied (if one is available) and muchthought must be used when deciding on what exactly the climatic inferences might be.

Acknowledgements

The author would like to thank Graham Elliott, Jukka Käyhkö and PeterWorsley for field assistance and for many useful discussions concerning theevidence for relict cryogenic mounds in the UK and elsewhere. Thanks also toHeather Browning for creating the figures from the original tortured diagrams.Finally the comments of an anonymous referee helped to clarify the text in anumber of places.

References

Åhman, R. (1976) The structure and morphology of minerogenic palsas in northern Norway, BiuletynPeryglacjalny 26, 25-31.

Ballantyne, C.K. and Harris, C. (1994) The periglaciation of Great Britain, Cambridge University Press,Cambridge, 330pp.

Bastin, B., Juvigné, E., Pissart, A. and Thorez, J. (1974) Étude d’une coupe dégagée à travers un rempart d’unecicatrice de pingo de la Brackvenn, Annales de la Société Géologique de Belgique 97, 341- 358.

Berry, F.G.Q. (1979) Late Quaternary scour-hollows and related features in Central London, Journal ofEngineering Geology 12, 9-29.

Bijlsma, S. (1983) Weichselian pingo remnants (?) in the eastern part of the Netherlands, Proceedings of the 4thInternational Conference on Permafrost (National Academy Press, Washington DC), 62-67.

Bout, P. (1986) Des cicatrices de pingos en Devès (Haute Loire), France, Biuletyn Peryglacjalny 31, 17-26.Boyé, M. (1957) Clots, laguées et lagunes de la Lande girondine. Comptes Rendus, Académie des Sciences, Paris

244, 1058-1060.Bryant, R.H. and Carpenter, C.P. (1987) Ramparted ground ice depressions in Britain and Ireland’ in

Boardman, J. (Editor) Periglacial processes and landforms in Britain and Ireland, (CambridgeUniversity Press, Cambridge), 183-190.

Bryant, R.H., Carpenter, C.P. and Ridge, T. (1985) Pingo scars and related features in the Whicham Valley,Cumbria’ in Boardman, J. (Editor) Field-guide to the periglacial landforms of northern England (QuaternaryResearch Association, Cambridge), 47-53.

Cailleux, A. (1957) Les mares du sud-est de Sjaelland (Danemark), Comptes Rendus, Académies des Sciences,Paris 245, 1074-1076.

Carpenter, C.P. and Woodcock, M.P. (1981) A detailed investigation of a pingo remnant in western Surrey,Quaternary Studies 1, 1-26.

Coxon, P. (1986) A radiocarbon dated early postglacial pollen diagram from a pingo remnant near Millstreet,County Cork, Irish Journal of Earth Sciences 8, 9-20.

Coxon, P. and O’Callaghan, P. (1987) The distribution and age of pingo remnants in Ireland, in Boardman,J.(Editor) Periglacial processes and landforms in Britain and Ireland, (Cambridge University Press,Cambridge), 195-202.

De Gans, W. (1982) Location, age and origin of pingo remnants in the Drentsche Aa Valley area (TheNetherlands), Geologie en Mijnbouw 61, 147-158.

De Gans, W. (1988) Pingo scars and their identification in Clark, M. J. (Editor) Advances in periglacialgeomorphology John Wiley and Sons, Chichester, 299-322.


De Gans, W. and Sohl, H. (1981) Weichselian pingo remnants and permafrost on the Drenthe Plateau (TheNetherlands), Geologie en Mijnbouw 60, 447-452.

Dylik, J. (1965) L’étude de la dynamique d’évolution des depressions fermées à Jozefow aux environs de Lødz,Revue de Géomorphologie Dynamique 15, 158-171.

Flemal, R.C. (1976) Pingos and pingo scars: their characteristics distribution and utility in reconstructing formerpermafrost environments, Quaternary Research 6, 37-53.

Flemal, R.C., Hinkley, K.C. and Hesler, J.L. (1969) Fossil pingo field in north-central Illinois, GeologicalSociety of America Abstracts with programs 6, 16.

Flemal, R.C., Hinkley, K.C. and Hesler, J.L. (1973) The Dekalb mounds; a possible Pleistocene (Woodfordian)pingo field in north-central Illinois, Geological Society of America Memoir 136, 229-250.

Gangloff, P. and Pissart, A. (1983) Evolution and géomorphologique et palses minérales prés de Kuujjuaq (FortChimo, Québec), Bulletin de la Société Géographique de Liège 19, 119-132.

Gurney, S.D. (1995) ‘A reassessment of the relict Pleistocene ‘pingos’ of west Wales: hydraulic pingos ormineral palsas, Quaternary Newsletter 77, 6-16.

Gurney, S.D. (1998) Aspects of the genesis and geomorphology of pingos: perennial permafrost mounds,Progress in Physical Geography 22, 307-324.

Gurney, S.D. and Worsley, P. (1996) Relict cryogenic mounds at Owlbury, near Bishop’s Castle, Shropshire,Mercian Geologist 14, 14-21.

Gurney, S.D. and Worsley, P. (1997) Genetically complex and morphologically diverse pingos in the Fish Lakearea of south west Banks Island, NWT, Canada, Geografiska Annaler 79A, 41 -56.

Hinkel, K.M. (1988) Frost mounds formed by degradation at Slope Mountain, Alaska, USA, Arctic and AlpineResearch 20, 76-85.

Holmes, G.W., Hopkins, D.M. and Foster, H.L. (1968) Pingos in central Alaska, United States Geological SurveyBulletin 1241-H, 40pp.

Hutchinson, J.N. (1980) Possible late-Quaternary pingo remnants in central London, Nature 284, 253-255.Hutchinson, J.N. (1991) Periglacial and slope processes, in Forster, A., Culshaw, M.G., Cripps, J.C., Little, J.A.

and Moon, C.F. (Editors) Quaternary Engineering Geology, Geological Society Engineering Geology SpecialPublication 7, 283-331.

Juvigné, E. and Pissart, A. (1979) Un sondage sur le plateau des Hautes Fagnes au Lieu-dit ‘La Brackvenn,Annales de la Société Géologique de Belgique 102, 277-284.

Lagerbäck, R. and Rodhe, L. (1986) Pingos and palsas in northernmost Sweden - preliminary notes on recentinvestigations, Geografiska Annaler 68A, 149-154.

Mackay, J.R. (1979) Pingos of the Tuktoyaktuk peninsula area, North West Territories, Géographie physique etQuaternaire 33, 3-61.

Mackay, J.R. (1998) Pingo growth and collapse, Tuktoyaktuk peninsula area, western arctic coast, Canada: Along-term field study, Géographie physique et Quaternaire 52, 271-323.

Maarleveld, G.C. (1965) Frost mounds, a summary of the literature of the past decade, Mededelingen van derGeologische Stichting 17, 7-20.

Marsh, B. (1987) Pleistocene pingo scars in Pennsylvania, Geology 15, 945-947.Matthews, J.A., Dahl, S.O., Berrisford, M.S. and Nesje, A. (1997) Cyclic development and thermokarstic

degradation of palsas in the mid-alpine zone at Leirpullan, Dovrefjell, southern Norway, Permafrost andperiglacial processes 8, 107-122.

Miller, D. J. (1990) Relict periglacial phenomena within the Tamar basin, west Devon and east Cornwall; theirsignificance with regard to Quaternary environmental reconstruction, Unpublished PhD Thesis, Universityof Exeter, 721pp.

Mitchell, G.F. (1971) Fossil pingos in the south of Ireland, Nature 230, 43-44.Mitchell, G.F. (1973) Fossil pingos in Camaross Townland, County Wexford, Proceedings of the Royal Irish

Academy B73, 269-282.Mitchell, G.F. (1977) Periglacial Ireland, Philosophical Transactions of the Royal Society of London B280, 199-

209.Mückenhausen, E. (1960) Eine besondre art von pingos aun Hohen Venn/Eifel, Eiszeitalter und Gegenwart 11, 5-

11.Müller, F. (1959) Beobachtungen uber pingos, Meddelelser om Grønland 153, 127pp.Nelson, F.E., Hinkel, K.M. and Outcalt, S.I. (1992) Palsa-scale frost mounds in Dixon, J.C. and Abrahams, A.D.

Periglacial geomorphology, Wiley, Chichester, 305-325.

228 GURNEY

Paris, F.P., Cleveringa, P. and De Gans, W. (1979) The Stokersdobbe: geology and palynology of a deep pingoremnant in Friesland (The Netherlands), Geologie en Mijnbouw 58, 33-38.

Picard, K. (1961) Reste von pingos bei Husum/Nordsee, Naturw Ver Schleswig-Holstein 32, 72-77.Pissart, A. (1956) L’origine périglaciaire des viviers des Hautes Fagnes, Annales de la Société Géologique de

Belgique 79, 119-131.Pissart, A. (1958) Les dépressions fermées dans le région Parisienne. Le problème de leur origine, Revue

Géomorphologie Dynamique 9, 73-83.Pissart, A. (1963) Les traces de pingos du Pays de Galles (Grande Bretagne) et du Plateau des Hautes Fagnes

(Belgique), Zeitschrift für Geomorphologie 7, 381-392.Pissart, A. (1965) Les pingos des Hautes Fagnes; les problèmes de leur genese, Annales de la Société

Géologique de Belgique 88, 277-289.Pissart, A. (1967) Les pingos de L’ile Prince-Patrick (76°N - 120°W), Geographical Bulletin 9, 189-217.Pissart, A. (1974) Les viviers des Hautes Fagnes sont des traces de buttes périglaciaires; mats s’agisaitil

réelement de pingos?, Annales de la Société Géologique de Belgique 97, 359-381.Pissart, A. (1983a) Remnants of periglacial mounds in the Hautes Fagnes (Belgium): structure and age of the

ramparts, Geologie en Mijnbouw 62, 551-555.Pissart, A. (1983b) Pingos et palses: un essai de synthése des connaissances actuelles, Abhandlungen der

Academie der Wissenschaften in Mathematische-Physikalische Klasse, Dritte Folge 35,48-69.

Pissart, A. (1985) Pingos et palses: un essai de synthése des connaissances actuelles, Inter-nord 17, 21-32.Pissart, A. (1987) Weichselian periglacial structures and their environmental significance: Belgium, the

Netherlands and northern France, in Boardman, J. (Editor) Periglacial processes and landforms inBritain and Ireland, Cambridge University Press, Cambridge, 77-85.

Pissart, A., Bastin, B. and Juvigné, E. (1975) L’origine des viviers des Hautes Fagnes: traces de pingos ou depalses?, Extrait de Hautes Fagnes 1975 1, 9-38.

Pissart, A. and Gangloff, P. (1984) Les palses minérales et organiques de la vallée de l’Aveneau, prés deKuujjuaq, Québec subarctique. Géographie physique el Quaternaire 38, 217-228.

Pissart, A. and Juvigné, E. (1980) Génese et âge d’une trace de butte périglaciaire (pingo ou palse) de laKonnerzvenn (Hautes Fagnes, Belgique), Annales de la Société Géologique de Belgique 103, 73-86.

Ploeger, P. L. and Groenman-Van Waateringe, W. (1964) Late Glacial pingo development in the Boorne regionnear Wijnjeterp, Province of Friesland, Netherlands, Biuletyn Peryglacjalny 13, 199-223.

Prince, H.C. (1961) Some reflections on the origin of hollows in Norfolk compared with those in the Paris region,Revue de Géomorphologie Dynamique 12, 110-117.

Prince, H.C. (1964) The origin of pits and depressions in Norfolk, Geography 49, 15-32.Railton, J.B. and Sparling, J.H. (1973) Preliminary studies on the ecology of palsas mounds in northern

Ontario, Canadian Journal of Botany 51, 1037-1044.Rousset, C. (1965) Traces de pingos sur les formations volcaniques du Massif Central Français, Comptes

Rendus, Académie des Sciences, Paris 261, 4461-4463.Seppälä, M. (1994) Snow depth and controls on palsa growth, Permafrost and periglacial processes 5, 283-288.Slotboom, R.T. (1963) Comparative geomorphological and palynological investigation of the pingos (viviers) in

the Hautes Fagnes (Belgium) and Mardellen in the Gutland (Luxembourg), Zeitschrift für Geomorphologie 7,193-231.

Sparks, B.W., Williams, R.B.G. and Bell, F.G. (1972) Presumed ground-ice depressions in East Anglia,Proceedings of the Royal Society of London A327, 329-343.

Stone, J.R. and Ashley, G.M. (1989) Fossil Pingos? : New evidence for permafrost in southern New Englandduring Late Wisconsinan deglaciation, Geological Society of America Abstracts with Programs 21, 69.

Stone, J.R. and Ashley, G.M. (1992) Ice wedge casts, pingo scars and the drainage of glacial Lake Hitchcock, inRobinson, P. and Brady, J.B. (Editors) 84th New England Intercollegiate Geological Conference, Guidebookfor fieldtrips in the Connecticut valley region of Massachusetts and adjacent states, University ofMassachusetts, Department of Geology and Geography 66, A-7, 305- 331.

Stone, J.R. and Ashley, G.M. and Peteet, D.M. (1991) Cross section through a post-lake Hitchcock surfacedepression: new 14C dates and evidence for periglacial origin, Geological Society of America Abstracts withPrograms 23, 135.

Taylor, J. (1987) Timescales of environmental change - an inaugural lecture, Wednesday 6th May, 1987University College of Wales, Aberystwyth, 45pp.


Trimble, D.E. (1978) Pingo scars in southwestern North Dakota, United States Geological Survey ProfessionalPaper 1100, 58.

Trotman, D.M. (1963) Peat deposits within a pingo remnant near Llangurig, Wales, Zeitschrift fürGeomorphologie 7, 168-171.

Van der Meulen, S. (1988) The spatial facies of a group of pingo remnants on the southeast Frisian till plateau(the Netherlands), Geologie en Mijnbouw 67, 61-74.

Washbum, A.L. (1983a) Permafrost features as evidence of climatic change, Earth Science Reviews 15, 327-402.Washbum, A.L. (1983b) Palsas and continuous permafrost, Proceedings of the 4th International Conference on

Permafrost, 1372-1377.Watson, E. (1971) The remains of pingos in Wales and the Isle of Man, Geological Journal 7, 381-392.Watson, E. (1972) Pingos in Cardiganshire and the latest ice limit, Nature 236, 343-344.Watson, E. and Watson, S. (1972) Investigation of some pingo basins near Aberystwyth, Wales, Report of the

24th International Geological Congress {Montreal) Section 12, 212-233.Watson, E. and Watson, S. (1974) Remains of pingo in the Cletwr basin, south-west Wales, Geografiska

Annaler 56A, 213-225.Wiegand, G. (1965) Fossile pingos in Mitteleuropa, Wurzburger Geographische Arbeiten 16, 152pp.Williams, P.J. and Smith, M.W. (1989) The frozen Earth. Fundamentals of Geocryology, Cambridge University

Press, Cambridge, 306pp.Williams, R.B.G. (1975) The British climate during the last glaciation ; an interpretation based upon periglacial

phenomena, in Wright, A. E. and Moseley, F. (Editors) Ice ages ancient and modern. Seel House Press,Liverpool, 95-120.

Worsley, P. and Gurney, S.D. (1996) Geomorphology and hydrogeological significance of the Holocene pingos inthe Karup Valley area, Traill Island, northern east Greenland, Journal of Quaternary Science 11, 249-262.

Worsley, P., Gurney, S.D. and Collins, P.E.F. (1995) Late Holocene ‘mineral palsas’ and associated vegetationpatterns: a case study from Lac Hendry, northern Québec, Canada and significance for European Pleistocenethermokarst, Quaternary Science Reviews 14, 179-192.

INVESTIGATIONS INTO LONG-TERMFUTURE CLIMATE CHANGES

P.E. BURGESS, J.P. PALUTIKOF*AND C.M. GOODESSClimatic Research Unit, University ofEast Anglia, Norwich NR4 7TJ, UK* Corresponding author

Abstract

A two-dimensional climate model (LLN-2D) has been used to investigate long-termclimate change (over periods of around years in the past and in the future) forCentral England. The model is forced by periodic changes in the distribution ofincoming radiation associated with periodic changes in the Earth’s orbit around the Sunand by changes in the atmospheric concentration of Future natural changes inconcentration are estimated using a regression equation. Eight anthropogenicscenarios have been constructed. The LLN-2D simulations indicate that anthropogeniceffects have the potential to disturb the climate system over very long time scales anddemonstrate the non-linearities that can operate between cause and effect.

1. Introduction

Over long time scales, defined here as those greater than years, climate change isof primary importance in determining the development of local and regional landformsand ecosystems. Thus, an understanding of climate change is essential in order tounderstand landform changes in the British Isles over the last two million years (theQuaternary period). It is also important for studies of long-term future development oflandforms, such as those required for safety assessments of underground radioactivewaste disposal in the UK (Thorne, 1995; Nirex, 1997). The research presented in thischapter has been funded by United Kingdom Nirex Limited and British Nuclear Fuelsplc. A knowledge of the types and sequences of climatic states likely to be experiencedin the UK over the next glacial-interglacial cycle (approximately the next 100,000 years(100 ka)) is relevant to quantitative radiological assessment studies in the UK.

It is widely, though not universally, accepted that the major cause of the glacial-interglacial cycles observed over the Quaternary period are periodic changes in theseasonal and latitudinal distribution of incoming radiation associated with periodic

231


232 BURGESS, PALUTIKOF AND GOODESS

changes in the Earth's orbit around the Sun (see Goodess et al., 1992a, b; Adcock et al.,1997; Goodess et al., 1999 and Palutikof et al., 1999 for reviews of orbital forcing, orMilankovitch, theory). The changes in insolation due to periodic variations in theEarth’s orbital parameters of obliquity, precession and eccentricity are predictable. Thisraises the possibility of modelling climate changes due to orbital forcing not only overlong periods in the past, but also into the future. A two-dimensional climate model hasbeen developed at the L'Institut d'Astronomie et de Géophysique de Georges Lemaître atthe Université Catholique de Louvain, Belgium (Gallée et al., 1991) which is able toestimate parameters such as temperature and ice sheet volumes at 1000 year time stepsand over 5° latitudinal bands for the Northern Hemisphere. The forcing inputs to themodel are insolation and atmospheric concentration at each time step.

The Louvain-la-Neuve model, referred to here as LLN-2D, has been used to investigatelong-term climate change (over periods of around years in the past and in the future)for central England (Burgess, 1998; Goodess et al., 1999). These modelling studies arethe subject of this chapter. First, we describe the characteristics of the LLN-2D model inSection 2. Second, the insolation and atmospheric forcing inputs due to naturalvariations are outlined in Section 3. Over the time scales of interest here,anthropogenically-induced changes in greenhouse-gas concentrations, principallyare also considered important (Goodess et al, 1992a). Thus, the development ofscenarios of anthropogenic changes in atmospheric concentrations over the next150 ka is described in Section 4. The LLN-2D model generates output (temperature andice volumes) for seven surface types (see Section 2) in 5° latitude bands. In order toobtain results applicable to a particular region, here central England, a method fordownscaling model output is required. A rule-based approach to downscaling, in whichthe temperature time series from the model are divided into a succession of discreteclimate states, is proposed in Section 5. Eight LLN-2D simulations using both naturaland anthropogenic forcing, together with insolation forcing, have been completed.The results from these simulations are summarized in Section 6, and compared withthose from a simulation using only natural and insolation forcing. Finally, thepotential applications and limitations of these modelling studies are discussed in Section7.

2. The LLN-2D Model

To achieve simulations over the time scales of interest here, any model must besimplified in comparison to the fully three-dimensional models used for climate changepredictions on time scales of a few hundred years. However, for the regional level ofdetail required, it is also essential that the model has greater resolution than the simpleone-dimensional models used for energy balance calculations. LLN-2D is described as2.5 dimensional because of the altitudinal, latitudinal and sectoral differentiationemployed for the surface models (for the Northern Hemisphere only). To be moreprecise, the model has components for land (snow covered and snow free), sea (openand sea-ice covered), and the North European, North American and Greenland ice

INVESTIGATIONS INTO LONG-TERM FUTURE CLIMATE CHANGE 233

sheets. The ice sheet sub-model uses a vertically integrated equation for ice massbalance which assumes plastic flow (Gallée et al., 1991). There are parameterizations toaccount for lateral calving from ice sheets, snow ageing processes, albedo effects oftaiga/tundra line shifts and terms for extension of sea ice into relatively warm waters.The sub-models in LLN-2D encompass many important interactions of the Earth-atmosphere-cryosphere systems.

The atmospheric model has a time stepping interval of three days for each month,synchronously passing the mid-monthly climatologies to the oceanic and land sub-models. This process is repeated at 1000 year intervals, at which point information isfed to the ice sheet model. LLN-2D is particularly powerful in the way that it calculatesexplicit snow mass balances, as opposed to the snow line parameterizations employed inmany other models.

It is possible to use values from ocean cores to give a measure of global ice volumevariation over time scales of the order of to years (Duplessy et al., 1988). LLN-2D has been shown to reproduce well the low frequency variations in ice volumethroughout the Late Quaternary and Holocene (125 ka Before Present to present, i.e. thelast glacial-interglacial cycle) when assumptions are made about Antarctic ice volumevariations over that period (Gallée et al., 1992). Furthermore, the model is known toreproduce present-day mid-latitude temperatures particularly well (Burgess, 1998).

The strengths of LLN-2D derive from the acceptable computational power required forintegration, representation of fast and slow feedback mechanisms and coupling of themajor elements of the Earth-atmosphere-cryosphere systems. Weaknesses of the modelinclude relatively crude spatial and temporal scales of resolution, a parameterizedhydrological scheme based on observed present-day relationships, and a single basinocean that cannot reproduce the thermohaline circulation.

3. Long-term Natural Changes in Climate Forcing

The aim of this research has been to derive estimates of millennial scale climate changeover the next 150 ka. In order to do this, scenarios of future climatic forcing arerequired. As discussed above, LLN-2D is forced by changes in seasonal, latitudinal andlong-term variations in the insolation received at the top of the Earth’s atmosphere, andby changes in atmospheric concentrations. The first of these, insolation changes, arecalculated during model integration (Berger, 1978; Berger and Loutre, 1997), leavingonly the need for the specification of the atmospheric forcing.

During the past 125 ka there have been significant variations in atmospheric asrecorded in the gas bubbles trapped in ice sheets (Barnola et al., 1987; Jouzel et al.,1993). These gas inclusions provide snapshots of past atmospheres. A complex web ofcause and effect in the carbon cycle is responsible for variations in atmosphericconcentrations, as recorded in these bubbles. The causes of these variations involve


interactions between the atmosphere, the terrestrial and oceanic biospheres, shelf andocean floor sediments and the geosphere. They occur on all time scales from seasonalcycles to millions of years, the nature of their impact being dictated by such factors astemperature, nutrient availability, sea level, gas exchange at the ocean surface andvarious feedbacks (Knox and McElroy, 1985; Sundquist, 1990; Kier, 1993; Francois etal., 1997; see also reviews in Adcock et al., 1997; Goodess et al., 1999). Ideally,carbon cycle modelling would provide the necessary atmospheric concentrationsrequired by LLN-2D as a forcing. However, although individual components of thiscycle are understood in some detail, the present state of carbon cycle modelling is notable to capture fully the complexity of the interactions over the full range of the timescales involved.

In the study presented here, a statistical approach to estimation of futureconcentrations is taken. The basis for this methodology lies in the observation that,during the Late Quaternary and Holocene, atmospheric concentrations varied fromhigh values of around 280 ppmv during interglacial periods to only 190 ppmv duringglacial periods. These high and low periods coincided with times at which NorthernHemisphere high latitude summer insolation receipts were, respectively, high(interglacial periods) and low (glacial periods). It should therefore be possible to linkatmospheric concentration changes to spatio-temporal changes in insolationreceipts. A purely statistical methodology such as this does not address the causallinkages between insolation and change. It assumes that sufficient informationabout those linkages can be captured in a regression analysis that relatesconcentrations to a small number of variables characterising insolation.

Insolation time series at different latitudes and months throughout the past 125 ka(independent variables) were regressed using stepwise multiple regression against the

time series (dependent variable) obtained from ice cores over the same period.After development and validation, the resultant equation for prediction of atmospheric

concentrations over millennial time scales from insolation was found to containstrong signals from high latitude areas in winter months. This is in line with certaintheories regarding the carbon cycle, which state that high latitude oceanic productivity isa strong control on oceanic draw-down and storage of atmospheric (Knox andMcElroy, 1985).

Validation of the predicted time series showed that the main shortcomings wereduring the warming since the end of the Last Glacial Maximum, with predicted valuesconsistently below observed values. This period is interesting in that it represents aperiod of rapid environmental change typical of the end of a glacial period and verydifferent from the slow englaciation over tens of thousands of years that preceded it (andwhich occupies most of the time frame over which the regression was developed).Factors such as rapid inundation of continental shelves and burial of carbon havepreviously been suggested as important factors in the dynamics of the carbon cycleduring such periods of rapid change.


Bearing in mind the known shortcomings of the regression equation, it was initialisedusing insolation time series over the next 150 ka to derive a time series of atmospheric

change under natural conditions. Figure 1 shows predicted and observed (Jouzel etal., 1993) atmospheric concentrations over the past 125 and predicted values overthe next 150 ka. The predicted future forcing was used to force the LLN-2Dintegration NAT (i.e. natural changes in atmospheric concentrations only).

4. Derivation of Long-term Anthropogenic Forcing

The next step was to derive scenarios of anthropogenic changes in atmospheric overthe next 150 ka. Relatively little modelling has been performed on anthropogenic effectsover long time scales, two notable exceptions being the studies by Walker and Kasting(1992) and by Sundquist (1990). Results from the latter study were used here becausethey include terms to account for the various calcium reactions leading to precipitationor dissolution of carbonates in sea-floor sediments. First, two scenarios of the change inemissions were selected from Sundquist’s work, one representing high rates of fossil fuelburning, the other low rates. From these scenarios, time series of anthropogenic


atmospheric concentrations up to ~10 ka AP (after present) were derived. Then,from 10 ka AP until the end of the simulation period (150 ka AP), four differentscenarios of the decline in atmospheric concentrations were defined. These arebounding and intermediate cases thought to reasonably represent the rate at which thenatural carbon cycle can sequester anthropogenic These scenarios specify totalabsorption of anthropogenic atmospheric at 30, 50, 100 and 150 ka AP. In thisway, a total of eight anthropogenic time series were defined, (2 emissions scenariosx 4 absorption scenarios). These were added to the natural change time seriesdescribed in Section 3 to define scenarios for the future (as shown in Figure 2).ANTH1, ANTH3, ANTH5 and ANTH7 represent low initial emissions scenarios,ANTH2, ANTH4, ANTH6 and ANTH8 represent high initial emissions scenarios, withthe different rates of decay outlined above.


5. Downscaling the LLN-2D Model Results

Using the output from three-dimensional global climate models, there are two maintechniques for deriving detailed sub-grid scale climatologies. The first involves the useof a high resolution Regional Climate Model nested within the coarse resolution globalmodel, and driven using the global model grid-scale climatologies as boundaryconditions. The second technique uses statistically-derived relationships between gridscale and local climatologies. The zonally-averaged nature of the LLN-2D atmosphericmodel means that neither of these techniques are appropriate.

A rule-based approach to downscaling was adopted in the studies presented here(Burgess, 1998). In order to derive rules for downscaling LLN-2D zonal output tocentral England, relationships between climatic-cryospheric output from LLN-2D andobserved indices of climate were established over the last glacial-interglacial cycle. Theclimate indices were derived from various floral, faunal, limnological and pedologicalpalaeo-evidence from around England (Burgess, 1998). By following this procedure,thresholds can be defined at which climate switches from e.g. boreal to temperateconditions. One example of the rules thus established is that if the February zonal landtemperature in the latitude band 55-60°N is greater than -20.5°C then a temperate state isassigned to England. The full hierarchy of rules is outlined in Table 1. In Figure 3, weshow the time series of climate states for England over the last 125 ka based on observedpalaeo-evidence, and the time series derived by applying the rules.


Once climate states have been assigned for England, analogue stations for each state canbe selected from around the world (see Table 2 for the stations used). Themeteorological observations from these sites were used to describe the monthlytemperature and precipitation regimes for each state (Burgess, 1998). Means andstandard deviations across a number of analogue stations for any climatic state give anidea of the possible range of variability that could be expected under such conditions.


6. Results

The climate change modelling studies presented here, based on scenarios of both naturaland anthropogenic atmospheric concentrations, estimate the broad patterns of futureclimate change over central England up to 150 ka in the future. In our discussion of theresults, we concentrate on four of the eight concentrations scenarios, as follows:

NAT: natural concentrations, as predicted by the regression equation described inSection 3;ANTH1: low fossil fuel-based emissions, anthropogenic atmospheric concentrationsof tailing off to zero at 30 ka AP;ANTH7: low fossil fuel-based emissions, anthropogenic atmospheric concentrationsof tailing off to zero at 150 ka AP; and,ANTH8: high fossil fuel-based emissions, anthropogenic atmospheric concentrationsof tailing off to zero at 150 ka AP.


Figure 4 shows the results from the LLN-2D model (hemispheric ice volumes andJanuary and July temperatures) generated by these four scenarios. Figure 5 shows theassociated succession of climate states over central England. In all but the most extremeemissions scenario (ANTH8), there is a deterioration in climate leading to a maximum inhemispheric ice volume at between 55-60 ka AP. This glacial maximum is restricted inits severity in comparison to the Last Glacial Maximum. In England it is represented bya tundra environment, and only restricted upland ice formation.


The exact pattern of climate change over central England associated with the generalcooling toward this glacial period is strongly dependent on the scenario used. Forthe natural case, NAT, the present temperate climate is succeeded after 1000 yearsby boreal conditions that become steadily colder and dryer (i.e. changing from maritimeto continental boreal conditions) towards about 45 ka AP. Thereafter tundra conditionspersist for around 20 ka. The lower emissions ANTH scenarios, ANTH1 and ANTH7,show a period of enhanced warming until about 5 ka AP, then return to temperateconditions for a further 10 ka. Warmer boreal conditions are maintained well beyond 50ka AP, followed by a period of rapid cooling coinciding with ice growth and culminatingfor ANTH1 in a brief glacial event at roughly 60 ka AP. The short lived nature of thisglacial event implies that it is restricted to highland locations such as in the LakeDistrict. Beyond the restricted glacial event at 65 ka AP in ANTH1, there is increasingharmony between different model integrations, regardless of the scenario chosen.A return to boreal conditions around 75 ka AP is followed by gradual and sustaineddecline to tundra and then glacial climate by 100 ka AP.


ANTH8, the higher initial emissions scenario, shows a very different evolution ofclimate. The Northern Hemisphere remains largely ice free until nearly 100 ka AP, afterwhich there is a short period of very rapid cooling. Over central England, this isreflected in enhanced warming conditions which persist until 20 ka AP, a completeabsence of the glacial event seen in ANTH1 at 65 ka AP, and only a brief onset ofglacial conditions at around 105 ka AP. Similarities with other, less severe, ANTHscenarios do not really develop until after 120 ka AP.

It is clear from this research that anthropogenic effects have the potential to disturb theclimate system over very long time scales. A somewhat surprising result is the presenceof a short glacial event over central England in ANTH1 at 65 ka AP, which is notpresent in NAT. In ANTH1, complete Northern Hemisphere deglaciation is observed ataround 5 ka AP. However, following this state of deglaciation, ice volumes in ANTH1develop so that by 65 ka AP they are greater than in NAT (even though NAT neverdemonstrates complete deglaciation). The greater ice volume in ANTH1 is attributableto rapid growth of the Fennoscandian and Laurentide ice sheets starting at around 50 kaAP (growth of the Greenland ice sheet is limited due to its geographical setting). Thisresult is interesting because it suggests that ice growth in a 'post-anthropogenic' worldcould be greater than that which would result from a purely natural evolution of climate.

In order to discover the reasons for the additional ice growth in ANTH1, a detailed studyof model behaviour was carried out, focusing on the relevant time periods. In particularthe response of bedrock to ice loading and unloading in LLN-2D was investigatedbecause it was considered that bedrock 'memory' effects might contribute to theanomalous behaviour of ANTH1 ice volumes. Actual model bedrock deflections inresponse to ice loading (expressed as metres below sea level) were compared to theequilibrium bedrock deflection depths that would be expected for the ice volume present(i.e. that depth which would occur if the bedrock had sufficient time to respond). Theratio of the two gives a measure of the continuing, time lagged, response of the bedrockto present and previous ice loading conditions. At 65 ka AP, the forcing conditionsfavour the decay of ice sheets. In NAT, full bedrock depression has been reached andstrong rebound takes place, with associated high ice decay rates (due to lateral calvingand desert altitude effects). In ANTH1, however, bedrock depression at 65 ka AP is stillin disequilibrium with the ice loading and this means weaker isostatic rebound and henceslower rates of ice melt. A larger ice volume develops in ANTH1 when insolationconditions favour ice growth, because ablation is slower and hence ice build up proceedsat an accelerated rate.

7. Discussion

LLN-2D simulations based on natural changes, including the NAT simulationdescribed here, show the onset of the next glaciation at about 55-60 ka AP. The addition


of anthropogenic effects (simulations ANTH1-ANTH8) can, in the more extremescenarios, prevent this glaciation occurring, so that the date of the next glaciation isdelayed until around 110 ka AP. Anthropogenic forcing also has a major impact atthe start of the simulations, causing the Greenland ice sheet to disappear by 2 ka AP. Itdoes not reappear until just before 20 ka AP at the earliest. Future ice-sheetdevelopment is sensitive to both the maximum concentration and the decay rate ofanthropogenic and there are major uncertainties in attempting to constructemissions scenarios over such long future time periods. Nonetheless, the group ofANTH1-ANTH8 simulations demonstrate that anthropogenic impacts may severelydelay the onset, and restrict the extent of, the next glaciation. They also provide anappropriate envelope of possible futures that can be considered for radiologicalassessment purposes (Goodess et al., 1999).

In order to use LLN-2D output for radiological assessment studies, it is convenient tocharacterise the model time series as a sequence of discrete climate states (temperate,boreal, tundra, glacial, subtropical or enhanced warming). The climatic conditions likelyto be experienced in the British Isles during each of these states can then be representedusing instrumental data from selected analogue sites (although there are probably nogood analogues at the present day for the British Isles during a glacial state). Theanalogue sites used for central England are listed in Table 2.

These two steps form the basis of the rule-based downscaling technique which has beendevised to provide estimates of long-term future climates for central England based onoutput from the NAT and ANTH1-ANTH8 simulations. Severe limitations are imposedby the model configuration – for example, because the model is only two-dimensional itis necessary to assume in the downscaling that the climates of the British Isles will movein the same sense, and with broadly the same intensity, as the zonal climate. However,these limitations cannot be overcome until such time as computing power and speedallows the application of three-dimensional climate models to these problems of verylong-term climate change. The use of three-dimensional models might also allow awider range of climate forcing mechanisms to be explicitly modelled. The LLN-2Dsimulations are forced by insolation and only, whereas climate varies on all timescales in response to a whole suite of random and periodic forcing factors (Goodess etal., 1992b). Further limitations arise from the model’s necessarily simplifiedrepresentation of the Earth-atmosphere-cryosphere system (Burgess, 1998). Forexample, the single ocean basin cannot reproduce important effects such as thethermohaline circulation, precluding the modelling of events such as the Younger Dryas.

At the present time, however, the downscaled output from the LLN-2D simulationsprovides a valuable source of information on possible climate change over very longtime scales. The simulations demonstrate clearly the non-linearities that can operatebetween cause and effect, such that an anthropogenic-emissions scenario may at certaintimes exhibit greater ice sheet development than a natural-emissions scenario. They alsoshow that the effects of anthropogenic emissions on climate can persist long after thoseemissions have reduced to zero, a feature depending on the length of the ‘decay’ period


selected to return atmospheric concentrations of back to their natural levels. Untilsuch time as three-dimensional climate models can be used to simulate change over thevery long time scales under investigation here, the results from the simpler two-dimensional models usefully inform our view of how future climates may develop.

Acknowledgements

This work was funded by United Kingdom Nirex Limited as part of the Nirex SafetyAssessment Research Programme and by British Nuclear Fuels plc. The authors wish tothank A. Berger and M-F. Loutre of the Université Catholique de Louvain, Belgium formaking the LLN-2D palaeoclimate model available for use by Paul Burgess in his PhDstudies.

References

Adcock, S.T., Dukes, M.D.G., Goodess, C.M. and Palutikof, J.P. (1997) A Critical Review of the ClimateLiterature Relevant to the Deep Disposal of Radioactive Waste, Nirex Science Report S/97/009, UnitedKingdom Nirex Limited, Harwell.

Barnola, J.M., Raynaud, D., Korotkevich, Y.S. and Lorius, L. (1987) Vostok core provides 160,000-yearrecord of atmospheric Nature, 329, 408-414.

Berger, A. (1978) Long-term variations of daily insolation and Quaternary climate change, J. Atmos. Sci., 35,2362-2367.

Berger, A. and Loutre, M-F. (1997) Long-term variations in insolation and their effects on climate, the LLNexperiments, Surv. Geophys., 18, 147-161.

Burgess, P.E. (1998) Future Climatic and Cryospheric Change on Millennial Timescales: An AssessmentUsing Two-dimensional Climate Modelling Studies, PhD Thesis, University of East Anglia, Norwich.

Duplessey, J., Labeyrie, L. and Blanc, P. (1988) Norwegian Sea deep water variations over the last climaticcycle: Palaeooceanographic implications, in H. Wanner and U. Seigenthaler (eds.), Long and Short TermVariability of Climate, Earth Science Series, Springer Verlag, New York, pp. 83-116.

Francois, R., Altabet, M.A., Yu, E.F., Sigman, D.M., Bacon, M.P., Frank, M., Bohrmann, G., Bareille, G. andLabeyrie, L.D. (1997) Contribution of Southern Ocean surface-water stratification to low atmosphericconcentrations during the last glacial period, Nature, 389, 929-935.

Gallée, H., van Ypersele, J.P., Fichefet, T., Tricot, C. and Berger, A. (1991) Simulation of the last glacialcycle by a coupled, sectorially averaged climate-ice sheet model. 1 The climate model, J. Geophys. Res.,96, 13139-13161.

Gallée, H., van Ypersele, J.P., Fichefet, T., Marsiat, I., Tricot, C. and Berger, A. (1992) Simulation of the lastglacial cycle by a coupled, sectorially averaged climate-ice sheet model. 2. Response to insolation andvariations, J. Geophys. Res., 97, 15713-15740.

Goodess, C.M., Palutikof, J.P. and Davies, T.D. (1992a) Studies of Climatic Effects Relevant to DeepUnderground Disposal of Radioactive Waste, Nirex Report NSS/R267, United Kingdom Nirex Limited,Harwell.

Goodess, C.M., Palutikof, J.P. and Davies, T.D. (1992b) The Nature and Causes of Climate Change.Assessing the Long Term Future, Studies in Climatology Series. Belhaven Press, London.

Goodess, C.M., Watkins, S.J., Burgess, P.E. and Palutikof, J.P. (1999) Assessing the Long-term FutureClimate of the British Isles in Relation to the Deep Underground Disposal of Radioactive Waste, NirexReport, United Kingdom Nirex Limited, Harwell, in press.

Jouzel, J., Barkov, N.I., Barnola, J.M., Bender, M., Chappellaz, J., Genthon, C., Kotlyakov, V.M., Lipenkov,V., Lorius, C., Petit, J.R. el al. (1993) Extending the Vostok ice-core record of palaeoclimate to thepenultimate glacial period, Nature, 364, 407-412.


Kier, R. (1993) Are atmospheric content and Pleistocene climate connected by wind speed over a polarMediterranean sea?, Glob. Planet. Change, 8, 59-68.

Knox, F. and McElroy, M. (1985) Changes in atmospheric Influence of marine biota at high latitudes, J.Geophys. Res., 89, 4629-4637.

Nirex (1997) Nirex 97: An Assessment of the Post-closure Performance of a Deep Waste Repository atSellafield, Nirex Science Report S/97/012, United Kingdom Nirex Limited, Harwell.

Palutikof, J.P., Goodess, C.M., Watkins, S.J. and Burgess, P.E. (1999). Developments in long-term climatechange, Prog. Env. Sci., 1, 89-96.

Sundquist, E.T. (1990) Long-term aspect of future atmospheric and sea-level changes, in R. Revelle(ed.), Sea Level Change, National Research Council Studies in Geophysics, National Academy Press.Washington, D.C., pp. 193-207.

Thorne, M.C. (ed.) (1995) Nirex Biosphere Research: Report on Current Status in 1994, Nirex ScienceReport S/95/003, United Kingdom Nirex Limited, Harwell.

Walker, J.C.G. and Kasting, J.F. (1992) Effects of fuel and forest conservation on future levels of atmosphericcarbon dioxide, Palaeogeog., Palaeoclim., Palaoecol, 97, 151-189.

GEOMORPHOLOGICAL AND CLIMATOLOGICALPERSPECTIVES ON LAND SURFACE–CLIMATE CHANGE

DOMINIC KNIVETON and SUE McLAREN

The climate system can be considered to consist of the atmosphere, oceans, biosphere,cryosphere (ice and snow) and lithosphere. In this book we have focused on theinteractions between the first and last of these, the atmosphere and its long-termexpression, the climate, and the lithosphere or more specifically, the land surface. Thereare numerous and often highly complex linkages between climate and land surfaceswhich are still relatively poorly understood. As well as trying to understand thecomplex processes that operate today, there are the added difficulties in trying tounderstand how these processes may have operated and changed in the past as well astrying to determine how they may change in the future.

Improvements in our understanding of how climates have changed in the past usingpalaeoenvironmental reconstructions of both land-based and ocean-based records arestill needed before it can be determined what causes the climate to alter. Only when thecauses of past climates are understood will it be possible to fully anticipate or forecastclimatic variations in the future (Bradley and Eddy, 1991); and to then identify whetherit will be possible to find out whether climate predictions can be used to predict futuregeomorphic, sedimentary and surface changes?

The cross disciplinary nature of the study of the interactions between the atmosphereand land surface can be exemplified by the contemporary debate about whethertectonics have been driven by or drive long term climate change (Whipple et al., 1999,Molnar and England 1990, Brozovic et al., 1997). The question of whether Quaternaryclimate change could have produced significantly increased topographic relief (as aresult of increased weathering and associated isostatic uplift) resulting in acceleratedrates of tectonism, can only be resolved by the contribution of climatologists definingwhat Quaternary climate change is as well as geomorphologists deciding whether these

247


1. Introduction

Department of Geography, UniversityOf Leicester, Leicester LE1 7RH

changes could significantly increase the volume of ‘missing mass’ between summits andridges and finding evidence to this effect.

In this chapter we highlight some of the different approaches that have been undertakento address the many issues in land surface-climate change. In particular we examinehow multidisciplinary approaches over different timescales (from to years) andfrom the scale of local catchment studies to global processes can improve ourunderstanding of the complex interactions between the surface and atmosphere.

2. Large scale research programmes on climate change

There are a number of reasons why scientists are interested in land surface and climateinteractions as well as change. As key components of the larger climate system, theatmosphere and land surface influence and are influenced by each other. The role of theland surface in relation to the atmosphere ranges from the purely physical, includinginducing aerodynamic drag on the atmosphere; biological, through the stomata responseto environmental changes and the control of energy and mass fluxes (Pitman et al.,1999); to geochemical, through for example, evaporation of water and dust inputs.While the role of the atmosphere on the land surface can also be defined in terms of aphysical interaction, through erosion and deposition; biological, through the control ofgrowing conditions for vegetation; and geochemical, with the addition of precipitationonto and into sediments and soils as well as the deposition of dust. Through theseinteractions the land surface can also provide evidence of past climates (Huang et al.,2000) thus helping to determine the relative magnitudes of human induced climatechange and natural climate variability.

In the technical summary of Climate Change 1995: The Science of Climate Change theIntergovernmental Panel of Climate Change identified the following scientific problemsas requiring the most urgent attention:

(i) the rate and magnitude of climate change and sea level rise:

the factors controlling the distribution of clouds and their radiativecharacteristics;the hydrological cycle, including precipitation, evaporation andrunoff;the distribution and time evolution of ozone and aerosols and theirradiative characteristics;the response of terrestrial systems to climate change and their positiveand negative feedbacks;the response of ice sheets and glaciers to climate;the influence of human activities on emissions;the coupling between the atmosphere and ocean, and oceancirculation;

248 KNIVETON AND McLAREN

PERSPECTIVES ON LAND SURFACE-CLIMATE CHANGE 249

the factors controlling the atmospheric concentrations of carbondioxide and other greenhouse gases;

(ii) the detection and attribution of climate change:

systematic observations of key variables, and development of modeldiagnostics relating to climate change;relevant proxy data to construct and test palaeoclimatic time series todescribe natural variability of the climate system;

(iii) regional patterns of climate change:

land-surface processes and their link to atmospheric processes;coupling of scales between global climate models and regional andsmaller scale models;simulations with higher resolution climate models.’

(IPCC WGI, 1995, p47)

As can be seen from this list the investigation of the relationship between the landsurface and atmosphere is common to all three research problems. The IPCC suggeststhat to resolve these issues requires systematic and sustained global observations ofrelevant parameters and allied research by individual investigators in a variety ofinstitutions as well as by co-ordinated international efforts. One of the key playersorganising large-scale terrestrial land surface experiments is the Global Energy andWater Cycle Experiment (GEWEX). Initiated by the World Climate ResearchProgramme (WCRP) in 1988, the remit of the project is to observe and model thehydrologic cycle and energy fluxes in the atmosphere, at the land surface, and in theupper oceans. While each project or experiment of GEWEX has its own objectives,the overall aim of this co-coordinated programme of research is to improve prediction ofglobal and regional climate change. Of particular interest with respect to land surface-atmosphere interactions are the Continental-Scale Experiments (CSEs). Theseexperiments aim to provide improved observations and coupled land-atmospheremodels by studying specific hydrological regions. An example of one of theseexperiments is the Continental-Scale International Project (GCIP), based in theMississippi River basin, which was designed to improve climate models by bridging thegap between small scales appropriate for modelling discrete processes over land andlarge scales practical for modelling the global climate system. In this book we focus lesson internationally co-ordinated projects, already well documented elsewhere (e.g.Pitman et al., 1999, Gash and Kabat 1999), rather choosing to concentrate on theresearch activities of smaller groups of geomorphologists and climatologists, who arealso contributing to the further understanding of interactions of the atmosphere and landsurface. Like the larger projects however these studies reveal a variety of approachesand a number of overriding issues that need to be addressed in order to improve theunderstanding of the linkages between surface and atmosphere.

250

3. Linking Climate Change with Land Surface Change

In the preface a number of issues were raised that have been covered to some degree bythe selection of papers presented in this book. These will now be summarised.

3.1 SPATIAL AND TEMPORAL SCALES OF GEOMORPHOLOGICALCHANGE AND CLIMATIC VARIABILITY

Both the atmosphere and land surface (as well as the processes that operate within andon) vary spatially from micro to global scales, and temporally over seconds to centuriesand even millennia. Figure 1, from Pitman et al., 1999, details the spatial and timescales of observational and modelling evidence providing support for the influence ofthe land surface on weather and climate. Interestingly, given the intended aim of anumber of the larger international land-atmosphere experiments to improve themodelling and prediction of the land-climate system there is relatively little overlap interms of spatial and temporal scale between observational evidence and modellingevidence. In brief, Pitman et al., (1999) attribute this to measurements being generallyon far too short a time scale to determine the role of the surface on the climate andmodelling tending to focus on large perturbations. In terms of the studies presented herea key issue concerns linking micro, local and even catchment scale processes to largerscale global phenomena both in terms of the role of the atmosphere on the surface andvice versa. In particular the upscaling of processes is a central theme of many of thepalaeoclimate studies. The issue of moving from one scale to another is also importantin the other direction, when moving from large scale to smaller scales. This isparticularly important when one considers the spatial scale of climate model output,which is typically at the resolution of 2-3° latitude/longitude grid squares. Examples ofdownscaling approaches to the atmosphere-surface interaction are illustrated by Wilbyand Dettinger, Schmidt and Dehn, and Burgess et al.

In Figure 2 we show the spatial and temporal context of the studies outlined in thisbook. Spatially research has varied from small regional scale to global scaleobservations. Catto and Bachhuber, for example, look at the development and change indune type as well as the periodic development of thin soils in the Estancia valley as aresult of changes in climate. In contrast Brooks and Legrand have observed dustvariability over the whole of North Africa. Modelling climate change has allowedscientists to increase the spatial area that they are studying to the size of a hemisphere oreven globally (e.g. Burgess et al.,). In terms of timescales the approaches vary fromshort periods (e.g. Viles and Goudie) through to changes over periods spanning the lastglacial and into the Holocene (e.g. Bachhuber and Catto and White et al.,).

KNIVETON AND McLAREN


(1: Gash & Nobre, 1997; 2: Andre et al, 1989; 3: Pielke & Avissar, 1990; 4: Pielke et al,1991; 5: Avissar & Chen, 1993; 6: Vidale et al, 1997; 7: Eastman et al, 1998; 8: Smithet al, 1992; 9: Bonan et al, 1992; 10: Polcher & laval, 1994; 11: Lean & Rowntree,1997; 12: Henderson-Sellars et al, 1993; 14: Nicholson et al, 1998; 15: Chase et al 1996;16: Zhao et al, 1999a; 17: Zhao et al, 1999b; 18:Eastman et al, 1999; 19: Lu, 1999; 20:Otterman et al, 1984; 21: Harvey, 1988; 22: Harvey, 1989; 23: Texier et al, 1997; 24:Claussen, 1997; 25: Claussen et al, 1999).


3.2 EFFECTS OF CHANGING CLIMATE ON GEOMORPHOLOGICAL ANDSEDIMENTOLOGICAL PROCESSES

Landforms and sediment accumulations represent the effects of variousgeomorphological processes that can be interpreted in the geological record from keycharacteristics such as texture, geochemistry, sedimentary structure, morphology and thecombination of certain landforms within a region. Climatic factors are known to havean influence on the nature and the rate of operation of geomorphic processes andsedimentation.

The rapid changes in climate resulted in the shifting/migration of climatic boundariesthroughout the Quaternary. Geomorphological processes changed, reflecting the degreeand type of climatic change. Preserved on many land surfaces today are the relictslandforms that formed under past different climatic conditions. Through the study ofsuch remnant landforms and sediments researchers try to piece together the informationpreserved, to interpret past conditions and to attempt to date when they occurred.Bachhuber and Catto, for example, identified within the palaeo-record changes inconditions in the Estancia Valley in central New Mexico, from dry playa through to afresh water lake at various times in the past 60,000 years. Bachhuber and Catto theninterpreted the changes in the lake levels in terms of climatic change. White et al.,.,have studied various geochemical and organic sediments to aid in the interpretation of


higher water tables and existence of palaeolakes in the Fezzan area of southern Libyaduring the late Quaternary.

Shorter term variations in the land surface due to changes in the climate are shown bythe work of Yair and Bryan in their assessment of the alteration of surface propertiesconnected with the atmospheric deposition of loess or sand in subtropical semi-arid andarid areas. Their study concludes that arid and semi-arid environments, although highlyadapted to extreme variations in rainfall, may be extremely sensitive to slight changes intheir surface properties, which alter their hydrological regime quickly and efficiently.The influence of the atmosphere on the land surface through short-term hydrologicaland weathering processes is illustrated by the work of Wilby and Dettinger and Vilesand Goudie, respectively.

Geomorphological responses to environmental change are understood at a range ofdifferent spatial and temporal scales. Observations and model simulations can be scaledup e.g. palaeo reconstructions or scaled down from global to regional and local scale.Different approaches are useful as they allow different types of data to be integrated.Downscaling has been much used by climatologists but to date has largely been ignoredby geomorphologists. However by simulating past climates using general circulationmodels and downscaling the results to smaller scales it might both be possible tounderstand the linkages between the land surface and atmosphere and interpret the landsurface observations.

3.3 EFFECTS OF SURFACE ON CLIMATE

Examples of observational and modelling evidence that provide support for the role ofthe land surface on the weather and climate at varying time and space scales wereshown in Figure 1. It is quite apparent from this body of literature that the land surfaceprimarily affects the atmosphere through the portioning of water and energy,particularly at short time scales. What however is less clear is the question about therole of humans altering the surface and thus influencing the climate and weather(Pitman et al., 1999). In Chapter 5, Adegoke and Carleton look at the meso-scale effectsof human modification of terrestrial vegetation and its affect on climate through albedoand the soil heat flux, using a combination of remotely sensed data and fieldwork. Theyconclude that the effects of modification are detectable within the climatic record of thelast 100 years. Remotely sensed data and geographical information systems (GIS) allowthe observation and analysis of surface-atmosphere processes over large, near globecoverages. Unlike point measurements remotely sensed data is often continuous inspace and together with GIS provide ideal tools for assessing issues of heterogeneityand scaling characteristics of surface-climate interactions (Carleton 1999). The use ofgeostatistical methods to explore issues of spatial variability in rainfall is shown byAgnew and Chappell. Their study examines the issue of whether the Sahel has indeeddried out over the last fifty years or whether the variations in rainfall are due to changesin the raingauge network.

254

While the use of remote sensing data has greatly enhanced the understanding of theEarth’s surface and atmosphere over large scales, there is a gap in information on thesurface atmosphere interaction over the longer term. Satellite based data has only beenavailable over the last twenty to thirty years. Longer term measurements are needed.Improved simulations of the climate at higher resolution will help develop theunderstanding of the surface-atmosphere interaction. Coupled with the improvement ofdownscaling and upscaling techniques future research will be able to ascertain theextent, magnitude and direction of the climate–land surface coupling. With thedistribution of three dimensional general circulation models (GCM) by such groups asthe UK Global Atmospheric Modelling Programme, geomorphologists andclimatologists will be able to alter the characteristics of the land surface to assesschanges in the climate; attempt to simulate changes in the climate which might result ingeomorphological change and even help understand the sometimes contradictingevidence of palaeo records (e.g. Dong et al., 1996, Price et al., 1996, Toumi et al., 1996,Thorpe et al., 1997, Price et al., 1998).

KNIVETON AND McLAREN

3.4 PREDICTION AND MANAGEMENT OF LAND SURFACE CHANGES AS ARESULT OF FUTURE CLIMATIC CHANGES.

Climate models allow the simulation of future climate over extended time scales. Theimpact of future climate changes on the surface is the subject of the work of Schmidtand Dehn, Wilby and Dettinger and Burgess et al., Schmidt and Dehn looked at theresponse of landslides to varying climatic conditions in New Zealand and Italy. Theirresults show that in Italy there is a significant decrease in activity due to a temperatureincrease in winter and reduced snow storage. However, they could not identify anyclear trend in rainfall. In New Zealand there were fewer events of high landslideprobability in winter due to decreases in rainfall. If changes in climate can be predictedthen management strategies designed to cope with landslide activity versus stability canbe developed.

The management of land surfaces and the features on and below them, under changingconditions has been recognised as being important by both Burgess et al.,; and Wilbyand Dettinger. It is clearly important that prediction and management of landformenvironments used for purposes such as waste disposal is possible. Burgess et al.,., lookat the long term future development of landforms which need to be understood forsafety assessments of underground radioactive waste disposal in the United Kingdom.In particular they suggest that under enhanced greenhouse gas scenarios glaciations willnot occur until 110 Ka AP. While control simulations indicate that the next glaciationwill be 55-60 Ka AP. The model reveals that the Greenland ice sheet is sensitive to bothmaximum concentration and decay rate of anthropogenic Limitations to thisapproach are that the model is only two-dimensional and grossly simplifies the Earth-atmosphere-cryosphere system. In all three studies mentioned in this sectiondownscaling of the climate model output is required to allow studies on a regional scaleto be conducted. Wilby and Dettinger use a regression-based downscaling model for


streamflow simulation under current and future climate scenarios in south west USA.Their model demonstrated a reasonable skill at reproducing observed area-averageprecipitation, temperature and streamflow. The downscaled data resulted in slightunderestimates of streamflow as a result of underestimates of spring rainfall andoverestimates of temperature in winter.

As stated at the beginning of this chapter successful prediction of future climate andland surface change requires an understanding of past ‘natural’ changes. A number ofstudies within the book have looked at elements of palaeoreconstructions. For instanceGurney shows that surface features such as pingos can be used to provide accurateindicators of past thermal conditions. While Bachhuber and Catto provide geologicevidence of rapid, multiple and high magnitude climate change during the last glacial(Wisconsinan) of North America. Considerable effort has been applied to piecetogether indirect or proxy evidence of longer term global and hemispheric climatechange (Briffa et al., 1998, Mann et al., 1998, Huang et al., 2000). Inevitably the moreevidence that is accumulated of palaeoclimate changes the greater degree of accuracycan be attached to these reconstructions.

4. Conclusion

It is apparent from both the research outlined in this book and other studies on thesubject that the primary difficulty faced in attempting to improve our understanding ofclimate and land surface change is reconciling the vast range in the spatial and temporalscales of processes operating between the atmosphere and surface with the observationsand tools available to study them. Inevitably in a dynamic environment when trying tostudy past climate and surfaces we are limited in the availability of observations at theglobal scale over extended time periods. Remotely sensed data now provide near globalcoverage of the surface and atmosphere. While modern geostatistical methods andsystems provide the tools with which to analyse these data at varying scales. Yet bydefinition remotely sensed measurements are removed from the object of observationthus restricting the detail of analysis. In addition they are of limited record lengthstretching back a mere 20-30 years, at best.

Models of the climate and surface potentially offer a powerful tool to understand thestrength, direction and extent of response of the surface and atmosphere to pertubationsin their and other climate system components states. Yet again models can only providepart of the solution to questions of climate and land surface change. Partially this isbecause of the level of understanding or rather lack of it of the underlying processesinvolved and of the scales over which these processes operate. There are also restraintsto the use of models because of the way in which processes are represented within themodel. Computing power restraints and an incomplete knowledge of processes meanthat climate models run at coarse resolution. The most advanced climate models operateat approximately 2 to 3°latitude/longitude resolution. This means that many processesbetween the surface and atmosphere are parametrised or simplified. While the output


from global climate models are also of coarse resolution impacting on their use tointerpret the effect of future or past simulated climate changes on the surface.

The development of downscaling and upscaling techniques has attempted to bridge thegap between the scale over which models operate and over which observations aremade. Indeed it could be said that it is the movement up and down scale anddevelopment of methods to do this that provides the interface between climatologicaland geomorphological disciplines.

In the future it can be seen that the resolution of climate models will be improved. Witha greater understanding of processes and improved model resolution it can be expectedthat a more detailed and realistic representation of surface and atmosphere processeswill be incorporated into climate models leading eventually to a fully coupled threedimensional land surface-atmosphere model. It can also be envisaged that longer datasets will be available from remote sensing and field observations, while improvedgeostatistical methods will be developed to integrate spares and non-continuous datasets.

References

Adegoke, J.O and Carleton, A. M., (2000) Warm season land surface-climate interactions in the United StatesMidwest from meso-scale observations, in McLaren, S and Kniveton, D. (eds), Linking Climate Changeto Land Surface Change, Kluwer Academic Press,

Agnew, C. and Chappell, A., (2000) Desiccation in the Sahel, in McLaren, S and Kniveton, D. (eds), LinkingClimate Change to Land Surface Change, Kluwer Academic Press,

Andre, J-C., Bougeault, P., Mahfouf, J-F., Mascart, P., Noilhan, J., Pinty, J-P., (1989) Impact of forests onmesoscale meteorology. Philosophical Transactions of the Royal Society of London, B 324, 407-422.

Avissar, R., and Chen, F., (1993) Development and analysis of prognostic equations for mesoscale kineticenergy and mesoscale (subgrid-scale) fluxes for large scale atmospheric models. Journal of AtmosphericScience, 50, 3751-3774.

Bachhuber, F. W. and Catto, N., (2000) Geologic evidence of rapid, multiple and high magnitude climatechange during the last glacial (Wisconsinan) of North America, in McLaren, S and Kniveton, D. (eds),Linking Climate Change to Land Surface Change, Kluwer Academic Press,

Bonan, G.B., Pollard, D., and Thonpson, S.L., (1992) Effects of boreal forest vegetation on global climate.Nature, 359, 76-718.

Bradley, R. S. and Eddy, J. A. (1991) Records of past global changes, in Bradley, R. S. (ed.), Global Changesof the Past. Boulder: University Corporation for Atmospheric Research, 5-9.

Brooks, N. and Legrand, M., (2000) Dust variability over Northern Africa and rainfall in the Sahel, inMcLaren, S and Kniveton, D. (eds), Linking Climate Change to Land Surface Change, KluwerAcademic Press,

Brozovic, N., Burbank, D., Meigs, A., (1997) Climatic limits on landscape development in the NorthwesternHimalaya. Science, 276, 572-574.


Burgess, P. E. Palutikof, J. P. and Goodess, C. M., (2000) Approaches to modelling long-term climatechange over the British Isles, in McLaren, S and Kniveton, D. (eds), Linking Climate Change to LandSurface Change, Kluwer Academic Press,

Catto, N. and Bachhuber, F. W., (2000) Aeolian geomorphic response to climate change: an example fromthe Estancia valley, Central New Mexico, U.S.A. , in McLaren, S and Kniveton, D. (eds), LinkingClimate Change to Land Surface Change, Kluwer Academic Press,

Chase, T.N., Pielke, R.A., Kittel, T.G.F., Nemani, R., and Running, S.W., (1996) Sensitivity of a generalcirculation model to global changes in leaf area index. Journal of Geophysical Research, 101, 7393-7408.

Chase, T.N., Pielke, R.A., Kittel, T.G.F., Nemani, R., and Running, S.W., (2000) Simulated impacts ofhistorical land cover changes on global climate. Climate Dyamics, 16, 93-105.

Claussen, M., Kubatzki, C., Brovkin, V., Ganopoloski, A., Hoelzmann, P., and Pachur, H.J., (1999) simulationof an abrupt change in Saharan vegetation at the end of the mid-Holocene. Geophysical ResearchLetters, 26, 2037-2040.

Claussen, M., (1997) Modeling biogeophysical feedback in the African and Indian Monsoon region. ClimateDynamics, 13, 247-257.

Dong B., P.J. Valdes and N.M.J. Hall, (1996) The changes of monsoonal climates due to Earth's orbitalperturbations and ice age boundary conditions. Palaeoclimates: Data and Modelling, 1, 203-240.

Eastman, J.L., Pielke, R.A., andMcDonald, D.J., (1998) Calibration of soil moisture for large eddy simulationsover the Fifearea. Journal of Atmospheric Science, 55, 1131-1140.

Eastman, J.L., Coughenour, M.B., and Pielke, R.A., (1999) The effects of and landscape change using acoupled plant meteorological model, Global Change Biology, submitted.

Gash, J., and Kabat, P., (1999) Land-surface experiments. IGBP Global Change Newsletter, 39,12-14.

Gash, J.H.C and Nobre, C.A., (1997) Climatic effects of Amazonian deforestation: Some results fromABRACOS. Bulletin of the American Meeorological Society, 78, 823-830.

Gurney, S., (2000) Relict cryogenic mounds in the UK as evidence of climate change, in McLaren, S andKniveton, D. (eds), Linking Climate Change to Land Surface Change, Kluwer Academic Press,

Harvey, L.D.D., (1988) A semianalytic energy balance climate model with explicit sea ice and snow physics.Journal of Clim, 1, 1065-1085.

Harvey, L.D.D., (1989) An energy balance climate model study of radaiative forcing and temperatureresponse at 18Ka. Journal of Geophysical Research., 94, 12873-12884.

Huang, S.P., Pollack, H.N., Shen, P.Y., (2000) Temperature trends over the past five centuries reconstructedfrom borehole temperatures. Nature, 403, 756-759.

Henderson-Sellers, A., Dickinson, R.E., Durbridge, T.B., Kennedy, P.J., McGuffie, K., and Pitman, A.J.,(1993) Tropical deforestation: Modelling local to regional-scale climate change., Journal of GeophysicalResearch., 98, 7289-7315.

IPCC WGI, (1995) Climate Change 1995: - The Science of Climate Change: Contribution of Working GroupI to the Second Assessment Report of the Intergovernmental Panel on Climate Change. Houghton, J.T.,Meira Fiho, L.G., Callander, B.A., Harris, N., Kattenberg, A., and Maskell, K., (eds.). CambridgeUniversity Press, New York, USA


Lean, J., and Rowntree, P.R., (1997) Understanding the sensitivity of a GCM simulation of Amazoniandeforestation to the specification of vegetation and soil characteristics. J. Climate, 10, 1216-1235.

Lu, L., (1999) Implementation of a two way interactive atmospheric and ecological model and its applicationto the central united States, PhD thesis, Colorado State University, fort Collins, Colorado.

Nicholson, S.E., Tucker, C.J., and Ba., M.B., (1998) Desertification, drought and surface vegetation: Anexample from the west African Sahel. Bulletin of the Ameican Meteorological Society., 79, 815-829.

Molnar, P., and England, P., (1990) late Cenozoic uplift of mountain ranges and global climate change:Chicken or egg? Nature, 346, 29-34.

Otterman, J., Chou, M-D., and Arking, A., (1984) effects of non tropical forest cover on climate. J. Appl. Met.,23, 762-767.

Pielke, R.A, and Avissar, R., (1990) Influence of landscape structure on local and regional climate. LandscapeEcology., 4, 133-155.

Pielke, R.A., dalu, G.A., Snook, J.S., lee, T.J., and Kittel, T.G.F., (1991) Nonlinear influence of mesoscaleland use on weather and climate. J. Climate, 4, 1053-1069.

Pitman, A., Piekle, Sr., R., Avissar, R., Claussen, M., Gash, J., Dolman, H., (1999) The role of the landsurface in weather and climate: Does the land surface matter? IGBP Global Change Newsletter, 39, 4-11.

Polcher, J., and Laval, K., (1994) A statistical study of the regional impact of deforestation on climate in theLMD GCM. Climate Dynamics, 10, 205-219.

Price, G.D., P.J. Valdes and B.W. Sellwood, (1998) A comparison of GCM simulated Cretaceousand climates; implications for the sedimentary record. Palaeogeography,

Palaeoclimatology, Palaeoecology, 142, 123-138.

Price, G.P., B.W. Sellwood and P.J. Valdes, (1996) Sedimentological evaluation of general circulation modelsimulations for the "greenhouse" earth: Cretaceous and Jurrasic case studies. Sedimentary Geology, 100,159-180.

Schmidt, M. and Dehn, M., (2000) Examining links between climate change and landslide activity usingGCMs, in McLaren, S and Kniveton, D. (eds), Linking Climate Change to Land Surface Change,Kluwer Academic Press,

Smith, E.A., Hsu, A.Y., crosson, W.L., field, R.T., Fritscen, L.J., Gurney, R.J., kanemasu, E.T., Kutstas, w.P.,nie., D., Shuttleworth, W.J., Stewart, J.B., Verma, S.B., Weaver, H.L., and Wesley, M.L., (1992) Area-averaged surface fluxes and their time-space variability over the FIFE experimental domain. Journal ofGeophysical Research., 97, 18599-18622.

Texier, D., de Noblet, N., Harrison, S.P., Haxeltine, A., Jolly, D., Joussaume, S., Larrif, F., prentie, I.C., andTarasov, P., (1997) Quantifying the role of biosphere-atmosphere feedbacks in climate change: Coupledmodel simulations for 6000 years BP and comparison with palaeodata for northern Eurasia and NorthernAfrica. Climate Dynamics, 13, 865-882.

Thorpe R.B., K.S. Law, S. Bekki and J.A. Pyle, (1996) Is methane driven deglaciation consistent with the icecore record? Journal of Geophysical Research, 101-D22, 28627-28635.


Toumi R., J.D. Haigh and K.S. Law, (1996) A tropospheric ozone- lightning climate feedback. Geophys. Res.Lett., 23, 1037-1040. Valdes, P.J., B.W. Sellwood and G.P. Price, 1996: The concept of Cretaceousclimate equability. Palaeoclimates, 1, 139-158.

Vidale, P.L., Pielke, R.A, barr, A., and steyaert, L.T., (1997) case study modelling of turbulaent andmesoscale fluxes over the BOREAS region. Journal of Geophyical Research, 102, 29167-29188.

Viles, H. and Goudie, A. S., (2000) Weathering, geomorphology and climatic variability in the CentralNamib Desert, in McLaren, S and Kniveton, D. (eds), Linking Climate Change to Land Surface Change,Kluwer Academic Press,

Whipple, K.X., Kirby, E., Brocklehurst, S.H., (1999) Geomorphic limits to climate –induced increases intopographic relief. Nature, 40, 39-43.

White, K., McLaren, S. Black, S. and Parker, A., (2000) Evaporite minerals and organic horizons insedimentary sequences in the Libyan Fezzan: implications for palaeoenvironmental reconstruction, inMcLaren, S and Kniveton, D. (eds), Linking Climate Change to Land Surface Change, KluwerAcademic Press,

Wilby, R. L. and Dettinger, M. D., (2000) Streamflow changes in the Sierra Nevada, California, simulatedusing a statistically downscaled General Circulation Model scenario of climate change, in McLaren, Sand Kniveton, D. (eds), Linking Climate Change to Land Surface Change, Kluwer Academic Press,

Xue, Y., (1996) The impact of desertification in the Mongolian and the Inner Mongolian grassland on theregional climate. J. Climate, 9, 2173-2189.

Yair, A. and Bryan, R. B., (2000) Hydrological response of desert margins to climate change, in McLaren, Sand Kniveton, D. (eds), Linking Climate Change to Land Surface Change, Kluwer Academic Press,

Zhao, M., Pitman, A.J., and Chase, T., (2000) The impact of land cover change on the atmosphericcirculation, Climate Dyn., submitted.

Zhao, M., Pitman, A.J., and Chase, T., (2000) Sensitivity of a general circlation model to global changes inleaf area index: A reassessment, J. Geophys. Res., submitted.

Index

A

aeolian activity · 166, 167, 170, 171,183, 185

Antecedent Daily Rainfall Index · 131

atmospheric 227-231, 234, 239,240

B

aerosols · 1, 3, 4, 5, 8, 11, 12, 13, 24,103, 118, 121, 126, 134, 242

Algeria · 7, 9Angola · 12

atmosphere · 2-4, 11-13, 15, 18, 21, 24,81, 82, 87, 93-95, 97,100, 102, 103,106, 107, 117, 120, 134, 137,228,229, 238, 242-248,251

biological weathering · 63British Isles · 117, 226, 238, 240, 249Burkina Faso · 28, 30-32, 45

C

California · 96-98, 114, 116-118, 167,187, 251

Chad · 7climate anomalies · 81, 83, 86, 89

climate change · 2, 26, 29, 30, 37, 46,47, 62, 97, 115-119, 121, 125, 127,134, 136, 137, 139, 140, 142, 163-167, 182, 184, 207, 226-228, 234,236, 238-244, 247, 249-251

climate change impact · 119climate model · 100, 117, 123, 124,

136, 226, 227, 239, 244, 246, 247,250

climatic forcing · 228Crop Moisture-Z Index 81cryogenic mounds · 205-207, 209-211,

214-218, 220-225, 250

D

deflation · 1, 2, 7, 8, 15, 16, 18, 19, 21-23, 141-143, 145, 150, 151, 166,167, 170-172, 174, 178, 180-185

deforestation · 82, 250Democratic Republic of Congo · 12desertification · 26-29, 46, 47, 49, 60,

61, 67, 82, 201, 251desiccation · 2, 3, 7, 8, 22, 23, 26, 29-

31, 33, 35, 37, 66, 139, 140, 145,150, 153, 158, 159, 163, 164, 168,180

dome dunes · 166, 170, 171, 175, 177-180, 182-185

downscaling · 96-98, 100-103, 108-111, 114-123, 125, 127, 134-137,227, 232, 236, 238, 244-248

downscaling techniques · 96, 98, 120,137

droughts · 29, 45, 81, 82, 93duricrusts · 199

262

dust · 1-16, 18-25, 49, 59, 61, 65, 80,186, 242, 244

dust indices · 1dust production · 1, 3, 6, 8, 9, 13, 14,

15, 18-23, 25dust sources · 3, 5, 7

E

ecosystems · 48, 49, 60, 226Egypt · 7, 200, 201, 202El Niño · 69, 120endemic species · 49environmental change · 188, 189, 200,

227, 230, 245environmental degradation · 26, 29Erg of Bilma · 7evaporation · 57, 59, 60, 68, 95, 162,242evapotranspiration · 83, 95, 106, 135,136, 182

F

feedbacks (see climatic forcing)Fezzan · 188, 189, 199-203, 245, 251fire · 11, 12, 192, 196, 202, 209floods · 81, 83, 151fog · 63, 66, 67, 69, 70, 78-80foggara · 194, 196, 199

G

Gabon · 12

general circulation models (GCMs) · 94-98, 100, 101, 114, 119-121, 125,126, 134, 136, 245, 246, 251

Great Man-Made River Project · 189,202

greenhouse gases · 96, 98, 103, 243gypsarenite · 143, 145, 151, 158, 168,

172-174, 176, 178, 180-183, 185,195, 196, 198, 199-204, 206, 210,214

gypsum · 63, 70, 74, 75, 80, 173, 174,179, 181, 187, 190, 194, 195, 200,203

gypsum crusts · 68

H

Holocene · 25, 141, 142, 151, 166, 168,170, 180, 183, 185, 186, 199, 201,202, 227, 228, 229, 244, 249

Horn of Africa · 8, 10

I

ice · 81, 139, 140, 206, 209-211, 214-217, 220, 221, 223, 224, 226-229,234-242, 247, 249-251

infiltration · 48-50, 59, 60Infra-Red Difference Dust Index

(IDDI) · 1, 3insolation · 4, 65, 77, 78, 107, 206,

227-230, 237-239Italy · 119, 124, 135, 136, 234, 247

pluvial lake · 139, 142, 145, 150, 160,166, 185

precipitation (see rainfall)

K

L

M

N

P

Kalahari · 12, 185

263

New Zealand · 119, 120, 128, 135, 136,247

Niger · 7-9, 11, 24, 28, 30-34, 36, 37,45, 46

O

landslide scenarios · 119landslides · 119, 120, 129, 132, 135,

247Libya · 7, 189, 201-203, 245lichen · 63, 65-71, 73-77, 79, 80loess · 48-51, 58-62, 166, 167, 172,

177-179, 183-186, 245

Mali · 7, 24, 28, 30-32, 45, 202Mauritania · 7, 31METEOSAT · 1, 3microorganisms · 63, 65, 66, 75, 76, 80Morocco · 7, 9, 80mudslide · 119, 120, 123-125, 134, 135

Namib Desert · 12, 63, 64, 66, 68, 69,77-80, 251

Negev · 49-51, 58-62, 65, 79, 80, 186Negev desert · 49-51, 58-62, 65, 66, 67Neolithic · 199, 200, 201New Mexico · 139-141, 163-168, 182,

185-187, 245, 249

ocean cores · 228optical stimulation luminescence · 180ostracods · 143, 146-149, 151-155, 158,

160, 176

Palmer Drought Severity Index-PDSI · 81, 83

palsas · 205, 206, 209, 211, 214, 215,217, 220, 221, 223-227

parabolic dunes · 166, 169, 171, 181,182

periglacial · 205, 206, 209, 210, 220,221, 224-227, 231, 233-236

permafrost · 205, 206, 209, 211, 212,215, 216, 221, 223, 225-227

pingos · 205, 206, 209-212, 214-217,219-221, 223-227, 247

playa · 139, 141-143, 153, 155, 156,160, 162-164, 166, 167-173, 178-183, 185, 186, 200, 201, 245

Pleistocene · 61, 142, 143, 166, 188,201, 202, 209, 211, 220, 225, 227,240

264

Q

Quaternary · 25, 50, 58, 61, 139, 140,145, 165, 166, 168, 185, 186, 187,189, 200, 201, 202, 203, 215, 223,224-229, 239, 241, 245

R

radiation . 4, 8, 65, 68, 101, 106, 107,116, 118, 219, 225

rainfall · 1-3, 8, 11, 15, 16, 18-38, 40-51, 54, 55, 57, 59-63, 69, 78, 79, 83,86, 93, 94, 97, 114-118, 129, 130,132, 133, 136, 142, 167, 181, 189,245, 246, 247, 249

raingauges · 26, 38remote sensing · 11, 24, 94, 189-191,

201, 246, 248rock-surface microenvironments · 63runoff · 15, 48-51, 54, 55, 57, 59-62,

83, 96-98, 106, 113-116, 118, 163,242

S

Sahara · 1, 2, 5, 7-16, 18, 20, 22-24, 28,29, 45, 49, 188, 189, 201-203

Sahel · 1-3, 5, 7-11,13-16,18-35, 37,38, 40, 41, 45-47, 82, 246, 249, 250

Sahel-Sahara · 1, 3, 6, 7, 9, 11, 13, 23salt weathering · 63, 65, 66, 69, 70, 79,

80snowpack · 97, 98, 106, 107, 111, 113,

114, 115, 118

soil degradation · 8, 13, 14, 22, 33, 34soil-dust cycle · 8stable isotope analysis · 188, 189Sudan · 7, 46, 47, 200 203surface properties · 48-50, 60, 106, 245

T

tafoni · 65, 70, 80thermoluminescence · 180

U

upscaling techniques · 246, 248U-Th dating · 193

V

variograms · 38, 40-42vegetation cover · 15, 16, 19, 49, 167,

184, 222

W

Wisconsinan · 139, 142, 143, 145, 146,148-151, 153, 155-164, 168, 170,172, 183-185, 216, 226, 247, 249

Advances in Global Change Research

1.

2.

3.

4.

5.

6.

P. Martens and J. Rotmans (eds.): Climate Change: An Integrated Perspective. 1999ISBN 0-7923-5996-8

A. Gillespie and W.C.G. Burns (eds.): Climate Change in the South Pacific: Impactsand Responses in Australia, New Zealand, and Small Island States. 2000

ISBN 0-7923-6077-XJ.L. Innes, M. Beniston and M.M. Verstraete (eds.): Biomass Burning and Its Inter-Relationships with the Climate Systems. 2000 ISBN 0-7923-6107-5M.M. Verstraete, M. Menenti and J. Peltoniemi (eds.): Observing Land from Space:Science, Customers and Technology. 2000 ISBN 0-7923-6503-8T. Skodvin: Structure and Agent in the Scientific Diplomacy of Climate Change. AnEmpirical Case Study of Science-Policy Interaction in the Intergovernmental Panel onClimate Change. 2000 ISBN 0-7923-6637-9S.J. McLaren and D.R. Kniveton (eds.): Linking Climate Change to Land SurfaceChange. 2000 ISBN 0-7923-6638-7

KLUWER ACADEMIC PUBLISHERS – DORDRECHT / BOSTON / LONDON

climate change to land surface change

Documents