BMRC RESEARCH REPORT NO. 133

NOVEMBER 2007

BUREAU OF METEOROLOGY RESEARCH CENTRE | AUSTRALIA

A.J. Hollis and V. Jemmeson (editors)

'Physical processes and modelling

of the water and carbon cycle' :

extended abstracts of presentations at the first annual CAWCR Modelling Workshop

27-29 November 2007

BMRC RESEARCH REPORT NO. 133

NOVEMBER 2007

'Physical processes and modellingof the water and carbon cycle' :

A.J. Hollis and V. Jemmeson (editors)

BMRCGPO Box 1289 Melbourne VICAustralia 3001www.bom.gov.au

All images reproduced in grayscale. A colour version of Research Report 133 is available online: http://www.bom.gov.au/bmrc/pubs/researchreports/researchreports.htm

extended abstracts of presentations at the first annual CAWCR Modelling Workshop

27-29 November 2007

iii

CONTENTS Foreword . . . . . . . . . . v Damien Barrett

Developing a hydrologic data assimilation scheme to integrate multiple satellite data sets in stream-flow forecasting. . . . . 1 C.S. Draper, J.P. Walker and P.J. Steinle

An evaluation of remotely sensed soil moisture over Australia . . . 5 J. Walker, D. Barrett, R. Gurney, J. Kalma, Y. Kerr, E. Kim and J. Le Marshall MoistureMap: a soil moisture monitoring, prediction and reporting system for sustainable land and water management . . . . . 9 Xu Liang Land surface modeling and data assimilation . . . . . 13 Adam Smith and Huqiang Zhang

Evaluating CABLE soil moisture predictions in the Murray Darling Basin . 15 Debbie Hudson and Oscar Alves The impact of land-atmosphere initialisation on dynamical seasonal prediction . 19 Garry Willgoose, Patricia Saco and Alice Howe

Vegetation as an intermediary in the coupling between hydrology-soil moisture-atmosphere . . . . . . 23 Ian N. Harman Canopy dynamics and the surface energy balance . . . . 25 Roger M. Gifford Terrestrial carbon cycle feedbacks in atmospheric and climate change . . 29 Helen Cleugh Carbon cycle observations: challenges and opportunities . . . 33 Derek Eamus, S. Fuentes, C. Macinnis-Ng, A. Palmer, D. Taylor, R. Whitley, I. Yunusa and M. Zeppel Woody thickening: a consequence of changes in fluxes of carbon and water on a warming globe? . . . . . . . . 37 Gab Abramowitz and Hoshin Gupta Model independence and the representation of model space as a source of uncertainty 41 Yingping Wang, Gabriel Abramowitz, Eva Kowalczyk, Rachel Law, Bernard Pak, Adam Smith and Huqiang Zhang Evaluating the Australian Community Land Surface Model for Australian ecosystems . 45 Peter Isaac, Jason Beringer, Lindsay Hutley and Stephen Wood Modelling Australian tropical savannas: current tools and future challenges . 47 Yaohui Li Observations and studies on arid and semi-arid land-atmosphere interaction in northwest China . . . . . . . . 53 Francis Chiew

Hydrological modelling in Australia . . . . . . 55 Michael J. Manton

Integrated hydrological studies . . . . . . 59 Jim Elliott

Integrating streamflow prediction and data into an operational system – the example of flood warming . . . . . . . 63 Huqiang Zhang and Liang Zhang Energy, water and carbon cycles simulated in a 51-year CABLE global offline experiment: towards an integrated modelling system . . . 67

iv

Albert van Dijk Modelling to assess water resources availability and use – how does it relate to weather and climate modeling? . . . . . . 71 Michael R. Raupach

Carbon, climate and humans: Australia in the Earth System . . . 75

v

FOREWORD The past year has seen a sharp increase in the attention given to climate, weather and related environmental challenges such as national water security, an effective response to ever-increasing greenhouse gas concentrations and adapting our industries and society to the climate changes that are now inevitable over the coming few decades. Numerical models that include all components of the earth system: land, atmosphere, oceans and ice and which resolve biogeochemical as well as physical processes are essential in order to provide the scientific evidence base for decision-making over the coming years and decades. The scientific challenges that lie ahead to meet the needs of Australia are significant and require the focused mobilisation of national resources and talent. Building upon the Australian Community Climate and Earth System Simulator initiative, this year the Bureau of Meteorology and CSIRO jointly established the Centre for Australian Weather and Climate Research. The Centre not only provides a focus for the ACCESS initiative, it also brings together efforts to improve understanding of the key processes simulated by numerical models. Australia will benefit from a remarkable array of applications arising from ACCESS and other science undertaken by Centre scientists including improved weather forecasting and environmental predictions, forecasts of ‘ocean weather’, more robust predictions of climate a season or two ahead and projections of climate change. This year’s workshop ‘Physical Processes and Modelling of the Water and Carbon Cycles’ is the first under the auspices of the new Centre but continues the Modelling Workshops established by the Bureau of Meteorology Research Centre, being the nineteenth of the series. It is also the final publication in the BMRC Research Report series. The subject of this year’s workshop especially brings the land surface into focus. Underlying the thinking for this year’s workshop is the recognition that the carbon and water cycles are closely coupled through a variety of processes. The papers presented at the workshop examine the challenges presented in modelling these processes from a variety of viewpoints including the acquisition and assimilation of necessary observations, the ecological as well as physiological processes that modulate these cycles and how carbon and water interact and feedback through the wider earth system. The workshop includes participants from research groups around Australia. We also welcome a number of contributors from overseas institutes to the workshop, with keynote presentations from Professors Praveen Kumar (University of Illinois), Xu Liang (University of Pittsburgh) and Colin Prentice (Bristol University). We are grateful for these expert contributions and to all the participants’ contributions to the debate and discussions.

I would like to thank the Local Organising Committee for the workshop, comprising Alan Seed (Chair), Kamal Puri, Yingping Wang, Huqiang Zhang, Andrew Hollis and Val Jemmeson.

Chris Mitchell Foundation Director Centre for Australian Weather and Climate Research: A partnership between the Australian Bureau of Meteorology and CSIRO November 2007

1

Developing a hydrologic data assimilation scheme to integrate multiple satellite data sets in stream-flow

forecasting

Damian Barrett CSIRO Land and Water, GPO Box 1666, Canberra, ACT, 2601

Introduction The aim of hydrologic forecasting is to predict at some future time state variables of rainfall, evapotranspiration, soil moisture content, runoff, drainage, stream flow and/or flood extent based on the evolution of those variables in time (determined by a ‘forward’ model and forcing data), and on the conditioning of those variables with observations. The conditioning involves selecting optimal values of state variables that take into account the relative errors introduced by both the model and observations to yield an analysis product that forms the basis of the forecast (McLaughlin 2002). This talk will present a prototype model-data assimilation scheme for forecasting profile soil moisture and stream inflows based on satellite data for the Murrumbidgee River catchment that is independent of stream gauge observations. The independence from flow observations allows direct assessment of the utility of this method for prediction of stream flow in ungauged basins. Current limitations to improved stream flow forecasting As water resources become increasingly limited, information on the future hydrologic status of rivers and catchments at timescales ranging from hours through days, to seasons and decades is increasingly important for a range of management and policy purposes. On short timescales rapid updating of flows is required for flood warning and implementation of emergency plans. On day to week timescales, hydrologic forecasts are required for release scheduling and river operations, as well as anticipating water demand for irrigation off-takes, and management of environmental flows to ecological assets. On seasonal to annual timescales, hydrologic forecasts are required for developing water accounts and water availability scenarios and assessing soil water stores in rain-fed agricultural systems. And, finally, on decadal timescales scenarios of water availability are required for developing land use and water allocation policy. The benefits of an improved capability for hydrologic forecasting are many-fold and include improved efficiencies of water use leading to reductions in water loss and reduced shortfalls on water orders, improved targeting of flows to environmental assets through better anticipation of natural flow events, basin-wide consistency in management practices and enhanced responses to flood threat. Currently, the primary limitations to improved forecasting of hydrologically relevant variables are:

• Uncertainties associated with rainfall distribution in space and time due to gaps in rainfall observation network and biases introduced through indirect methods of rainfall measurement;

• Spatial variation in soil and vegetation biophysical properties as parameters in hydrological models due to limited datasets on landscape properties; and,

• Inadequate knowledge of antecedent soil moisture as initial conditions for catchment models which determines the partitioning of rainfall between runoff and infiltration into soil water stores.

2

The inadequate characterization of spatial and temporal correlations between precipitation, surface physical properties and antecedent soil moisture is the primary source of uncertainty in flow prediction from catchment models. To address this problem, it is necessary to exploit new information sources; particularly, spatially extensive datasets of observations to constrain state variables in catchment models (eg Renzullo et al. 2007). This improvement in initializing catchment models coupled with improved precipitation forcing from NWP models is the basis for improved flow prediction on day to week timescales. Role of remote sensing in hydrologic forecasting Satellite observations provide an important and rich source of spatially contiguous information on the radiometric properties of the earth’s atmosphere and land surface at update frequencies from minutes to days. Routine products are now available from passive and active microwave, thermal, optical, infrared and gravity sensors that provide information on surface layer soil moisture, surface temperature, snow cover and depth, water storage and evapotranspiration. Hydrologic forecasting from satellite observations provides the opportunity for flow prediction in ungauged basins in remote areas and in internal, ungauged sub-catchments. This approach does not, however, preclude the integration of flow observations in the assimilation scheme where they are available. The challenge to utilizing satellite observations in hydrologic forecasting is two-fold: Firstly, errors and artifacts that are introduced by the earth’s atmosphere, sub-pixel variability, and sensor calibration and drift problems need to be identified and removed. Secondly, observed reflectance or radiance data need to be related directly to state variables and/or parameters in the hydrologic model. Currently, rainfall-runoff models do not represent well satellite observations due to differences in scale and interpretation between the observation and model state variables. For example, rainfall-runoff models do not directly represent the surface layer moisture observed by passive microwave sensors. This is a major research challenge and requires the development of suitable observations models to better relate satellite radiance and reflectance data to hydrologic variables. A hydrologic model-data assimilation scheme The hydrologic forecasting prototype scheme for the Murrumbidgee River catchment comprises the following components (Barrett et al. 2007):

• A ‘forward’ model, M, comprising a six-layer soil water balance scheme, Penman-Monteith actual evapotranspiration function and infiltration/runoff module;

• Model state variables of surface soil moisture (to 2.5cm depth; 1Sθ ) and soil profile

average soil moisture content ( zθ ); • Model parameters of soil physical properties, leaf area index, and canopy

micrometeorology; • Forcing data comprising climate inputs of rainfall, daily maximum and minimum

temperatures, vapour pressure, and shortwave downward radiation; • Satellite observations of reflectance in red and near infra-red (MODIS Terra bands 1 & 2

at ~10:30am overpass), thermal (AVHRR NOAA-18 bands 4 & 5 at ~1:30pm overpass) and microwave wavelengths (AMSR-E Aqua 6.9 GHz at ~1:30pm overpass);

• ‘Observation’ models of (1) the surface energy balance relating soil profile average moisture content and land surface temperature, TS, via the dependency of canopy

3

conductance on moisture content; and, (2) a microwave radiative transfer model relating surface brightness temperature, TB, to soil dielectric constant via the dependency of the dielectric constant on surface soil moisture content; and,

• a ‘3-D’ variational assimilation scheme that generates optimal estimates of state variables ( 1Sθ and zθ ) based on satellite observations and their modeled counterparts and the relative errors from the model and observations.

Observation errors were derived from published specifications for each satellite sensor while model errors were obtained from the spatial covariance of f a

k kT T− ; where subscript k refers

to observations type (land surface temperature or brightness temperature) and the super-script refers to forecast at 1-day lead time (f) or analysis (a). The scheme utilizes satellite observations when they are available (cloud-free conditions and within the image swath) or modeled values only when observations are absent (subject to covariances imparted by the model errors). Initial verification work has examined the relationship between predicted inflows from storm events, calculated as integrated surface runoff across sub-catchments, and peak flow/base flow separation based on gauge observations. Figure 1 shows an example of the modeled and observed land surface temperatures and the subsequent ‘innovation’ by the assimilation scheme expressed as the difference in modeled and analysis soil profile average moisture contents. In a band extending from the lower left hand to upper right hand corners across the Murrumbidgee catchment, modeled land surface temperatures are hotter than observed by satellite whereas in the upper left hand corner (in the hottest part of the scene) land surface temperatures are close to observed. The cooler observations indicate that actual profile average soil moisture content (due to a rainfall event 3 days prior) is greater than suggested by the model as shown by the innovations ( m a

z zθ θ− ). This discrepancy could result from errors in the rainfall forcing, from lateral flows not represented by the soil moisture model or from surface-groundwater linkages leading to increase in water table height in the affected areas. While this prototype assimilation scheme demonstrates the utility of these methods in hydrologic forecasting of short term stream flow, further work is required to elucidate the mechanisms associated with lateral flows, understand the error covariances of observations and model (particularly bias removal), examine the errors associated with forcing data (particularly rainfall), and improve the efficiency of the system for potential operational applications. Figure 1: Modeled (m) and observed (o) land surface temperature (TS K) and modeled less analysis (a) deviations of soil profile average volumetric water content ( zθ ) for the Murrumbidgee River catchment (outlined).

osT m a

z zθ θ−( )m ms zT H θ=

270 310oK -0.1 0.102/10/05 (272)

4

References Barrett, D.J., Kuzmin, V.A., Walker, J.P. McVicar, T.R. and Draper, C. 2007. Improving stream

flow forecasting by integrating satellite observations, in situ data and catchments models using model-data assimilation methods. eWater CRC Technical Report. 60 pp. In Review.

McLaughlin, D. 2002. An integrated approach to hydrologic data assimilation: interpolation, smoothing, and filtering. Adv Water Resources Res. 25, 1275-1286.

Renzullo, L.J., Barrett, D.J., Marks, A.S., Hill, M.J., Guerschman, J.P., Mu, Q., Running, S.W. 2007. Multi-sensor model-data fusion for estimation of hydrologic and energy flux parameters., Rem. Sen. Environ., In Press.

5

An evaluation of remotely sensed soil moisture over Australia

Draper, C. S.1,2, J.P. Walker1, and P.J. Steinle2

1 Department of Civil and Environmental Engineering, University of Melbourne. 2 Centre for Australian Weather and Climate Research, Bureau of Meteorology.

Introduction Soil moisture is an important control over atmospheric evolution, since it controls the partition of incoming radiation into latent and sensible heating. To accurately model these land surface fluxes, atmospheric models must ultimately have accurate soil moisture fields. Yet soil moisture is typically initialised indirectly in Numerical Weather Prediction (NWP) models, frequently resulting in unrealistic model soil moisture. For example, both the soil moisture and land-surface fluxes in the Australian NWP system (LAPS) have been observed to be unrealistic (e.g., Ellett et al (2005); Draper and Mills, (2007)). The soil moisture in LAPS is initialised using a background field based on antecedent precipitation and climatological evaporation (following Pescod et al (1994)), which is then incremented (following the scheme developed by Viterbo and Beljaars (1995) at ECMWF) according to low-level forecast humidity errors. A similar scheme is used in the ACCESS model, as run in global mode by the U.K. Met. Office, based on warm-running soil moisture, which is incremented according to low-level forecast humidity and temperature errors (and requires periodic relaxation towards a climatology).

Novel remote sensing technologies offer the potential to replace these indirect soil moisture initialisation schemes with techniques that utilise observed surface soil moisture. It has already been demonstrated that assimilating remotely sensed soil moisture into land surface models can be benefit modelled soil moisture (e.g., Reichle and Koster, 2005). Currently, the most advanced techniques for remote sensing soil moisture over Australia are based on data from the Advanced Microwave Scanning Radiometer – Earth Observing System (AMSR-E) instrument, on NASA’s Aqua satellite. AMSR-E provides global coverage of passive microwave brightness temperatures in two days or less (except for regions of dense vegetation or frozen ground cover), with a nominal spatial resolution of 25 km. These observed brightness temperatures can be related to near-surface (~ 1cm at C-band and a few mm at X-band) soil moisture using a geophysical emissivity model.

This paper presents an evaluation of AMSR-E derived near-surface soil moisture over Australia, within the context of its potential for assimilation into Australian NWP systems. If the remotely sensed data is shown to be more accurate than existing initialized soil moisture fields, then its assimilation is expected to readily benefit modelled soil moisture. Surface ( 0 – 7 cm) soil moisture from the newly initialised LAPS model has then been adopted as the benchmark against which the AMSR-E derived soil moisture is assessed. Two different AMSR-E soil moisture retrieval models are considered here: one developed collaboratively by Vrije Universiteit Amsterdam (VUA) and NASA (VUA-NASA), following Owe et al (2007), and one developed at NASA, following Njoku and Chan (2003). The VUA-NASA retrieval algorithm has been separately applied to C-band (6.92 GHz) and X-band (10.65 GHz) AMSR-E brightness temperatures (referred to below as VUA-NASA-C and VUA-NASA-X). Due to radio frequency interference in C-band frequencies across North America, the NASA product (NASA-X) is based on the higher frequency X-band AMSR-E data only, which is considered to be less appropriate for soil moisture sensing.

6

Methods and data Timeseries of AMSR-E and LAPS soil moisture have been evaluated against in-situ soil moisture data from ten sites within the Murrumbidgee Soil Moisture Monitoring Network (MSMMN; see http://www.oznet.unimelb.edu.au for details), in southeast Australia. Two of these sites multiple monitoring stations (five at Kyeamba Creek and four at Adelong Creek), and a simple average has been used at each of these. There are differences between the spatial scale of soil moisture obtained from models, remote sensing, and ground-based monitoring stations (or calculated from the average of a modest number of stations in the case of Kyeamba and Adelong). While these differences will lead to differences in the behaviour of soil moisture from each source, their temporal behaviour should still be similar. Consequently, inter-comparison of soil moisture from LAPS, AMSR-E, and the MSMMN has focussed on comparison of their temporal dynamics. To better enable this comparison, the LAPS and AMSR-E soil moisture timeseries have been re-scaled so that their range over 2005 matches that of the MSMMN data range at each location (the range is defined here as difference between the 5th and 95th percentile).

Prior to the timeseries comparison, the LAPS and AMSR-E soil moisture have been visually compared to maps of (24-hour) antecedent precipitation across Australia, to check that the broad spatial patterns described by each are realistic. This comparison is based on the Bureau’s daily 0.25° rain gauge analysis (Weymouth et al, 1999).

Results

Figure 1 shows maps of the AMSR-E and LAPS soil moisture, together with the previous day’s precipitation for 6 January, 2005. All of the soil moisture panels show the expected climatological distribution of soil moisture across Australia, with a superimposed region of wetter soil associated with a rain-band south of the Gulf of Carpentaria. The VUA-NASA C- and X- band retrievals are similar, while NASA-X shows comparatively dry soils, with little variability. This is typical of the NASA-X soil moisture, which has a strong tendency to remain low. All three AMSR-E panels in

Figure 1: Maps of soil moisture on 6 January, 2005: (left to right) VUA-NASA-C, VUA-NASA-X, NASA-X, and LAPS. The black lines are 20 mm precipitation contours from the previous 24 hours. The soil moisture colour range in each panel describes the range of each product (0 – 0.5 vol/vol from AMSR-E, and 0.17 – 0.32 vol/vol for LAPS.

7

Figure 1 show several small in-land moist regions which are not associated with recent rainfall. These false moist regions are persistent through time and are caused by high surface salinity. For example, the yellow and green region to the north of Spencer Gulf in South Australia is the Lake Torrens salt pan. In contrast to AMSR-E, the LAPS soil moisture shows a broader region of wet soil, which is associated with rain several days previously (indicating a longer memory in the LAPS soil moisture). Additionally, the LAPS soil moisture has less fine detail than the AMSR-E fields (more so than can be attributed simply to their different resolutions).

Figure 2 compares the soil moisture timeseries from AMSR-E and LAPS to MSMMN data for the Adelong and Kyeamba Creek catchments (these are the two MSMMN sites with multiple monitoring stations). As with the spatial comparison above, the greatest differences between the AMSR-E soil moisture retrievals are the result of using different retrieval algorithms, rather than different observation frequencies. Again, the VUA-NASA products appear more realistic. All three AMSR-E soil moisture timeseries show the rapid soil moisture increase brought on by heavy rains in June, while the subsequent dry-down (and intermittent precipitation-induced increases) are well represented by only VUA-NASA-C, and to a slightly lesser extent by VUA-NASA-X. There is an artificial drift upwards in the three AMSR-E soil moisture products during the dry autumn months, when the MSMMN data is steady. Possible reasons for this drift are currently being investigated. In comparison to the AMSR-E fields, the LAPS soil moisture is extremely noisy, with the noise having a similar amplitude to the seasonal signal.

The relative performance of the LAPS and AMSR-E soil moisture fields was consistent across the ten MSMMN sites (only Kyeamba and Adelong are shown here). In summary, VUA-NASA-C has the best fit to the MSMMN data, with an average correlation coefficient across the ten MSSMN sites of 0.79, with VUA-NASA-X performing nearly as well, with an average of 0.77. In contrast, NASA-X and LAPS showed much poorer predictive skill, with average correlation coefficients of 0.54 of 0.58, respectively. The RMSE statistics (assuming the MSMMN data to be the truth) indicate a similar result, with low average RMSE for VUA-NASA-C and VUA-NASA-X (0.031 and 0.034 vol/vol), and a substantially higher value for NASA-X and LAPS (0.048 and 0.043 vol/vol ).

Discussion and conclusions The VUA-NASA soil moisture derived from either C- or X- band AMSR-E brightness temperatures appear to be realistic, and both offer substantial improvement over the current LAPS soil moisture initialisation scheme. Both show a strong correlation with the MSMMN soil moisture timeseries, and a good spatial agreement with precipitation data. The agreement between the VUA-NASA soil moisture products and the MSMMN is remarkably good, given the spatial differences between remotely sensed and ground-based (point) measurements. While the difference is minimal, VUA-NASA-C performed slightly better than VUA-NASA-X with better overall statistics. The C-band product is also theoretically superior, and it is recommended that this product be used in Australian applications. In contrast to the VUA-NASA products, the soil moisture produced by NASA from the AMSR-E X-band is less realistic. Since the VUA-NASA X-band product has

Figure 2: Comparison of LAPS and AMSR-E derived soil moisture to MSMMN observations for 2005, at Adelong (upper) and Kyeamba (lower) Creek catchments.

8

compared favourably, the poor performance of the NASA product should not be contributed to its use of a sub-optimal microwave frequency.

While the broad spatial behaviour of AMSR-E and LAPS soil moisture has been checked for realism by comparison to precipitation, quantitative assessment here has been limited to a handful of locations, and these results do not necessarily extrapolate to other regions. However, the main findings regarding the relative performance of the different soil moisture products was consistent across all of the MSMMN sites, and was also supported by comparison to precipitation data. The superior performance of both VUA-NASA products over the NASA-X product also concurs with the findings of Wagner et al (2007), based on soil moisture data from Spain.

An experimental EnKF system is being developed to assimilate the VUA-NASA C-band soil moisture into the ACCESS NWP model. While assimilating soil moisture data into the model is expected to improve the modelled soil moisture, improved atmospheric forecasts may not automatically follow. Hurdles to achieving this improvement include inconsistencies between the definition of soil moisture in the observations and in the model, and inaccuracies in the land surface model physics (specifically, the parametrisations describing the dependencies between soil moisture and the land surface fluxes). However, assimilating realistic soil moisture data into an NWP model will enable better identification, and hence treatment, of these inaccurate model physics. The assimilation is (at the least) a first step towards improved land surface flux forecasts, and hence improved low-level atmospheric forecasts.

References Draper. C.S, and Mills, G.A, 2007: The atmospheric water balance over the semi-arid Murray-

Darling River basin, Journal of Hydrometeorology, in press. Ellett, K. M., J. Walker, M. Rodell, J. Chen, and A. Western, 2005: GRACE gravity fields as a new

measure for assessing large-scale hydrological models. Proceedings. MODSIM 2005 International Congress on Modelling and Simulation, Modelling and Simulation Society of Australia and New Zealand, Melbourne, Australia, 2911–2917.

Njoku, E., T. Jackson, V. Lakshmi, T. Chan, and S. Nghiem. (2003), Soil moisture retrieval from AMSR-E. IEEE Transactions on Geoscience and Remote Sensing, 41(2), 215–229.

Owe, M., R. A. M. de Jeu, and T. Holmes (2007), Multi-sensor historical climatology of satellite-derived global land surface moisture, Journal of Geophysical Resources, in press.

Pescod, N. (1994), A four-parameter, three-layer model of soil moisture based on hydraulic properties of the soil in the absence of vegetation, in J. Jasper and P. Meighen (eds), Bureau of Meteorology Research Report, No 46. Parametrisation of Physical Processes: Papers Presented at the 5th BMRC Modelling Workshop, 101–106.

Reichle, R. and R. Koster. (2005), Global assimilation of satellite surface soil moisture retrievals into the NASA catchment land model. Geophysical Research Letters, 32.

Viterbo, P. and A. Beljaars. (1995), An improved land surface parameterization scheme in the ECMWF model and its validation. Journal of Climate, 8, 2716–2745.

Wagner, W., V. Naeimi, K. Scipal, R. de Jeu, and J. Martinez-Fernandez. (2007), Soil moisture from operational meteorological satellites. Hydrogeology Journal, 15, 121–131.

Weymouth, G., G. Mills, D. Jones, E. Ebert, and M. Manton. (1999), A continental-scale daily rainfall analysis system. Australian Meteorological Magazine, 48(3), 169–179.

9

MoistureMap: A soil moisture monitoring, prediction and reporting system for sustainable land and water management

J. Walker1, D. Barrett2, R. Gurney3, J. Kalma4, Y. Kerr5, E. Kim6 and J. LeMarshall7

1 Department of Civil and Environmental Engineering, University of Melbourne, Australia 2 CSIRO Land and Water, Australia

3 NERC Environmental Systems Science Center, University of Reading, United Kingdom 4 Discipline of Civil, Surveying and Environmental Engineering, University of Newcastle, Australia

5 Biospheric Processes, CESBIO, France 6 Hydrospheric and Biospheric Sciences Laboratory, NASA Goddard Space Flight Center, United States

7 Bureau of Meteorology Research Center, Bureau of Meteorology, Australia

Introduction Accurate knowledge of current and future spatial variation in surface and root zone soil moisture at high resolution is critical for achieving sustainable land and water management. The fundamental limitation is that spatial and temporal variation in soil moisture is not well known, nor easy to measure or predict. Consequently, the recently funded project described here seeks to develop a prototype soil moisture monitoring, prediction and reporting system (MoistureMap) for Australia, with the Murrumbidgee as the demonstration catchment. The system will provide current and future soil moisture information and its uncertainty at 1km resolution, by combining weather, climate and land surface model predictions with soil moisture data from the European Space Agency’s Soil Moisture and Ocean Salinity (SMOS) satellite scheduled for launch in October 2008; the first-ever dedicated microwave soil moisture mission.

The unique feature of this project as compared to existing projects, such as the “Australian Water Availability Project” of the Australian Bureau of Rural Sciences and the “Short Term Climate Outlooks” of the Centre for Ocean, Land and Atmospheres in the USA, is that microwave observations offer a direct measure of surface soil moisture, rather than the indirect measure that thermal observations and energy balance algorithms or rainfall forecasts provide. It is widely recognised that passive microwave is the most promising remote sensing method for soil moisture measurement [Njoku et al., 2002].

While passive microwave remote sensing holds the most promise for cost-effective surface soil moisture mapping, it will not directly provide high resolution observations or information about the deeper soil moisture. Such information can be gleaned from models, but it is limited by the fidelity of model physics, parameter estimates, and atmospheric forcing data. Consequently, remotely sensed observations and model predictions must be combined by data assimilation [Walker and Houser, 2005], with point measurements used for verification. MoistureMap will provide near-real-time monitoring and prediction of soil moisture and its uncertainty at 1km resolution, by assimilating (i) low-resolution passive microwave-derived surface soil moisture (SMOS) and (ii) high-resolution thermal infrared skin temperature (MODIS) observations into a land surface model using weather and climate model ensemble forecasts from the Bureau of Meteorology

Approach The key elements of this project are outlined in Fig. 1. Specifically, it will develop and test innovative techniques for monitoring, prediction and reporting of 1km resolution soil moisture content from extensive ground-, air- and space-based measurements for Australian conditions. Data collected in the field program will consist of: (i) long-term monitoring of soil moisture profiles and supporting data; (ii) air-borne radiobrightness, thermal infrared, shortwave infrared and visible data; and (iii) satellite radiobrightness, thermal infrared, shortwave infrared and visible data from the SMOS and MODIS sensors. While this proposal will rely heavily upon the comprehensive NAFE data sets collected from recent air-borne field experiments in the Murrumbidgee (and Goulburn) Catchments [Walker et al., 2007], additional air-borne and ground-based data collection coincident with SMOS overpasses will be needed to verify the results obtained from real SMOS data. Consequently, MoistureMap and the five projects which underpin its development all leverage off the same air-borne and ground-based monitoring data to be collected.

Study Regions and Data Sets The field work required by this project will be undertaken in three study areas with distinctly different characteristics. These are the Simpson Desert and Wet Tropics World Heritage Area for SMOS calibration studies, and the Murrumbidgee Catchment for SMOS verification, and demonstration and verification of MoistureMap and the projects that underpin it.

10

SMOS Radiobrightness and Supporting Measurements: The SMOS satellite will have a spatial resolution of about 50km, with a 6:00am/pm overpass once every 2 to 3 days [Kerr et al, 2001]. A unique feature of SMOS is its vertical and horizontal polarisation measurements of radiobrightness at a range of incidence angles for each pixel. Thermal infrared, shortwave infrared and visible data will be provided by the MODIS satellite, which has daily 1km resolution and a 10:30am/pm overpass time. Landuse classification and landcover (vegetation) information will also be provided from MODIS. Thermal data at 4km resolution coincident with SMOS overpasses will be obtained from the hourly MTSAT-1R satellite.

Air-borne Radiobrightness and Supporting Measurements: Air-borne measurements will be made using the Polarimetric L-band Multibeam Radiometer (PLMR) and a thermal infrared imager, together with visible and shortwave infrared measurements, coincident with SMOS overpasses. An integral part of this proposal are calibration campaigns in the Simpson Desert and Wet Tropics, and verification campaigns for a 100km wide transect along the length of the Murrumbidgee Catchment (Fig. 2) under different seasonal and climatic conditions. This 1km resolution air-borne data is a critical component for verification of the 1km MoistureMap predictions of near-surface soil moisture.

Simpson Desert and Wet Tropics World Heritage Area: Deserts, tropical forests and the Antarctic are purported to have spatially and temporally consistent microwave emission characteristics, making them suitable ground-based targets for passive microwave calibration [Ferrazzoli et al., 2002; Macelloni et al., 2006]; a hypothesis that Project 1 will test. Consequently, a 100km area of the Simpson Desert and a 30km area of the Wet Tropics World Heritage Area just north of Port Douglas (the largest expanse of tropical forest that still exists in Australia) have been chosen for the SMOS calibration study; funding to test the Antarctic site is still being sought. Additional to the air-borne data described above, there will be coincident ground and high resolution air-borne characterisation across a small area of these sites. Such data will include soil moisture and temperature, and vegetation biomass, temperature, structure and roughness using ground measurements and a full wave-transform LIDAR.

Data Assimilation

Pro

ject

1SM

OS

Cal

ibra

tion

and

Ver

ific

atio

n

SMO

S Si

mul

ator

1km LanduseLandcover Data

Project 4Multi-angle Retrieval

Rainfall and ET Data

BoM

Rea

l-tim

e O

bser

ving

N

etw

ork

BoM

7-da

y N

WP

Fore

cast

BoM

Seas

onal

Clim

ate

Fore

cast

50km

SM

OS

Dat

a

1km

MO

DIS

Dat

a

Project 5Soil Property Estimator

Project 2Mixed-pixel Retrieval

Project 3Multi-sensor Retrieval

Project 4Multi-angle Retrieval

MoistureMapA soil moisture monitoring, prediction and reporting system for land management

1km

Soi

l Moi

stur

e P

redi

ctio

n M

odel

Fig. 1: MoistureMap and its underpinning projects. Project 1 provides calibrated and verified SMOS data for the Australian environment, Projects 2 to 4 develop components of an advanced soil moisture retrieval algorithm (SMOS Simulator) for Australian conditions, and Project 5 derives Australian soil hydraulic properties for soil moisture retrieval and modelling. MoistureMap delivers the current soil moisture information by fusion of model predictions and observations, and future soil moisture information from ensemble weather and climate forecast propagation.

Fig. 2: The Murrumbidgee Catchment showing location of soil moisture monitoring sites, variation in topography, and transect (100km pixels) to be used for verification purposes.

11

Murrumbidgee Catchment: The Murrumbidgee Catchment has been instrumented and monitored for soil moisture and supporting data for more than 5 years. The existing network of monitoring sites (Fig. 2), data sets, and detailed knowledge of the catchment provide an ideal basis for the field work and data requirements of this research. Moreover, the diverse climatic, topographic and land cover characteristics make it an excellent demonstration test-bed for SMOS Simulator and MoistureMap developments. No intensive ground data collection is planned during the air-borne transect campaigns, as they will be underpinned by the monitoring network.

National Air-borne Field Experiment (NAFE) Dataset: NAFE’06 was undertaken in the Murrumbidgee Catchment during November 2006 [Walker et al., 2007], providing an ideal data set for SMOS soil moisture retrieval and data assimilation developments. The study area included the Yanco area (which contains the Coleambally Irrigation Area) and a 40km x 50km area containing the Kyeamba Catchment (and the city of Wagga Wagga). PLMR and the thermal infrared imager were flown to provide 1km resolution passive microwave data across the Yanco area every 2 to 3 days (and Kyeamba catchment weekly) for a period of 3 weeks, together with extensive ground monitoring. This both simulated a SMOS pixel and provided 1km soil moisture data for surface soil moisture assimilation and downscaling techniques to be tested with remote sensing data which is consistent with that from MODIS and SMOS. Additionally, multi-angular passive microwave data were collected for SMOS algorithm development. Consequently, this data set provides an ideal test-bed for initial development of MoistureMap and Projects 2 to 5, with final testing to be conducted using SMOS, Murrumbidgee Monitoring Network and air-borne transect data.

SMOS Calibration and Verification, the SMOS Simulator, and MoistureMap The campaigns described above provide the necessary ground-, air- and space-borne data for SMOS calibration and verification studies (Project 1), and verification data of both the SMOS Simulator (Projects 2 to 5) and MoistureMap.

Project 1 – SMOS Calibration and Evaluation: This project will test the hypothesis that desert, rainforest and Antarctic sites are suitable for ground-based SMOS calibration. Aircraft and supporting ground data, together with radiobrightness modelling, will be used to understand spatial and temporal variability in microwave emission from these targets. The campaign observations for the desert (and Antarctic if funded) sites will be compared to SMOS overpass data; as the rainforest site is not suitable for direct comparison with SMOS, knowledge gleaned from the Wet Tropics site will be explored for its transferability to the larger Amazon rainforest. A timeseries of SMOS data will also be analysed for the Simpson, Amazon and Antarctic sites. Finally, SMOS radiances and official SMOS soil moisture products will be compared with coincident PLMR radiances and derived soil moisture content for the Murrumbidgee transect.

Project 2 – Mixed-Pixel Retrieval: This project will test the hypothesis that high resolution optical data can be used to account for the mixed-pixel response that will confound passive microwave retrieval of soil moisture throughout much of Australia. In addition to agricultural and natural variations in land cover, most agriculturally significant parts of Australia contain urban areas and/or standing water across 50km scales; it has been shown that as little as 3% standing water coverage leads to more than 4%v/v error in derived soil moisture [Walker et al., 2006] – the target accuracy for SMOS soil moisture retrieval. Consequently, methods will be developed to account for crop, pasture, urban and water (farm dams, rice paddies, etc) emission contributions to the overall SMOS microwave response. The wide range of landcover characteristics in the Murrumbidgee will provide a thorough test of SMOS retrieval algorithms. The retrieval algorithms will be developed using NAFE’06 data and verified across the Murrumbidgee Catchment using Murrumbidgee transect flight data.

Project 3 – Multi-Sensor Retrieval: This project will test the hypothesis that more accurate soil moisture information can be derived from SMOS if vegetation and soil temperature information are derived from other coincident remote sensing observations at higher resolution. Current state-of-the-art retrieval algorithms require ancillary information on soil properties, surface soil temperature and vegetation water content in order to estimate soil moisture from horizontally polarised passive microwave data. The current approach has been to estimate the vegetation water content using vertically polarised passive microwave data, and soil temperature from 37GHz microwave data [Owe et al., 2001] or model predictions, but there will be no 37GHz observations coincident with SMOS and model estimates are inaccurate [Betts et al., 2003]. However, preliminary analysis of the NAFE’05 dataset has shown that vegetation and temperature data can be estimated with sufficient accuracy from independent visible, shortwave infrared and thermal infrared data, with an overall absolute soil moisture retrieval error better than 6% v/v [Maggioni et al., 2006]. This project will further develop and test algorithms for estimating vegetation water content and soil temperature using NAFE data, apply them to SMOS using visible and shortwave infrared data from MODIS and coincident thermal infrared data from MTSAT1R, and verify the retrieved moisture with Murrumbidgee transect flight data.

Project 4 – Multi-Angle Retrieval: This project will test the hypothesis that more accurate soil moisture information can be derived from SMOS if we take advantage of its unique multi-incidence angle capability.

12

Microwave emission from the land surface varies as a function of viewing angle. Consequently, there is great potential for retrieving more accurate ancillary data (soil properties, vegetation water content, surface temperature etc.) required by current state-of-the-art retrieval algorithms, and/or reducing the error in derived soil moisture content by simultaneously using all the information that exists in multi-angle observations. A retrieval algorithm will be developed using the multi-angle data collected during the NAFE campaigns, applied to SMOS data, and tested using Murrumbidgee transect flight data.

Project 5 – Soil Property Estimator: This project will test the hypothesis that more accurate soil hydraulic properties can be derived from remotely sensed passive microwave and/or time series soil moisture observations, leading to more accurate soil moisture monitoring and prediction. Specifically, this project will develop techniques to estimate soil wilting point, porosity, hydraulic conductivity and characteristic curve parameters. Methods for soil property estimation will include (i) temporal analysis of timeseries surface soil moisture remote sensing information, (ii) model calibration using, and model assimilation of, timeseries surface soil moisture observations, and (iii) multi-incidence angle and multi-sensor retrieval approaches leveraging developments in Projects 3 and 4.

MoistureMap: This is the framework through which the basic research of Projects 1 to 5 is transferred to a specific application. MoistureMap’s development consists of three key components: (i) the soil moisture prediction model, (ii) observed and forecast meteorological data, and (iii) the data assimilation component which contains the SMOS Simulator.

The soil moisture prediction model to be used is CSIRO Atmosphere Biosphere Land Exchange (CABLE), a column model based on Richards’ equation that simulates water and energy fluxes between a vertical profile of six soil layers, land surface, vegetation and the atmosphere [Kowalczyk et al., 2006]. This model is ideally suited to the assimilation requirements of this project due to its prediction of surface and root zone soil moisture together with surface soil and vegetation temperature, which are necessary for radiance and thermal data assimilation. Moreover, its grid-based structure makes the 1km application relatively straight forward.

Near-real-time observational data will be provided by WRON and the BoM’s existing observational grid-based data streams, and forecast data will be derived from ACCESS. ACCESS will also be run retrospectively from 2005 to enable testing and development with NAFE data, and will provide the meteorological ensemble forecast data out to 7-days and 3-months [Kamal Puri, pers. comm.].

The SMOS Simulator will be based on algorithm developments from Projects 2 to 4, together with soils data from Project 5. This will simulate radiobrightness data for comparison with SMOS observations and ultimate assimilation in CABLE. We will use a Bayesian framework to combine the SMOS and MODIS observations with the soil moisture knowledge embodied in an ensemble of CABLE simulations. Initial MoistureMap development will be based on NAFE data followed by extensive verification using Murrumbidgee Monitoring Network and Murrumbidgee transect data.

References Betts AK, JH Ball and P Viterbo, 2003. Evaluation of the ERA-40 surface water budget and surface temperature for the

Mackenzie River basin. J Hydromet. 4(6), 1194-1211. Ferrazzoli, P, L Guerriero and JP Wigneron, 2002. Simulating L-band emission of forests in view of future satellite

applications. IEEE Trans. Geosci. Rem. Sens., 40(12), 2700- 2708. Kerr, YH, P Waldteufel, JP Wigneron, J Martinuzzi, J Font and M Berger, 2001. Soil moisture retrieval from space: the

Soil Moisture and Ocean Salinity (SMOS) mission. IEEE Trans. Geosci. Rem. Sens., 39(8): 1729-1735. Kowalczyk, EA, YP Wang, RM Law, JL McGregor and G Abramowitz, 2006. CSIRO Atmosphere Biosphere Land

Exchange model for use in climate models and as an offline model. CSIRO Technical Report. Macelloni, G, Brogioni, M, Pampaloni, P, 2006. DOMEX 2004: An experimental campaign at Dome-C Antarctica for

the calibration of space-borne low–frequency microwave radiometers, IEEE Trans. Geosci. Rem. Sens., 44(10): 2642-2653.

Maggioni, V, R Panciera, JP Walker, M Rinaldi and V Paruscio, 2006. A multi-sensor approach for high resolution airborne soil moisture mapping. Proceedings 30th Hydrology and Water Resources Symposium.

Njoku, EG, WJ Wilson, SH Yueh, SJ et al., 2002. Observations of soil moisture using a passive and active low-frequency microwave airborne sensor during SGP99. IEEE Trans. Geosci. Rem. Sens., 40(12), 2659-2673.

Owe, M, R de Jeu, and JP Walker, 2001. A methodology for surface soil moisture and vegetation optical depth retrieval using the microwave polarization difference index. IEEE Trans. Geosci. and Rem. Sens., 39(8), 1643-1654.

Walker, JP and PR Houser, 2005. Hydrologic data assimilation, In: A. Aswathanarayana (Ed.), Advances in Water Science Methodologies, A.A. Balkema, The Netherlands, 230pp.

Walker, J.P., Merlin, O., Panciera, R. and Kalma, J.D., 2006. National Airborne Field Experiments for soil moisture remote sensing, Proceedings 30th Hydrology and Water Resources Symposium.

Walker, JP, and 62 others, 2007. The National Airborne Field Experiment data sets. Proceedings MODSIM 2007 International Congress on Modelling and Simulation.

13

Land surface modeling and data assimilation

Xu Liang

Department of Civil and Environmental Engineering

University of Pittsburgh

Pittsburgh, PA 15261

Abstract Hydrologic processes, such as evapotranspiration, surface and subsurface runoff, groundwater

recharge, surface water and groundwater interactions, snow, and river network routing, are

critical components of the land surface water and energy budgets. Realistic simulations of these

processes in climate and hydrologic coupled models are essential for correctly representing land-

atmosphere feedback, providing accurate climate predictions (such as flood, drought),

understanding global water cycle, planning and managing water resources, and studying

environment sustainability. Past applications of land surface modeling in the land-atmosphere

system have not always been successful. Challenges arise from identifying and representing

dominant hydrologic processes for applications over large spatial domains, and from considering

subgrid spatial variability associated with the hydrologic processes. In this talk, several steps

towards addressing these challenges will be discussed. These steps are grouped into two

categories (1) representations of important hydrologic processes, and the consideration of

subgrid spatial variability associated with these processes; and (2) utilization of remote sensing

observations in conjunction with model simulations through a new approach of data assimilation.

This new data assimilation approach explicitly considers spatial correlation structures, error

propagation, and dissimilar spatial resolutions. Significance of these steps will be presented

through examples.

15

Evaluating CABLE soil moisture predictions in the Murray Darling Basin

Adam Smith1, 2, and Huqiang Zhang2

1. Water Division, Bureau of Meteorology, Melbourne, Victoria, Australia. 2. CAWCR, Bureau of Meteorology, Melbourne, Victoria, Australia.

Introduction The CSIRO Atmospheric Biosphere Land Exchange (CABLE) model is the land surface model (LSM) used in the Australian Community Climate Earth System Simulator (ACCESS). ACCESS is to be used by both the Australian Bureau of Meteorology and the Commonwealth Scientific and Industrial Research Organisation (CSIRO) for operational numerical weather prediction (NWP) and climate studies (both regional and global). Currently VB95 (Viterbo and Beljaars, 1995), an older version of the European Centre for Medium-Range Weather Forecasting (ECMWF) LSM the Tiled ECMWF Scheme for Surface Exchanges over Land (TESSEL) (Van den Hurk et al. 2000), is used operationally at the Bureau of Meteorology for NWP. Richter et al. (2004) assessed the sensitivity of VB95 to soil and vegetation parameters via offline simulations at 10 locations in the temperate Murrumbidgee River Catchment (NSW, Australia). Similarly, using observed soil moisture data at the same 10 locations in the Murrumbidgee Catchment (located within the Murray Darling Basin) this work assesses the impact of replacing the default CABLE soil parameters (taken from the Zobler (1986) global 1° resolution soil map) with parameters derived from the 1:2,000,000 scale Digital Atlas of Australian Soils.

Figure 1: The Murrumibidgee River Catchment located within the Murray Darling Basin. Yellow dots are soil moisture monitoring sites used to evaluate CABLE. Kyeamba and Adelong Catchments contain further monitoring sites not used in this evaluation.

16

Model The CSIRO Atmosphere Biosphere Land Exchange (CABLE) land surface model is a 3rd generation model in the classification of Pitman (2003). CABLE has 6 layers for solution of soil moisture and heat using Richard’s and the heat equation respectively (with layer depths from the surface of 2.2., 8, 23.4, 64.3, 172.8 and 360 cm), and a 3 layer snowpack model that solves for albedo at the surface, as well as temperature, density and thickness of each layer, additionally permafrost (frozen soil) is modelled (Kowalczyk et al. 2006). CABLE has a single (above ground) two “big” leaf (shaded and sun-lit) canopy for calculation of stomatal conductance, photosynthesis and leaf temperature (for each “big” leaf) (Wang & Leuning 1998; Wang 2000), and a turbulence model to calculate within canopy air temperature and humidity (Raupach et al. 1997).

Data Forcing data is needed to run a land surface model offline (i.e. uncoupled from its parent atmospheric model). This data was derived from 15 automatic weather station (AWS) sites, 11 within and 4 in the near vicinity of the Murrumbidgee River Catchment. The AWS data is complemented by manual observations of cloud cover and sunshine hours, and in-situ and satellite observations of radiation (Siriwardena et al. 2003). This offline forcing data set begins in January 2000 and is updated every six months, for this study January 2000 to December 2006 is used.

Other ancillary data that is required by CABLE includes Leaf Area Index (LAI), dominant vegetation and soil type. Mean monthly LAI is calculated from remotely sensed 0.05° monthly average woody and herbaceous fractional cover and LAI for the period 1981-1994 (Lu et al. 2003). Dominant vegetation and soil types are given by the (CABLE default) global 2° resolution maps of 13 vegetation (Potter et al. 1993) and 9 soil classes (Zobler 1986), where both of these datasets have been aggregated from the 1° originals. The Zobler (1986) map is in turn based on the 1:5,000,000 FAO/UNESCO Soil Map of the World.

The soil moisture monitoring sites used to evaluate CABLE are located in the Murrumbidgee Catchment (Fig. 1). This catchment is part of the Murray Darling Basin, contains the Australian Capital Territory (ACT) and has land use varying from the Snowy Mountain Hydro-Electric Scheme in the high elevation of the south east of the catchment to the Colleambally and Murrumbidgee Irrigation Areas (CIA and MIA) in the planes of the west.

Default Global Soil Dataset Results CABLE simulates soil moisture in the top 90 cm fairly well. In particular the temporal dynamics in the 0-7 & 0-30 cm observations are well represented at most sites, however the wilting point (minimum soil moisture) in CABLE simulations is consistently too high (see top two panels of Figure 2). There is a bias in all CABLE simulations of the (lower) 30-60 and 60-90 cm soil moisture (Figure 2 is one example). This is due to the fact that the soils of the Murrumbidgee sites are a duplex type with a clear transition between an A and a B horizon at a depth of about 30 cm. CABLE represents the entire soil profile as a single soil type, due in part to the soil moisture based formulation of Richard’s Equation employed by the model to simulate water flux.

High Resolution Atlas of Australian Soils Results Replacing the default soil parameters with those derived from the Atlas did not result in any model improvement, in fact at the majority of sites (6 of 10) model performance was severely degraded (Figure 3 is one example). This is due to the parameters used for the Atlas runs being created as a depth weighted average of the A and B horizon parameters. This gave most weighting to the B horizon, because the A horizon has a depth of 20 or 30 cm, whereas the B horizon depth ranges from 50 to 110 cm. Upon further inspection it was found that the B horizon parameters had been flagged as unreliable at the 6 sites where the Atlas parameters resulted in degraded model performance. This unreliability was determined by the calculated b (or ? S) being greater than 26 (or less than -0.120 bar) (McKenzie et al. 2000). Using these B horizon parameters consequently led to an unrealistically high estimate of b (and low ? S) being input to the model. Clapp and Hornberger (1978) give a typical b value of 11.4 (with a standard deviation of 3.7) for clay however the B horizon b value for West Wyalong is 55.6! The depth averaged b parameters were consistent with Richter et al. (2004) and in the range 6 to 20.6, whereas the A horizon parameters are consistent with the Clapp and Hornberger (1978) estimates (an observed range of 5.1 at Canberra to 11.3 at Balranald).

17

Figure 2: Kyeamba soil moisture using the default Zobler (1986) soil dataset. Green line is predicted and red line is observed soil moisture. Top panel is 0 - 8 cm prediction and 0 - 7 cm observation, 2nd panel is 0 - 23.4 cm prediction and 0 - 30 cm observation, 3rd panel is 23.4 - 64.3 cm prediction and 30 - 60 cm observation, and bottom panel is 64.3 - 172.8 cm prediction and 60 - 90 cm observation.

Figure 3: Same as Figure 2 but for the Australian Atlas soil dataset.

18

Conclusions Overall the performance of CABLE in simulating soil moisture (with the default Zobler (1986) soil parameters) is quite satisfactory. Replacing the default parameters with ones derived from the Australian Atlas of Soils resulted in degraded model performance at the majority of sites. This is due to unreliable B horizon parameters being used to create a depth weighted average, it is believed that using only the A horizon parameters will result in considerable improvement of the Atlas results. The default predictions matched the variability in observed soil moisture quite well (particularly for the surface 0-7 and 0-30 cm soil moisture) however a bias was clear in the deeper soil moisture predictions (30-60 and 60-90cm) due to the (duplex) soil profile being represented by one soil type in the model. In comparison to the default simulations the soil moisture predictions from the Atlas reduced the deeper soil moisture (30-60 and 60-90 cm) bias (due to the B horizon being included in the Atlas parameter estimates), however this improvement was swamped by the woeful simulation of the surface (0-7 and 0-30 cm) soil moisture (due to the b parameter being far too high). At this stage it is recommended to continue using the default (Zobler 1986) global soil dataset for the higher spatial resolution requirements of NWP and regional climate simulations.

References Clapp, R.B. and Hornberger, G.M. 1978. Empirical equations for some soil hydraulic properties. Water

Resour. Res., 14, 601-604. Kowalczyk, E.A., Wang, Y.P., Law, R.M., Davies, H.L., McGregor, J.L. and Abramowitz, G. 2006. The

CSIRO Atmosphere Biosphere Land Exchange (CABLE) model for use in climate models and as an offline model, CSIRO Marine and Atmospheric Research Paper 013, 37 pp.

Lu, H., Raupach, M.R., McVicar, T.R. and Barrett, D.J. 2003. Decomposition of vegetation cover into woody and herbaceous components using AVHRR NDVI time series, Remote Sens. Environ., 86, 1-18.

McKenzie, N.J., Jacquier, D.W., Ashton, L.J. and Cresswell, H.P. 2000. Estimation of soil properties using the Atlas of Australian Soils, CSIRO Land and Water Technical Report 11/00, 23 pp.

Pitman, A.J. 2003. The evolution of, and revolution in, land surface schemes designed for climate models. Int. J. Climatol., 23, 479-510.

Potter, C.S., Randerson, J.T., Field, C.B., Matson, P.A., Vitousek, P.M., Mooney, H.A. and Klooster, S.A. 1993. Terrestrial ecosystem production: A process model based on global satellite and surface data. Global Biogeochem. Cycles, 7, 811-841.

Raupach, M.R., Finkele, K. and Zhang, L. 1997. SCAM (Soil-Canopy-Atmosphere Model): Description and comparison with field data, CSIRO Centre for Environmental Mechanics Technical Report 132, 81 pp.

Richter, H., Western, A.W. and Chiew F.H.S. 2004. The effect of soil and vegetation parameters in the ECMWF land surface scheme, J. Hydrometeor., 5, 1131-1146.

Siriwardena, L., Chiew, F., Richter, H. and Western, A. 2003. Preparation of a climate data set for the Murrumbidgee River catchment for land surface modelling experiments, Cooperative Research Centre for Catchment Hydrology Working Document 03/1, 50 pp.

Van den Hurk, B.J.M.M., Viterbo, P., Beljaars, A.C.M., and Betts, A.K. 2000. Offline validation of the ERA40 surface scheme, ECMWF Technical Memorandum 295, 42 pp.

Viterbo, P., and Beljaars, A.C.M. 1995. An improved land surface parameterization scheme in the ECMWF model and its validation, J. Clim., 8, 2716-2748.

Wang Y.P. and Leuning, R. 1998. A two-leaf model for canopy conductance, photosynthesis and partitioning of available energy I. Model description and comparison with a multilayered model. Agric. Forest Meteor., 91, 89-111.

Wang Y.P. 2000. A refinement to the two-leaf model for calculating canopy photosynthesis. Agric. Forest Meteor., 101, 143-150.

Zobler, L. 1986. A world soil file for global climate modelling. NASA Technical Memorandum 87802, 32 pp.

19

The impact of land-atmosphere initialisation on dynamical seasonal prediction

Debbie Hudson and Oscar Alves

Centre for Australian Climate and Weather Research – A partnership between the Bureau of Meteorology and CSIRO, Melbourne, Victoria, Australia

Introduction This study presents a new Atmosphere and Land Initialisation scheme (ALI) for POAMA (Predictive Ocean Atmosphere Model for Australia). POAMA is a coupled ocean/atmosphere model seasonal forecast system, and the first version, POAMA-1 (Alves et al., 2003), has been run operationally by the Bureau of Meteorology since 2002. For atmospheric initialisation, the POAMA-1 system uses data directly from the NWP forecast system for the real-time forecasts, and data from an AMIP-style atmosphere-only simulation (referred to here as BAM AMIP) for hindcasts. The initial conditions for the hindcasts thus contain observed atmospheric information which is related to sea-surface temperature, but they do not capture the true intra-seasonal state. This is true also for the initialisation of the land surface, which uses an AMIP-style climatology for both the hindcasts and forecasts. A new version of POAMA, POAMA1.5, is currently undergoing operational trials and uses atmospheric and land initialization from the ALI scheme. ALI involves the creation of a new reanalysis dataset using the atmospheric model of POAMA. This dataset is then used for the initial conditions of the hindcasts and forecasts. This process means that real atmospheric information, as well as land surface conditions that are in balance with this atmospheric forcing, are incorporated into the initial conditions. This has lead to improvements in skill at both seasonal and intra-seasonal scales. Atmosphere-Land Initialisation Scheme (ALI) Motivation for developing ALI The rationale behind developing a new initialisation scheme is that, firstly, real atmospheric initial conditions are important for intra-seasonal forecasting. Results from the current POAMA system appear to support this hypothesis. The MJO is important for Australian climate variations primarily through impacts on the onset and breaks of the summer monsoon and tropical cyclone genesis (e.g. Hendon and Liebmann 1990; Hall et al. 2001; Wheeler and Hendon 2004). Secondly, real atmospheric initial conditions may be important for seasonal forecasting. There are indications that the MJO may affect the development of ENSO events, interannual variability of monsoon rainfall, as well as rainfall over higher latitudes via teleconnections (e.g. Hendon et al. 1999, Zhang et al. 2001; Hendon et al. 2007). Information of the MJO in the initial conditions thus has the potential to improve our ability to forecast these events. There may also be benefits for intra-seasonal/seasonal forecasting from improving the initialisation of the land surface, primarily due to soil moisture memory in the earth-atmosphere system (e.g. Zhang and Frederiksen 2003; Koster et al. 2004). Description of ALI ALI uses a forecast-analysis (or nudging) scheme to produce the atmospheric and land surface initial conditions for the hindcasts and real-time forecasts. In this approach, an offline version of the POAMA atmospheric model (BAM) is nudged towards “reality”, or an “analysis”, provided by ERA-40 for the hindcasts and the NWP forecast system in real-time. The model’s forecast of u-wind, v-wind, atmospheric temperature and humidity is compared directly to the analysis at six-hour intervals, and a fractional difference between the forecast and the analysis is added repeatedly to the evolving model atmospheric state. The land surface is initialised indirectly via the nudged atmosphere, such that the soil moisture and temperature evolve to become consistent with the atmospheric forcing. (This approach has also been used by the Climate Change Prediction Program, CCPP, and Atmospheric Radiation Measurement, ARM, Program

20

in the CCPP-ARM Parametrization Testbed to provide initial conditions for the evaluation of climate models in NWP-mode.) The atmospheric model (forced by observed weekly sea-surface temperatures) is run in forecast-analysis mode from 1980 to real-time, allowing an initial spin-up year for the land surface. The output from this simulation, called BAM-reanalysis, is used as initial conditions for the hindcasts and forecasts. This approach of initialising the land surface does not attempt to correct biases in the land surface model, but it may offer improvements over using climatological land initial conditions. Advantages of ALI

• ALI introduces more realistic atmosphere and land initial conditions into the hindcasts. Hindcasts will now capture true intra-seasonal atmospheric states.

• ALI allows greater consistency between the hindcasts and real-time forecasts, thus allowing better use of the hindcasts to assess intra-seasonal and seasonal forecast skill.

• ALI reduces the shock to the system compared to using ERA-40 directly for the hindcasts (also, the use of ERA-40 directly reduces consistency between hindcasts and real-time forecasts).

• ALI reduces the sensitivity to changes in the NWP forecast system (acts as a buffer). Assessment of ALI: BAM Reanalysis compared to BAM AMIP This section evaluates the BAM Reanalysis simulation, described above, and compares it to the BAM AMIP simulation, which is used to initialise the hindcasts of the current operational system (POAMA-1). As expected, BAM Reanalysis is able to capture the intra-seasonal variability of the MJO, as given by the ERA-40 reanalysis (where BAM AMIP is unable). In general, BAM Reanalysis exhibits higher spatial and temporal correlations and smaller temporal RMSE than BAM AMIP (for example, see Table 1). Note that the variables in Table 1 are not directly nudged in the creation of BAM Reanalysis. The nudging process to create BAM Reanalysis worked well and produces more realistic initial conditions than BAM AMIP. Table 1: Correlations (r), bias and RMSE over Australia for BAM AMIP and BAM Reanalysis for monthly data from 1980 to 2002. For the bias, the units are hPa for mean sea level pressure, K for 2m-temperature and mm/day for precipitation. Soil moisture is normalized. The observed data are the ERA-40 reanalysis, except for precipitation where Global Precipitation Climatology Project (GPCP) data are used. The ± values shown for the spatial correlations represent 1 standard deviation of the monthly data for the analysis period.

MSLP 2m TEMP PRECIP NORMALISED SOIL MOISTURE

BAM AMIP 0.81±0.20 0.95±0.02 0.53±0.19 0.20±0.26 SPATIAL

CORRELATION BAM REANALYSIS 0.98±0.01 0.98±0.01 0.74±0.11 0.53±0.21

BAM AMIP 0.78 0.968 0.56 0.33 TEMPORAL

CORRELATION

BAM REANALYSIS 0.996 0.999 0.96 0.82

BAM AMIP -1.13 0.52 -0.37 - TEMPORAL BIAS

BAM REANALYSIS -2.3 0.59 -0.76 -

BAM AMIP 2.03 1.16 0.97 0.49 TEMPORAL RMSE BAM

REANALYSIS 0.29 0.28 0.36 0.27

Will the use of ALI produce improved skill? The hindcast dataset extends from 1980-2006, with 9 month forecasts starting on the 1st of every month. There are two experiments considered here:

• p15a: BAM AMIP initial conditions (3 member ensemble) • p15b: ALI (BAM Reanalysis) initial conditions (10 member ensemble)

21

Improved skill at seasonal and intra-seasonal scales Initial results indicate improvements in skill at both seasonal (e.g. Figures 1 and 2) and intra-seasonal (e.g. Figure 3) scales. Skill varies a great deal as a function of forecast start month and region of analysis (e.g. Figure 3). Initial investigations of the simulation of the MJO in p15b are promising (e.g. Figure 4).

Fig. 1: Sea-surface temperature anomaly correlation skill from p15a (left) and p15b (right) for lead times of 1 (top), 2 (middle) and 3 months (bottom).

Fig. 2: NINO3 (left) and NINO3.4 (right) sea-surface temperature anomaly correlation skill as a function of lead time (months) for all forecast start months (1980-2001). The dashed line is for persistence, the grey line for p15a and the black line for p15b.

Skill for precipitation MDB: 1-14 days

-0.4

-0.3

-0.2

-0.1

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DECAn

om

aly

corr

elat

ion

p15a persistence p15b

Skill for precipitation TROP: 1-14 days

-0.3

-0.2

-0.1

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC

An

om

aly

corr

elat

ion

p15a persistence p15b

Skill for precipitation MDB: 15-28 days

-0.3

-0.2

-0.1

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DECAn

om

aly

corr

elat

ion

p15a persistence p15b

Skill for precipitation TROP: 15-28 days

-0.3

-0.2

-0.1

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DECAn

om

aly

corr

elat

ion

p15a persistence p15b Fig. 3: Precipitation anomaly correlation skill for SE Australia (left) and Tropical Australia (right) based on forecast start month for the average of the first 2 fortnights of the forecast (1-14 days and 15-28 days), for persistence (clear), p15a (grey) and p15b (black).

p15a p15b

22

Fig. 4: Case study showing the phase diagrams (Wheeler and Hendon, 2004) of MJO propagation for a forecast start date of 1/3/1997 for the observations and 3 ensemble members of p15b (Figure courtesy of Harun Rashid).

References Alves, O., Wang, G., Zhong, A., Smith, N., Tzeitkin, F., Warren, G., Shiller, A., Godfrey, S., and Meyers, G.

2003. POAMA: Bureau of Meteorology Operational Coupled Model Forecast System. National Drought Forum, 15/16 April, Brisbane.

Hall, J.D., Matthews, A.J. and Karoly, D.J. 2001. The modulation of tropical cyclone activity in the Australian region by the Madden-Julian oscillation. Monthly Weather Review, 129,2970-2982.

Hendon, H.H., Zhang, C. and Glick, J.D. 1999. Interannual Variation of the Madden–Julian Oscillation during Austral Summer. Journal of Climate, 12,2538-2550.

Zhang, C., Hendon, H.H., Kessler, W.S. and Rosati, A.J. 2001. Meeting summary: A workshop on the MJO and ENSO. Bulletin of the American Meteorological Society, 82,971–976.

Hendon, H.H., and Liebmann, B. 1990. A composite study of onset of the Australian summer monsoon. Journal of the Atmospheric Sciences, 47,2227–2240.

Wheeler, M.C. and Hendon, H.H. 2004. An all-season real-time multivariate MJO Index: Development of an index for monitoring and prediction. Monthly Weather Review, 132,1917-1932

Hendon, H.H., Wheeler, M.C. and Zhang, C. 2007. Seasonal dependence of the MJO-ENSO relationship. Journal of Climate, 20,531-543

Fennessy, M.J. and Shukla, J. 1999. Impact of initial soil wetness on seasonal atmospheric prediction. Journal of Climate, 12,3167-3180.

Zhang, H. and Frederiksen, C.S. 2003. Local and nonlocal impacts of soil moisture initialization on AGCM seasonal forecasts: A model sensitivity study. Journal of Climate, 16,2117-2137.

Koster, R.D., Suarez, M.J., Liu, P., Jambor, U., Berg, A., Kistler, M., Reichle, R., Rodell, M. and Famiglietti, J. 2004. Realistic initialisation of land surface states: Impacts on subseasonal forecast skill. Journal of Hydrometeorology, 5, 1049-1063.

Observed Ensemble member #1

Ensemble member #3

Ensemble member #2

23

Vegetation as an intermediary in the coupling between hydrology-soil moisture-atmosphere.

Garry Willgoose, Patricia Saco and Alice Howe

The University of Newcastle, New South Wales, Australia Climate models have had a patchy record in terms of predicting the hydrology

response to climate. Hindcasts have commonly had the hydrology mass balance in

error by factors of up to 3. This is typically attributed to scale effects (i.e. spatial

variability), and while these scale effects are unquestionably important, we argue that

the fundamental approach to predicting hydrology from climate model forcing is

problematic in arid and semi-arid climates. Hydrology is simply the difference

between rainfall and evapotranspiration but in dry climates this difference can be 10%

or less. Errors that are minor for either rainfall or evapotranspiration are significant,

and even dominant, in the runoff calculation. Furthermore the transpiration response

to water limitation is highly nonlinear, and dependent on soil and vegetation

properties that are difficult to estimate. Current approaches with dynamic vegetation

models are beginning to address the most glaring deficiencies in SVATS at the

individual plant level, however, they still ignore the ecological adaptations that occur

at the patch and hillslope level. These adaptations (mostly patchy and banded

vegetation) aim to maximise the amount of water captured on the hillslope, and these

adaptations are ubiquitous in arid regions worldwide. While for the atmospheric water

balance these effects are minor they appear to dominate the terrestrial water balance

for water limited landscapes. Our group has been using nonlinear dynamical models

of vegetation growth on the hillslope to understand the self-organising behaviour of

the interaction between vegetation growth, infiltration and water scarcity. We will

summarise this work, argue the relevance of this work for the sub-grid representation

of the terrestrial-atmosphere interaction and present some ideas on the use of

ecosystem optimality as a parsimonious way forward.

25

Canopy dynamics and the surface energy balance

Ian N. Harman

Centre for Australian Weather and Climate Research – A partnership between the Australian Bureau of Meteorology and CSIRO, F.C. Pye Laboratory, Canberra,

A.C.T., Australia. The exchange of momentum, heat and mass between the surface and the atmosphere is a key component of any numerical weather prediction or climate model. The majority of models in use formulate this transfer using ideas and results from the study of rough wall boundary layers. However, it has long been known that many of these results fail in the proximity of, in particular tall, canopies (e.g. Raupach, 1979). Underlying this failure are differences in the structure of turbulence within and just above canopy. Indeed the majority of the turbulent transport from a canopy is by coherent structures within the turbulence which are now thought to resemble mixing layer eddies and not boundary-layer eddies (Raupach et al., 1996). The layer of the boundary layer where these structures dominate, and standard rough wall boundary layer theory fails, is known as the roughness sublayer (RSL). While schemes accounting for the failure of rough wall boundary layer theory have been proposed (e.g. Physick and Garratt, 1995) it is only recently that the impacts of the mixing layer eddies have been fully accounted for within a model simple enough to be used as the basis of a surface exchange scheme (Harman and Finnigan, 2007). Wind speed and scalar profiles within and above a canopy The approach of Harman and Finnigan (2007) is to consider the profiles of wind speed and scalar concentration within and just above a horizontally homogeneous canopy. The profiles above the canopy are given by modified rough wall boundary layer theory, while the profiles within the canopy are given by a balance between the flux divergence and the source/sink of momentum and scalars. These sources and sinks are given by simple, invariant, functions of wind speed and the concentration difference. These model components are valid across a range of diabatic stability provided the turbulence is dominated by mechanical instabilities. Additionally, the displacement height in the rough wall boundary layer theory is given a physical interpretation as the centre of the turning moment arising because of the drag on the boundary layer (Jackson, 1981). The mixing layer eddies are incorporated through the use of their controlling length scale, the vorticity thickness at canopy top, as the basis of the modifications to rough wall boundary layer theory. Coupling of the model components is achieved by ensuring continuity of the mean values and the vertical fluxes. This coupling implies that parameters in the two components are directly related and thereby reduces the degree of empiricism in the model. Model predictions (solid line) and observed (ensemble) profiles of wind speed and potential temperature from the CSIRO research site at Tumbarumba are shown if Fig. 1. The roughness sublayer is most easily viewed as the layer where the model predictions and the standard rough wall boundary layer approach (dashed lines) differ in neutral conditions (panels b) and e)). The figure clearly shows the ability of the coupled approach to successfully predict the profile within the (upper) canopy and into the boundary layer aloft and across a range of diabatic stability. Similar agreement exists for profiles of the water vapour concentration profile – agreement for the carbon dioxide concentration profile is less successful, most likely because of the complex structure of the sources/sinks of carbon dioxide. For this canopy the RSL is 20-30m deep. More importantly the rough wall boundary layer approach also has the incorrect variation as the diabatic stability of the boundary layer changes.

26

unstable near-neutral stable

a) b) c)

d) e) f) Figure 1. Model predictions for the wind speed and potential temperature profiles within and above a canopy incorporating the roughness sublayer and its variation with stability (solid line) and for a surface layer with constant displacement height and roughness length (dashed lines). Observations are taken over a 6 month period from the CSIRO research site at Tumbarumba, with crosses marking the median and triangles the inter-quartile range of an ensemble (at least 150 hourly averages in each ensemble a) and d) unstable conditions, Lc/L<-0.5; b) and e) near-neutral conditions, -0.05 <Lc/L< 0.05; and c) and f) stable conditions Lc/L>0.5. Lc=1/(cd a) quantifies the density of the canopy with cd a leaf-level drag coefficient and a the leaf-area per unit volume. L, the Obukhov length, quantifies the diabatic stability of the boundary layer, and

ch is the canopy height.

There are two of important implications of the coupled canopy-RSL approach. First, greater fluxes are predicted for the same canopy to air differences. Second, key parameters within many surface exchange schemes, including the displacement height and roughness lengths, can no longer be considered invariant and be specified a priori. Indeed, these parameters are now viewed as integral measures of the impact of the canopy on the boundary layer. Since this impact depends on the vertical profile of the wind speed and scalar concentration within the canopy then these parameters vary with any process which affects the profiles. In particular, the displacement height can vary by a factor of 2 and the roughness length for momentum by a factor of 5 over a reasonable range of diabatic stabilities (Harman and Finnigan 2007). Impact on the surface energy balance The recognition of the role of the properties of the flow within the canopy in determining the character of the flow and scalar profile above the canopy has the potential to influence the surface exchange schemes within numerical weather predictions and climate model. These schemes are

27

usually either directly formulated using rough wall boundary layer theory, or can be posed in an equivalent manner. The influence arises through two routes; first, the flux-gradient relationships used within these models in the formulation of the turbulent fluxes have to be revised, and secondly, through the systematic variation of key parameters with diabatic stability.

Figure 2. Diurnal evolution of the surface energy balance from the four models (see text for details). Thick solid line, net radiation; thin solid line, heat storage in canopy and substrate; dashed line, sensible heat; and dashed-dotted line, latent heat. Here we illustrate the potential impacts by considering their impact within a simple model for a horizontally homogeneous boundary-layer overlying a canopy. The model components are the 1-D boundary layer of Busch et al. (1976) and the canopy energy balance model of Watanabe (1994) which includes a description of the stomatal resistance in terms of solar radiation, temperature, humidity deficit and soil moisture. These two models are coupled together in four ways. Model CAN uses standard rough wall boundary layer theory with constant displacement height and roughness lengths. Model CAN+S uses this same theory coupled to within canopy models for the wind speed and scalar profiles - the displacement height and roughness lengths then have a prescribed variation with diabatic stability. Model CAN+R includes a roughness sublayer but maintains the displacement height and roughness length at the values in neutral conditions. Finally model CAN+RS uses the full roughness sublayer theory to couple the models together. The examples shown in Figures 2 and 3 are results from a realistic canopy ( 10mc ch L= = ) at mid-latitudes on the equinox. Figure 2 shows the diurnal evolution of the surface energy balance in these four models. Energy closure is assured through formulation of the models. There are two key differences between the four models. First, is the large increase in daytime sensible heat (dashed line) in the CAN+R and CAN+RS as compared to CAN and CAN+S. This is despite a negative feedback which occurs through interaction with the atmospheric temperature. Second, is the additional increase in the daytime sensible heat when the impact of diabatic stability on the displacement height and roughness length is included. These changes are accomplished through non-simple adjustments in all the other terms notably the heat storage (thin solid line) and latent heat (dash-dotted line). Together these two effects imply a near-doubling of the daytime Bowen ratio between models CAN and CAN+RS.

28

Figure 3. Comparison of reference level potential temperature (left) and accumulated evaporation (right) from the four cases. Thin solid line, CAN; dashed line, CAN+S, dash-dot line, CAN+R, thick line CAN+RS. This striking (and qualitatively robust) result can be explained by considering the different pathways for the turbulent transfer of heat and moisture from a canopy. The transfer of heat is primarily controlled through aerodynamic processes and therefore is (modelled to be) more efficient when the effects of the roughness sublayer and mixing layer eddies are recognised. The transfer of moisture however is primarily controlled through stomatal processes which, here, are not directly affected by the mixing layer eddies. The net result is therefore a preference towards energy transfer via sensible rather than latent heat when the impact of the mixing layer eddies is included. The same argument also explains the further change in the partitioning of energy when the impacts of diabatic stability on the surface exchange parameters are included. Accompanying these differences in the surface energy balance are differences in the state variables. Figure 3 shows the evolution of the 15m (above canopy) potential temperature and the accumulated evaporation over the day. These show that, even for these parameters and modest forcing, the differences between the standard CAN model and the full CAN+RS model are noticeable. Temperature differences of over 1K (greater differences occur nearer the canopy) and a reduction in total evaporation of almost 20% are significant in either numerical weather prediction or climate modelling. Together these results suggest further research is required to confirm these impacts within a more complex surface exchange model. References Busch, N.E., Chang, S.W. and Anthes, R.A. 1976. A multi-level model of the planetary boundary

layer suitable for use with mesoscale dynamic models. J. Appl. Meteorol., 15, 909-919. Harman, I.N. and Finnigan, J.J. 2007. A simple unified theory for flow in the canopy and roughness

sublayer. Boundary-Layer Meteorol., 123, 339-363. Jackson, P.S. 1981. On the displacement height in the logarithmic velocity profile. J. Fluid. Mech.,

111, 15-25 Physick, W.L. and Garratt, J.R. 1995. Incorporation of a high-roughness lower boundary into a

mesoscale model for studies of dry deposition over complex terrain. Boundary-Layer Meteorol., 74, 55-71

Raupach, M.R. 1979. Anomalies in flux-gradient relationships over forest. Boundary-Layer Meteorol., 16, 467-486

Raupach, M.R., Finnigan, J.J. and Brunet, Y. 1996. Coherent eddies and turbulence in vegetation canopies: The mixing layer analogy. Boundary-Layer Meteorol., 78, 351-382.

Watanabe, T. 1994. Bulk parameterization for a vegetated surface and its application to a simulation of nocturnal drainage flow. Boundary-Layer Meteorol., 70, 13-35.

29

Terrestrial carbon cycle feedbacks in atmospheric and climatic change

Roger M. Gifford

CSIRO Plant Industry, Canberra, ACT 2601, Australia

Introduction The terrestrial carbon cycle is influential in the climate system over a wide range of timescales from sub-daily to æons (Gifford 1991). The influences occur via numerous ecological attributes and processes that are driven by the plant requirement to convert atmospheric carbon dioxide into plant dry matter. These requirements involve interactions of the terrestrial carbon cycle with the energy, water, mineral and geological cycles of the earth with repercussions for local and global temperature and rainfall on the range of timescales. This short paper focuses on some aspects of the terrestrial carbon cycle that interact with climate over the decadal to century timescale pertinent to predictions of human induced global climate change via greenhouse forcing. The terrestrial carbon sink – potential causes There are three primary global effects of the increasing atmospheric CO2 concentration that impact the terrestrial carbon cycle directly: 1) the CO2 fertilising effect on vegetation growth, 2) the stomatal conductance effect and 3) the enhanced greenhouse effect. All three effects have the potential to impact the global climate via their influence on the carbon cycle. The CO2 fertilising effect may be a major contributor to the terrestrial CO2 sink that is currently absorbing almost one third of the human-induced emissions. The stomatal closure effect must be leading to some combination of reduced terrestrial evaporation and increased terrestrial surface daytime temperature, the balance between which at any one site depends on both vegetation properties and atmospheric conditions. Estimation of the global terrestrial sink has been problematic. It has been calculated by the difference between calculated emissions from fossil fuel burning plus cement production plus net deforestation and the estimated ocean sink. Given the magnitude of such estimates it is important to understand exactly what is causing it in order to appreciate when it will saturate and cease to provide a free ecosystem service of helping to mop up human emissions. The dominant potential causes discussed are the CO2 fertilising effect mentioned, enhanced vegetation growth due to greenhouse warming, stimulation of vegetation growth and C storage by fertilisation from enhanced N-deposition, and enhanced uptake of CO2 by previously deforested land recovering CO2 in the process of growing back to mature stands of trees. None of them are well quantified with observational evidence. However, whatever the balance of contributions to the terrestrial sink from these sources, the net effect has been a remarkable stability of the fraction of the human-emissions that have been taken up by the land. Over the last 35 years of rapid atmospheric CO2 and temperature increase the fraction has remained close to 27% as a running average. Model estimates of the future of the terrestrial C sink Digital Global Vegetation Models (DGVMs) include equations intended to represent the impacts of the CO2 fertilising effect and atmospheric warming from the greenhouse effect. Most do not, however, yet deal with the effects of N-deposition and the accumulation of C during regrowth of cleared forest. While incorporation of results of elevated CO2 experiments on plant growth over several years into global carbon cycle models can readily account for a terrestrial C sink of 2-3 Gt C yr-1 from that effect (Gifford 1991), comparison of several DGVMs (Cramer et al 2001;

30

Friedlingstein et al. 2006) provides a wide range of future CO2 emissions from the land among models. The models evaluated by the C4MIP inter-comparisons had the land-uptake of CO2 anthropogenic emissions declining between the present and 2100, but one (CSM1) had the cumulative land-borne fraction over 250 years (1850 to 2100) declining from 0.26 without the terrestrial carbon cycle coupled to only 0.25 with the C cycle coupled. In contrast the Hadley model (HadCM3LC) had the land-borne fraction declining from 0.31 without the terrestrial C cycle to only 0.05 with the terrestrial C-cycle. By 2100 the modelled net CO2 flux for the land ranged from an sink of 11 Gt C yr-1 to a net source of 6 Gt C yr-1. The primary reason for this huge contrast in the effect of the terrestrial carbon cycle on model outcome is that we do not have the information base for quantifying feedbacks of the response of the carbon cycle to atmospheric CO2 increase and warming. The gross terrestrial sink for CO2 has been increasing for over a century and is now estimated at 2-3 Gt C yr-1 (Houghton 2007). The gross sink represents the balance between a large CO2 uptake by primary production and a large CO2 output by respiration of plants and microbes. The accelerating trend for the increase is closely matched by the trend for atmospheric CO2 concentration. The fraction of global emissions taken up by the land has been stable in terms of running average for the last 30 years. This is qualitatively consistent with a CO2 fertilising effect on terrestrial carbon stocks being the cause but there are other possibilities that would probably correlate with the increasing atmospheric CO2 concentration; notably stimulation of terrestrial C stocks by enhanced N-deposition and re-establishment of forest on land that was deforested earlier, and a potential positive effect of global warming on forest vegetation growth. Feedbacks moderating the tendency for the global terrestrial sink to become a global source Unfortunately, there is considerable uncertainty about the magnitude of the baseline NPP, the impact of concurrent warming, and whether vegetation responses to elevated CO2 concentration become down-regulated over time in the field. For example, an intercomparison of 18 models of Australian baseline NPP, produced an almost order-of-magnitude range of estimates from 0.38 Gt C yr-1 to 3.3 Gt C yr-1 (Roxburgh et al. 2005). Additionally the modelled distribution of NPP around the continent was not consistent even among models that produced similar mid-range values. An evaluation of the above-ground NPP of major world biomes found that it was independent of growing season temperature where growing season was defined as the months having an average temperature above 0oC (Kerkhof et al. 2005). Thus NPP is likely to increase with global warming – all else equal - as the growing seasons get longer. Concerns about the real-world applicability of findings about vegetation growth responses to elevated CO2 from enclosed experiments (growth cabinets etc) have been diminished by analyses of many-year-long free air CO2 enrichment experiments in forests that have been summarised to find an average stimulation of growth of 23% by 200ppmv CO2 across a wide range of productivities (Norby et al . 2005). The proposed “progressive nitrogen limitation - PNL” to the CO2 fertilising effect (Luo et al. 2004) did not apply (Finzi et al. 2007). Rather the hypothesis that in the long run the nitrogen cycle would follow the CO2-stimulated C-cycle (Gifford 1992), applied. On the respiratory output side of the terrestrial C-balance, which can lead the biosphere to be either a source or a sink of CO2, a simple view has been extant in the modelling community that respiration is highly temperature sensitive and hence global warming will lead to reduction in stocks of plant and soil organic matter and hence conversion of the terrestrial sink into a source leading to accelerated global warming (eg Cox et al. 2000). Is this consistent with ecological understanding? First, with respect to plant respiration, its high positive sensitivity to experimental warming is very short lived, down-regulation taking only a few days (Gifford

31

1995). For fully acclimated whole-plant respiration the respiration:photosynthesis ratio is quite stable over a wide range of growth temperature because respiration and photosynthesis are, in the long term, coupled (Gifford 2003). For heterotrophic (ie microbial) respiration during litter decomposition, there is also a coupling between litter input from plants and its decomposition. That coupling occurs for several reasons. But even without consideration of that coupling there are two aspects of soil organic matter (SOM) decomposition that compensate in the longer term for the initial short term sensitivity of SOM respiration to higher temperatures in lab incubations. The first relates to the huge range of turnover times of fractions within the total SOM pool. When a fast turnover fraction is oxidised by microbes, about half of the C it is converted into a longer turnover time component as the other half emits as CO2. Thus as warming accelerates decomposition, there is an acceleration of the conversion of the fast turnover pools into slow turnover pools thereby leading to a redistribution of the fractions – more slow turnover and less fast turnover. This shift in distribution acts as a negative feedback reducing the overall sensitivity of SOM oxidation to warming (Ågren and Bosatta 2002). A second soil aspect is less well understood. It is that the presence of labile organic matter (fast pools) can act as a “primer” for accelerated microbial oxidation of the slower decomposing pools (Subke 2004). Thus, when warming leads to a decrease in the ratio of labile to non-labile SOM, the priming effect will decline with the slower oxidation of the less labile SOM pools, again acting as a negative feedback. Possibly the strongest feedback moderating the impact of warming on SOM decline is the coupling of soil and plant processes by the N-cycle. Most natural vegetation grows faster with added soluble N to the soil. That is, N-availability is one of the factors that co-limit productivity. In the soil most of the N is in unavailable organic form that is converted to plant-available nitrate and ammonium ions by microbes during litter & SOM decomposition. Plants take up that N and incorporate it into plant organic matter. SOM has a typical C:N ratio of about 12. Vegetation has a wide range of C:N ratios averaging about 100. Thus when SOM, having 12 carbons associated with each N, releases its N which plants take up and convert to plant organic matter having 100 carbons associated, there is a huge increase in the net carbon stored by the ecosystem. Thus any warming that accelerates that process, accelerates C storage essentially by shifting some soil N into plant N. This is a powerful negative feedback in the terrestrial carbon cycle. The final feedback of warming that modulating any tendency to acceleration of SOM decomposition with global warming, is the feedback through species composition change with climate change. This seems to be occurring in numerous parts of the world where woody shrubs are replacing herbaceous species. In Australia we have much woody weed invasion of tropical grasslands (Gifford and Howden 2001) that may in part be caused by climate warming. It is also happening in cold climates including the Arctic. In these cold climates, warming is causing grass and sedge communities, having fast decomposing litters, to be replaced by woody shrubs having slow decomposing litters (Cornelissen et al. 2007). This is a negative feedback that tends to offset the faster decomposition tendency resulting from atmospheric warming. References Ågren GI, Bosatta E 2002. Reconciling differences in the predictions of temperature

responses of soil organic matter. Soil Biology and Biochemistry 34: 129-132. Cox PN, Betts RA, Jones CD, Spall AS, Totterdell IJ 2000. Acceleration of global

warming due to carbon-cycle feedbacks in a coupled climate model. Nature 408: 184-187.

32

Cornelissen JHC, and 50 others. Global negative feedback to climate warming responses of leaf litter decomposition rates in cold biomes. Ecology Letters 2007. 10:619-627.

Cramer W and 16 others 2001. Global response of terrestrial ecosystem structure and function to CO2 and climate change: results from six dynamic global vegetation models. Global Change Biology 7:357-373.

Finzi AC, Norby RJ, Calfapietra C, Gallet-Budynek A, Gielen B, Holmes WE, Hoosbeek MR, Iversen CM, Jackson RB, Kubiske ME, Ledford J, Liberloo M, Oren R, Polle A, Pritchard S, Zak DR, Schlesinger WH, Ceulemans R 2007. Increases in nitrogen uptake rather than nitrogen-use efficiency support higher rates of temperate forest productivity under elevated CO2. Proceedings National Academy Sciences 104: 14014-14019.

Friedlingstein P, and 28 others 2006. Climate-carbon cycle feedback analysis: Results from the C4MIP model intercom parison. J of Climate 19: 3337-3353.

Gifford RM 1991. Implications of CO2 effects on vegetation for the global carbon budget. In: The Global Carbon Cycle. NATO ASI Series 1 Vol 15. pp159-199. Ed M. Heimann, Springer-Verlag, Berlin, Heidelberg.

Gifford RM 2003. Plant respiration in productivity models: conceptualisation, representation and issues for global terrestrial carbon-cycle research. Functional Plant Biology 30: 171-186.

Gifford RM, Howden M 2001. Vegetation thickening in an ecological perspective: Significance to national greenhouse gas inventories and mitigation policies. Environmental Science and Policy 4: 59-72.

Gifford, R.M. 1994. The global carbon cycle: A viewpoint on the missing sink. Aust J Plant Physiol 21:1-15.

Gifford, R.M. 1995. Whole plant respiration and photosynthesis of wheat under increased CO2 concentration and temperature: long-term and short-term distinctions for modelling. Global Change Biology 1:385-396.

Luo Y, Su B, Currie WS, Dukes JS, Finzi A, Hartwig U, Hungate B, McMurtrie RE, Oren R, Parton WJ, Pataki DE, Shaw MR, Zak DR, Field CB 2004. Progressive nitrogen limitation of ecosystem responses to rising atmospheric carbon dioxide Bioscience 54: 731-739.

Norby RJ, DeLucia EH, Gielen B, Calfapietra C, Giardina CP, King JS, Ledford J, McCarthy HR, Moore DJP, Ceulemans R, De Angelis P, Finzi AC, Karnosky DF, Kubiske ME, Lukac M, Pregitzer KS, Scarascia-Mugnozza GE, Schlesinger WH, Oren R 2005. Forest response to elevated CO2 is conserved across a broad range of productivity. Proceedings of the National Academy of Sciences of the United States of America 102, 18052-18056.

Roxburgh SH, Barrett DJ, Berry SL, Carter JO, Davies ID, Gifford RM, Kirschbaum MUF, McBeth BP, Noble IR, Parton WG, Raupach MR, Roderick ML. 2004. A critical overview of model estimates of net primary productivity for the Australian continent. Functional Plant Biology 31: 1043-1059.

Subke J-A, Hahn V, Battipaglia G, Linder S, Buchanon N, Cotrufo MF 2004. Feedback interactions between needle litter decomposition and rhizoshere activity. Oecologia 139: 551-559.

33

Carbon cycle observations: challenges and

opportunities

Helen Cleugh

Centre for Australian Weather and Climate Research – A partnership between CSIRO and the Australian Bureau of Meteorology

Summary The cycling of carbon in the terrestrial biosphere is a key component of the climate system. Papers by Canadell et al (2006 and 2007) and Friedlingstein et al (2006) have clearly demonstrated how the trajectory of atmospheric CO2 levels, and hence climate, depends on the ocean and land sinks, which currently take up about half of all anthropogenic CO2 emissions but there are signs that this sink strength may be weakening. The task of understanding and modelling the feedbacks between terrestrial carbon (and water) cycles, climate and human activities (including land management) are therefore a priority for climate science. And the need for quality observations of the terrestrial carbon and water budget, including fluxes, stores and atmospheric concentrations, to support model simulations and to monitor the response of the coupled carbon - climate system, must also be a priority. This presentation will describe, and present key results from, the approach used to quantify the terrestrial carbon and water cycles in Australian landscapes at multiple space and time scales - combining atmospheric measurements, modelling and remote sensing. This approach is used to explore the elements and strategy for any future carbon cycle measurement program Background The underlying cause of global warming is the perturbation to the global carbon cycle caused by anthropogenic emissions of greenhouse gases (GHGs), which result from accelerating per-capita GDP and energy use over the last 100 years (Raupach et al, 2007). The response of the climate system is well known: atmospheric CO2 concentrations are increasing at almost 2 ppm per year; the planet has warmed by 0.74oC, and sea levels have risen by an estimated 17 cm, over the last century. The future trajectory of the global climate depends on how the carbon cycle – especially the land and oceanic uptake – responds to this changing climate which will in turn influence the airborne fraction of CO2. The results of Friedlingstein et al (2006) illustrated that differences in the rate of carbon uptake by the land resulted in a spread of climate responses (temperature) that were as large as the range of IPCC emission scenarios. Recent publications by Canadell et al (2007) and Le Quere et al (2006) have suggested that the earth’s natural sinks may be weakening – thus revealing sooner-than-expected the serious concerns raised by Canadell et al (2006) about the

34

vulnerability of these carbon sinks, and the impact of these carbon – climate feedbacks on climate change projections. The risks and research challenges posed by these results have been recognised by the Australian climate science community, who identified the terrestrial carbon cycle as a research priority, for example at a 2006 Workshop on Climate Change Science priorities where it was agreed that: ”We know that the carbon cycle and its strength affects atmospheric concentrations of greenhouse gases. We also know that as climate changes, so too will various components of the cycle. Science has an important role in assessing the likely future changes to Australian terrestrial and marine environments and their influences”, while the 2005 “Blueprint for Terrestrial Carbon Cycle Research” (hereafter, the Blueprint) articulated three research priorities: i)carbon budget dynamics (patterns of sources and sinks; processes and observations) across Australia; ii) the vulnerability of terrestrial carbon sinks especially to the effects of drought, fire and pests/diseases; and iii) representing the interactions between the carbon cycle and climate system in ACCESS. Underpinning these stated priorities is the need for a strategic and integrated program of carbon (and energy, water and other biogeochemical cycles) budget observations – identified as a “cross-cutting” priority in the Blueprint. Such observations are needed to monitor trends and variability in the earth system; to compare against model predictions; and to provide the process understanding needed for model parameterisations. A terrestrial carbon cycle measurement program to address these needs must be multi-scale and have a clear strategy for combining a range of measurements (discrete/continuous; in situ/remotely sensed etc.) with models. The challenge for the Australian science community now is to articulate a coherent and integrated research program that implements these priorities, including a specification of what such a measurement program should comprise, what it needs to deliver, and how it could be achieved. Terrestrial carbon and water budget measurements in Australian ecosystems: looking back and looking forward Given this context, this paper provides an overview of the key results of a measurement program, initiated by the CSIRO Biosphere Working Group in 1999, and funded by CSIRO and the AGO, that has provided long-term measurements of the net carbon and water fluxes in several Australian ecosystems along with detailed terrestrial carbon budget measurements and modelling and some land-based, high precision CO2 concentration measurements. These measurements, and the results, illustrate the following key pointers for future observation programs: a) Constraint of multiple measurements: Measurements taken at a range of scales

and of multiple components of the carbon budget, which are then integrated with models, have enabled us to place much greater confidence in any one measurement. This was a design strategy in our measurement program, and the results reinforce the value of this for any future integrated and quality-assured measurement program is to be embarked upon.

35

b) Carbon (and water) budgets at ecosystem to continental scales: These measurements have been used to calibrate and test remote sensing approaches for estimating the carbon fluxes at regional and continental scales (Leuning et al, 2005). Using remote sensing to provide continuous space/time distributions of the carbon fluxes, along with other recent developments such as NCAS and trend analyses of vegetation indices, mean that we have the necessary elements in place to begin the task of constructing time series of the carbon (and water) balances for specified ecosystems, regions and the continent.

c) Combining terrestrial carbon budget with atmospheric measurements: Combining

an estimated terrestrial carbon source – sink distribution with atmospheric CO2 concentration measurements (terrestrial and baseline) is needed for a complete observation-based assessment of the regional carbon budget dynamics for parts, or maybe all, of the Australian continent. Such information not only provides us with an estimate of the regional and continental carbon fluxes, but also the affect and perhaps attribution due to climate and land management drivers.

d) Continuous, ecosystem-scale measurements of fluxes: For parameterising and

testing land surface models used in climate models, such as CABLE in ACCESS, and to capture trends and variability at all timescales, it is imperative to have continuous measurements. Net carbon and water fluxes obtained from flux towers are temporally continuous and are spatially averaged across a plant canopy. The research has developed theoretically-sound methods to reduce the bias in NEE estimates from flux tower measurements (van Gorsel et al, 2006). Fluxnet is a global network of > 400 flux towers sampling a large diversity of bioclimates which provides a data set of ecosystem-scale, continuous net carbon fluxes. Current work demonstrates the potential of this large and global database to improve and parameterise land surface schemes for climate models, such as CABLE.

Drawing from these results and points, the presentation will conclude with a strawman view of what an integrated measurement program would comprise and what it could deliver.

36

References

Canadell, J.G., Pataki, D., Gifford, R., Houghton, R.A., Lou, Y., Raupach, M.R.,

Smith, P., Steffen, W. (2007), in Terrestrial Ecosystems in a Changing World, eds Canadell JG, Pataki D, Pitelka L (IGBP Series. Springer-Verlag, Berlin Heidelberg), pp 59-78.

Canadell, J.G., Le Quéré, C., Raupach, M.R., Field, C.B., Buitenhuis, E.T., Ciais, P.,

Conway, T.J., Gillett, N.P., Houghton, R.A., Marland, G. (2007) Contributions to accelerating atmospheric CO2 growth from economic activity, carbon intensity, and efficiency of natural sinks. PNAS:0702737104

Le Quéré, C., Rödenbeck, C., Buitenhuis, E.T., Thomas, J., Conway, T.J.,

Langenfelds, R., Gomez, A., Labuschagne, C., Ramonet, M., Nakazawa, T., Metzl, N. et al. (2007) Science, 316 doi:10.1126/science.1136188.

Leuning, R., Cleugh, H.A., Zegelin, S.J. and Hughes, D. (2005). Carbon and water

fluxes over a temperate Eucalyptus forest and a tropical wet/dry savanna in Australia: measurements and comparison with MODIS remote sensing estimates. Agricultural and Forest Meteorol., 129: 151-173.

Friedlingstein P, and 28 others (2006) Climate-carbon cycle feedback analysis:

Results from the C4MIP model intercom parison. J. Climate, 19: 3337-3353. Raupach, M.R., Marland, G., Ciais, P., Le Quéré, C., Canadell, J.G., Klepper, G.,

Field, C.B. (2007) Global and regional drivers of accelerating CO2 emissions. PNAS:0700609104

van Gorsel, E., Cleugh, H.A., Keith, H., Suni, T. and Leuning, R. (2007). Nocturnal

carbon efflux: Reconciliation of eddy covariance and chamber measurements using an alternative to the u*-threshold filtering technique. Tellus B. doi:10.1111/j.1600-0889.2007.00252.x.

37

Woody thickening: a consequence of changes in fluxes of carbon and water on a warming globe?

Derek Eamus, S Fuentes C Macinnis-Ng, A Palmer, D Taylor, R Whitley, I Yunusa, M Zeppel

Institute for Water and Environmental Resource Management, University of Technology Sydney,

PO Box 123, Broadway, NSW, 2007, Australia

Abstract

Understanding patterns, rates and controls of water and CO2 exchange between land surfaces and the atmosphere is central to the sciences of meteorology, ecology, hydrology, ecophysiology, forestry and related endeavours. Measurements involving sapflow sensors, eddy covariance and remote sensing have contributed substantially to our understanding of these issues. In this talk, we apply a combination of methods in order to apply a soil-plant-atmosphere model to the question: what is causing the globally observed phenomenon of woody thickening? The density of woody plants in arid and semi-arid regions is increasing regionally and globally (Fensham et al., 2005, Hoffman et al., 1999, Bowman et al. 2001, Burrows et al. 2002). This can be deduced from analyses of tree-ring widths, forest inventory data, aerial photo-interpretation and from long-term monitoring sites (Spiecker et al., 2003). Potential causes of woody thickening have been extensively discussed in the past. Mechanisms that have been proposed include the (a) Walther model, which invokes competition for water and nutrients among the deeper roots of woody plants versus the shallower roots of shrubs and grasses; (b) a role for changes in the timing, intensity and frequency of fire; and (c) changes in herbivory by large herbivores. Such thickening may have a large impact on regional CO2 budgets, atmospheric CO2 concentration and ecosystem function and regional water budgets. We propose an alternative mechanism to explain woody thickening based upon changes in water and carbon fluxes within the soil-plant-atmosphere continuum resulting from a change in global atmospheric conditions. Such a mechanism is global in reach, appears consistent with a number of phenomena and has several testable predictions, which we briefly discuss. In this talk we present:

(a) the bio-physical conceptual basis of the model; (b) supporting evidence from evapo-transpiration rates, run-off, soil moisture and changing

atmospheric conditions associated with a warming global environment; (c) results of a modelling analyses using the SPA model of Williams et al. (2001) as applied to an

Australian open woodland.

The conceptual framework

The following observations constitute the a priori foundations for the model:

1. The concentration of CO2 in the atmosphere ([CO2]a) has been increasing since the start of the industrial revolution.

2. This rise in [CO2]a has two effects: (i) it increases rates of photosynthesis of woody plants, typically of about 30 to 50 %. Photosynthesis is enhanced more in woody shrubs (+45 %) than grasses (+38 %) or trees (+25 %) in response to CO2 enrichment; and (ii) stomatal conductance of woody plants is

38

decreased by about 20 %. C4 grasses show a smaller response to CO2 enrichment than C3 plants such as shrubs and trees.

3. The growth rate of trees and shrubs is enhanced by increased [CO2]a because of the stimulation of photosynthesis and decreased photorespiration. Importantly, the proportional increase in tree growth is larger under xeric than mesic conditions.

4. Pan evaporation rates have declined globally, including across Australia.

We propose that these observations may explain the phenomenon of woody thickening. It is useful to note that there are three key predictions from this conceptual model. First, long-term trends in tree water-use-efficiency should be increasing; second, run-off should increase where woody thickening is occurring; finally enriched CO2 studies should reveal an enhanced plant water status. These three predictions are discussed later.

Supporting evidence

1) A decrease in pan evaporation rates has been recorded for the northern hemisphere, Australia and New Zealand (Roderick and Farquhar, 2002, Roderick and Farquhar, 2004). In water-limited ecosystems, decreasing pan evaporation rates can best be explained by decreased wind speed and decreased solar radiation receipt at ground level (global dimming) because of increased cloud cover and atmospheric aerosol content. Vapour pressure deficit has decreased for water-limiting ecosystems of Africa, Australia and the Indian sub-continent (Nemani et al., 2003) ; this is another measure of a decreasing pan evaporation rate.

2) Because of a decrease in evaporative demand for water either soil moisture content must increase in the short-term or excess moisture must be lost through increased run-off and in the longer-term by increased water use by vegetation.

3) The prediction that increased run-off as a consequence of climate change is supported from both observational and simulation studies (Labat et al., 2004; Probst and Tardy, 1987). The high [CO2]a, of recent decades, compared to the levels observed in the 18th century, has also been recently invoked to explain this increased run-off (Gedney et al., 2006 ) through the observed response of stomatal conductance to [CO2]a, (Eamus and Ceulemans 2001; Medlyn et al. 2001).

4) There is evidence of global soil moisture increasing with positive soil moisture trends observed during the 20th century in the Ukraine, Mongolia and the western USA, for example (Robock et al., 2000; Robock et al., 2005, Hamlet et al. 2007, Hirabayashi et al. 2005). While models of global warming predict summer soil desiccation, there is no evidence for this even in regions that have been warming over the past 50 years. For arid and semi-arid regions of the southern hemisphere, however, evidence of increases in soil moisture is limited. Elevated moisture levels across land-use gradients have been documented in sparsely grassed, deep sand-dunes of the southern Kalahari. This higher soil moisture status has promoted the success of C3 shrubs and trees, including Acacia mellifera and Rhigozum trichotomum.

5) A second prediction from our model is that a reduced stomatal conductance in response to CO2 enrichment and a concomitant enhancing of soil moisture stores (for at least some months of the year), will result in a more positive plant water status. There is ample evidence that an improved water status is observed under CO2 enriched conditions (Eamus et al. 1995). Wullschleger et al. (2002) reviews many of the studies published between 1990 and 2000 and shows that improved water status is frequently observed in studies of CO2 enriched environments. These differences were larger during drought than wetter conditions, as expected.

39

Testing the conceptual model

In order to test the conceptual model presented here, we ran the SPA model under a range of scenarios, which include different levels of atmospheric CO2 concentration, rainfall, VPD and temperature, either in isolation (ie one factor at a time) or in combination. The SPA model is a detailed process based mechanistic model that has been successfully tested and validated across a range of diverse ecosystems, including Arctic tundra (Williams et al. 2000), Brazilian tropical rainforests (Williams et al. 1998, Fisher et al. 2006) and temperate Ponderosa pine forests (Williams et al. 2001). The SPA model predicts, amongst other parameters, carbon and water fluxes, leaf water relations and changes in soil moisture and is previously untested in Australian ecosystems. We have recently successfully applied the SPA to a temperate open woodland in NSW. Key results of the climate scenarios we applied to the open woodland can be summarised thus:

I. When CO2 concentration was reduced to 80 % of ambient, stomatal conductance (gs) and annual tree sapflow increased but GPP decreased. When CO2 levels were increased above ambient, gs and annual tree sapflow declined by between 10 and 60 % and leaf water potential was increased; most importantly GPP increased by almost 50 % when CO2 levels were increased above ambient.

II. As annual rainfall ranged between 50 % and 150 % of current levels for this site (about 680 mm per y), gs was unaffected. When rainfall was reduced to 50 % of current levels, but annual tree sapflow decreased by 20 %. When rainfall was increased by 50 %, annual sap flow was increased by 25 % and leaf water potential was increased. GPP increased by about 15 % for an increase in rainfall of 50 %.

III. As VPD was varied between - 25 % (wetter air) and + 25 % (drier air), gs was unaffected. Annual tree sapflow was reduced when atmospheric water content increased (VPD declined) but increased by 20 % when atmospheric water content declined by 25 %. Simultaneously, leaf water potential declined. GPP decreased by about 15 % (ie GPP declined as the air became drier). This was attributed to a decline in leaf water potential and the impact that this had on C uptake on dry days.

IV. There was minimal response in any variable to a temperature increase of between 1 and 4 oC. V. When atmospheric CO2 concentration was increased to 550 µmol mol-1 in combination with decreased

VPD (wetter air) of 25 % and temperature increased by 2 oC, gs showed a minor decline but annual tree sapflow increased by 5 % and leaf water potential increased. Most importantly, GPP increased by 25%.

Conclusion

Clearly, we have a simple yet powerful explanation for the trend of increasing woody thickening observed in arid and semi–arid regions over the past 50-100 years. As pan evaporation rates have declined, the availability of soil moisture has increased (as evidenced by increased water potentials), effectively equivalent to increased rainfall. Simultaneously there is a decrease in stomatal conductance resulting from increased atmospheric [CO2]a levels which were shown also to reduce tree sapflux. This has all resulted in an increased ecosystem-scale GPP which is translated into woody thickening. A final prediction from our conceptual model is that water-efficiency should have been increasing in woody plants for the past century because of the increase in photosynthesis and the decline in stomatal conductance arising from increased [CO2]a. In support of this prediction is the observation that GPP was increased and tree water use decreased in our final simulation. It is pertinent to note that long-term increase in WUE of trees that has been reported over the past 100 years using stable isotope analyses of tree rings (Hietz et al., 2005).

References

Bowman, D.M.J.S., Walshe, A., Milne, D.J, 2001. Forest expansion and grassland contraction within a Eucalyptus savanna matrix between 1941 and 1994. Global Ecol. and Biogeog. 10, 535-548.

40

Burrows, W.H., Henry, B.K., Back, P.V., Hoffman, M.B., Tait, L.J., Anderson, E.R., Menke, N., Danahar T., Carter, J.O., McKeon, G.M., 2002. Growth and carbon stock change in eucalypt woodlands in northeast Australia: ecological and greenhouse sink implications. Global Chnge. Biol. 8, 769 – 784.

Eamus, D., Berryman, C.A., Duff, G.A., 1995. The impact of CO2 enrichment on water relations in Maranthes corymbosa and Eucalyptus tetrodonta. Aust. J. of Bot. 43: 273-282. Eamus, D., Ceulemans, R., 2001. Effects of greenhouse gases on the gas exchange of forest trees, in:

Karnosky, D., Ceulemans, R., Scarascia-Mugnozza, G.E., Innes, J.L. (Eds. ) The Impact of CO2 and other Greenhouse Gases on Forest Ecosystems. CABI Publishing, United Kingdom. pp. 17-56

Fensham, R.J., Fairfax, R.J., Archer, S.R., 2005. Rainfall, land use and woody vegetation cover change in semi-arid Australian savanna. J. of Ecol. 93, 596-606.

Fisher, R.A., Illiwma, M.,, Vale, R.l.D, Costa, A.L.D and Meir, P. 2006. Evidence from Amazonian forests is consistent with isohydric control of leaf water potential. Pl., Cell and Environ. 29, 151-165. Gedney, N., Cox, P.M., Betts, R.A., Boucher, O., Huntingford, C., Stott, P.A., 2006. Detection of a direct

carbon dioxide effect in continental river runoff records. Nature 439, (7078), 835-838. Hamlet, A F., Mote, P.W., Clark, M. P., Lettenmaier, D. P., 2007 Twentieth-century trends in runoff, evapotranspiration, and soil moisture in the western United States. J. of Climate 20 (8): 1468-1486. Hietz, P., Wanek, W., Dunisch, O., 2005. Long-term trends in cellulose delta C-13 and water-use efficiency of tropical Cedrela and Swietenia from Brazil. Tree Phys. 25, 745-752. Hirabayashi Y, Kanae S, Struthers I, Oki T A 2005. 100-year (1901-2000) global retrospective estimation of the terrestrial water cycle. J. of Geophysical Res.-atmospheres 110 (D19): Art. No. D19101. Hoffman, M.T., O’Connor, T.G., 1999. Vegetation change over 40 years in the Weenen/Muden area,

KwaZulu-Natal: evidence from photo-panoramas. Afr. J. of Range and Forage Sci. 16, 71-88. Labat, D., Godderis, Y., Probst, J.L., Guyot, J.L., 2004. Evidence for global runoff increase related to climate warming. Adv. in Water Resources 27, 631-642. Medlyn B.E, Barton C.V.M, Broadmeadow M.S.J, Ceulemans R, De Angelis P, Forstreuter M, Freeman M,

Jackson S.B, Kellomaki S, Laitat E, Rey A, Roberntz P, Sigurdsson B.D, Strassemeyer J, Wang K, Curtis P.S, Jarvis P.G 2001. Stomatal conductance of forst species after long-term exposure to elevated CO2 concentration: a synthesis. New Phytologist 149, 247-264.

Nemani, R.R., Keeling, C.D., Hashimoto, H., Jolly, W.M., Piper, S.C., Tucker, C.J., Myneni, R.B., Running, S.W., 2003. Climate-driven increases in global terrestrial net primary production from 1982 to 1999.

Science 300, 1560-1563. Probst, J.L., Tardy, Y., 1987. Long-range streamflow and world continental runoff fluctuations since the

beginning of this century. J. of Hydrology 94 (3-4), 289-311. Robock, A., Vinnikov, K.Y., Srinivasan, G., Entin, J.K., Hollinger, S.E., Speranskaya, N.A., Liu, S., Namkhai, A. 2000. The global soil moisture data bank. Bulletin of the Am. Met. Soc. 81, 1281-1299. Robock, A., Mu, M.Q., Vinnikov, K., Trofimova, I.V., Adamenko, T.I., 2005. Forty-five years of observed soil moisture in the Ukraine: No summer desiccation (yet). Geophysical Research Letters 32, L03401, doi:10.1029/2004GL021914. Roderick, M.L., Farquhar, G.D., 2002. The cause of decreased pan evaporation over the past 50 years. Science 298, 1410-1412. Roderick, M.L., Farquhar, G.D., 2004. Changes in Australian pan evaporation from 1970 to 2002. Int. J. Climatol. 24, 1077-1090. Roderick M.L., and, Farquhar G.D. 2005. Changes in New Zealand pan evaporation since the 1970s. Int. J. of Climatology 25 (15): 2031-2039. Williams, M., Malhi, Y., Nobre A., Rastetter, E.B. and Grace, J. 1998 Seasonal variation in net carbon exchange and evapotranspiration in a Brazillian rainforest: a modelling analyses. Pl. Cell and Environ 21, 953-968. Williams, M., Bond, B.J. and Ryan, M.G. (2001) Evaluating different soil and hydraulic constraints on tree function using a model and sap flow data from ponderosa pine. Pl. Cell and Environ 24, 679-690. Wullschleger, S.D., Tschaplinski, T.J., Norby, R.J., 2002. Plant water relations at elevated CO2 – implications for water-limited environments. Pl. Cell and Environ. 25, 319-331.

41

Model independence and the representation of model space as a source of uncertainty

Gab Abramowitz1,2 and Hoshin Gupta3

1UNSW Climate Change Research Centre 2Centre for Australian Weather and Climate Research

3University of Arizona

This talk explores the idea of a "model space" as the space of all possible models for a given prediction problem. While we may be familiar with the notion of uncertainty in model parameters, states or inputs, uncertainty associated with the model we use is clearly a more difficult problem. It’s easy for us to conceptualise the parameter, state or input spaces of a model because they are real number spaces (Figure 1). Uncertainty in these spaces can be easily defined with probability density functions (PDFs) and, through either frequentist or Bayesian techniques, propagated through the model to the (real number) output space. A well known example of this is climateprediction.net (Stainforth et al, 2005), where thousands of perturbed parameter runs of HadCM3 were used to estimate the climate response to anthropogenic CO2 forcing. While a pdf of model output behaviour obtained in this way might represent the probability of a particular model outcome, it is clearly not equivalent to the probability of the event actually occurring. This distinction is precisely because of model space uncertainty.

Figure 1: A typical systems representation of a model showing input, parameter,

state and output spaces as real number spaces.

The most common approach to overcoming this issue is the use of multi-model ensemble simulations (e.g., Houghton et al, 2001; Gillet et al, 2002). The rationale is that different modelling groups produce quite different models and so provide independent estimates, but this independence is rarely if ever quantified (Tebaldi and Knutti, 2007). Models may be different for a variety of reasons:

• Different processes may be included. This can be thought of as a difference in perceptual models – the processes we see or imagine are part of the natural system.

• Different process interactions may be included: This can be thought of as a difference in conceptual models – the nature of the relationship between processes.

• Different symbolic representation of processes which are included – that is, differences in the translation from conceptual model to mathematical model.

• Differences in numerical or scale implementation, precision – the numerical model. Even with these potential differences in mind, it’s hard to judge how independent models need to be. The hope is that an ensemble samples the model space broadly enough so that in most measures of performance, the multi-model ensemble should be unbiased relative to observations. To date however, we have no technique to define dependence or independence of models, and so no way of knowing what ‘spanning’ the model space may mean.

42

We attempt to formalise the idea of the model space by proposing a distance measure, or metric, for this space. That is, a way of measuring the distance between two models. If we can successfully cast the model space as a metric space in the mathematical sense, this may allow us to treat the model space in a similar statistical framework to real number spaces such as parameter, input and state space. To begin with however, we simply use this distance measure as a proxy for model dependence. To illustrate the proposed model space metric, we use land surface models (LSMs), specifically the prediction of net ecosystem exchange of CO2 (NEE) as a function of site/grid cell meteorology. Measured values of meteorological data are taken from 13 flux tower sites in Europe and the USA from a variety of biomes, totalling about 40 site years, and used to drive the LSMs. The first step is to divide all the time steps of meteorological data into discrete groups (henceforth ‘nodes’), each containing similar conditions. A very simple way to do this, for example, might be air temperature > (or <) 15C and relative humidity > (or <) 50% - this would provide us with four climatologically divided nodes. For this work we chose to use a self-organising map (SOM, Kohonen, 1989) to divide meteorological data into 9 nodes. The mean values of four meteorological variables for each of the 9 nodes are shown in Figure 2.

Figure 2: Average values for four LSM input variables: SW down, air temperature, specific humidity and windspeed for time steps belonging to each node of a 9-node Self Organising Map. This meteorological data, used to drive the LSMs, was sourced from 13 eddy covariance flux tower sites across two continents.

For each time step belonging to one of these nodes, we extract the corresponding NEE values from the time series of predictions by two LSMs: CABLE (Kowalczyk et al, 2006) and ORCHIDEE (Krinner et al, 2005). For each LSM, we construct a PDF of the NEE predictions associated with each node. These PDFs are shown for each node in Figure 3.

43

Figure 3: Probability density functions of CABLE's and ORCHIDEE's prediction of NEE for each SOM node (conditions and the number of data of each node are shown in previous figure). Note that abscissa units are µmol/m2/s and that nodes have different scales to show relative model differences.

The grey region in each panel of Figure 3 represents the overlap of CABLE’s and ORCHIDEE’s PDF of NEE predictions for the particular set of meteorological conditions represented by a SOM node. If we represent this overlap region for the nth node as On and the number of time steps belonging to the nth node as an, then we define the distance between two models m1 and m2 as:

d(m1,m2) =1−an

A•On

n=1

N

∑

where A is the total number of time steps (here 726384) and N is the total number of nodes (here 9). That is, we simply add all areas associated with the shaded regions, weighted by the number of time steps belonging to each node. Since these are PDFs, this sum is between 0 and 1. While we have not proved it to be the case, it seems plausible that this measure of the model space may satisfy the metric space axioms: d ≥ 0 and it will only be zero when all density functions overlap perfectly, so d?m1,m2?=0 if and only if m1=m2. It should also be clear that d?m1,m2?=d?m2,m1?. No immediate counter examples to the triangle inequality d?m1,m2??d?m2,m3?=d?m1,m3? seem apparent either. Using this metric, any binary decision about which models are dependent or independent is simply a matter of choosing a 'dependence radius' – models separated by less than this radius could be considered to be dependent, those separated by more to be independent. In most multi-model ensemble applications, however, this information would be used as a separate model property, alongside performance, in the construction of model weights. The process might proceed as follows:

1. Construct model weights using a (collection of) performance measures. 2. Reduce or increase a model’s weight according to its distance to all other models. For

example: w1 = P1.(Σ d(m1,mi))/(w1+…+wn) where P1 is the performance of model 1.

44

The conceptualisation of the model space as a metric space presented here may give us a much needed tool that could allow us to assess the relationship between model spread and prediction uncertainty. While we are clearly at a very early stage of understanding model independence, it should be obvious that it needs to be a separate consideration to model performance when constructing an unbiased model ensemble. Gillett, N. P., F. W. Zwiers, A. J. Weaver, G. C. Hegerl, M. R. Allen, and P. A. Stott, 2002.

Detecting anthropogenic influence with a multi-model ensemble, Geophysical Research Letters, 29, 1970, doi:10.1029/2002GL015836.

Houghton, J. T., Y. Ding, M. Nogua, D. Griggs, P. Vander Linden, K. Maskell (eds.) 2001.

Climate Change 2001: The Scientific Basis. Contribution of Working Group I to the Third Assessment Report of the Intergovernmental Panel on Climate Change, Cambridge University Press, Cambridge, U.K.

Kohonen, T. 1989. Self-Organization and Associative Memory. Springer-Verlag. Kowalczyk, E. A., Y. P. Wang, R. M. Law, H. L. Davies, J. L. McGregor and G.

Abramowitz, 2006. The CSIRO Atmosphere Biosphere Land Exchange (CABLE) model for use in climate models and as an offline model CSIRO Marine and Atmospheric Research paper 013, www.cmar.csiro.au/e-print/open/kowalczykea_2006a.pdf 264

Krinner, G., N. Viovy, N. de Noblet-Ducoudre, J. Ogee, J. Polcher, P. Friedlingstein, P. Ciais,

S. Sitch and I. C. Prentice, 2005. A dynamic global vegetation model for studies of the coupled atmosphere-biosphere system. Global Biogeochemical Cycles, 19(GB1015).

Stainforth, D. A., T. Aina, C. Christensen, M. Collins, N. Faull, D. J. Frame, J. A.

Kettleborough, S. Knight, A. Martin, J. M. Murphy, C. Piani, D. Sexton, L. A. Smith, R. A. Spicer, A. J. Thorpe & M. R. Allen, 2005. Uncertainty in predictions of the climate response to rising levels of greenhouse gases, Nature, 433, pp.403-406.

Tebaldi, C. and R. Knutti, 2007. The use of the multimodel ensemble in probabilistic climate

projections, Philosophical Transactions of the Royal Society A, 365, 2053-2075, doi:10.1098/rsta.2007.2076.

45

Evaluating the Australian Community Land Surface Model for Australian ecosystems

Yingping Wang1, Gabriel Abramowitz, Eva Kowalczyk1, Rachel Law, Bernard Pak1, Adam Smith2, Huqiang Zhang2

yingping.wang@csiro.au

1. CSIRO Marine and Atmospheric Research, PMB #1, Aspendale, Victoria 3195 Australia

2. Australian Bureau of Meteorology, Melbourne, Victoria 3000, Australia

The Australian Community Climate and Earth System Simulator (ACCESS) initiative is in the process of building a fully coupled global climate model which includes a fully interactive carbon cycle. One of the key priorities is to build a model able to perform IPCC 5th Assessment Report (AR5) experiments. The strategy for ACCESS in terms of land-biosphere modelling for the AR5 is to couple the CSIRO Atmosphere Biosphere Land Exchange (CABLE), CASA-CNP and LPJ DGVM models. The coupled system has to be calibrated against available observations and stable under present climate conditions when coupled to ACCESS atmospheric model by mid 2009. CABLE is a model of biosphere-atmosphere exchange which includes the aerodynamic and radiative interaction of a canopy with the bare ground below and atmosphere above and the treatment of turbulence inside the canopy. CABLE is a one layer two-leaf canopy model as described in Wang and Leuning (1998) and it was formulated on the basis of a multilayer model of Leuning et al. (1995). CABLE was coupled to the CSIRO global circulation model but it is available to be used as an offline model by the Australian research community. The main features of CABLE are: - a coupled model of stomatal conductance, photosynthesis and the partitioning of absorbed net radiation into latent and sensible heat fluxes - the model differentiates between sunlit and shaded leaves i.e. two-big-leaf submodels for calculation of photosynthesis, stomatal conductance and leaf temperature - the radiation submodel calculates the absorption of beam and diffuse radiation in visible and near infrared wavebands, and thermal radiation - the vegetation is placed above the ground allowing for full aerodynamic and radiative interaction between vegetation and the ground - the plant turbulence model by Raupach et al. (1997) is used to calculate the within canopy air temperature and humidity - a multilayer soil model is used; Richards equations are solved for soil moisture and heat conduction equation for soil temperature - the snow model computes temperature, density and thickness of three snowpack layers. For a full description of the model see Kowalczyk at al. (2006). At present we are developing a global terrestrial biogeochemical model for carbon, nitrogen and phosphorus (CASACNP) including symbiotic nitrogen fixation (see Wang, Houlton and Field, 2007). Our model is similar to CENTURY, but includes biochemical phosphorus mineralization and symbiotic nitrogen fixation, both processes are important to the soil nutrient cycling in the tropical forest and savannah. An offline version of the model is being calibrated. Coupling of the LPJ (Lund-Potsdam-Jena) dynamic vegetation model to CABLE and CASACNP will be led by the Australian universities.

46

References E.A. Kowalczyk, Y.P. Wang, R.M. Law, H.L. Davies, J.L. McGregor and G. Abramowitz, ‘CSIRO

Atmosphere Biosphere Land Exchange (CABLE) model for use in climate models and as an offline model’, CMAR Research Paper 013, 2006.

Leuning, R, Kelliher, FM, de Pury, DGG and Schulze, ED 1995. Leaf nitrogen, photosynthesis, conductance and transpiration: scaling from leaves to canopy. Plant, Cell and Environment 18:1183-1200.

Raupach, MR, Finkele, C, Zhang, L. 1997. SCAM: description and comparison with field data. CSIRO Centre for Environmental Mechanics Technical Paper No. 132. 81p.

Wang, Y.P. and Leuning, R. (1998) A two-leaf model for canopy conductance, photosynthesis and partitioning of available energy I. model description and comparison with a multi-layered model. Agricultural and Forest Meteorology, 91:89-111.

Wang, YP, Houlton, B. and Field, CB. 2007. A model of biogeochemical cycles of carbon, nitrogen and phosphorus including symbiotic nitrogen fixation and phosphatase production. Global Biogeochemical Cycles, 21, GB1018, doi:10.1029/2006GB002797.

47

Modelling Australian Tropical Savannas: current tools and future challenges

Peter Isaac1, Jason Beringer1, Lindsay Hutley2 and Stephen Wood1

1 School of Geography and Environmental Science, Monash University, Melbourne, Victoria, Australia

2 School of Science and Primary Industry, Charles Darwin University, Darwin, Northern Territory, Australia

Introduction Savanna ecosystems cover approximately 20% of the Earth's surface and 25% of the Australian

landmass. Australian tropical savanna is a strong sink of carbon with a typical net ecosystem productivity of 4.3 tCha-1yr-1 for an undisturbed system (Beringer et al., 2007). Processes controlling the fluxes of mass and energy over tropical savanna are not well understood and this has hampered attempts to include these biomes in global modelling studies (Law et al., 2006).

The Australian tropical savanna presents many unique features to experimentalists and modellers alike. It is a mix of C3 (typically overstorey) and C4 (typically understorey) plants. It is a highly dynamic ecosystem with large annual variations in the fluxes driven by the response of the vegetation, particularly the understorey, to a bi-modal rainfall pattern (wet season/dry season) of the tropics. There are also large inter-annual variations driven by climate and disturbance by fire.

We have recently conducted a sensitivity study to answer two fundamental questions about the modelling of the Australian tropical savanna. The questions are: • Is it necessary to use a multi-layer canopy model to resolve the seasonal dynamics of the canopy in order

to replicate the seasonal variation in the mass and energy fluxes? • To what extent is it necessary to resolve the dynamics of the canopy morphology, particularly the ratio of

C4 leaf area index (LAI) to total LAI, in order to replicate the seasonal variation in the fluxes? This paper will briefly describe a multi-year project aimed at improving our understanding of the

processes governing spatial and temporal variations in the mass and energy fluxes over tropical savanna in Northern Australia. We present observations from a long-term flux tower site to show the important drivers of the annual cycle in the fluxes over tropical savanna. The results of a model sensitivity study to determine the complexity of model and inputs required to capture the annual cycle are then presented. We finish with some conclusions regarding the models and some thoughts on future directions.

The Project Groups from Monash University, Charles Darwin University, the University of California, Davis,

Airborne Research Australia and MetAir AG will collaborate over the next three years in a detailed study of the patterns and processes of carbon, water and energy cycles across Northern Australia (the Top End). The major aims of the study are: • To characterise the spatial variability of carbon, water and energy cycles across a range of ecosystems in

the Top End and the processes responsible for this variability. • To extend this knowledge from point to regional scales by integrating process-based models and

measurements from a range of platforms. Six flux towers have been installed over different ecosystems at three geographic locations in the

Northern Territory. Measurements include the four components of the radiation budget, eddy-covariance measurements of sensible heat, latent heat, CO2 and momentum fluxes, ground heat flux, air temperature, relative humidity and soil moisture and temperature at 10 and 50 cm. Rainfall is measured at Howard Springs, the main site 50 km south-east Darwin. An Intensive Observation Period (IOP) will take place at the end of the dry season in September 2008. Two research aircraft from Airborne Research Australia will provide observations of the surface fluxes at regional scales using low-level and box-budget flight techniques as well as providing fine-scale Normalised Difference Vegetation Index (NDVI) measurements and digital photographs of the surface. Land Surface Models (LSM), off-line and coupled to a mesoscale model, will be

48

used throughout the project to investigate the processes responsible for modulating the fluxes of mass and energy and to identify areas of particular interest for examination during the IOP.

Cycles and Drivers The Howard Springs site has been in operation since 2001 over a 16 m canopy of open forest

savanna. A second tower system was recently installed to measure the understorey and soil contributions to the total fluxes. An automated chamber system for measuring soil respiration has also been operating at this site for the past two months.

The climate at Howard Springs is characterised by warm temperatures (annual average 26.4 C), high humidity (annual average 70%) and a bi-modal distribution of annual rainfall (wet season/dry season).

0 3 6 9 12Month

0

0.5

1

1.5

2

2.5

LAI (total)LAI (US)C4 fraction

0 3 6 9 12Month

-50

0

50

100

150

200

Net radiationLatent heat fluxCO2 flux

-5

0

0 3 6 9 12Month

0

0.1

0.2Soil moisture (10cm)

0 3 6 9 12Month

0

100

200

300

400

Rainfalla) b) c) d)

mmµmolm-2s-1 m3m-3

Wm-2

Fig. 1: Monthly averages of a) total (open circles) and understorey LAI (filled triangles) and the C4 fraction (filled squares), b) net radiation, latent heat flux and CO2 flux, c) soil moisture and d) rainfall at Howard Springs.

Figure 1 shows the monthly average LAI and C4 fraction, net radiation (Fn), latent heat (Fe) and CO2 (Fc) fluxes, volumetric soil moisture at 10cm depth and rainfall at the Howard Springs site for the year 1/6/2004 to 31/5/2005. C4 fraction is the ratio of the C4 understorey LAI to the total canopy LAI.

The rainfall plot shows the distinct wet (November to March) and dry (April to October) seasons and these are also evident in the soil moisture. The annual cycles of rainfall and soil moisture drive the annual cycle in LAI and C4 fraction. The annual cycle in total canopy LAI is driven mainly by the understorey LAI since the annual variation in overstorey LAI (not shown) is less than 30%. Understorey LAI reaches a maximum at the end of the wet season after which the grassy understorey senesces with understory LAI reaching a minimum toward the end of the dry season. The annual cycle in the understorey LAI drives a similar cycle in the C4 fraction.

The annual cycles in LAI and C4 fraction are expected to be major drivers of the seasonality in the fluxes of mass and energy over the tropical savanna. We have used two LSMs of differing complexity to predict monthly fluxes at the Howard Springs site in order to assess the extent to which the seasonality of the model predictions depend on these drivers. The period modelled was from 1/6/2004 to 31/5/2005.

Models CABLE (CSIRO Atmosphere-Biosphere-Land Exchange model, Kowalczyk et al. 2006) is a coupled

assimilation/transpiration model that treats the vegetation as two "big leaves", one sunlit and one shaded . A mixed C3/C4 canopy is modelled by specifying that a fraction of the "big leaves" follows the C4 pathway. CABLE accepts a seasonally varying LAI and has been modified to accept a seasonally varying C4 fraction for this study. Soil moisture is modelled over 6 layers using the Richards equations with soil properties assigned on the basis of soil type (9 types).

A SINE curve was fit to the total LAI and C4 fraction (soild lines in Figure 1a). The fitted curves reproduce the observed seasonality at a small expense to the maximum and minimum values. The curves were used to calculate total LAI and C4 fraction as functions of day number.

ACASA (Advanced Canopy-Atmosphere-Soil Algorithm, Pyles et al. 2006) is a relatively complex model in comparison to CABLE but uses the same equations for coupled assimilation/transpiration and soil moisture. The canopy is split into 101 levels for calculating radiation transfer and into 11 layers for calculating the subsequent fluxes of mass and energy. Soil moisture is modelled over 15 layers using the

49

Richards equations with soil properties assigned on the basis of soil type (16 types). The model contains a full third-order turbulence sub-model and is able to provide predictions of the fluxes within and immediately above the canopy. ACASA does not explicitly model the C4 photosynthetic pathway but this can be achieved by specifying the appropriate parameter values for those canopy layers dominated by the C4 vegetation (Collatz et al. 1992). This has not been done for this study. As a result, ACASA is expected to under-estimate Fe and Fc when the C4 fraction is high (November to March).

The total canopy LAI and the LAI profile through the canopy were allowed to vary each month. Total LAI was set to the observed values. The LAI profile was estimated by separating the total LAI into under- and overstorey components on the basis of the observed C4 fraction. The resulting component LAI was distributed over the profile based on the vertical extent of each component. The overstorey was assumed to occupy the top half of the canopy with no seasonal variation while the height of the understorey varied from 0.5 m at the end of the dry season to 2 m at the end of the wet. The LAI of each component was distributed evenly across all layers occupied by that component.

Results

Default Parameters Figure 2 shows the monthly averages of observed and modelled Fe and Fc. With the exception of

LAI and C4 fraction, the models were used with their default parameter sets. This was done to assess the ability of the models to predict the observed fluxes over tropical savanna without site-specific tuning. This is the similar to the way such models are used in General Circulation Models (Law et al. 2006).

0 3 6 9 12Month

0

40

80

120

160

200

(Wm

-2)

ObsCABLEACASA

0 3 6 9 12Month

-4

-2

0

2

4

6

8

(µm

olm

-2s-1

)

a) Fe b) Fc

Fig. 2: Monthly averages of a) Fe and b) Fc from observations (filled circles), CABLE (open diamonds) and ACASA (open triangles). Both models used default parameter sets and observed LAI and C4 fraction.

Neither CABLE nor ACASA predict the magnitude of Fe well over the whole annual cycle. CABLE does well during the wet season but poorly in the dry, whereas ACASA does poorly in the wet season but well in the dry. Time series plots of the fluxes (not shown) suggest that during the wet season ACASA seeks to maintain Fe at the expense of the sensible heat flux (Fh), often to the point of driving Fh to values of -250 Wm-2 in the late afternoon and early evening. This feature remained in the ACASA predictions of Fe and Fh despite the modification of input parameter values to better reflect site-specific conditions, see next section.

Both models capture the seasonal variation of Fe. CABLE predicts both the magnitude and seasonal variation of Fc well using the default parameters and the observed LAI and C4 fraction. In contrast, ACASA does poorly in predicting Fc except for three months towards the end of the dry season.

Using the default parameter values, neither model performed well enough to assess their ability to capture the observed seasonal variation in Fe and Fc with any confidence.

Examination of time series showing the diurnal variation in both observations and model predictions (not shown) revealed several areas where the default parameter sets could be improved without undertaking a detailed site-specific tuning exercise for each model. First, CABLE uses a default value of 1 for the C4 fraction of savanna vegetation whereas observations show that this varies from 0.13 to 0.6 at Howard Springs. Second, both models used a value of 0.135 m3m-3 for the soil moisture at wilting point whereas observations show that the monthly average soil moisture at 10 cm depth can reach 0.08 m3m-3 during the dry season and yet Fe remained significant. Third, both models assumed that most root material is located in the top 0.5 m of the soil column whereas evidence suggests that the tree species in Australian tropical

50

savanna have roots to a much greater depth (Hutley, unpublished data). Finally, the Fc predictions from ACASA are dominated by night-time respiration during the wet months but the dry season values are much closer to the observations (not shown). ACASA includes a soil microbial respiration term that is strongly dependent on soil moisture. These results suggest that this term may not be appropriate for soil conditions in the tropical savanna of the Top End.

Modified Parameters On the basis of the above, the default parameter sets were modified as follows. First, CABLE was

modified to accept the C4 fraction as an input variable. Second, the value of the maximum rate of Rubisco-limited carboxylation, vcmax, used by CABLE for savanna was considered too low (Wang, pers. comm.) and was doubled. This aspect is further discussed in the Conclusions. Third, the soil moisture at wilting point was set to 0.08 m3m-3, the minimum monthly average at 10 cm, and the distribution of root material changed so that only 30% of roots were above 0.5 m. Finally, the soil microbial respiration term in ACASA was disabled.

Figure 3 shows the monthly averages of observed and modelled Fe and Fc obtained using the modified parameter sets and the observed annual cycle in LAI and C4 fraction.

CABLE does well at predicting both the magnitude and seasonal variation of Fe. Modification of the default parameter set did not improve the performance of ACASA for Fe. The CABLE prediction of Fc with the modified parameter set is not as good as with the default values and CABLE significantly under-estimates Fc during the dry months to the point of predicting the tropical savanna is a net source of CO2 for August, September and October. The observations indicate that the savanna is a net sink all year.

The prediction of Fc by ACASA using the modified parameter set is much better than with the default values and most of this improvement is due to the disabling of the soil microbial respiration.

0 3 6 9 12Month

0

40

80

120

160

200

(Wm

-2)

ObsCABLEACASA

0 3 6 9 12Month

-4

-3

-2

-1

0

1

(µm

olm

-2s-1

)

a) Fe b) Fc

Fig. 3: Monthly averages of a) Fe and b) Fc from observations (filled circles), CABLE (open diamonds) and ACASA (open triangles). Both models used the observed LAI and C4 fraction and modified parameter sets, see text for details.

CABLE and ACASA perform significantly better when their parameter sets are modified to suit site-specific conditions. The improvement in performance for both models is sufficient to warrant addressing the original aims of this study. That is, a) is a multi-layer model necessary to resolve the seasonal dynamics of mass and energy fluxes over tropical savanna and b) is it necessary to resolve seasonal changes in LAI and C4 fraction in order to accurately model the fluxes of mass and energy over tropical savanna?

Constant LAI and C4 Fraction We addressed these questions by running CABLE and ACASA with constant LAI and C4 fraction.

The total canopy LAI was set to the annual average of 1.5 and the C4 fraction was set to 0.5, representative of most of the year.

Figure 4 shows the monthly averages of Fe and Fc predicted by CABLE and ACASA with constant LAI and C4 fraction. Holding LAI and C4 fraction constant throughout the year makes little difference to the model's prediction of the magnitude or seasonal variation in Fe. In contrast, assuming constant values for LAI and C4 fraction has a slight effect on Fc predicted by ACASA but it has a significant effect on Fc predicted by CABLE. The ACASA predictions of Fc under-estimate the observations for most months but follow the seasonal variation in the observations. The CABLE predictions of Fc no longer match the observations in either magnitude or seasonal variation.

51

0 3 6 9 12Month

40

80

120

160

200

(Wm

-2)

ObsCABLEACASA

0 3 6 9 12Month

-3

-2

-1

0

1

(µm

olm

-2s-1

)

a) Fe b) Fc

Fig. 4: Monthly averages of a) Fe and b) Fc from observations (filled circles), CABLE (open diamonds) and ACASA (open triangles). Both models used constant LAI and C4 fraction and modified parameter sets, see text for details.

Conclusions This study allows a number of conclusions to be reached and suggests some areas of future work. First, increased model complexity does not guarantee robustness, simple and complex models both

do equally badly when used with inappropriate parameter values. The saving grace of simple models is that they require fewer input parameters and, for the models considered here, run at least two orders of magnitude faster.

Second, simple adjustment of soil and vegetation parameters based on observations markedly improved performance for both simple and multi-layer models.

Third, the performance of a "big leaf" model was similar to that of a multi-layer model when the model parameter sets were tuned appropriately and the models used the observed LAI and C4 fraction.

Fourth, a "big leaf" model (CABLE) was more sensitive to the correct specification of seasonal variation in LAI and C4 fraction than a multi-layer model (ACASA).

The immediate challenges in modelling mass and energy fluxes over Australian tropical savanna relate more to a lack of knowledge about model parameters than to a lack of knowledge about the processes involved. Our modification in this study of the default parameter sets used by each model dealt mainly with quantities that can be measured directly eg LAI, C4 fraction, vcmax, soil properties in general but particularly soil moisture at wilting point and the distribution of root material within the soil column. These quantities can, and perhaps should, be routinely measured. Our results show that the effort required will be rewarded with much more realistic model results.

For the longer term, a major challenge remains the use of observations from flux towers, aircraft and satellites to inform and constrain LSMs. We believe that integrating observations and models offers the best approach for accurate assessment of mass and energy budgets at regional scales, especially over a dynamic land surface such as the tropical savanna.

References Beringer, J., Hutley, L. B., Tapper, N. and Cernusak, L. A. 2007. Savanna fires and their impact on net

ecosystem productivity in North Australia. Global Change Biol., 13, 990-1004 Collatz, G. J., Ribas-Carbo, M. and Berry, J. A. 1992. Coupled photosynthesis-stomatal conductance model

for leaves of C4 plants. Aust. J. Plant Physiol., 19, 519-538 Kowalczyk, E. A., Wang, Y. P., Law, R. M., Davies, H. L., McGregor, J. L. and Abramowitz, G. 2006. The

CSIRO Atmosphere-Biosphere-Land Exchange (CABLE) model for use in climate models and as an offline model. CSIRO Marine and Atmospheric Research paper 013, November 2006.

Pyles, R. D., Weare, B. C. and Paw U, K. T. 2000. The UCD Advanced-Canopy-Atmosphere Soil Algorithm (ACASA): comparisons with observations from different climate and vegetation regimes. Q. J. Roy. Meteorol. Soc., 126, 2951-2980

Law, R. M., Kowlaczyk, E. A. and Wang, Y. P. 2006. Using atmospheric CO2 data to assess a simplified carbon-climate simulation for the 20th century. Tellus, 58B, 427-437.

53

Observations and studies on arid and semi-arid

land-atmosphere interaction in Northwest China

Yaohui Li

Institute of Arid Meteorology, China Meteorological Administration, Lanzhou, Gansu

province, China

Introduction

Institute of Arid Meteorology (IAM), China Meteorological Administration (CMA), is one of the national specialized research institutes in studying basic meteorological system and serving for the public, with primary focuses on aridity monitoring, prediction and warning, proper utilization of water resources, sand-dust storm research and prediction and so on. In this presentation, some key observation and research activities conducted by IAM in land-atmosphere interaction and its potential applications in arid and semi-arid regions will be discussed.

Objects of the observation and experiment

• To recearch on the land surface process and land-atmosphere interaction in arid and semi-arid regions;

• To research the characteristics and formation mechanism of atmosphere boundary layer in arid and semi-arid regions;

• To research the energy exchange and water cycle rules of agro-ecosystem in arid and seim-arid regions;

• To research the methods of drought monitoring, drought index, formation and evaluation of drought disaster in arid and semi-arid regions;

• To research the remote sensing monitoring in soil-plant- atmosphere system;

Observation and experiment regions and station nets

Main observation regions includ arid area of Hexi Corridor and semi-arid area of middle-east region of Gansu Province, China. Arid climate observation systems (ACOS) covers almost all over the northwest part of China. Observation station nets consist kinds of experiment bases and meteorological stations, such as Dingxi Arid Meteorological and Ecological Environment Experimental Site (DAMES) of Institute of Arid Meteorology, China Meteorological Administration.

54

Observation items and instruments

Eddy Covariance Observation Systems • Gradients of Temperature, Humidity and Wind • Doppler weather radars upper-air observation • Surface radiation balance observation • Soil temperature, moisture and heat flux observations • Lysimeter • Crop ecological observation • Surface Remote-sensing • General weather data observation • Water vapor GPS/MET observation system • Atmospheric composition observation

Some preliminary results

• The reason that soil become dry on loess plateau in the eastern Gansu • Soil water “respiration” • Momentum, sensible heat flux, latent heat flux and bulk transfer coefficient in arid and

semi-arid region of northwest China • Characteristics of surface energy in farmland, grassland and gobi in northwest China • Characteristics and formation mechanism of atmosphere boundary layer in arid and

semi-arid region of northwest China • Impacts of climate warming on crops in arid and semi-arid region of northwest China.

55

Hydrological modelling in Australia

Francis Chiew

CSIRO Land and Water, GPO Box 1666, Canberra, ACT, 2601 Introduction There are many types of hydrological models, which may operate over different spatial scales and time steps and are developed for various applications. These range from models for estimating catchment and continental water availability to models for planning and managing water resources across regulated river systems. The need for hydrological models is increasing both in terms of coverage and functionality. The types of people that require access to hydrological models and in particular model results is also increasing. Hydrological models need to be more robust, transparent and defensible as they are increasingly relied on to make informed decisions on sharing and managing Australia’s limited water resources. The different needs of hydrological modelling are so complex that it is beyond the scope of any organisation to continue developing models on its own. The way forward is for modelling experts across Australia to work together in model development in a collaborative environment. There is no better opportunity to do this than now, with the various research partnerships across Australia (e.g., eWater CRC, WIRADA, CACWR) and the strong national support (e.g., NWC, BoM, AWRIS, DEWR, state water agencies) for providing water information by combining hydrological models and integrated data systems. There will continue to be different types of hydrological models because they are required for different purposes. However, to provide a consistent language and interpretation for water accounting, water forecasting, policy development and legislation across river basins, states and territories, the water fluxes and system states simulated by the different models must have a consistent or similar meaning across time and space scales. For example, simulations from a higher resolution model when aggregated up should match the same simulation from a lower resolution model. To develop a realistic vision for future hydrological modelling development, the various modelling communities need to appreciate the different model types and purposes. This is particularly important where hydrological models are used together with other models, like climate models. In some applications, the different models need to be linked directly because feedbacks between models are very important. In other applications, the different modelling can essentially be carried out separately with outputs from one model used to drive the other model. This talk will present the different types of hydrological models, grouped into key modelling applications in Australia (summarised in the next section), and discuss the different hydrological modelling requirements for different time scales and objectives (summarised in the last section).

56

Typical features of main types of hydrological models Daily conceptual rainfall-runoff models • Used to estimate daily runoff from daily rainfall and potential evapotranspiration, mainly to

extend streamflow records, to estimate catchment, regional or continental water availability, and to estimate catchment inflows into river system models.

• Increasingly adapted and used to estimate climate and development impact on runoff. • 5–20 parameters, generally calibrated against observed streamflow data, and regionalisation

method used to transfer parameter values for use in ungauged catchments. • Essentially lumped catchment or subcatchment representation (although some models use

distribution functions to represent spatial variability and sub-daily rainfall characteristics). • Literally thousands of conceptual rainfall-runoff models in the literature, commonly used

models in Australia include Sacramento, SIMHYD, AWBM and IHACRES. Sub-daily hydrological models • Used mainly to estimate catchment runoff for forecasting (e.g., flood forecasting and river

operations). • Sub-daily rainfall (intensity, amount and spatial distribution) is the key driver of the models. • Accurate knowledge of antecedent soil water conditions is important. The use of remotely-

sensed data can improve estimation of antecedent conditions, but to meaningfully assimilate remotely-sensed data, models require direct representation of surface layer moisture observed by satellites.

• Except for loss functions used in flood forecasting, and models used in experimental and research studies, sub-daily hydrological models have not been widely used in Australia for water applications and catchment runoff forecasting.

River system models • Used by state water agencies and MDBC (e.g., REALM, IQQM, MSM-Bigmod) to simulate

river flows and water uses across regulated river systems for water accounting and planning. • Link-node network used to represent river system, and key model components include: rainfall-

runoff models to estimate catchment inflows; routing algorithms to rout flows; models of surface-groundwater interaction in river reaches and floodplains; empirical functions of irrigation and environmental water demand; and system and reservoir operating rules that describe water sharing and allocation. Models run on either a daily, weekly or monthly time step.

• There is a hydrological modelling initiative in Australia involving research institutions and industry partners, which is attempting to truly integrate rainfall-runoff, river system and groundwater models (both in terms of modelling algorithms and in system-wide calibration and model parameterisations) to improve water accounting and estimation of climate and development impacts on water uses throughout managed river systems.

Land surface models • Used in numerical weather prediction (NWP) and global climate models (GCMs) to model land

surface processes. • Lumped parameterisations over climate model grids of 20–500 km2, with many parameters

whose value need to be specified for all grids across the globe.

57

• Unlike the above model types, accurate modelling of atmosphere-surface feedbacks, in particular partitioning of net radiation into latent and sensible heat fluxes, is important. Because land surface models are linked to atmospheric models, they run over time steps of less than 30 minutes. Simulation of runoff and runoff components are usually poor.

• Numerous land surface models have been developed and used by various climate modelling groups worldwide. The ACCESS climate modelling initiative between CSIRO, BMRC and universities uses the CABLE scheme as the land surface model.

Hydrological modelling requirements for different time scales and objectives Up to 2–3 days Forecast based on current hydroclimate information (hydrologic memory) • Sub-daily catchment hydrological modelling. • Sub-daily river routing models for floods and sub-daily river operation model for river

operation. • Real-time data ingestion, data assimilation and states update/correction. • Sub-daily rainfall from gauges and radar, and initial states are very important. 3–4 days to several weeks Probabilistic forecast based on current hydroclimate information and NWP model forecast (hydrologic memory and climate/weather memory) • Same as above, but with NWP model forecast to extend forecast lead time. • Real-time data ingestion and data assimilation into NWP land surface model and/or catchment

hydrological model (either the one model or some reconciliation of the two models). • Initial states are very important. Months/seasons Probabilistic forecast based on seasonal hydroclimate forecast • Daily catchment rainfall-runoff modelling and river system modelling using climate forecast

inputs. • Seasonal hydroclimate forecast from statistical methods and/or statistical–climate model hybrid. • Start from known initial system state. Several years Probabilistic prediction based on climatology or low-skill long-term climate forecast • Ensemble hydrological modelling starting from current system state to provide probabilistic

indication of future water availability (for example, given the current dry conditions and low system storages, water availability and water allocation are likely to remain low for several more years even with average rainfall conditions).

More than 10 years Scenario modelling for planning (modelling requirements for historical water accounting are similar) • Daily catchment hydrological modelling and river system modelling.

58

• Scenario modelling for future water availability and impact on water uses throughout river systems (impact of climate change, development/management, etc…).

• For climate change impact assessment, projections from GCMs are used, either to modify historical climate series to obtain future climate series or to statistically or dynamically downscale to obtain future climate series at the catchment scale to drive the hydrological models.

• Initial state is not important.

59

Integrated hydrological studies

Michael J. Manton School of Mathematical Sciences, Monash University, Clayton, Victoria, Australia Introduction In recent years, water availability has become a problem in many countries of the world. Climate change, affecting precipitation, temperature and evaporation, is one cause of these problems. However, in most cases, climate change has been an additional stress that compounds the impacts of long-running stresses due to human activities. Activities, such as urbanisation and irrigated agriculture, have adversely affected water availability in northern China, eastern Australia and northern Chile as well as many other regions of the world (IPCC, 2007). As surface water becomes a limiting factor in water availability, it is common for groundwater to be seen as an alternative perhaps unlimited source. However, Hu et al. (2007) show that the extraction of groundwater in the Heihe river of north China has affected the sustainability of the basin. Similar problems are found in Chile and Australia. The range of ecological, economic and cultural issues arising as water use increases raises the question of whether current practices (including current regional population centres) are sustainable in the future. Programs to support water sustainability studies The scale and scope of the problems associated with water availability are so large that it is not practical for the problems to be solved by a single group, and so a number of national and international programs have been set up to provide a collaborative framework for addressing water issues. The Global Energy and Water Cycle Experiment (GEWEX) was established over twenty years ago under the World Climate Research Programme (WCRP). Two of the aims of GEWEX were to promote climate-related satellite missions and to promote cooperation between meteorologists and hydrologists. We are finally moving towards the first goal with CEOS (2006) preparing a response to the GCOS Implementation Plan. The CEOS plan provides international mechanisms to help ensure that consistent, long-term and homogeneous climate records can be collected from space in the future. Satellite data are becoming increasingly important for observing aspects of the water cycle, including precipitation, atmospheric moisture, snow cover, glaciers, lake area, and even soil moisture and groundwater. The Gravity Recovery and Climate Experiment (GRACE) satellites measure vertically-integrated terrestrial water storage, leading to estimates of the variability of groundwater for the first time. There has also been progress on the second goal of GEWEX, but there is still some way to go when we consider the full range of interactions in the water cycle. Indeed when we consider the interactions that relate to the sustainable use of water, the links between meteorology and hydrology may not appear to be the most important. The critical place of water in all community activities, including agriculture, energy generation from fossil fuels, heavy industry and urban living, means that we really need to consider the connections between all these sectors in order to properly assess the sustainability of water availability in a region.

60

The Murray Darling Basin (MDB) project was established in 2002 as an Australian contribution to the GEWEX Regional Hydroclimate Projects. The MDB project involves collaboration between the Bureau of Meteorology, CSIRO and the universities, and one of its aims is to promote the application of hydro-meteorological modelling and prediction. The project therefore provides a mechanism for promoting collaboration across the various communities interested in the sustainability of water availability. Basis of integrated studies While fully-integrated models are important to estimate the overall sustainability of a region, it is critically important to ensure that the basic physical interactions are understood and modelled correctly. Studies of the impacts of water variability on specific sectors or regions will be flawed unless the physical interactions are represented at an appropriate scale and scope and with an appropriate accuracy. The key interaction is the exchange of water between the atmosphere and the land surface, which determines the basic availability of water for most human activities and ecological systems. However, a specific issue is the degree to which groundwater needs to be included in a regional water budget. It is often believed that the time scale for groundwater variations is too long for climate-related studies to include the dynamics of groundwater. Indeed the poor representation of the lower boundary condition for the surface water components of early climate models led to considerable uncertainties in the overall land-surface water balance. On the other hand, human activities have been found to be influencing the availability of groundwater on time scales of years. Moreover, recent studies of GRACE data (Rodell et al., 2007) demonstrate that groundwater variations on seasonal time scales are a significant component of the overall water budget on continental scales. In Australia, the National Water Initiative is developing procedures to enable detailed water budgets to be carried out across all the major river basins. For example, the National Water Commission (2007) report describes the results of a water availability analysis for the period 2004-2005. The relevance of groundwater interactions on annual scales is apparent from this analysis. Example studies Under the auspices of the MDB project, a workshop was held in May 2007 to document existing modelling in Australia by the various groups involved in water budget modelling for the Murray Darling Basin and to consider the potential benefit of enhanced collaboration between those groups. It was noted that there are significant differences between the nature of the models used for operational purposes in Australia and those being developed under research programs. The promotion of technology transfer within operational agencies would seem to remain as a challenge in Australia. An outcome of the MDB workshop was the proposal to develop a project for the intercomparison of products, initially focused on evapotranspiration, from the different models used for water budget studies. Such an approach has been used successfully at the international level to enhance large-scale climate models (through the series of Model Intercomparison Projects of WCRP) as well as specific products such as sea-surface temperature (SST) analyses (through the GCOS SST and Sea Ice Working Group). Support for the water budget modelling project is currently being considered by the relevant agencies.

61

Activities to promote the continuous improvement of the representation of the basic hydro-meteorology provide a sound basis for the development of integrated studies that account for the interactions between the climate, the hydrology, the ecology and human activities across a region. For example, the Goulburn River basin in northern Victoria is an important region for irrigated agriculture, and it is being impacted by the large-scale climate trends of south-eastern Australia. Moreover, it may become directly linked with the water sources for the Melbourne urban area. Integrated studies in such regions would be very relevant to future policy development on water sustainability in Australia. Moreover, there is scope for such studies to be established in cooperation with overseas studies, such as the continuing work in China on the Heihe river basin. References CEOS. 2006. Satellite observation of the climate system: The Committee on Earth Observation Satellites (CEOS) response to the Global Climate Observing System (GCOS) Implementation Plan (IP), 50pp. Hu, L.T., Chen, C.X., Jiao, J.J. and Wang, Z.J. 2007. Simulated groundwater interaction with rivers and springs in the Heihe river basin. Hydrol. Processes, 21, 2794-2806. IPCC. 2007. Fourth Assessment Report, Working Group 2, Chapter 3, p128. National Water Commission (2007) Australian water resources 2005: A baseline assessment of water resources for the National Water Initiative, Level 2 assessment. May 2007, pp459. Rodell, M., Chen, J., Kato, H., Famiglietti, J., Nigro, J. and Wilson, C. 2007. Estimating ground water storage changes in the Mississippi River basin (USA) using GRACE. Hydrogeology J., doi: 10.1007/s10040-006-0103-7.

63

Integrating streamflow prediction and data into an operational system – the example of flood warning

Jim Elliott

Water Division, Bureau of Meteorology Introduction The aim of this paper is to discuss the issues associated with transforming the predictive capacity of some form of natural process modeling into an operational system. This is done through the example of flood warning, which involves using streamflow prediction modeling to forecast future flood levels and to use that forecast information to provide a flood warning service. There are considerations in this example that are peculiar to flood forecasting and warning, but the aim is to use flood warning as a case study to illustrate the need for an integrated approach when transforming any predictive modeling capacity into an operational system. An operational system for flood warning An operational system is essentially an information system designed to support the operation of a particular business. In the case of flood warning, it is the system that supports the production and dissemination of warning and other flood information aimed at reducing the damaging impacts of floods. A schematic of the Bureau of Meteorology operational flood forecasting system is shown in Figure 1. This system supports the transformation of data and other inputs into products that

Fig. 1. Schematic of Bureau of Meteorology Operational Flood Forecasting System contribute to this damage reduction, by providing agencies and people at risk with advance information about future flood conditions. An obvious requirement of a flood warning system is to be able to accurately predict future flood conditions with sufficient lead time to enable effective protection action to be taken. Clearly streamflow prediction and the data required to operate the prediction model are key technical inputs to an effective system; but they are not the

64

system. Streamflow prediction must work in harmony with all other system components in an integrated way if the objectives of the operational system are to be achieved efficiently and effectively. Some of the influences that have shaped this integration for the Bureau operational flood forecasting system are discussed in the following in terms of a number of general characteristics of many operational systems in order to illustrate this point. System purpose The outcome to be achieved from the system should be kept in clear view. What is the system trying to achieve? Who are the target group and what do they need from the system? In the case of flood warning the aim is to reduce the risk to loss of life and property damage by providing information both directly to those at risk, as well as to agencies involved in flood emergency management. Accurate streamflow predictions are crucial some of the time during an event, but not at others. For example long lead time, but uncertain forecasts can be very helpful in the early stages of an event when the benefits of being prepared early can outweigh the cost should the flood not eventuate. Flash flood warning is another example when effective warning response will only require a flood or no flood prediction rather than specific levels of flooding. This is not saying that accurate streamflow predictions are not important, but by focusing more on the outcome from (or purpose of) the system, their relative importance in the context of other components of the operational system can be better assessed. Output products What output products does the system need to produce to achieve the outcome? The product needs to match the requirements of the recipient (response agency, public) and may not always require a precise streamflow prediction. Although the current forecasting system does routinely produce forecast streamflow hydrographs, the more common output product is a worded warning message based on this hydrograph; often with some interpretation and indication of degree of certainty. Much less commonly, the hydrograph can be shared with selected clients judged to be able to interpret it appropriately. The issue of uncertainty is relevant here too, as it is not clear that warning clients are yet able to fully utilize the quantitative information on uncertainty being generated by modern prediction systems. The choice, and investments to be made in the streamflow prediction component needs to take consideration of this range of needs. Data issues The primary data inputs to the flood forecasting and warning system are rainfall and river level observations collected through the real-time data collection network. This network utilizes a range of technologies and cooperative arrangements and the assimilation of these data into a useable data set is a significant component of an operational flood forecasting system. This data set must serve the needs (timeliness, format, etc) of the streamflow prediction system used as well as other operational products. The issues involved with integrating these data into an operational system include dealing with missing data, interpolating data into regular intervals, blending multiple data streams from a single site, voiding obviously erroneous data, etc. Some of this can be done immediately prior to, and during a flood event, but data preparation must be as up-to-date as possible to enable the system to go operational with as little delay as possible. Much of this can be automated but not all, and so the system needs to provide the capability for users to intervene in order to adjust data in particular situations.

65

Model issues While there is a wide choice of models available for streamflow prediction, the characteristics of the flood warning problem make some more suited to the task than others, and the operation of the model needs to be integrated with other characteristics of the operational system. As indicated above, the warning requirement can be such that a refined model may not be justified to achieve the system outcome and care should be taken not to over-model. Simple models can give perfectly adequate results in many cases and have the advantage that users can more readily relate to a model built on simple concepts, which can facilitate user-intervention during critical stages of certain events in order to “correct” the model. While objective updating techniques can be used here, the ability to inject the experience of the operator through manual adjustments to the model can often be superior. Ideally the modelling system should be set up so that on-line verification is supported such as to provide an objective basis for considering when to implement new or improved models. It is important that such a verification system be designed around performance statistics that best match the use of the model output in the warning context. For example the normal model evaluation statistics for streamflow simulation that focus on overall hydrograph reproduction are less relevant for flood forecasting, where the ability to forecast the rising limb or the exceedance of particular thresholds can be most important. The ability of the system to offer users a choice of models can be an advantage, particularly where it is found that certain models work best in particular flood situations. Forecasting models need to operate at incremental time steps to match the way the data becomes available rather than in a “simulation” mode more common for hydrological models. Many research models are not developed in this mode and so some modification may be necessary before the model can be integrated into the operational system. Re-calibration (or verification) of the model after each event is necessary to ensure the information from each new event is “captured” in the model. To facilitate this, the Bureau system is set up so that calibration is undertaken in the same “system” that is used operationally. This ensures that the model calibration is adjusted in a manner consistent with its proposed operation. Operational and user requirements The circumstances under which the system needs to be operated needs to be considered. Does the demand for outputs from the system follow a regular predictable pattern, or is it random? For a flood warning system, the demand is unpredictable and so the system must be in a constant state of readiness. This requires that data must be continually up to date and the streamflow prediction model ready to operate at the required time interval to match the flooding pattern. The requirements of system users are also an important input to the system design. In the case of flood forecasting, the system has many users in all of the national forecasting offices of the Bureau. As discussed above, the requirement to operate the system is not always predictable and so a web-based design has been followed so that users can access the system from home as well as the office. Remote operation allows the forecaster to be able to respond to a quickly developing situation as well as monitoring a flood situation remotely when conditions are suitable. Having the same “national” system in all forecasting offices means that users moving between offices when (for example) workloads in one part of the country are particularly high, can become productive more quickly.

66

Other inputs and interactions In the case of flood forecasting, inputs of meteorological forecasts are critical at both the early developing stages of a flood as well as during the event. Although confidence in the quality of rainfall forecasts from numerical weather modeling systems is increasing, this input can still be quite qualitative and the operational system must be designed to make best use of the information in this form. The current system does this by providing for the sensitivity of the flood forecast to different meteorological forecasts to be tested in order to evaluate the impact of the uncertainty before issuing the warning. Increasingly meteorological forecasting involves more quantitative information on uncertainty and the streamflow prediction system needs to be structured such that this information can be sensibly used. As discussed above, whether or not the ultimate clients can utilise the detail in this information is another issue. Feedback Most operational systems have some form of feedback. In the case of flood forecasting this includes the observations of flood levels that are used to adjust or update the streamflow prediction model in order to improve its performance for later forecasts. Feedback from response agencies in terms of the relationship between current flood levels and whether critical threshold levels (eg. levee over-topping) are being approached is also important, as this can provide more specific accuracy and timeliness targets for the streamflow prediction model at that particular time of the event. Observations in the field can also prove invaluable during an event as a check that automated observation systems are functioning correctly. Support and maintenance The support and maintenance required for operational systems depends on its role and the consequences of outages. This aspect is often overlooked and under funded. In the case of flood forecasting, clearly the emergency nature of the operation of the system means that a robust system is required. This is not just confined to considerations such as IT infrastructure and its support, but includes building in features such as redundant data inputs and flows and choosing streamflow prediction models that are robust to missing data or work adequately with degraded data inputs. Maintenance of the metadata that supports the overall operation of the system is also important. Experience with the current flood forecasting operational system is that the effort required to maintain the integrity of the data inputs, metadata, responding to system configuration changes etc significantly outweighs the effort involved in streamflow prediction model calibration and maintenance. Conclusion Building an effective operational system involves many more considerations than just the operation of some form of prediction model. Issues such as the overall outcome expected from the system, its operating environment, characteristics of the operators and operations, data inputs and the level of operational robustness expected are also critical. Some of the influences of these factors on the design and development of the Bureau’s operational flood forecasting system have been discussed as a case study to illustrate this point.

67

Energy, Water and Carbon Cycles Simulated in a 51-Yr CABLE Global Offline Experiment: towards an integrated

modelling system

Huqiang Zhang1, Liang Zhang2

1Centre for Australian Weather and Climate Research, Melbourne, Australia 2 Institutes of Arid Meteorology, CMA, Lanzhou, China

Introduction In the development of the Australian Climate Community Climate Earth System Simulator (ACCESS), the CSIRO Atmosphere Biosphere Land Exchange (CABLE) model (Kowalczyk et al., 2007) is being developed and/or improved to realistically represent the land-surface energy, water and carbon cycles in the global climate system. Reviewing the development of land-surface modelling, hydrological modelling and carbon cycle modelling in the last several decades, one can see the trend towards integrating these “separate” components into a systematic modelling approach for a wide range of weather, climate and hydrological applications. For instance, the historical role of land-surface models (LSMs) used in weather and climate model was to provide correct surface fluxes to their host model. Soil moisture (SM) and runoff were not their primary concern. In some cases, SM was treated as a “rubbish bin” in such a way as SM “nudging” in NWP which used SM as an adjustable model parameter to minimise errors in near-surface meteorological variables due to deficiencies anywhere in the NWP model. This clearly becomes questionable when SM is a prime variable of interest for (i) the likely impacts of SM on dynamical seasonal forecasts and climate variability/predictability; (ii) the need of accurate SM information for river flow/flood forecasting, water resource and agriculture management; (iii) runoff acting as freshwater input in coupled model needed to be realistically simulated in fully coupled climate models. Similarly, the historical role of hydrological models was primarily to provide correct rain-runoff relationship and simulating surface fluxes right did not have the same importance as getting runoff right in its modelling framework. Nevertheless, evaporation is a significant water loss term in surface water budget and modelling (potential) evaporation in hydrological models has evolved from simple statistical approaches to Penman-Monteith type of physically-based calculations. It is also true that carbon cycle modelling is closely linked to water and energy cycle simulations. (Arora, 2002) as a large part of the ecosystem’s responses to weather and climate are water and energy limited. As part of the process in the CABLE development, in this study, we force the model with a 50-year global meteorological forcing dataset to study features of land-surface processes simulated in the model. We focus on exploring the characteristics of interannual and longer time-scale variations of land-surface processes and the connections between energy, water and carbon cycles simulated by the model. In addition, this offline simulation offers us as the first step to study the potential contribution of land-surface processes to climate variability and predictability, to be complemented by coupled modelling experiments in future. A river routing scheme is being employed to integrate the model runoff outputs to allow us compare the model results with some observed river discharge data over a number of river basins, demonstrating its potential for some hydrometeorological applications with further model development and calibrations. This paper presents some preliminary results from this study. Model and forcing data An offline version of CABLE is used in this study with its detailed model features being documented in Kowalczyk et al. (2007). In this study, the global soil and vegetation data come from

68

the model default datasets and interpolated to 1 by 1 degree resolution. The forcing data is developed by Ngo-Duc et al (2005) by combining and interpolating NCEP/NCAR reanalysis with CRU precipitation and temperature data and surface radiation component from SRB (Surface Radiation Budget). It is named NCC forcing data. This 6-hrly 1o by1o forcing dataset has been used in validating several land-surface model simulations of continental scale land-surface processes. In this offline experiment, we used year 1949 forcing data for a ten-year spin-up and then we run the model continuously with the forcing data for the period of 1950 to 2000. To validate the model numerical stability with 6-hourly forcing data, we have compared surface evaporation simulated with hourly in situ forcing data at Tumbarumba available for the period of 2002-03 and the ones simulated by using the global forcing data at the grid point of 36oS and 148oE for the period of 1999-2000. The similarity in both magnitudes and seasonal cycles from the two runs (not shown) suggest that results from the 6-hourly forcing experiments are valid in allowing us investigating characteristics of land-surface processes in the model simulations at monthly, seasonal and interannual time scales. Preliminary results Surface climatology: In the analysis, we try to evaluate the model surface climatologies of key variables controlling energy, water and carbon cycles. Nevertheless, such efforts have been limited by the lack of high quality global observational datasets. In recent years, there are a number of land-surface data assimilation projects which provide model-based land-surface climate by constraining a number of land-surface models with a combination of a large range of observational forcing data sources. Figure 1 shows the comparison of CABLE surface evaporation climatology with the ones produced by three land-surface schemes participating the Global Land-surface Data Assimilation Project (GLADS) for the same period of 1979-2000 (http://ldas.gsfc.nasa.gov/GLADS). These three models are representative of current land-surface models being used in weather and climate applications: with CLM from NCAR being widely used in climate research communities, NOAH model being used operationally in the NCEP operational systems and MOSAIC model being the first model explicitly considering land-surface sub-grid heterogeneity,. Results suggest that the CABLE surface evaporation climatology using NCC forcing offers similar features as seen in the three models using GLADS forcing data, except that comparing with MOSAIC and NOAH, CABLE tends to underestimate surface evaporation in the middle latitudes boreal forests in NH summer. Comparison of other variables leads to similar conclusions. Fig. 1: Comparison of CABLE surface evaporation climatology (mm d-1) with three GLADS model results. Interannual variability: With the completion of 51-yr CABLE offline simulations, it allows us to explore the characteristics of variability in land-surface processes simulated by the model. This is an important issue as a number of modelling studies suggested the potential impacts of some slowly varying land-surface processes in affecting climate variability and predictability in several regions. Another important aspect in CABLE evaluation is to assess whether it has skills in simulating some observed

69

variations of land-surface/water processes in the last several decades. Changes in stream flow have been highlighted as significant climate change consequences as reported in Climate Change in Australia by CSIRO and the Bureau of Meteorology in 2007. Figure 2 demonstrates that CABLE, as an integrated modelling system, is able to reproduce the nonlinear relationship between rainfall decline and a sharp reduction in surface runoff area-averaged over the SW of Western Australia (115-120oE, 36-30oS). Results also suggest that evaporation, as a significant water loss term, contributes to such discrepancy and therefore call for an integrated modelling framework in studying the likely consequences of climate change on hydrological processes and water management.

Fig. 2: With the rainfall forcing (a), CABLE simulated total runoff over the region in SWWA(b), the rain-runoff relationship(c), and the modulation due to surface evaporation (d). Interactions between energy, water and carbon cycles As an integrated modelling system, it is critical to understand the interactions between energy, water and carbon cycles simulated by CABLE before being implemented into the more complex Earth system simulator. Such analyses will also provide some insight view on the path of future model development and necessary complexities needed. As an example, Figure 3 shows the correlations between model-simulated Net Primary Product (NPP) and soil moisture (averaged in the model first 4 layers). Clearly, SM can have significant impacts on modelling carbon cycle. Results suggest the importance of improving soil hydrological process modelling while making significant efforts in introducing dynamical vegetation and other components in the model for simulating ecosystem-climate interactions.

Fig 3: Correlations between NPP and soil moisture simulated by CABLE. Potential hydrological applications Water cycle/hydrology is one of the dominant processes in the Earth’s climate system. Earlier studies showed that without correct modelling of surface runoff partitioning, water and energy

a b

c d

70

fluxes between the land-surface and atmosphere cannot be realistically simulated, no matter how complex the land-surface schemes are. Such processes can further affect weather and climate at a wide range of scales, from NWP, seasonal, climate variability/predictability to climate change. In CABLE development, efforts are being made in this area to enhance its potential hydrological applications. In this study, a simple river routing scheme named TRIP (Oki et al, 1999) has been coupled to the outputs from the CABLE 51-yr offline experiment to explore the model skill in simulating river discharge over large river basins. Figure 4 shows the averaged river discharge simulated by the CABLE experiment for the period of 1950 to 2000. Results are being compared to observed datasets available over some river basins to asses the potential of using CABLE for some hydrological applications in future. Fig 4: River discharge climatology from a preliminary CABLE-TRIP experiment (×106 kg s-1). Conclusion: A 51-yr CABLE global offline experiment has been conducted to study the model simulations of energy, water and carbon cycles. Very preliminary results are presented in this study which demonstrate the potential of this scheme in contributing towards integrated studies of regional and global climate system. This study also calls for a balanced development of the scheme for being used in ACCESS due to complex feedbacks among the energy, water and carbon cycles. Acknowledgment: Mr. Liang Zhang’s visit is supported by an Australia-China climate change bilateral project conducted between BMRC/CAWCR and CMA. References: Arora, V, 2002: Modelling vegetation as a dynamics component in soil-vegetation-atmosphere transfer schemes and hydrological models, Rev. Geophys.,40(2), 1006, doi:10.1029/2001RG000103. Kowalczyk, E.A., Y.P. Wang, R.M. Law, J.L. McGregor, and G. Abramowitz, 2007: CSIRO Atmosphere Biosphere Land Exchange model for use in climate models and as an offline model, CSIRO Marine and Atmospheric Technical report. Ngo-Duc, T., J. Polcher and K. Laval, 2005: A 53-year forcing data set for land surface models. Journal of Geophysical Research, 110, D06116, doi:10.1029/2004JD005434 Oki, T., T. Nishimura, and P. Dirmeyer, 1999: Assessment of annual runoff from land surface models using Total Runoff Integrating Pathway (TRIP), J. Meteo. Soci. Janpan, 77, 235-255.

71

Modelling to assess water resources availability and use – How does it relate to weather and climate modelling?

Albert van Dijk

CSIRO Land and Water, Canberra, ACT

Why is water resources information needed? Recent droughts have painfully exposed the scarcity and often unsustainable level of reliance on surface and groundwater resources in Australia. The new responsibilities of the Bureau of Meteorology reflect a wide appreciation that to manage a scarce resource like water, information on its use and availability in space and time needs to be provided better, faster and cheaper. Some possible examples by which water information can contribute to more efficient and sustainable water management are listed in Table 1, along with suggested key issues with regards to the techniques, challenges and avenues open to deliver this information. The listed applications probably only represent a subset of all conceivable needs, however. Below, the role of modelling in developing water resources information is explored, with emphasis on linkages to weather and climate modelling.

Modelling of past and current water resources availability In retrospective and monitoring applications, the greatest technical challenge requiring modelling is probably to overcome the sparse and partial observation of diffuse or highly distributed water stores and fluxes. Examples include surface and groundwater pumping and diversion (e.g. floodplain harvesting), and diffuse water exchanges between soil, groundwater and surface water systems. Extended metering networks can reduce some of the uncertainty in distributed fluxes, but inference and modelling will continue to be required to interpolate and extrapolate observations in space and time.

Satellite instruments do not directly meter water fluxes and alone do not provide an accurate alternative to on-ground observations, particularly in Australia’s drier environments. Currently, the apparent error of various available satellite-based ET and streamflow estimation methods are in the order of 50–100 mm/y (taking streamflow and flux tower records as a reference; Van Dijk and Mattersdorf, 2007; Raupach et al., 2007; Zhang et al., submitted; Guerschman et al., in prep.). This is typically within 10% of annual average ET, but because net influx into the surface and groundwater systems is the difference between two large terms (rainfall minus ET) this relative error becomes similar or larger than the flux itself. Uncertainty is the accumulation of errors in on-ground rainfall and meteorological observations, their interpolation in space, and the model used to combine these with satellite data.

Despite their indirect nature, the good spatial and temporal coverage of satellite observations does make them an ideal data source to integrate with streamflow and metering records in model-data assimilation (MDA) applications. MDA can also help to reconcile inconsistencies in on-ground observations by explicitly considering gauging and metering errors that are presently largely left unaddressed.

72

Table 1: Examples of potential water information uses, with the suggested most suitable technology, key obstacles and avenues to overcome these. This assumes that all existing data, technology and science can be deployed to their full potential - in reality, pervasive additional obstacles exist associated with modelling technology and data management and delivery.

Purpose Information ideally required

Time scales

Most suitable technology (in theory)

Key obstacle Possible solutions

Accounting of regulated water use against entitlement

Up to date and accurate water use records

Recent Full and ongoing metering and gauging

Infrastructure costs, acceptance/implementation

cheaper metering, up-scaling methods, earth observation

Accounting of unregulated and environmental water use

Up to date and accurate water use records

Recent Water use estimation based on remote sensing and modelling

Accuracy of remote sensing based ET estimates

Integration of multiple data sources (satellite, on-ground, radar)

Compliance with water affecting land use regulations

Ongoing robust data on land use by land holder

Present Ongoing satellite observation

Costs of imagery and accurate classification and interpretation

Merge complementary data, automated analysis methods

Decision making for more efficient day-to-day water use

Forecasts of rainfall and crop water demand

Days/ weeks

Weather, rainfall and flow forecasts

Uncertainty in short-medium term weather and rainfall forecasts

Improve weather and rainfall forecasting skill

Decision making for efficient water management by suppliers, supporting water trade

Forecasts of water orders, water in store and flows at key points

Days/ months

Weather and rainfall forecasts combined with surface water gauging in modelling

Uncertainty in short-medium term rainfall forecasts and outlooks. Unsuited current hydrological modelling techniques

Improve rainfall forecasting skill, introduce model-data assimilation in hydrological modelling

Early response to emerging threats to water availability

Detected trends in water resources, use or climate response

Months/ years

Comprehensive and continuous statistical analysis of observed and inferred data

Too short and incomplete data on surface and groundwater resources

Correlation analysis, systems modelling

Improved water resources planning and sharing

Capacity to test options under latest climate predictions

Months/ years

River and groundwater planning models updated with current state and predicted climate

Uncertainty in climate change predictions and downscaling

Improve climate modelling and downscaling

Comparison of modelling approaches Despite some commonalities between hydrological models and weather and climate models through near-surface processes, there are important discrepancies. Water resources modelling requires coupling of descriptions to vertical and lateral fluxes in the landscape (between available

73

and unavailable soil water and groundwater flows), the river system (between impoundments, irrigation areas and floodplains, including routing, regulation, diversion, and breakout flows) and the groundwater system (between layered or adjoining aquifers and localised extractions). Exchanges between these components also need to be represented (e.g. groundwater recharge and surface-groundwater exchanges). The different model components vary in their representation of space and time. For example, spatial structure can vary from a relatively simple one-dimensional node-link network to describe a river, to a complex three-dimensional raster to describe a multi-layered groundwater system. In an efficient modelling framework, the spatial resolution also varies as a function of the importance of hydrological processes. For example, two-thirds of all surface water resources in the Murray-Darling Basin is generated from only 15% of the area, whereas irrigated areas and water bodies represent less than 2% and 1% of the total basin area, respectively. These components need to be represented in greater detail than the remaining 82%. In summary, the strong heterogeneity of the hydrological system and processes requires a modelling paradigm and spatial structure that is largely incompatible with the homogenous, large scale, and grid based structure of land surface schemes in weather and climate models.

Model linkages in forecasting and prediction Forecasting of water availability can obviously only be as good as the weather and rainfall forecasts that feed into it and this is the strongest (but one-way) link between weather and flow forecasting models. Similarly, the large uncertainty in projections of rainfall patterns due to global change and the transformation of these into data useable in hydrological modelling are by far the greatest source of uncertainty in projections of water resources availability and use.

There are some (potentially important) feedbacks between the climate and water system through changes in the surface water and energy balance. It seems unlikely that explicit inclusion of the hydrology of irrigation areas and wetlands would markedly improve weather or climate forecasts due to the relatively small size of the fluxes involved. For example, the amount of water in the Murray-Darling Basin returned to the atmosphere from irrigation areas, water bodies and wetlands in second instance (i.e. derived from surface water or groundwater resources) is equivalent to approx. 20 mm/y or 4% of total evapotranspiration from the Basin (Kirby et al., 2006).

A more important area of overlap between hydrological and weather forecasting is the influence of dry land soil moisture status on both weather and streamflow processes. Success so far has been modest, but in principle soil moisture observations should help to better forecast catchment response to (forecasted) rainfall. The nature of this relationship depends on the relative importance of different runoff generating processes, and can be a complex combination of groundwater level and unsaturated zone, root zone and surface wetness, with strong spatial heterogeneity. This leads to discrepancies between weather and streamflow forecasting with regard to the type of soil moisture information required and the way it is interpreted in the model. In addition, the intra-storm rainfall intensity distribution is typically an equally or more important control on runoff response and therefore would also need to be forecasted with greater skill before the benefits of catchment and soil wetness data can be realised. This information can be developed through integration of point-gauged rainfall intensity and totals, rainfall radar, and satellite observations (e.g. TRMM).

The greatest link between the climate and hydrological systems is probably the role of longer-term land cover dynamics in both systems. An important uncertainty in predicting climate change impacts on water resources is the change in the transformation of rainfall into streamflow or groundwater due to changes in vegetation function. Ecohydrology can change due to the effects of average and extreme weather and CO2 fertilisation (dieback and recovery from droughts,

74

bushfires, and associated changes in agricultural land use, water harvesting and water conservation). This in turn may have repercussions for the surface water and energy balance and the carbon balance, and therefore there is a potentially strong linkage between climate and hydrology through vegetation dynamics. In low to medium rainfall areas (e.g. <1500 mm/y), evapotranspiration rates usually exceeds water resource generation rates several times, and therefore a relatively small ecohydrological change can have a significant impact on water resources. This is complicated by the episodic nature of streamflow generation and groundwater recharge in this rainfall regime, which requires reliable predictions of changes in the distribution of rainfall as much as or more than changes in average conditions.

Summary

At first glance, weather and climate modelling appears to have many areas of overlap with hydrological modelling to develop water resources information. Indeed, in many cases the quality and usefulness of water resources information critically depends on weather or climate information, and importantly often require greater temporal detail and skill than presently achieved. Research needs include (1) accurate and timely spatial estimates of recent meteorology, rainfall and rainfall intensity through integration of multiple data sources; (2) greater skill in rainfall forecasts, and forecasts of rainfall intensity characteristics; and (3) reduced quantitative uncertainty in changes in rainfall and its temporal distribution due to global climate change. These are one-way data linkages.

The requirements of modelling to assess water resource availability and use are too specific and different to be addressed using weather and climate modelling approaches. Closer scrutiny suggests that most overlapping processes are too different, and feedbacks too insignificant, to warrant a common modelling approach. In particular: (1) small bias and uncertainty in rainfall and evapotranspiration in climate and weather modelling can be greatly amplified in hydrology, requiring models that can use a multitude of on-ground point data in water resources applications; (2) the nature of the role of soil and catchment wetness in atmospheric and hydrological modelling requires different modelling paradigms; and (3) the influence of (changes in) river or groundwater derived ET on atmospheric processes appears negligible on the larger scale.

However, an area of true and important overlap relates to potential changes in dry land ecohydrological processes due to climate variability or change, which could impact on local water resources generation, the surface water and energy balance and global carbon cycle alike.

References Kirby, M. et al. (2006). The Shared Water Resources of the Murray-Darling Basin. Report to the Murray-

Darling Basin Commission, Part 1. CSIRO Water for a Healthy Country.

Raupach, M.R., Briggs, P.R., King, E.A., Paget, M. and Trudinger, C.M. (2007). Australian Water Availability Project (AWAP). Final Report for Phase 2, CSIRO Marine and Atmospheric Research.

Van Dijk, A.I.J.M. and Mattersdorf, G. (2007). Comparison of MODIS-based scaling of potential evapotranspiration with on-ground observations. EGU 2007, Geophysical Research Abstracts, 9: 11692.

75

Carbon, climate and humans: Australia in the Earth System Michael R. Raupach

CSIRO Marine and Atmospheric Research, Canberra, Australia

Abstract: Climate change and its hydrological impacts will have far-reaching consequences for Australia (along with other Mediterranean-climate regions in both hemispheres) because of mid-latitude drying trends. To address the principal driver, uncontrolled anthropogenic CO2 emissions, a rule is proposed for allocation of a biophysically limited cumulative global emission of CO2.

The new global ecology Humans have transformed the environment in which evolution occurs, so that the earth now functions as a coupled biophysical-human system (Steffen et al. 2004; Field and Raupach 2004). The industrial revolution brought about a dramatic acceleration of this planetary transformation (Figure 1). Growth in per-capita Gross Domestic Product (GDP), a surrogate for per-capita human energy use, accelerated in the few decades 1800-1850 from negligible to its present growth rate, a doubling every 45 y. Economic growth, with its associated energy and ecological demands, is a very recent phenomenon.

Global population and GDP

100

1000

10000

100000

0 500 1000 1500 2000

Pop

ulat

ion

(mill

ion)

, GD

Ppp

p ($

billi

on)

PopulationGDPppp

Global per capita GDP

100

1000

10000

0 500 1000 1500 2000

Per

cap

ita G

DP

ppp

($/p

erso

n)

Figure 1: Changes over the past two millennia in global population, GDP (in Geary-Khamis international 1990 dollars, a unit with constant purchasing power) and per capita GDP. Source: Historical Statistics for the World Economy: 1-2003 AD, by Angus Maddison, Groningen Growth and Development Centre (http://www.ggdc.net/maddison/).

Climate change induced by rising concentrations of greenhouse gases (GHGs) is a direct consequence of this growth pattern and its dependence on fossil energy. The dominant contribution comes from CO2 emissions from fossil fuels, which make up about 2/3 of all GHG emissions in terms of 100-year radiative forcing (IPCC 2007). The result is a rapid and currently uncontrolled buildup of CO2 and other GHGs in the atmosphere, in turn provoking other positive feedbacks such as the ice-albedo feedback, climate feedbacks from cold and hot wetland releases of CO2 and methane, terrestrial water-carbon feedbacks from increased drought, ocean circulation changes, etc.

CO2 emissions Figure 2 (from Raupach et al. 2007) shows that global fossil fuel emissions are growing faster than even fossil-fuel-intensive emissions scenarios, let alone stabilisation trajectories. Drivers include the growth of emissions with income (Figure 3) and the properties of carbon intensities (Figure 4).

Australia, with 0.32% of world population, contributed 1.43% of fossil-fuel CO2 emissions in 2004 (104 MtC y- 1 or 381 MtCO2 y- 1). Australia's carbon intensity of energy (fossil fuel burned per unit of energy produced) is 20% higher than the world average, and 25 to 30% higher than values for the USA, Europe and Japan (Figure 4). Australia's carbon intensity of GDP (fossil fuel burned per dollar of GDP) is

76

25% higher than the world average and nearly double that of Europe and Japan (Figure 4). For 1980-2004, the average growth rate of Australian emissions was about twice that for the world, twice that for the USA and Japan, and five times that for Europe. The rate of improvement (decline) in the carbon intensity of GDP for Australia is lower than in the USA and Europe.

Recent emissions

1850 1900 1950 2000 2050 2100

CO

2 E

mis

sion

s (G

tC y

-1)

0

5

10

15

20

25

30Actual emissions: CDIAC450ppm stabilisation650ppm stabilisationA1FI A1B A1T A2 B1 B2

5

5.5

6

6.5

7

7.5

8

8.5

9

9.5

10

1990 1995 2000 2005 2010F

oss

il F

uel

Em

issi

on

(G

tC/y

)

actual emissions: CDIACactual emissions: EIA450 ppm stabilisation650 ppm stabilisationA1FIA1BA1TA2B1B2

Figure 2: Observed global CO2 emissions from fossil fuels, compared with IPCC SRES emissions scenarios and stabilisation trajectories. Data are from EIA and CDIAC, and are normalised to same mean for 1990-1999 because of small accounting differences between the two data sets. Left panel shows period 1850-2100; Right panel shows 1990-2010, including an update and revision of the CDIAC data to 2006. For details see (Raupach et al. 2007).

Per capita emissions, F/P [tC/y/person]

0

1

2

3

4

5

6

0 5 10 15 20 25 30

Income (g=Gp/P)

F/P

USAEUJapanD1FSUChinaIndiaD2D3WorldAustraliaCanadaTaiw anKyotoA1

Carbon intensity of energy, F/E [gC/MJ]

0

5

10

15

20

25

US

A

EU

Japa

n D1

FS

U

Chi

na

Ind

ia D2

D3

Wo

rld

Au

stra

lia

Can

ada

Taiw

an

Kyo

toA

1

Carbon intensity of GDPppp, F/Gp [gC/$]

0

50

100

150

200

250

300

350

400

US

A

EU

Japa

n D1

FS

U

Ch

ina

Ind

ia D2

D3

Wo

rld

Au

stra

lia

Can

ada

Tai

wan

Kyo

toA

1

(Left) Figure 3: Per capita CO2 emissions (F/P) plotted against per-capita income defined with GDP-PPP (GP/P), for 9 regions spanning the world: USA, EU, Japan, other developed nations (D1), FSU, China, India, other developing nations (D2), least developed nations (D3). Also shown are the world average, three D1 nations (Australia, Canada, Taiwan) and the average for Kyoto Annex 1 nations. Points linked by lines represent values in 1980, 1992 and 2004 (left to right). Data from (Raupach et al. 2007).

(Below) Figure 4: Carbon intensity of energy and carbon intensity of GDP-PPP, both in 2004, for regions in Figure 3.

77

Global and local targets Let us consider connections between global and local actions, using Australia as an example. Many impacts of climate change (on water resources, agriculture, terrestrial ecosystems, coral reefs, human health, coastlines, …) will be very significant for Australia (Steffen et al. 2006), perhaps proportionally more than for other developed nations. Adaptation is possible and important, but there is a point beyond which a global GHG mitigation strategy is essential for Australia's long-term national interest. This is clearly apparent in the water challenge faced by southern Australia (along with other Mediterranean-climate regions in both hemispheres) arising from mid-latitude drying trends superimposed on a globally wetter and warmer future climate (IPCC 2007).

Important elements of any mitigation strategy are (1) a global GHG emissions reduction target, and (2) a concept of how the target should be shared. These elements are needed whether the main mechanism for achieving future emissions reductions is an international Kyoto-style treaty with negotiated targets, or cooperative behaviours evolving naturally from the unilateral responses of nations repeatedly playing a common climate game (Liebreich 2007).

The global emissions target: A broad, long-range target for avoiding "dangerous" climate change is necessary because of biophysical, social, technological and policy inertias in carbon-climate-human feedbacks (Field and Raupach 2004). Recent analyses (eg Schneider and Lane 2006) have emphasised that this target is a value judgement.

There has been recent widespread affirmation (eg by the UK Stern Review in 2006) of the 1995 European Union target of a 2 degC temperature rise above preindustrial levels. About 0.8 degC has already occurred and about an equal increment is committed in the future by climate-system inertia (IPCC 2007). To keep the probability of exceeding this 2 degC target to less than 50%, GHG concentrations must stabilise at 440 (±70) ppm CO2eq (Meinshausen 2006, Fig 28.5 and Table 28.1) or at less than about 400 ppm CO2. At current CO2 growth rates, 400 ppm CO2 will be reached within a year or two of 2017, and 440 ppm CO2eq has arguably been passed already. Hence, this target seems unrealistic.

Let us express a long-term (~100 y) target simply as a maximum allowable cap on cumulative total CO2 emissions (from fossil fuels and land use) from 2000 to 2100. An illustrative cap is 500 GtC, which is consistent with stabilisation at approximately 500 to 550 ppm CO2 (Meinshausen 2006, Fig 28.3) and a stabilisation temperature around 2.5 to 3 degC above preindustrial under a median climate sensitivity to CO2 doubling. A lower cap would provide more allowance for the possibility of positive climate and carbon-climate feedbacks, which would imply a higher climate sensitivity.

Even a 500 GtC cap is a severe challenge. In the six years 2001-2006, 55 GtC has already been emitted. Total emissions were 9.85 GtC y- 1 in 2006, with a growth rate above 3% y- 1 (Raupach et al. 2007; Canadell et al. 2007). If this growth rate continues, a 500 GtC cap will be reached around 2035.

Sharing future emissions: By focusing on cumulative emissions, the problem of defining targets is changed from determination of flux trajectories to allocation of a stock, the non-renewable resource represented by the remaining capacity of the planet to bear CO2 emissions from fossil fuels. How should this resource be divided among nations? This is a highly negotiable question to which it is not possible to give a simple answer, but it is possible to place the answer within simple bounds.

Cumulative emissions through the 21st century could be allocated according to (a) current shares of global annual emissions (1.43% for Australia), or (b) population (0.32% for Australia). Option (a) locks in present global inequalities in the use of fossil fuels for development and wealth creation (Figure 3), while (b) immediately removes all present inequalities. Neither is politically acceptable: developing nations require differentiation of targets and would not accept (a), and (b) is unacceptable to developed nations. Therefore let us consider a middle option, (c), in which the share of nation receives a share midway between those from options (a) and (b), determined using data for a reference year (present figures use 2004). This would provide a cumulative emission of 0.87% of 500 GtC = 4400 MtC, for Australia, which

78

will last about 42 years from 2000 at a constant (2004) emission rate. Stated formally, the share of globally capped 21st century cumulative emissions for nation i is wFi/F + (1−w)Pi/P, where Fi and Pi are emissions and population for nation i at a reference time, F and P are global totals at that time, and w is a globally uniform weight (0 ≤ w ≤ 1). Options (a), (b) and (c) correspond to w = 1, 0 and 1/2, respectively, and option (c) can be varied by choosing w other than 1/2.

Climate change as a tragedy of the commons: Climate change is now widely seen as a "Tragedy-of-the-Commons" problem (Hardin 1968). Solutions for these problems involve the natural or engineered emergence of "adaptive governance in complex systems". Known prerequisites (Dietz et al. 2003) are the ability for users of the commons to (1) monitor their common resource, (2) devise rules for sharing and nurturing it, (3) induce compliance with the rules, (4) resolve rather than escalate conflicts, and (5) adapt to change. All of this relies on a store of "social capital" (Pretty 2003): the understandings, institutions and basic trust which enable users of the commons to work together. From this perspective, international negotiations are a critical step in building the global social capital needed for the sharing of emissions. This is true whether emissions reductions are sought by treaty (Kyoto-style) or by the emergence of cooperative behaviours from repeated plays of a climate game (Liebreich 2007).

Summary: This paper explores (1) placing a firm cap on global cumulative 21st century emissions, and (2) apportioning this stock according to a weighted mix of population share and emissions share in a reference year. One general advantage of a stock approach is that nations have flexibility in choosing an emissions trajectory consistent with the cumulative constraint, limited only compliance requirements.

References Canadell JG, Le Quere C, Raupach MR, Field CB, Buitenhuis ET, Ciais P, Conway TJ, Gillett NP, Houghton RA, Marland G

(2007) Contributions to accelerating atmospheric CO2 growth from economic activity, carbon intensity, and efficiency of natural sinks. PNAS:0702737104

Dietz T, Ostrom E, Stern PC (2003) The struggle to govern the commons. Science 302:1907-1912

Field CB, Raupach MR (2004) The Global Carbon Cycle: Integrating Humans, Climate, and the Natural World. Island Press, Washington, 526 pp

Hardin G (1968) Tragedy of Commons. Science 162:1243-&

IPCC (2007) Climate change 2007: the physical science basis. Summary for policymakers. IPCC Secretariat, Geneva

Liebreich M (2007) How to save the planet: be Nice, Retaliatory, Forgiving and Clear. White Paper, New Energy Finance Ltd., London, 1 pp

Meinshausen M (2006) What does a 2C target mean for greenhouse gas concentrations? A brief analysis based on multi-gas emission pathways and several climate sensitivity uncertainty estimates. In: Schellnhuber HJ, Cramer W, Nakicenovic N, Wigley TML, Yohe G (eds) Avoiding Dangerous Climate Change. Cambridge University Press, Cambridge, pp 265-279

Pretty J (2003) Social capital and the collective management of resources. Science 302:1912-1914

Raupach MR, Marland G, Ciais P, Le Quere C, Canadell JG, Klepper G, Field CB (2007) Global and regional drivers of accelerating CO2 emissions. PNAS:0700609104

Schneider SH, Lane J (2006) An overview of 'dangerous' climate change. In: Schellnhuber HJ, Cramer W, Nakicenovic N, Wigley TML, Yohe G (eds) Avoiding Dangerous Climate Change. Cambridge University Press, Cambridge, pp 7-23

Steffen WL, Love G, Whetton P (2006) Approaches to defining dangerous climate change: an Australian perspective. In: Schellnhuber HJ, Cramer W, Nakicenovic N, Wigley TML, Yohe G (eds) Avoiding Dangerous Climate Change. Cambridge University Press, Cambridge, pp 219-225

Steffen WL, Sanderson A, Tyson PD, Jager J, Matson PA, Moore BI, Oldfield F, Richardson K, Schellnhuber HJ, Turner BLI, Wasson RJ (2004) Global Change and the Earth System: a Planet Under Pressure. Springer, Berlin, 336 pp

BMRC Research Reports Holland, G.J., Berzins, I.A. and Merrill, R.T. 1985. The cyclone game: simulation of a tropical cyclone warning centre. BMRC

Research Report No. 1, Bur. Met. Australia.

McBride, J.L. and Holland, G.J. 1986. The BMRC Australian monsoon experiment: objectives and scientific basis. BMRC Research

Report No. 2, Bur. Met. Australia.

Spillane, K.T. and Lourensz, R.S. 1986. The hazard of horizontal windshear to aircraft operations at Sydney Airport. BMRC Research

Report No. 3, Bur. Met. Australia.

Spillane, K.T. 1987. Convection in an inversion limited boundary layer. BMRC Research Report No. 4, Bur. Met. Australia.

Glowacki, T.J. and Seaman, R. 1987. A general purpose package for univariate two-dimensional data checking and analysis. BMRC

Research Report No. 5, Bur. Met. Australia.

Williams, M. 1987. Relations between the Southern Oscillation and the troposphere over Australia. BMRC Research Report No. 6,

Bur. Met. Australia.

Karoly, D.J., Kelly, G.A.M. and Le Marshall, J.F. 1987. The Australian Southern Hemisphere climatology data tape. BMRC Research

Report No. 7, Bur. Met. Australia.

Keenan, T.D. and Martin, S.C. 1987. AMEX Radar Atlas. Volume 1 AMEX Phase 1. BMRC Research Report No. 8, Bur. Met.

Australia.

Keenan, T.D. and Martin, S.C. 1987. AMEX Radar Atlas. Volume II AMEX Phase II. BMRC Research Report No. 9, Bur. Met.

Australia.

Lajoie, F.A. 1988. Circulation asymmetries around a tropical cyclone and its future direction of motion. BMRC Research Report No.

10, Bur. Met., Australia.

Pike, D.J., Leslie, L.M., Mills, G.A., Glowacki, T., Hubbert, G. and McIntosh, P. 1988. Report on performance of regional forecast

models January-June 1987. BMRC Research Report No. 11, Bur. Met. Australia.

Hart, T.L., Bourke, W.P., McAvaney, B.J., Forgan, B.W. and McGregor, J.L. 1988. Atmospheric general circulation simulations with

the BMRC global spectral model: the impact of revised physical parametrisations. BMRC Research Report No. 12, Bur.

Met. Australia.

Drosdowsky, W. 1988. Lag relations between the Southern Oscillation and the troposphere over Australia. BMRC Research Report

No. 13, Bur. Met. Australia.

Pike, D.J., Leslie, L.M., Mills, G.A., Glowacki, T., Hubbert, G. and McIntosh, P. 1989. Report on performance of regional forecast

models July-December 1987. BMRC Research Report No.14, Bur. Met. Australia.

Bourke, W. 1989. Working Group on Numerical Experimentation survey of model boundary condition data sets. BMRC Research

Report No. 15, Bur. Met. Australia.

Hess, G.D. and Guymer, A.E. 1989. Verification of a Markov technique for the short-term prediction of rainfall at eight Australian

stations. BMRC Research Report No. 16, Bur. Met. Australia.

Pike, D.J., Leslie, L.M., Mills, G.A., Glowacki, T., Hubbert, G. and McIntosh, P. 1989. Report on performance of regional forecast

models January-June 1988. BMRC Research Report No.17, Bur. Met. Australia.

Blomley, J.E., Smith, N.R. and Meyers, G. 1989. An oceanic subsurface thermal analysis scheme. BMRC Research Report No. 18,

Bur. Met. Australia.

Naughton, M.J. and Balgovind, R.C. 1990. BMRC Implementation of the NCAR CCM Modular Processor. BMRC Research Report

No. 19, Bur. Met. Australia

Hess, G.D. and Spillane, K.T. 1990. A survey of dust devils in Australia. BMRC Research Report No. 20, Bur. Met. Australia.

Jasper, J.D. (editor). 1990. 'The role of sea surface temperatures in numerical modelling of the atmosphere': papers presented at the

first BMRC modelling workshop, July 1989. BMRC Research Report No. 21, Bur. Met. Australia.

Pike, D.J., Leslie, L.M., Mills, G.A., Glowacki, T., Hubbert, G. and McIntosh, P. 1990. Report on performance of regional forecast

models July-December 1988. BMRC Research Report No.22, Bur. Met. Australia.

Pike, D.J., Leslie, L.M., Mills, G.A., Glowacki, T., Hubbert, G. and McIntosh, P. 1991. Report on performance of regional forecast

models January-December 1989. BMRC Research Report No. 23, Bur. Met. Australia.

Colman, R.A. and McAvaney, B.J. 1991. Experiments using the BMRC general circulation model with a heat balance ocean. BMRC

Research Report No. 24, Bur. Met. Australia.

Rikus, L. 1991. The role of clouds in global climate modelling. BMRC Research Report No. 25, Bur. Met. Australia.

Davidson, N.E., Puri, K., Bowen, R., Fraser, J., Wadsley, J, and Wong, H. 1991. BMRC real time analyses of the tropospheric

circulation during TCM90. BMRC Research Report No. 26, Bur. Met. Australia.

Jasper, J.D. (editor) 1991. 'Data assimilation systems': papers presented at the second BMRC modelling workshop, September 1990.

BMRC Research Report No. 27, Bur. Met. Australia.

Lavery, B.M., Davidson, N.E., Karoly, D.J. and McAvaney, B.J. 1991. A climatology of the western Pacific region based on the

Australian tropical analysis system. BMRC Research Report No. 28, Bur. Met. Australia.

McAvaney, B.J., Fraser, J.R., Hart, T.L., Rikus, L.J., Bourke, W.P., Naughton, M.J. and Mullenmeister, P. 1991. Circulation statistics

from a non-diurnal seasonal simulation with the BMRC atmospheric GCM: R21L9. BMRC Research Report No. 29, Bur.

Met. Australia.

Colman, R.A., McAvaney, B.J., Fraser, J.R. and Dahni, R.R. 1992. Mixed layer ocean and thermodynamic sea ice models in the

BMRC GCM. BMRC Research Report No. 30, Bur. Met. Australia.

May, P.T. 1992. Wind profiler measurements in the tropics. BMRC Research Report No. 31, Bur. Met. Australia.

Pitman, A.J., Yang, Z.-L., Cogley, J.G. and Henderson-Sellers, A. 1992. Description of bare essentials of surface transfer for the

Bureau of Meteorology Research Centre AGCM. BMRC Research Report No. 32, Bur. Met. Australia.

Jasper, J.D. and Meighen, P.J. (eds). 1992. 'Modelling weather and climate': papers presented at the third BMRC modelling workshop,

November 1991. BMRC Research Report No. 33, Bur. Met. Australia.

Power, S.B., Moore, A.M., Post, D.A., Smith, N.R. and Kleeman, R. 1992. On the stability of North Atlantic deep water formation in

a global ocean general circulation model. BMRC Research Report No. 34, Bur. Met. Australia.

Power, S. and Kleeman, R. 1992. Multiple equilibria in a global ocean general circulation model. BMRC Research Report No. 35, Bur.

Met. Australia.

Nicholls, N. 1993. Climate change and the El Niño B Southern Oscillation: report of the workshop on climate change and the El Niño

B Southern Oscillation held at the Bureau of Meteorology Research Centre Melbourne, Australia 31 May and 4 June 1993.

BMRC Research Report No. 36, Bur. Met. Australia.

Power, S.B., Colman, R.A., McAvaney, B.J., Dahni, R.R., Moore, A.M. and Smith, N.R. 1993. The BMRC coupled

atmosphere/ocean/sea-ice model. BMRC Research Report No. 37, Bur. Met. Australia

McAvaney, B.J. and Colman, R.A. 1993. The AMIP experiment: the BMRC AGCM configuration. BMRC Research Report No. 38,

Bur. Met. Australia.

Jasper, J.D. and Meighen, P. J. (eds). 1993. 'Modelling severe weather': papers presented at the fourth BMRC modelling workshop,

26-29 October 1992. BMRC Research Report No. 39, Bur. Met. Australia.

Mullenmeister, P. and Hart, T. 1994. The UNIX verification programs. BMRC Research Report No. 40, Bur. Met. Australia.

Jones, D.A. 1994. A numerical vortex finding, tracking and statistics package. BMRC Research Report No. 41, Bur. Met. Australia.

Power, S. 1994. A report on two BMRC workshops: modelling and observing sea-ice in a coupled environment, November 19th, 1993

and physical oceanography in Melbourne, December 10th, 1993. BMRC Research Report No. 42, Bur. Met. Australia.

Bender, L.C. and Leslie, L.M. 1994. Evaluation of a Third Generation Ocean Wave Model for the Australian Region. BMRC Research

Report No. 43, Bur. Met. Australia.

Keenan, T., Holland, G., Rutledge, S., Simpson, J., McBride, J., Wilson, J., Moncrieff, M., Carbone, R., Frank, W., Sanderson, B.,

Tapper, N. and Hallett, J. 1994. Science Plan B Maritime Continent Thunderstorm Experiment (MCTEX). BMRC Research

Report No. 44, Bur. Met. Australia.

Pudov, V.D. and Holland, G.J. 1994. Typhoons and Ocean: results of experimental investigations. BMRC Research Report No. 45,

Bur. Met. Australia.

Jasper, J.D. and Meighen, P.J. (eds). 1994. 'Parametrisation of physical processes': papers presented at the fifth BMRC modelling

workshop, November 1993. BMRC Research Report No. 46, Bur. Met. Australia.

Dahni, R.R. and Stern, H. 1995. The development of a generalised UNIX version of the Victorian Regional Office's operational

analogue statistics model. BMRC Research Report No. 47, Bur. Met. Australia.

Power, S., Kleeman, R., Colman, R. and McAvaney, B. 1995. Modelling the surface heat flux response to long-lived SST anomalies

in the North Atlantic. BMRC Research Report No. 48, Bur. Met. Australia.

Seaman, R.S. 1995. Bias detection in the Australian sea level pressure observing network. BMRC Research Report No. 49, Bur. Met.

Australia.

May, P.T., Holland, G.J., Keenan, T.D., Le Marshall, J.F., Colquhoun, J., Buckley, B., Love, G., Elliot, J., Dear, J. and Potts, R. 1995.

Towards a comprehensive, very-short-range forecasting system for the Greater Sydney Region: observing system. BMRC

Research Report No. 50, Bur. Met. Australia.

Sanderson, B., Tang, Y.M., Holland, G., Grimshaw, R. and Woodcock, F. 1995. A tropical cyclone maximum envelope of waters

(MEOW) technique. BMRC Research Report No. 51, Bur. Met. Australia.

Meighen, P.J. and Jasper. J.D. (eds). 1996. 'Data assimilation in meteorology and oceanography': papers presented at the sixth annual

BMRC modelling workshop, October 1994. BMRC Research Report No. 52, Bur. Met. Australia.

Keenan, T.D. and Manton, M.J. 1996. Darwin climate monitoring and research station: observing precipitating systems in a monsoon

environment. BMRC Research Report No. 53, Bur. Met. Australia.

Meighen, P.J. and Jasper. J.D. (eds). 1996. 'Numerical methods': papers presented at the seventh annual BMRC modelling workshop,

4-6 December 1995. BMRC Research Report No. 54, Bur. Met. Australia.

Ebert. E.E. 1996. Results of the 3rd algorithm intercomparison project (AIP-3) of the global precipitation climatology project (GPCP).

BMRC Research Report No. 55, Bur. Met. Australia.

McAvaney, B.J. and Hess, G.D. 1996. The revised surface fluxes parametrisation in BMRC: formulation. BMRC Research Report No.

56, Bur. Met. Australia.

Keenan, T., Puri, K., Mills, G. and Bowen, R. 1996. Regional BMRC analyses of atmospheric circulation during MCTEX. BMRC

Research Report No. 57, Bur. Met. Australia.

Hess, G.D., McBride, J.L., Drosdowsky, W. and Whitby, F. 1996. A meteorological investigation into the Qantas flight 69 incident of

severe turbulence. BMRC Research Report No. 58, Bur. Met. Australia.

Keenan, T.D. and Glasson, K. 1996. 'MCTEX C-band polarimetric atlas and data summary'. BMRC Research Report No. 59, Bur.

Met. Australia.

Keenan, T.D. and Le Marshall, J. 1996. Satellite data summary and GMS imagery during MCTEX. BMRC Research Report No. 60,

Bur. Met. Australia.

Meighen, P.J. and Jasper. J.D. (eds). 1997. 'Symposium on climate prediction and predictability': papers presented at the eighth

modelling workshop, 12-14 November 1996. BMRC Research Report No. 61, Bur. Met. Australia.

Potts, R., Monypenny, P. and Middleton, J. 1997. An analysis of winds at Sydney Kingsford Smith airport and their impact on runway

availability. BMRC Research Report No. 62, Bur. Met. Australia.

Sanderson, B. 1997. A barotropic ocean model for calculating storm surges. BMRC Research Report No. 63, Bur. Met. Australia.

Meighen, P.J. and Jasper. J.D., (eds). 1997. 'Improving short-range forecasting': abstracts of presentations at the ninth annual BMRC

modelling workshop, 8-10 October 1997. BMRC Research Report No. 64, Bur. Met. Australia.

Drosdowsky, W. and Chambers, L. 1998. Near global sea surface temperature anomalies as predictors of Australian seasonal rainfall.

BMRC Research Report No. 65, Bur. Met. Australia.

Power, S.B., Tseitkin, F., Colman, R.A. and Sulaiman, A. 1998. A coupled general circulation model for seasonal prediction and

climate change research. BMRC Research Report No. 66, Bur. Met. Australia.

Power, S., Tseitkin, F., Mehta, V., Lavery, B., Torok, S. and Holbrook, N. 1998. Decadal climate variability in Australia during the

20th century. BMRC Research Report No. 67, Bur. Met. Australia.

Keenan, T., Kondratiev, V., Buis, B. and Christmas, R. 1998. The Bureau of Meteorology Research Centre (BMRC) portable

automatic weather station: description and operation. BMRC Research Report No. 68, Bur. Met. Australia.

Meighen, P.J. (ed.). 1998. 'Coupled climate modelling': abstracts of presentations at the tenth BMRC modelling workshop, 12-13

October 1998. BMRC Research Report No. 69, Bur. Met. Australia.

Jones, D.A. 1998. The prediction of Australian land surface temperatures using near global sea surface temperature patterns. BMRC

Research Report No. 70, Bur. Met. Australia.

Tang, Y.M., Smith, N. and Greenslade, D. 1998. 'Comparison of model and observed surface winds', (report prepared for AMSA

project). BMRC Research Report No. 71, Bur. Met. Australia.

Potts, R.J., Keenan, T. and May, P. 1999. Radar characteristics of storms in the Sydney area. BMRC Research Report No. 72, Bur.

Met. Australia.

Greenslade, D. 1999. The assimilation of ERS-2 altimeter data into the Australian wave model. BMRC Research Report No. 73, Bur.

Met. Australia.

Wang, Y. 1999. A triply-nested movable mesh tropical cyclone model with explicit cloud microphysics B (TCM3). BMRC Research

Report No. 74, Bur. Met. Australia.

Jasper, J.D. and Meighen, P.J. (eds) 1999. 'Parallel computing in meteorology and oceanography': abstracts of presentations at the

eleventh annual BMRC modelling workshop, 9-11 November 1999. BMRC Research Report No. 75, Bur. Met. Australia.

Mills, G.A. 2000. A synoptic/diagnostic study of the 1998 Sydney-Hobart yacht race storm - a warm cored extratropical cyclone.

BMRC Research Report No. 76, Bur. Met. Australia.

Wang, G., Kleeman, R., Smith, N. and Tseitkin, F. 2000. Seasonal predictions with a coupled global ocean-atmosphere model. BMRC

Research Report No. 77, Bur. Met. Australia.

Timbal, B. and McAvaney, B.J. 2000. A downscaling procedure for Australia. BMRC Research Report No. 78, Bur. Met. Australia.

Greenslade, D.J.M. 2000. Upgrades to the Bureau of Meteorology's ocean wave forecasting system. BMRC Research Report No. 79,

Bur. Met. Australia.

Jasper, J.D. and Meighen, P.J. (eds) 2000. 'Model systematic errors': Extended abstracts of presentations at the twelfth annual BMRC

modelling workshop (co-sponsored by WCRP/WGNE), 16-20 October 2000. BMRC Research Report No. 80, Bur. Met.

Australia.

Viviand, J., Lirola, S., Timbal, B., Power, S. and Colman, R. 2000. Impact of soil moisture on climate variability and predictability.

BMRC Research Report No. 81, Bur. Met. Australia

Mills, G.A. 2001. Impact of screen-level moisture observations in a regional data assimilation system. BMRC Research Report No. 82,

Bur. Met. Australia.

Zhong A., Colman, R., Smith, N., Naughton, M., Rikus, L., Puri, K. and F. Tseitkin. 2001. Ten-year AMIP 1 climatologies from

versions of the BMRC Atmospheric model. BMRC Research Report No. 83, Bur. Met. Australia.

Jasper, J.D. and Meighen, P.J. (eds) 2001. 'Understanding the climate of Australia and the Indo-Pacific region': Extended abstracts of

presentations at the thirteenth annual BMRC modelling workshop, 14-16 November 2001. BMRC Research Report No. 84,

Bur. Met. Australia.

Keenan, T., Joe, P., Wilson, J., Collier, C., Golding, B., Burgess, D., May, P., Pierce, C., Bally, J., Crook, A., Seed, A., Sills, D.,

Berry, L., Potts, R., Bell, I., Fox, N., Ebert, E., Eilts, M., O'Loughlin, K., Webb, R., Carbone, R., Browning, K., Roberts, R.

and Mueller, C. 2002. World Weather Research Programme forecast demonstration project: overview and current status.

BMRC Research Report No. 85, Bur. Met. Australia.

Chambers, L., Li, F. and Nicholls, N. 2002. Seasonal climate forecasts for south-west Western Australia. BMRC Research Report No.

86, Bur. Met. Australia.

Voldoire, A., Timbal, B. and Power, S. 2002. Statistical-dynamical seasonal forecasting. BMRC Research Report No. 87, Bur. Met.

Australia.

Lemus-Deschamps, L., Colman, R., Fraser, J. and Zhong, A. 2002. The model and climatological data comparison system (MACCS).

BMRC Research Report No. 88, Bur. Met. Australia.

Zhang, H., Henderson-Sellers, A., Irannejad, P., Sharmeen, S., Phillips, T. and McGuffie, K. 2002. Land-surface modelling and

climate simulations: results over the Australian region from sixteen AMIP II models. BMRC Research Report No. 89, Bur.

Met. Australia.

Hollis, A.J. and Meighen, P.J. (eds) 2002. 'Modelling and predicting extreme events'; Extended abstracts of presentations at the

fourteenth annual BMRC modelling workshop, 11-13 November 2002. BMRC Research Report No. 90, Bur. Met. Australia.

Wu, Z-J., Colman, R., Power, S., Wang, X. and B. McAvaney. 2002. The El Niño Southern Oscillation response in the BMRC

Coupled GCM. BMRC Research Report No. 91, Bur. Met. Australia.

Chambers, L.E. 2003. South Australian rainfall variability and trends. BMRC Research Report No. 92, Bur. Met. Australia.

Meighen, P.J. and Hollis, A.J. (eds) 2003. 'Current issues in the parameterization of convection’: Extended abstracts of presentations

at the fifteenth annual BMRC modelling workshop, 13-16 October 2003. BMRC Research Report No. 93, Bur. Met.

Australia.

Sun, X., Manton, M.J. and Ebert, E.E. 2003. Regional rainfall estimation using double-kriging of raingauge and satellite observations.

BMRC Research Report No. 94, Bur. Met. Australia.

Zhong, A., Alves, O., Schiller, A., Tseitkin, F. and Smith, N. 2004. Results from a preliminary version of the ACOM2/BAM coupled

seasonal forecast model. BMRC Research Report No. 95, Bur. Met. Australia

Zhang, H. 2004. A version of the BAM’s AMIP2 simulation over the Australian region: contrasting its performance with other AMIP2

models. BMRC Research Report No. 96, Bur. Met. Australia.

Greenslade, D.J.M. 2004. A validation of ERS-2 Fast Delivery Significant Wave Height. BMRC Research Report No. 97, Bur. Met.

Australia.

Colman, R. 2004. Coupled modelling in BMRC: present status and future directions. BMRC Research Report No. 98, Bur. Met.

Australia.

Dare, R.A. 2004. The BMRC Bulk Explicit Microphysics (BEM) scheme. BMRC Research Report No. 99, Bur. Met. Australia.

Chambers, L.E. 2004. The impact of climate on Little Penguin breeding success. BMRC Research Report No. 100, Bur. Met.

Australia.

Dare, R.A., Bowen, R.M. and Ebert, E.E. 2004. Assessment of 24 and 48 hour precipitation forecasts using a bulk explicit

microphysics scheme in the LAPS model. BMRC Research Report No. 101, Bur. Met. Australia.

Gunasekera, D. 2004. Economic issues relating to meteorological services provision. BMRC Research Report No. 102, Bur. Met.

Australia.

Mills, G.A. 2004. Verification of operational NWP model cool-season tornado threat area forecasts in Australia. BMRC Research

Report No. 103, Bur. Met. Australia.

Hollis, A.J. (ed.) 2004. ‘The past, present and future of numerical modelling’: Extended abstracts of presentations at the sixteenth

annual BMRC modelling workshop, 6-9 December 2004. BMRC Research Report No. 104, Bur. Met. Australia.

Kepert, J.D., Greenslade, D.J.M. and Hess, G.D. 2005. Assessing and improving the marine surface winds in the Bureau of

Meteorology numerical weather prediction systems. BMRC Research Report No. 105, Bur. Met. Australia.

Moise, A.F., Colman, R.A. and Zhang, H. 2005. Coupled model simulations of Australian surface precipitation and temperature and

their response to global warming in 18 CMIP2 models. BMRC Research Report No. 106, Bur. Met. Australia.

Wang, G., Alves, O. and Smith, N. 2005. BAM3.0 tropical surface flux simulation and its impact on SST drift in a coupled model.

BMRC Research Report No. 107, Bur. Met. Australia.

Colman, R., Deschamps, L., Naughton, M., Rikus, L., Sulaiman, A., Puri, K., Roff, G., Sun, Z. and Embery, G. 2005. BMRC

Atmospheric Model (BAM) version 3.0: comparison with mean climatology. BMRC Research Report No. 108, Bur. Met.

Australia.

Colman, R. 2005. The BMRC climate feedback climate analysis system. BMRC Research Report No. 109, Bur. Met. Australia.

Lemus-Deschamps, L., Sisson, J., Li, Z., Hudson, D. and Colman, R. 2005. The Model and Climatological data Comparison System:

Version 2 (MACCS 2). BMRC Research Report No. 110, Bur. Met. Australia.

Hollis, A.J. (ed.) 2005. ‘Hydrometeorological applications of weather and climate modelling’: Extended abstracts of presentations at

the seventeenth annual BMRC modelling workshop, 3-6 October 2005. BMRC Research Report No. 111, Bur. Met.

Australia.

Gunasekera, D. and co-authors. 2005. Economic value of fire weather services. BMRC Research Report No. 112, Bur. Met. Australia.

Power, S.B., Haylock, M., Colman, R. and Wang, X. 2005. Asymmetry in the Australian response to ENSO and the predictability of

inter-decadal changes in ENSO teleconnections. BMRC Research Report No. 113, Bur. Met. Australia.

Entel, M.B., Smith, N.R., Davidson, N.E., Warren, G.R. and Hanstrum, B.N. 2005. Enhancements to the Bureau of Meteorology’s

ocean surge forecasting system using operational TC-LAPS. BMRC Research Report No. 114, Bur. Met. Australia.

Engel, C. 2005. Hourly Operational Consensus Forecasts (OCF). BMRC Research Report No. 115, Bur. Met. Australia.

Wu, Z.-J., McAvaney, B. and Zhang, H. 2006. Evaluating tropical diurnal variation of clouds and OLR in AGCMs: AMIP II models.

BMRC Research Report No. 116, Bur. Met. Australia.

Wain, A.G. and Mills, G.A. 2006. The Australian Smoke Management Forecast System. BMRC Research Report No. 117, Bur. Met.

Australia.

Wu, Z.-J. and McAvaney, B.J. 2006. Sampling methods for climate model calculated brightness temperatures. BMRC Research

Report No. 118, Bur. Met. Australia.

Finkele, K., Mills, G.A., Beard, G. and Jones, D.A. 2006. National daily gridded soil moisture deficit and drought factors for use in

prediction of Forest Fire Danger Index in Australia. BMRC Research Report No. 119, Bur. Met. Australia.

Huang, X. and Mills, G.A. 2006. Objective identification of wind change timing from single station observations. BMRC Research

Report No. 120, Bur. Met. Australia.

Zhong, A., Alves, O., Hendon, H. and Rikus, L. 2006. On aspects of the mean climatology and tropical interannual variability in the

BMRC Atmospheric Model (BAM 3.0). BMRC Research Report No. 121, Bur. Met. Australia.

Marshall, A.G., Alves, O., Hendon, H.H. and Wheeler, M.C. 2006. A wave-number frequency spectral analysis of intraseasonal

variability in the standard BMRC atmosphere general circulation model. BMRC Research Report No. 122, Bur. Met.

Australia.

Hollis, A.J. and Kariko, A.P. (eds) 2006. ‘The Australian Community Climate and Earth System Simulator (ACCESS) – challenges

and opportunities’: Extended abstracts of presentations at the eighteenth annual BMRC modelling workshop, 28 November

– 1 December 2006. BMRC Research Report No. 123, Bur. Met. Australia.

Power, S.B. 2006. Simple analytic solutions of the Linear Delayed-Action Oscillator Equation relevant to ENSO theory. BMRC

Research Report No. 124, Bur. Met. Australia.

Power, S.B. and Pearce, K. 2006. Climate change research in the Bureau of Meteorology: abstracts of presentations at a workshop

held on 10 February 2006. BMRC Research Report No. 125, Bur. Met. Australia.

Greenslade, D.J.M., Simanjuntak, M.A., Burbidge, D. and Chittleborough, J. 2007. A first-generation real-time tsunami forecasting

system for the Australian region. BMRC Research Report No. 126, Bur. Met. Australia.

Draper, C.S. 2007. The atmospheric water balance over the Murray-Darling Basin. BMRC Research Report No. 127, Bur. Met.

Australia.

Huang, X. and Mills, G.A. 2007. Classifying objectively identified wind changes using synoptic pressure cycle phases. BMRC

Research Report No. 128, Bur. Met. Australia.

Kounkou, R., Mills, G.A. and Timbal, B. 2007. The impact of anthropogenic climate change on the risk of cool-season tornado

occurrences. BMRC Research Report No. 129, Bur. Met. Australia.

Beggs, H. 2007. A high-resolution blended sea surface temperature analysis over the Australian region. BMRC Research Report No.

130, Bur. Met. Australia.

Power, S. and Pearce, K. (eds) 2007. Tropical cyclones in a changing climate: research priorities for Australia. Abstracts and

recommendations from a workshop held on 8 December 2006 at the Bureau of Meteorology in Melbourne. BMRC Research

Report No. 131, Bur. Met. Australia.

Chambers, L., Webber, E., Mavromatis, A., Keatley, M. and Hughes, L. 2007. National Ecological Meta Database. BMRC Research

Report No. 132, Bur. Met. Australia.

Hollis, A.J. and Jemmeson, V. (eds) 2007. ‘Physical processes and modelling of the water and carbon cycle’: Extended abstracts of

presentations at the first annual CAWCR modelling workshop, 27-29 November 2007. BMRC Research Report No. 133,

Bur. Met. Australia.