Linking palaeoenvironmental data and models to understand the past and to predict the future

Download Linking palaeoenvironmental data and models to understand the past and to predict the future

Post on 30-Oct-2016

215 views

Category:

Documents

3 download

Embed Size (px)

TRANSCRIPT

<ul><li><p>oeuh</p><p>oroFe</p><p>UKy o</p><p>Review TRENDS in Ecology and Evolution Vol.21 No.12provide a sound theoretical basis for the sustainable man-agement of ecosystems and for measures adopted byhuman societies to cope with environmental change. Inrecent years, a research agenda has emerged that focuseson the combined use of palaeoenvironmental data drawnfrom environmental archives, such as lake sediments,peat deposits, floodplain stratigraphy and tree-rings,and dynamic simulation models based on mathematicalexpressions for biological and abiotic processes. Severalrecent developments have led to this synergy.</p><p>First, recent analyses of instrumental andpalaeoenvironmental records show that anthropogenicdisturbance in the form of global warming, disruption ofnutrient cycles, soil erosion, atmospheric pollution, UVradiation, land-cover change, habitat destruction andspecies invasions, has brought about considerable changeto ecosystems and their rates of change over the past 250years [1]. This has increased the need for models that</p><p>to validate complexity theory, but could be the only way oftesting the ability of dynamic models to simulate rarelyoccurring thresholds.</p><p>We highlight here a holistic view of environmentalchange that combines dynamic modelling and palaeoeco-logical methods [6] in a methodological framework wherepalaeoenvironmental reconstruction, hypothesis genera-tion and testing, and model development and validationare linked over a wide range of spatial and temporal scales,and levels of system complexity [7]. The main interactionsbetween modellers and palaeoecologists to date havefocussed on climate per se, especially in terms of usingpalaeoenvironmental records, such as pollen diagrams, asa response to climate forcing [8,9] and, hence, as a palaeo-climate proxy. However, at smaller spatial and temporalscales, particularly during the period of major anthropo-genic activity, models should be able ideally to simulate theinteractions among climate, ecosystems and humanactivities. Therefore, we also summarize the recent devel-opments in simulating the long-term (decadescenturies)</p><p>Corresponding author: Anderson, N.J. (n.j.anderson@lboro.ac.uk).Available online 26 September 2006.</p><p>www.sciencedirect.com 0169-5347/$ see front matter 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.tree.2006.09.005Linking palaeoenvirand models to undand to predict the fN. John Anderson1, Harald Bugmann2, Jo1Department of Geography, Loughborough University, Loughb2 Forest Ecology, Department of Environmental Sciences, Swiss3Department of Geography, University of Liverpool, Liverpool,4Department of Biology and Environmental Sciences, Universit</p><p>Complex, process-based dynamic models are used toattempt to mimic the intrinsic variability of the naturalenvironment, ecosystem functioning and, ultimately, topredict future change. Palaeoecological data provide themeans for understanding past ecosystem change andare the main source of information for validating long-term model behaviour. As global ecosystems becomeincreasingly stressed by, for example, climate change,human activities and invasive species, there is an evengreater need to learn from the past and to strengthenlinks between models and palaeoecological data. Usingexamples from terrestrial and aquatic ecosystems, wesuggest that better interactions between modellers andpalaeoecologists can help understand the complexity ofpast changes. With increased synergy between the twoapproaches, there will be a better understanding of pastand present environmental change and, hence, animprovement in our ability to predict future changes.</p><p>IntroductionWith the possibility of significant environmental changeoccurring over the coming decades, there is a pressing needto optimise our understanding of ecological change to helpnmental datarstand the pastture</p><p>n A. Dearing3 and Marie-Jose Gaillard4</p><p>ugh, UK, LE11 3TUderal Institute of Technology Zurich, 8092 Zurich, Switzerland, L69 7ZTf Kalmar, SE-39182 Kalmar, Sweden</p><p>produce ecological scenarios over the following years,decades and centuries. Second, the belief that the way inwhich ecosystems respond today might be a legacy of thehistory of the system [2] implies thatmodelling our presentand future will require a starting point in the past. Thus,where relevant timescales are greater than the length oftime over which instrumental data occur, palaeoenviron-mental data usually represent the only means for drivingand testing simulation models. Third, following Deeveys[3] adage of coaxing history to conduct experiments, awealth of information about ecological change can begained frommodelling the past to test post hoc hypotheses.Comparing model outputs for the past with palaeoenvir-onmental data can help untangle the relative roles ofmultiple stressors on an ecosystem, for example, climate,human activity and disease on vegetation change [4], orcan help identify the role of internally generated ecologicalchange, for example in aquatic ecosystems [5]. Fourth, theincreased theoretical awareness of the complexity ofecosystems, particularly the likelihood of sudden unanti-cipated responses, has focused attention on the evidencefor nonlinear change contained in palaeoenvironmentaltime series. Such time series not only provide the means</p></li><li><p>Scandinavia over the past 8000 years, and by Cowlinget al. [14] who focused on vegetation dynamics at two sitesin Scandinavia at a higher temporal resolution over thepast 1500 years.</p><p>Finally, a few studies have addressed the interactionsbetween direct climatic forcing, indirect climatic effects viadisturbance regime (e.g. wildfires), and direct human inter-ventions (i.e. management). For example, Keller et al. [15]used a forest model to evaluate the relative effects ofchanges in climate, forest fires and direct human impactson the decreasing abundance of European silver fir Abiesalba in the southern Alps that has been observed duringthe late Holocene. They found that a combination of directhuman activities (e.g. selective cutting), as well as a changein the wildfire regime must be invoked to explain thedynamics reflected in pollen records, thus indicating thatmodels can be used successfully for disentangling the</p><p>represent the integrated response of ecosystems or evenlandscapes, rather than local-scale variables that are typicallysimulated by dynamic models. In this context, dynamic modelsthat explicitly resolve the key driving variables in a mechanisticmanner have been used successfully to simulate vegetationdynamics during the Holocene [61]. Most of these studies focusedon forest succession and used forest gap models [62]; four modes ofsuch applications can be distinguished: Model validation: until the 1990s, the major emphasis was on</p><p>using pollen data from lake profiles to test the applicability ofmodels under climatic conditions (e.g. [6365]). Thus, palaeoeco-logical data were used in a one-way manner to validateecological models.</p><p> Projecting the future: model applications have long been used toprovide scenario assessments of the future behaviour ofecosystems under changing environmental conditions (e.g.[66,67]). Such an application requires previous model validationefforts, where palaeoecological data have an important role.</p><p> Synthesizing research findings: models can be used to integrateand synthesize data across disciplines and even temporal orspatial scales [68,69]; we are not aware of such applications in thefield of palaeoecology, but believe that such applications holdmuch promise.</p><p> Supporting data interpretation: dynamic models can be used toassist in the interpretation of palaeoecological data by performingscenario runs to test hypotheses about key driving factors for theobserved palaeo-proxies, such as pollen frequencies or macro-fossil densities.</p><p>Review TRENDS in Ecology and Evolution Vol.21 No.12 697ecological change at patch, stand, water body andcatchment and/or landscape scales, providing examplesof dynamic models developed for terrestrial and aquaticecosystem communities, and abiotic processes.</p><p>Terrestrial systemsChanges in vegetation structure, both natural andanthropogenic-driven, have considerable impact on localand regional climate, hydrology and biogeochemicalcycling. Given the impact of catchment changes on lakeand stream chemistry, and the recent emphasis onintegrated ecosystem science, terrestrial palaeoecologyand modelling are increasingly required to address thecomplexities and consequences of these changes.</p><p>Using models to understand vegetation dynamics atstandlandscape scalesDynamic models are models that describe the rates ofchange of variables characterizing a system rather thanthe state of the system itself, often by the use of differenceor differential equations. They have been combined suc-cessfully with palaeoecological data since the early 1980s.Whereas the early focus was on model validation (Box 1),attention has since shifted to the application of models forquantitatively evaluating hypotheses of the causes of vege-tation changes inferred from palaeo-records. This hasproved particularly successful for separating the directeffects of climate on vegetation dynamics from those ofnatural disturbances or anthropogenic impacts. Forexample, models have been used to identify periods ofthe past for which climate reconstructions need additionalattention: Hall and McGlone [10] simulated vegetationdynamics for islands of New Zealand for a period in therecent past [700800 yr before present (BP)] and in the firsthalf of the Holocene (70008000 yr BP). They interpreteddeviations between model outputs and palaeoecologicaldata in the more distant period as an indication thatclimatic conditions must have been significantly differentfrom those that are usually assumed to have prevailedat that time.</p><p>Other researchers have addressed the problem ofdisentangling direct human impacts (e.g. selectivelogging) from climatic impacts in palaeo-records. Forexample, Heiri et al. [11] evaluated the changes of uppertree-line elevation in the central Swiss Alps during theentireHolocene (Figure 1) using a forest successionmodel.Simulations yielded tree-line fluctuations of approxi-mately 100 m (i.e. between elevations of 2375 and2600 m), confirming results of an earlier palaeobotanicalstudy that had inferred decadalcentennial-scale Holo-cene fluctuations in tree-line altitude as themain driver ofthe observed changes in the pollen profile, rather thanchanges in pollen productivity [12]. The results of bothstudies showed that therehas beena stronghuman impacton tree-line elevation since 4500 yr BP, as climaticforcing alone could not be invoked to interpret the palaeoe-cological findings. Similar analyses about the relativeimportance of climatic versus direct anthropogenic effectson long-term ecosystem dynamics were conducted by</p><p>Bradshaw et al. [13], who evaluated the changes of thesouthern range limit of Norway spruce Picea abies in</p><p>www.sciencedirect.comBox 1. Models of ecological processes and palaeoecological</p><p>research</p><p>Quantitative models range from simple, static relationships thatrepresent a black box view of relationships between systemvariables (e.g. pollen biomass or diatom nutrient transfer functions[37]), to highly sophisticated, dynamic and mechanisticapproaches where the rates of change of the variables of thesystem (rather than their state) are described; such quantitativemodels can therefore be used to study temporal changes in complexphenomena. A model is a deliberate simplification of reality (allmodels are wrong, but some models are useful [60]) and thescientific question posed largely determines the value of a givenmodel. All scientists rely on various forms of models, qualitative orquantitative. Hence, the common divide between modellers andnon-modellers is inappropriate.</p><p>In palaeoecology, a dynamic view of ecological processes isessential, but difficult because the temporal resolution ofpalaeoecological data tends to be relatively low (rarely annual andhardly ever subannual). In addition, the measured variables oftencomplexity inherent in the long-term dynamics ofecosystems.</p></li><li><p>lts fr</p><p>ndep</p><p>ent</p><p>e si</p><p>site</p><p>atio</p><p>698 Review TRENDS in Ecology and Evolution Vol.21 No.12Geomorphic and landscape-response models</p><p>Figure 1. Alpine tree-line changes. (a) Macrofossil data and (b) simulation resu</p><p>Switzerland, 2350 m a.s.l.). A chironomid-based temperature reconstruction that is i</p><p>composition and tree-line position simulated in this study showed general agreem</p><p>however, palaeobotanical evidence indicated a lowering of the tree-line, whereas th</p><p>changes in temperature alone can account for changes in tree-line elevation at this</p><p>interpretations that there has been strong human influence on Alpine tree-line elevPalaeoenvironmental records exist for flood frequency andmagnitude, soil erosion, sediment yield, fluxes of carbonand nutrients and macroscale changes in river channels ata variety of spatial scales [16]. Learning about thefunctioning of modern systems from studies of these pastgeomorphic records takes many forms [17,18].</p><p>Some attempts have been made to model thesepalaeoenvironmental data in terms of abstract systemsconcepts. For example, Dearing and Zolitschka [19]analysed the magnitudefrequency characteristics of highresolution time series of sediment accumulation in LakeHolzmaar, Germany, and found that the early Holocenefluvial system conformed to a power function model ofself-organized criticality. In this state, most small changesin system behaviour are the result of nonlinear internalprocesses, but these are punctuated by disproportionatelylarge responses to large external forcing. Latercombinations of climate and human impact in thecatchment substantially modified this apparent naturalstate, giving rise to more simple relationships betweenerosion, land use and climate.</p><p>However,most linkswithmodels are concernedwith thedevelopment, forcing and testing of a steadily increasingnumber of dynamic process-based models [2022]. Two-dimensional models include hydrological models that aredriven by meteorological time series that compute water,soil or sediment discharge based on basic catchment prop-erties, such as size, relative relief and land-cover change.When compared with regional reconstructions of valleyinfilling throughout the Holocene [23] or high-resolutionlake sediment proxies of flooding over recent centuries [2],</p><p>www.sciencedirect.comthey have the potential to unravel climate and human</p><p>om the FORCLIM model ([72]) for the upper tree-line site Gouille Rion (Valais,</p><p>endent of vegetation proxies was used to drive the model. The changes in species</p><p>with palaeobotanical data between 11 000 and 4500 yr BP. In the late Holocene,</p><p>mulation projected continuous forest cover up to an altitude of 2400 m a.s.l. Thus,</p><p>only for the first half of the Holocene, and the results corroborate palaeoecological</p><p>n since at least 4500 yr BP. Modified with permission from [11].impacts on key ecological processes.Progress in linking data and models for disso...</p></li></ul>