Emissions of greenhouse gases are expected toraise global mean temperature over the next century

by 10ndash35 degC (Houghton et al 1995 1996) Ecologistsfrom around the world have begun experiments toinvestigate the effects of global warming on terrestrialecosystems the aspect of global climate change thatattracts the most public attention (Woodwell andMcKenzie 1995 Walker and Steffen 1999) The effort tounderstand response to warming builds on a history ofinvestigations of the effects of elevated CO

2 on plants andecosystems (Koch and Mooney 1996 Schulze et al 1999)There are important differences however betweenincreases in atmospheric CO

2 and temperature changeboth in the temporal and spatial patterns of change and inhow they affect ecosystems The scientists involved intemperature change research have had to face newtechnical and conceptual challenges in designing andinterpreting their experiments (Schulze et al 1999) In thispaper we describe these challenges and present aconceptual framework for interpreting experimentalresults and predicting effects of warming on ecosystems

October 2000 Vol 50 No 10 BioScience 871

Articles

Gaius R Shaver (e-mail gshavermbledu) is a senior scientist at The Ecosystems Center Marine Biological Laboratory Woods Hole MA 02543Josep Canadell (e-mail PepCanadelldwecsiroau) is GCTE executive officer at the GCTE International Project Office CSIRO Wildlife amp Ecology

Canberra ACT 2601 Australia F S Chapin III (e-mail fschapinlteruafedu) is a professor at the Institute of Arctic Biology University of Alaska

Fairbanks AK 99775 Jessica Gurevitch (e-mail jgurvtchlifebiosunysbedu) is an associate professor in the Department of Ecology and

Evolution SUNYndashStony Brook Stony Brook NY 11794-5245 John Harte (e-mail jhartesocratesberkeleyedu) is a professor in the Energy and

Resources Group Department of Environmental Science Policy and Management University of California Berkeley CA 94720 Greg Henry (e-

mail ghenrygeogubcca) is an assistant professor in the Department of Geography University of British Columbia Vancouver BC V6T 1Z2

Canada Phil Ineson (e-mail pi2yorkacuk) is a professor in the Department of Biology University of York York YO10 5YW United Kingdom Sven

Jonasson (e-mail svenjbotkudk) is a professor in the Department of Plant Ecology University of Copenhagen Copenhagen Denmark Jerry

Melillo (e-mail jmelillombledu) is a senior scientist at The Ecosystems Center Marine Biological Laboratory Woods Hole MA 02543 Louis

Pitelka (e-mail pitelkaalumcesedu) is director of and professor at the Appalachian Laboratory University of Maryland Center for Environmental

Sciences Frostburg MD 21532 Lindsey Rustad (e-mail rustadmaineedu) is a forest ecologist at the Northeastern Research Station USDA

Forest Service Pownal ME 04069 copy 2000 American Institute of Biological Sciences

Global Warming andTerrestrial Ecosystems AConceptual Framework forAnalysisGAIUS R SHAVER JOSEP CANADELL F S CHAPIN III JESSICA GUREVITCH JOHN HARTE GREG HENRY PHILINESON SVEN JONASSON JERRY MELILLO LOUIS PITELKA AND LINDSEY RUSTAD

ECOSYSTEM RESPONSES TO GLOBAL

WARMING WILL BE COMPLEX AND VARIED

ECOSYSTEM WARMING EXPERIMENTS

HOLD GREAT POTENTIAL FOR PROVIDING

INSIGHTS ON WAYS TERRESTRIAL

ECOSYSTEMS WILL RESPOND TO

UPCOMING DECADES OF CLIMATE

CHANGE DOCUMENTATION OF INITIAL

CONDITIONS PROVIDES THE CONTEXT FOR

UNDERSTANDING AND PREDICTING

ECOSYSTEM RESPONSES

Projections of global warming from General Circula-tion Models (GCMs) are now familiar to both scientistsand nonscientists Knowing however that the mean glob-al temperature will increase by 10ndash35 degC tells us littleabout how temperatures will change in a particular loca-tion or how the ecosystems in that location will respondFor instance temperature increases are likely to be greaterat higher latitudes (Houghton et al 1995 1996) where ini-tial conditions are below 0 degC most of the year and grow-ing season temperatures are only a few degrees higherEven a small increase in average temperature at higher lat-itudes could increase both the length of the unfrozen peri-od and degree-day accumulations by large percentages(Billings 1987) The implications of a given increase intemperature will also depend on the initial temperature ata specific location because the rate of many biologicalprocesses in relation to temperature typically peaks atsome intermediate temperature An increase from an ini-tially low temperature may cause an increase in photosyn-thesis for example while an increase from an initially high

temperature might cause photosyn-thesis to decrease (Larcher 1995)Finally even as the average tempera-ture increases some parts of theglobe are actually expected to experi-ence cooling (Houghton et al 19951996) The complexities of these pat-terns of change in temperature con-trast with the increase in atmospher-ic CO2 for which there is no largetemporal or spatial variation eitherin current levels or in the rate or pat-tern of increase

Direct and indirect effectsof warming onecosystemsThe direct effects of warming onecosystems will also be more com-plex than the direct effects ofincreased CO

2 because temperatureimpacts virtually all chemical andbiological processes whereas thedirect influence of CO

2 is almostentirely limited to leaves (photosyn-thesis stomatal aperture and per-haps respiration Koch and Mooney1996) For both warming and CO

2however the greatest obstacles tounderstanding lie in the web of indi-rect effects resulting from interac-tions among processes affecteddirectly by environmental changeThese interactions lead to feedbacksthat are sometimes positive and

sometimes negative so the responses to temperaturechange or CO

2 can be expected to vary among ecosystemsin both magnitude and direction depending on the prop-erties of the dominant species interactions among speciesand the initial physical and chemical environment

The effects of a temperature increase on the carbonbudget of an ecosystem provides an example of this com-plexity (Figure 1) Net ecosystem production (NEP)defined as the overall carbon balance of an ecosystem oversome time period (Woodwell and Whittaker 1968Mooney et al 1999) has two major components net pri-mary production (NPP) and heterotrophic respiration(R

h) NPP the principal input of carbon to the ecosystemis the net result of CO

2 fixation by photosynthesis andCO

2 loss by plant respiration The product of NPP is neworganic matter which accumulates first in plants as livingbiomass and is eventually transferred to soils as litter andto animals and decomposer organisms as food R

h repre-sents the loss of carbon from the ecosystem by respirationof animals and decomposers the products of R

h include

872 BioScience October 2000 Vol 50 No 10

Articles

Figure 1 Direct and indirect effects of temperature on net primary production (NPP)heterotrophic respiration (Rh) and net ecosystem production (NEP) the majorcomponents of carbon exchange between terrestrial ecosystems and the atmosphere(Woodwell and Whittaker 1968) NPP removes carbon from the atmosphere whereasRh returns carbon to the atmosphere NEP is the sum of these two opposing fluxes andrepresents the net exchange of carbon between an ecosystem and the atmosphereFrom the perspective of the ecosystem NEP is positive when NPP is larger than Rh andis negative when NPP is smaller than Rh In this diagram arrows represent chains ofcause and effect by which temperature affects NPP and Rh but are not intended toindicate the magnitude or sign of an effect Their purpose is simply to show thattemperature influences NPP and Rh directly and indirectly whereas changes in NEPare the net result of the effects on NPP and Rh (SOM soil organic matter)

CO2 and other inorganic carbon products (eg CH4)Both NPP and R

h are affected directly by temperaturechange Both are usually increased by warming althoughR

h often increases more rapidly in the short term (Wood-well 1995) In addition warming can affect NPP and R

h

indirectly by altering the ecosystemrsquos moisture regimenitrogen availability length of its growing season orspecies composition (Figure 1) Warming-driven changesin moisture nitrogen or species composition may alsohave intermediate effects on other ecosystem processes orstates (eg litter quality and quantity which affect both R

h

and nitrogen mineralization) leading to multistep indi-rect effects including losses of carbon through fire orleaching and changes in the balance of NPP and R

h (Melil-lo et al 1990 Chapin et al 1997)

For many ecosystems the indirect effects of a tempera-ture increase on carbon balance are likely to be moreimportant than the direct effects Nutrient-limited tundraand northern forest ecosystems for example are muchmore responsive in the short term (1ndash10 years) to changesin nutrient availability which is likely to increase with soilwarming than to the direct effect of increased tempera-ture (Clark and Rosswall 1981 Chapin 1983 Jonasson andShaver 1999) In many dry ecosystems increased evapora-tive water loss at higher temperatures resulting in dryersoils may strongly limit soil and plant processes so thatpotential temperature-driven increases in process rates arenot achieved (Mooney et al 1999 Saleska et al 1999)Finally changes in temperature may alter speciesrsquo compet-itive interactions and the activity of herbivores andpathogens (Mooney et al 1991 Smith et al 1997deValpine and Harte in press) leading to changes in lightwater and nutrient limitations with complex long-termeffects on carbon turnover NPP and Rh (Herbert et al1999)

Variable time scales of responses towarmingWarming will affect essentially all ecosystem processes andorganic matter pools but at different rates Because theprocesses and pools are linked to each other through bio-geochemical cycles the magnitude and even the directionof the net change in the ecosystem may vary over time Forexample NPP is affected by processes operating over awide range of time scales from the very short-termresponses of leaf-level photosynthesis to the long-termchanges in storage and turnover of soil nitrogen stocks(Figure 2) Although in most ecosystems the immediateeffect of warming may be to increase NPP throughincreased photosynthesis in the longer-term the uptakeand accumulation of nitrogen in plant biomass (particu-larly in forests) might reduce the possibility for furtherincreases in NPP by reducing nitrogen availability Even-tually changes in species composition or litter qualitymight lead to further increases or decreases in NPPthrough changes in plant biomass (and nutrient) turnover

or changes in litter decomposition and nitrogen mineral-ization In some ecosystems (eg boreal forests) theincreasing fuel load associated with biomass increase mayincrease the likelihood of fire and therefore the proportionof stands with low carbon storage on the landscape Thusthe principal long-term mechanisms of regulation of NPPand overall carbon balance with respect to temperaturemay be very different from the short-term mechanisms(McKane et al 1995 Braswell et al 1997 Rastetter et al1997)

The time scales of response of individual processesshould also vary among ecosystems adding further com-plexity to long-term prediction of changes caused bywarming (Figure 2) For example in ecosystems where thedominant species in the vegetation are long-lived treeschanges in species composition might have relativelyminor effects on NPP in the first few decades of warmingbut changes in biomass allocation within the existingdominants might have major effects Principal controlsover nitrogen availability in such forests might lie in con-trols over turnover of existing large soil nitrogen pools Onthe other hand the changes in species composition in anannual grassland with a relatively small soil organic matter

October 2000 Vol 50 No 10 BioScience 873

Articles

Figure 2 Time scales of response to temperature changeby various ecosystem processes and components Each ofthe processes or components shown in the figure affect netecosystem production either directly or indirectly Forconvenience they are grouped into categories vegetationsoils and other The intent is to show how differentprocesses and components respond to temperature changeat different rates hence the overall ecosystem response(the result of the individual responses and theirinteractions) may be very different in the long-termversus the short-term Many other processes andcomponents could be added to this figure (Psphotosynthesis Rs respiration SOM soil organicmatter)

pool could be more important to NPP and changes in soilnitrogen turnover less important within the first decades

Ecosystem warming experiments if continued over sev-eral years can capture much but not all of this changingsequence of responses to temperature change (Figure 2)Changes in basic metabolism (such as photosyntheticacclimation to temperature) may take place over less than1 year and will be missed by annual sampling schedulesbut these short-term responses may have less net impacton long-term change in ecosystem properties than thewarming-induced rearrangement of ecosystem carbonand nutrient stocks and species composition which takesplace on a time scale of 1ndash100 years Some changes espe-cially those occurring over very long time scales (eg soilprofile development and organic matter accumulation) orlarge spatial scales (changes in fire regime or long-distancemovements of large herbivores or timberlines) will not bepicked up by experiments and must be investigated in oth-er ways

Examples of warming responses CasestudiesDespite the complexities of ecosystem responses to warm-ing an important goal of ecosystem science should be topredict which kinds of ecosystems are more or less respon-sive to warming and to identify the characteristics ofecosystems that cause them to be more or less responsiveOur present ability to do so is limited but evidence indi-cates that ecosystem responses to warming are stronglyaffected by initial conditions including

874 BioScience October 2000 Vol 50 No 10

Articles

Figure 3 Changes in net ecosystem production (NEP the net exchange of carbon between an ecosystem and the atmosphere)over time in response to warming There is a net carbon uptake by the ecosystem when NEP is positive when NEP isnegative there is a net carbon loss The different lines represent qualitative differences in response (and predicted long-termresponses) to experimental warming by different ecosystems as discussed in the text Harvard Forest lines A B Toolik Lakelines C D Great Dun Fell lines D E Colorado subalpine meadow line C Abisko line C

Figure 4 Harvard Forest Soil Warming ExperimentMassachusetts showing early spring greening of forestfloor In this experiment electric heating cables areburied 10 cm beneath the soil surface and 10 cm apartalong lines indicated by the string laid out on thesurface Photo Kathy Newkirk

Stocks and initial turnover rates of labile carbon inthe soil

Stocks and initial turnover rates of labile nitrogen inthe soil

Relative size of the carbon pools of plants and soil

Dominant form of available nitrogen in the soil (egorganic nitrogen ammonium nitrate)

Soil water and precipitation regime

Chemical composition and turnover rates of liveplant tissues and litter

Longevity of individuals and population turnoverrates of dominant plant species

The effects of these initial conditions on the response ofecosystems to warming can be illustrated by comparingchanges in NPP NEP or R

h over time in contrastingecosystems that have been manipulated experimentallyThis approach has already led to useful insights as shownby the following examples and illustrated in Figure 3Nitrogen redistribution model (Harvard ForestMassachusetts United States) At Harvard Forest incentral Massachusetts (a temperate deciduous forestecosystem) a long-term soil warming experiment usedelectric heating cables buried in the soil to elevate soil tem-peratures at a depth of 10 cm by 5 degC above ambient (Fig-ure 4 Peterjohn et al 1994 Melillo et al 1995) At this sitethe important initial conditions include

Relatively small labile soil carbon pool

Large slow-turnover vegetation carbon pool

Relatively large soil nitrogen pool

Ammonification-dominated nitrogen mineralizationsequence

These initial conditions constrain how the Harvard Forestecosystem responds to soil warming over time such that athree-phase transient pattern of changes in carbon storageresults (Figure 3 line A Melillo et al in press) DuringPhase I the labile carbon pool is rapidly oxidized bydecomposing organisms leading to a short period of netcarbon loss In Phase II nitrogen is transferred from thesoil which has a low CN ratio to the vegetation whichhas a high CN ratio This redistribution of nitrogenresults in net carbon storage at the ecosystem levelalthough there is a loss of soil carbon (Rastetter et al 19911997) During Phase II there is no significant loss of nitro-gen from the ecosystem because the system has an ammo-nium economy If the system had a nitrate economy or amixed ammoniumndashnitrate economy nitrogen could belost from the system by leaching or denitrification and soreduce the nitrogen available to promote carbon storageEventually the labile and metastable nitrogen pools in thesoil decrease to the point where very little nitrogen can beredistributed from the soil to the plants At this point theecosystem may begin to lose carbon overall (ie NEPbecomes negative) because of increased respiratory lossesfrom plants and soils (Figure 3 line A) This decline may

be reversed if external nitrogen inputs (deposition or fixa-tion) are available or if plant biomass is lost due to distur-bance such as a hurricane or insect attack and is decom-posed releasing nitrogen once again (Figure 3 line B)Nitrogen redistribution with variable moistureregime (Toolik Lake Alaska United States) In amoist tundra ecosystem at Toolik Lake in northern Alaskaresponses to warming were studied using small green-houses to increase air temperature during the growing sea-son by 3ndash5 degC (Figure 5) The soil organic matter pool ofthis ecosystem is larger and contains even more labile car-bon and nitrogen than the soils of many northern tem-perate forests (Giblin et al 1991) Decomposition andnitrogen mineralization in tundra is strongly limited bylow soil temperatures and high soil moisture (Nadelhofferet al 1992) Thus if soil organic matter turnover isincreased due to warming there is a high potential forredistribution of nitrogen from soils (with low CN ratio)to vegetation (with high CN ratio) resulting in net car-bon storage with little or no net change in ecosystemnitrogen stocks (Shaver et al 1992) This response howev-er is strongly interactive with soil moisture with a rela-tively small temperature response if the soil is saturated(Nadelhoffer et al 1992)

Experimental results (Chapin et al 1995 Shaver et al1998) and modeling (McKane et al 1997a 1997b Rastet-ter et al 1997) indicate a multiphase response of NEP towarming in this tundra ecosystem that is quite similar tothat at Harvard Forest as long as soil moisture does notchange greatly (Figure 3 line C) Similar results have alsobeen obtained in the long-term monitoring and modeling

October 2000 Vol 50 No 10 BioScience 875

Articles

Figure 5 The Toolik Lake Field Greenhouse Experimentin moist tussock tundra vegetation Alaska showingmoderately increased abundance of deciduous shrubsand total vegetation mass The photograph was taken in1999 after 11 summers of warming Photo Gus Shaver

studies by Oechel and coworkers (eg Vourlitis andOechel 1997 1999 Waelbroeck et al 1997) The initialresponse to warming is a large increase in R

h leading to adecrease in NEP for the first 1ndash3 years (Phase I) Increasedsoil respiration however is accompanied by increased

nitrogen mineralization leading to increased plant nitro-gen uptake and increased NPP after 3ndash10 years (Phase II)Further increases in NPP are limited by other factors (eglight) Increased NPP leads to increased litter productionwhich eventually leads to higher R

h and NEP returns tonear zero (Phase III) after approximately 50ndash100 years

Changes in NEP in response to warming of moist tun-dra may be very different if there is an interaction between

warming and soil moisture (Billings 1987 McKane et al1997b Vourlitis and Oechel 1999) Warming combinedwith decreased soil moisture will cause a long-term loss ofboth carbon and nitrogen because of large increases in Rh

combined with losses of mineralized nitrogen by drainagefrom the system especially in areas where thaw depthincreases greatly (In such areas nitrogen losses mayinclude large amounts of dissolved organic nitrogen theproduct of incomplete mineralization) In this scenarioincreases in nitrogen uptake and NPP in Phase II areinsufficient to compensate for nitrogen and carbon lossesin leaching and respiration so that even though theecosystem eventually returns to an equilibrium whereNPP equals R

h (NEP = 0) over 50ndash100 years there is a netloss of carbon to the atmosphere (Figure 3 line D)Partial decay of large carbon pool (Great DunFell United Kingdom) In an upland grassland systemat Great Dun Fell in the United Kingdom a combinationof approaches was used to assess the impact of globalwarming on ecosystem carbon stores and soil water quali-ty In one intact vegetationsoil monoliths were moveddown an altitudinal gradient (Ineson et al 1998a) inanother in situ warming using electric heating cables (Ine-son et al 1998b) was applied Both approaches were testedin several different vegetationsoil systems including amineral acidic brown earth and a peaty stagnohumic gleyResults after 3 years of the heating treatments were quali-tatively similar to those predicted for the arctic tundraunder warming combined with reduced soil moisture(Figure 3 line D) but the net carbon losses were even larg-er

All systems at Great Dun Fell showed an increase inplant biomass under the warmer conditions but majorreductions in the total carbon content of the peaty gley(approximately 10 reduction in 3 years) demonstratedthe vulnerability of the carbon stored in this particularsoil The increases in temperature which stimulated themineralization of organic matter and release of nutrientssignificantly improved plant growth but these increases incarbon fixation were insufficient to replace the carbon lostthrough increased decomposition These increases indecomposition were caused by generally greater soil bio-logical activity and in particular by population increasesand changes in vertical distribution of the enchytraeidworm Cognettia sphagnetorum (Briones et al 1998) Theselarge losses in total soil carbon stocks clearly cannot con-tinue eventually the soils are expected to stabilize at alower carbon content analogous to the pattern shown inFigure 3 (line E) and by the native soils at the lower end ofthe same altitudinal gradientSpecies changes affect ecosystem carbon storage(subalpine meadow Colorado United States) Ina subalpine meadow ecosystem in Colorado Harte and col-leagues have used overhead infrared heaters to warm bothsoil and vegetation year-round since 1991 (Figure 6) caus-ing significant changes in NPP and NEP (Harte et al 1995

876 BioScience October 2000 Vol 50 No 10

Articles

Figure 6 Infrared heaters in place above subalpinemeadow vegetation Colorado Photo Taylor Ricketts

Figure 7 Smaller plastic greenhouses and burlap-coveredshade frames at the subarctic heath site Abisko SwedenPhoto Sven Jonasson

Saleska et al 1999) In contrast to the previous examples aprincipal mechanism driving the changes in Colorado is theeffect of the vegetation on species growth rates and func-tional type composition (Harte and Shaw 1995 deValpineand Harte in press) A 12 loss of soil carbon in the top 15cm of the heated plot soils was observed after 4 years oftreatment (Scott Saleska and John Harte University of Cal-iforniandashBerkeley unpublished data) This initial loss of soilorganic matter was not driven by a warming-inducedincrease in heterotrophic respiration because the positiveeffect of a temperature increase on R

h was cancelled by thenegative effect of soil drying Instead this initial loss of soilorganic matter was driven mainly by the effect of warmingon the rate of litter input to the soil In particular warminginduced a shift from more productive forbs to less produc-tive shrubs However litter produced by forbs is more easi-ly decomposed than litter produced by the shrubs whichhas a higher ligninnitrogen ratio Thus the loss of soilorganic matter is likely to be a transient effect with therecalcitrant litter input from shrubs causing an increase insoil organic matter over the long term and ultimately in anoverall pattern of change in NEP similar to that in Figure 3(line C)

By comparing responses observed in the warmingexperiment to changes in microclimate soil organic mat-ter and vegetation species composition along a naturalelevation gradient Harte and colleagues have shown thatthe combined use of manipulations and gradient studiescan provide a unified picture partitioned into short andlong-term components of how climate change influencessoil organic matter The research shows how the impacts ofwarming on carbon cycling can be dominated by the indi-rect effect of a warming-induced shift in relative speciescomposition on the quantity (short term) and quality(longer term) of litter input to the soilExploitation of available space (Abisko Sweden)In low-biomass ecosystems where the surface of the soil ispartially bare warming brought about by increasing plantcover may cause increases in NEP In tundra heath and fellfield ecosystems at Abisko Sweden warming with smallplastic-covered greenhouses caused such an increase inplant cover (Figure 7) Plant biomass driven in part bytemperature-related increases in plant nitrogen uptakeincreased almost twofold within 5 years (Jonasson et al1999)

The response of biomass to warming at Abisko was sim-ilar to that in the tussock tundra at Toolik Lake althoughthe response at Abisko was greater and NPP increasedwithout any pronounced changes in species compositionOverall the response followed the curve represented inFigure 3 (line C) initial declines in carbon stocks were notobserved because the experiment was not sampled imme-diately The longer term (more than 10 years) response isexpected to level off and reach an equilibrium at a net car-bon accumulation rate close to zero when the availablespace has been occupied by plants

Initial increases in biomass in response to warming neednot be a continuous process over time particularly not inthe northernmost ecosystems where the plants operate closeto the limit of their cold tolerance For instance an increasein biomass similar to that at Abisko took place over severalyears in a poorly vegetated high arctic semi-desert howev-er after an unusually mild and rainy growing season muchof the increase was lost in warmed and watered plots Theloss in biomass presumably occurred because winter hard-ening in the experimental plots was delayed and the plantssuffered frost damage when the winter set in (Robinson etal 1998) reflecting a possible effect of stochastic ldquoextremeeventsrdquo Such events are likely to be of great importance inclimatic regimes close to the tolerance limits of plantsUnder such regimes carbon pools in the living plant bio-mass may fluctuate between phases of accumulation andabrupt decline never stabilizing at zero net accumulationwhile soil organic matter is built up stepwise during theperiods of plant die-back

Experiments data and synthesis needsPast warming experiments have used a variety of heatingmethods including electric resistance heating infraredirradiation reciprocal transplants and open- and closed-top field greenhouses (Table 1) Some experiments havewarmed only portions of the ecosystem Others havewarmed the ecosystem for only part of each day or year Inmost cases the choice of warming method is constrainedby logistical and engineering considerations (eg avail-ability of electric power difficulty of enclosing an entireforest and the initial interests of the investigators) No onemethod perfectly simulates the expected climatic changesbut all have yielded useful results that with cautious inter-pretation have begun to reveal similarities differencesand patterns of response among the ecosystems studiedAlthough there is considerable intellectual appeal todesigning future experiments that use a common method-ology and realistic simulation of expected change it also istrue that different warming methods illuminate differentaspects of the overall warming response To optimizeintercomparisons future research should include a mix ofcommon manipulations in diverse ecosystem typesFuture research should also include manipulationsdesigned to provide focused insights into individualecosystems or key processes

Several networks and consortia are already working tomeet the need for a global understanding of ecosystemresponses to climate change (Ingram et al 1999) TheInternational Tundra Experiment (ITEX) uses passivelywarmed open-top chambers in 26 different arctic andalpine tundras to compare the effects of warming on plantgrowth and flowering (Henry and Molau 1997 Arft et al1999) A broader-based Network of Ecosystem WarmingStudies (NEWS) is being developed as part of the GlobalChange and Terrestrial Ecosystems (GCTE) Core Projectof the International GeospherendashBiosphere Program

October 2000 Vol 50 No 10 BioScience 877

Articles

(IGBP) Most warming experiments to date however havetaken place in north temperate boreal and arctic ecosys-tems which are cool and moist and where NPP is strong-ly nutrient-limited Although the processes that dominatethe warming responses of these ecosystems are importanteverywhere they may not dominate in other ecosystemtypes In warm and dry regions including Mediterraneanor desert ecosystems and in tropical humid ecosystemsindirect warming effects acting through changes in evapo-transpiration and soil dryness will play a major regulatoryrole We also might expect the direct effects of warming onNPP to be relatively smaller in these warm systemsbecause temperatures are already near optimum for pho-tosynthesis NPP might even decrease due to overheatingin the warmest ecosystems or microenvironments Addi-tional experiments and networks are needed to expand therange of ecosystems studied including studies of vegeta-tion with contrasting plant life histories (eg annual ver-sus perennial grasslands woodlands and evergreen versusdeciduous vegetation)

New methods are also needed for the quantitative syn-thesis of data from different studies of temperature effectson ecosystems Comparisons among different simulationmodels have also yielded important insights (Schimel et al1997) Some relatively new simulation models are

designed to facilitate comparisons among ecosystems andto predict their responses to temperature change within acommon model structure (eg Rastetter et al 1991Williams et al 1997) New statistical methods can also beused to search for and explain general patterns in diversedata sets For example we might want to know the averageeffects of experimental warming treatments on NPPacross a group of studies and the extent to which thoseeffects differ in biomes at different latitudes or with differ-ent moisture regimes Ecologists have begun to use met-analysis statistics to synthesize experimental data frommany independent studies to answer these kinds of ques-tions (Arnqvist and Wooster 1995 Gurevitch and Hedges1999) including plant responses to warming in ITEX (Arftet al 1999)

Finally there is a need to extrapolate experimentalresults over longer time scales and over larger landscapesand regions An excellent way to do so is to combineexperimentation with measurement of ecosystem proper-ties and processes along environmental transects such asthe IGBPndashGCTE Terrestrial Transects (Steffen et al1999) The transects are assumed to mirror longer-term(decades to centuries) responses of vegetation and soils toenvironmental conditions whereas warming experimentsreflect responses over the shorter-term (years to decades)

878 BioScience October 2000 Vol 50 No 10

Articles

Table 1 Comparison of currently used methods for experimental warming of terrestrial ecosystems

Method Mechanism of warming Advantages Disadvantages Examples

Field greenhouse ldquoGreenhouse warmingrdquo (ie Simple and inexpensive Little or no temperature Chapin et al 1995reflection of reradiated infrared requires no electrical power control and large tempera- Shaver et al 1998energy and reduced advective ture variability other arti- Jonasson et al 1999energy loss) facts include altered light

wind humidity and precipi- tation regimes

Passive Open-top Same as above Same as above Little or no temperature Marion et al 1997Chamber (OTC) control altered wind and Arft et al 1999

humidity only small areas can be manipulated uniformly

Active Open-top Same as above Precise control of air temp- Altered wind regime Norby et al 1997Chamber (OTC) plus warming by advection of erature or temperature humidity and

electrically heated forced difference may be combined evapotranspirationwarm air with CO2 control

Active Soil Warming Warming by conduction from Precise control of soil Altered soil moisture Peterjohn et al 1994buried electrical resistance temperature or temperature regime no effect on Ineson et al 1998bcables difference may be combined aboveground temperatures Hartley et al 1999

with greenhouses or OTCs

Electric infrared heat Warming by increased infrared Precise control of energy Warming depends entirely Harte et al 1995radiation input direct simulation of on radiation no change in Harte and Shaw 1995

global change in energy advective energy inputsbalance

Reciprocal or one-way Transplantation of plants Comparison of relatively Disturbance effects Ineson et al 1998atransplantation soils or whole plantndashsoil natural temperature gradients multiple environmental Grogan and Chapin1999 systems to warmer or cooler changes make it difficult

locations to assign specific causes to responses observed

Combined information from experiments and transectsmay thus be useful in distinguishing relatively short-termtransient changes in the experiments from longer-termnear-equilibrium changes For example global analyses ofprimary production (NPP Figure 1) and litter (the majorcontributor to R

h Figure 1) have shown that both are rea-sonably well-predicted by actual evapotranspiration(Lieth and Whittaker 1975 Meentemeyer 1984)Although short-term responses of NPP and R

h to experi-mental warming may not have the same relationshipswith actual evapotranspiration as do undisturbed ecosys-tems convergence toward the predicted relationshipmight be a useful measure of expected future change inthe experiments

Ultimately experimental and analytical approachesshould be combined with large-scale monitoring of vari-ables and processes that experimental studies have identi-fied as critical indicators of ecosystem change The combi-nation of global monitoring with ongoing experimentaldata acquisition metanalysis modeling and other datasynthesis efforts would provide an early warning system todetect and potentially mitigate these ecosystem responses

Field stations long-term ecological research sites andnature reserves already exist in many relatively unman-aged landscapes Collecting simple baseline data and mon-itoring annually or even less frequently would not be veryexpensive although data quality may be highly variable(eg Stohlgren et al 1995) Efforts to coordinate data andestablish protocols for data sharing might be daunting butthe resulting information would be immensely valuableresponses of species communities and biogeochemicalprocesses to climate change could potentially be detecteddecades earlier than they would be otherwise

Interactions with other drivers of globalchangeChanges in temperature are now occurring simultaneous-ly with other types of global change It will not be possibleto fully understand and predict ecosystem responses totemperature change without taking into account the inter-actions with the other components of global environmen-tal change (Koch and Mooney 1996 Sellers et al 1997Vitousek et al 1997 Luo and Mooney 1999) For exampleincreases in atmospheric CO

2 may modify plant tissuechemistry and ecosystem water balance both of whichmay in turn modify responses to warming Length of thegrowing season may increase with warming or it maydecrease as drought duration increases altering NPP NEPand other processes Many regions of the globe are experi-encing increased levels of nitrogen deposition and thenature of the nitrogen economy of an ecosystem is animportant factor in determining responses to other envi-ronmental factors (Aber et al 1998)

Changes in land use and disturbance regime may alsoinfluence responses to warming For instance land clear-ing over very large areas may modify local temperature or

precipitation regimes beyond what would happen withclimate change alone Similarly changes in fire regimescould be caused by and affect ecosystem responses towarming The spread of invasive species will also be affect-ed by both warming and land use change

Some interactions between warming and other globalchanges are predicted more readily than others For exam-ple outbreaks of insects or pathogens that kill the dominantplants and increase fire probability would lead to dramati-cally different results than the phased changes in response towarming alone We must expect some surprises

Clearly it will not be possible to incorporate all impor-tant dimensions of global change into experiments It willrarely be feasible to include more than one or two factorsin addition to temperature (eg CO2 concentration ornitrogen level) as experimental factors (Schulze et al1999) However through the use of combinations ofexperiments gradient studies remote sensing and simu-lation modeling it should be possible to make substantialprogress toward developing a mechanistic understandingof the most important interactions (Canadell et al 2000)

Although much current global change research focuseson feedbacks from terrestrial ecosystems to climatethrough greenhouse gases important feedbacks also existthrough changes in vegetation structure surface energybalance and water balance Leaves in the canopy act assurfaces for energy and mass exchanges with the atmos-phere and changes in canopy density and transpiration inresponse to warming will change the energy and massfluxes (McFadden et al 1998) One of the most importantand obvious effects of changes in the canopy is its albedo(ie the fraction of incoming solar radiation that isreflected) Decreasing albedo through increased leaf den-sity will increase net radiation (solar radiation absorbedby the system) The increased net radiation must be dissi-pated through one or a combination of three major ener-gy pathways or fluxes sensible heat (actual warming of theair) latent heat (evapotranspiration) and ground heat(warming the soil) If increased evapotranspirationdecreases soil moisture and species changes result in bet-ter stomatal control of water loss the result will beincreased sensible heat flux to the atmosphere Foley et al(1994) illustrated the importance of such a change duringthe last temperature optimum in the Holocene approxi-mately 6000 years ago Latitudinal treeline advanced sig-nificantly greatly reducing albedo The increased soil tem-perature and sensible heat flux from this vegetationchange was as great as the solar forcing believed to beresponsible for initiating the climate change Similarlyfurther loss of soil moisture in dry systems will decreaselatent heat flux with more energy dissipated as sensibleheat Current understanding of these types of feedbackson climate change (ie those operating through changesin surface energy and water balance) is less advanced thanthat of feedbacks through greenhouse gas fluxes andshould be a particularly high priority for future research

October 2000 Vol 50 No 10 BioScience 879

Articles

ConclusionsBecause temperature has a direct effect on virtually allecosystem processes responses of terrestrial ecosystems toglobal warming will be even more complex and difficult topredict than responses to increased atmospheric CO

2con-

centrations Nonetheless a useful conceptual frameworkfor analysis comparison and prediction of responses towarming is beginning to emerge (see box this page) Fornow the principal applications of this framework are as astatement of the major dimensions of complexity of theecosystem warming response and as a reminder of themajor constraints on long-term predictions based on rela-tively small-area short-term experimental studies A keyneed is to further develop our understanding of how ini-tial climate biogeochemical states and species character-istics might be used to generate long-term predictions ofrelative sensitivity to climate change Such a frameworkwould serve as a heuristic guide to new research

We also need improved models of the temporalsequence of ecosystem responses to temperature changebecause it is increasingly clear that long-term responses towarming and increased CO

2may differ greatly in both

magnitude and direction from initial responses Ecosys-tem-level experiments are a useful tool for developing thisunderstanding a broader array of experiments in con-trasting ecosystem types is needed including both whole-system manipulations and more focused experimentaltreatments Also needed are the transect studies long-term monitoring and the development of research net-works that will allow spatial extrapolation and validationof predictions based on intensive experimental studies

AcknowledgmentsThis paper contributes to the Global Change and Terres-trial Ecosystems (GCTE) core project of InternationalGeospherendashBiosphere Program (IGBP) and to the activi-ties of the Network of Experimental Warming Studies(GCTE-NEWS) The work was initiated with the supportof the National Center for Ecological Analysis and Synthe-sis a center funded by NSF (Grant DEB-94-21535) theUniversity of CaliforniandashSanta Barbara the CaliforniaResources Agency and the California Environmental Pro-tection Agency Additional support was provided by NSFgrant DEB-9730110 by NASA through a grant to GCTEand by USDA Forest Service

References citedAber J McDowell W Nadelhoffer K McGill A Berntson G Kamaka M

McNulty S Currie W Rustad L Fernandez I 1998 Nitrogen saturation

in forest ecosystems Hypotheses revisited BioScience 48 921ndash934Arft AM et al 1999 Response patterns of tundra plant species to experi-

mental warming A meta-analysis of the International Tundra Experi-

ment Ecological Monographs 69 491ndash512

Arnqvist FR Wooster D 1995 Meta-analysismdashsynthesizing research find-

ings in ecology and evolution Trends in Ecology and Evolution 10236ndash240

880 BioScience October 2000 Vol 50 No 10

Articles

Conceptual framework for analysis of ecosystem response to changing temperature

A Complexity Spatial and temporal variation Global tempera-

ture changes will not be uniform with mostregions warming by varying amounts but otherscooling These changes will occur over a range ofnatural temperature regimes and annual temper-ature variations that are greater than the expect-ed average temperature change

Direct and indirect effects Because temperatureaffects virtually all ecosystem processes indirecttemperature effects resulting from interactionsamong processes are likely to be very importantespecially in the long term

Variable time constants of response Diverseecosystem processes have different time constantsfor response to temperature Thus initial overallecosystem responses to change are not reliableindicators of long-term responses because theyrepresent only part of the total response

B Responses Initial response a function of initial conditions The

same temperature change applied to differentecosystems will elicit different responses depend-ing on initial position on the temperatureresponse surface (direct effects) and on initialbiogeochemical conditions and composition(indirect effects)

Changing sequence of controls The dominant con-trols over ecosystem response to temperature willchange over time as different processes change atdifferent rates This changing sequence will notnecessarily be the same in all ecosystems

Changing direction as well as magnitude ofresponse Initial losses or gains of carbon by theecosystem may be reversed over time as indirecteffects and slowly changing components of theecosystem come into play

Species changes can affect biogeochemistryChanges in species composition and abundancewill affect ecosystem functions (net primary pro-duction carbon and nutrient cycling energyfluxes) through changes in tissue chemistrydemography (both parts and individuals)turnover rates and vegetation structure (physiog-nomy leaf area index)

Response is predictable (at least qualitatively) frominitial conditions Key variables include initialtemperature regime initial carbon and nutrientpool sizes and turnover rates and initial popula-tion dynamics of vegetation

Projections of global warming from General Circula-tion Models (GCMs) are now familiar to both scientistsand nonscientists Knowing however that the mean glob-al temperature will increase by 10ndash35 degC tells us littleabout how temperatures will change in a particular loca-tion or how the ecosystems in that location will respondFor instance temperature increases are likely to be greaterat higher latitudes (Houghton et al 1995 1996) where ini-tial conditions are below 0 degC most of the year and grow-ing season temperatures are only a few degrees higherEven a small increase in average temperature at higher lat-itudes could increase both the length of the unfrozen peri-od and degree-day accumulations by large percentages(Billings 1987) The implications of a given increase intemperature will also depend on the initial temperature ata specific location because the rate of many biologicalprocesses in relation to temperature typically peaks atsome intermediate temperature An increase from an ini-tially low temperature may cause an increase in photosyn-thesis for example while an increase from an initially high

temperature might cause photosyn-thesis to decrease (Larcher 1995)Finally even as the average tempera-ture increases some parts of theglobe are actually expected to experi-ence cooling (Houghton et al 19951996) The complexities of these pat-terns of change in temperature con-trast with the increase in atmospher-ic CO2 for which there is no largetemporal or spatial variation eitherin current levels or in the rate or pat-tern of increase

Direct and indirect effectsof warming onecosystemsThe direct effects of warming onecosystems will also be more com-plex than the direct effects ofincreased CO

2 because temperatureimpacts virtually all chemical andbiological processes whereas thedirect influence of CO

2 is almostentirely limited to leaves (photosyn-thesis stomatal aperture and per-haps respiration Koch and Mooney1996) For both warming and CO

2however the greatest obstacles tounderstanding lie in the web of indi-rect effects resulting from interac-tions among processes affecteddirectly by environmental changeThese interactions lead to feedbacksthat are sometimes positive and

sometimes negative so the responses to temperaturechange or CO

2 can be expected to vary among ecosystemsin both magnitude and direction depending on the prop-erties of the dominant species interactions among speciesand the initial physical and chemical environment

The effects of a temperature increase on the carbonbudget of an ecosystem provides an example of this com-plexity (Figure 1) Net ecosystem production (NEP)defined as the overall carbon balance of an ecosystem oversome time period (Woodwell and Whittaker 1968Mooney et al 1999) has two major components net pri-mary production (NPP) and heterotrophic respiration(R

h) NPP the principal input of carbon to the ecosystemis the net result of CO

2 fixation by photosynthesis andCO

2 loss by plant respiration The product of NPP is neworganic matter which accumulates first in plants as livingbiomass and is eventually transferred to soils as litter andto animals and decomposer organisms as food R

h repre-sents the loss of carbon from the ecosystem by respirationof animals and decomposers the products of R

h include

872 BioScience October 2000 Vol 50 No 10

Articles

Figure 1 Direct and indirect effects of temperature on net primary production (NPP)heterotrophic respiration (Rh) and net ecosystem production (NEP) the majorcomponents of carbon exchange between terrestrial ecosystems and the atmosphere(Woodwell and Whittaker 1968) NPP removes carbon from the atmosphere whereasRh returns carbon to the atmosphere NEP is the sum of these two opposing fluxes andrepresents the net exchange of carbon between an ecosystem and the atmosphereFrom the perspective of the ecosystem NEP is positive when NPP is larger than Rh andis negative when NPP is smaller than Rh In this diagram arrows represent chains ofcause and effect by which temperature affects NPP and Rh but are not intended toindicate the magnitude or sign of an effect Their purpose is simply to show thattemperature influences NPP and Rh directly and indirectly whereas changes in NEPare the net result of the effects on NPP and Rh (SOM soil organic matter)

CO2 and other inorganic carbon products (eg CH4)Both NPP and R

h are affected directly by temperaturechange Both are usually increased by warming althoughR

h often increases more rapidly in the short term (Wood-well 1995) In addition warming can affect NPP and R

h

indirectly by altering the ecosystemrsquos moisture regimenitrogen availability length of its growing season orspecies composition (Figure 1) Warming-driven changesin moisture nitrogen or species composition may alsohave intermediate effects on other ecosystem processes orstates (eg litter quality and quantity which affect both R

h

and nitrogen mineralization) leading to multistep indi-rect effects including losses of carbon through fire orleaching and changes in the balance of NPP and R

h (Melil-lo et al 1990 Chapin et al 1997)

For many ecosystems the indirect effects of a tempera-ture increase on carbon balance are likely to be moreimportant than the direct effects Nutrient-limited tundraand northern forest ecosystems for example are muchmore responsive in the short term (1ndash10 years) to changesin nutrient availability which is likely to increase with soilwarming than to the direct effect of increased tempera-ture (Clark and Rosswall 1981 Chapin 1983 Jonasson andShaver 1999) In many dry ecosystems increased evapora-tive water loss at higher temperatures resulting in dryersoils may strongly limit soil and plant processes so thatpotential temperature-driven increases in process rates arenot achieved (Mooney et al 1999 Saleska et al 1999)Finally changes in temperature may alter speciesrsquo compet-itive interactions and the activity of herbivores andpathogens (Mooney et al 1991 Smith et al 1997deValpine and Harte in press) leading to changes in lightwater and nutrient limitations with complex long-termeffects on carbon turnover NPP and Rh (Herbert et al1999)

Variable time scales of responses towarmingWarming will affect essentially all ecosystem processes andorganic matter pools but at different rates Because theprocesses and pools are linked to each other through bio-geochemical cycles the magnitude and even the directionof the net change in the ecosystem may vary over time Forexample NPP is affected by processes operating over awide range of time scales from the very short-termresponses of leaf-level photosynthesis to the long-termchanges in storage and turnover of soil nitrogen stocks(Figure 2) Although in most ecosystems the immediateeffect of warming may be to increase NPP throughincreased photosynthesis in the longer-term the uptakeand accumulation of nitrogen in plant biomass (particu-larly in forests) might reduce the possibility for furtherincreases in NPP by reducing nitrogen availability Even-tually changes in species composition or litter qualitymight lead to further increases or decreases in NPPthrough changes in plant biomass (and nutrient) turnover

or changes in litter decomposition and nitrogen mineral-ization In some ecosystems (eg boreal forests) theincreasing fuel load associated with biomass increase mayincrease the likelihood of fire and therefore the proportionof stands with low carbon storage on the landscape Thusthe principal long-term mechanisms of regulation of NPPand overall carbon balance with respect to temperaturemay be very different from the short-term mechanisms(McKane et al 1995 Braswell et al 1997 Rastetter et al1997)

The time scales of response of individual processesshould also vary among ecosystems adding further com-plexity to long-term prediction of changes caused bywarming (Figure 2) For example in ecosystems where thedominant species in the vegetation are long-lived treeschanges in species composition might have relativelyminor effects on NPP in the first few decades of warmingbut changes in biomass allocation within the existingdominants might have major effects Principal controlsover nitrogen availability in such forests might lie in con-trols over turnover of existing large soil nitrogen pools Onthe other hand the changes in species composition in anannual grassland with a relatively small soil organic matter

October 2000 Vol 50 No 10 BioScience 873

Articles

Figure 2 Time scales of response to temperature changeby various ecosystem processes and components Each ofthe processes or components shown in the figure affect netecosystem production either directly or indirectly Forconvenience they are grouped into categories vegetationsoils and other The intent is to show how differentprocesses and components respond to temperature changeat different rates hence the overall ecosystem response(the result of the individual responses and theirinteractions) may be very different in the long-termversus the short-term Many other processes andcomponents could be added to this figure (Psphotosynthesis Rs respiration SOM soil organicmatter)

pool could be more important to NPP and changes in soilnitrogen turnover less important within the first decades

Ecosystem warming experiments if continued over sev-eral years can capture much but not all of this changingsequence of responses to temperature change (Figure 2)Changes in basic metabolism (such as photosyntheticacclimation to temperature) may take place over less than1 year and will be missed by annual sampling schedulesbut these short-term responses may have less net impacton long-term change in ecosystem properties than thewarming-induced rearrangement of ecosystem carbonand nutrient stocks and species composition which takesplace on a time scale of 1ndash100 years Some changes espe-cially those occurring over very long time scales (eg soilprofile development and organic matter accumulation) orlarge spatial scales (changes in fire regime or long-distancemovements of large herbivores or timberlines) will not bepicked up by experiments and must be investigated in oth-er ways

Examples of warming responses CasestudiesDespite the complexities of ecosystem responses to warm-ing an important goal of ecosystem science should be topredict which kinds of ecosystems are more or less respon-sive to warming and to identify the characteristics ofecosystems that cause them to be more or less responsiveOur present ability to do so is limited but evidence indi-cates that ecosystem responses to warming are stronglyaffected by initial conditions including

874 BioScience October 2000 Vol 50 No 10

Articles

Figure 3 Changes in net ecosystem production (NEP the net exchange of carbon between an ecosystem and the atmosphere)over time in response to warming There is a net carbon uptake by the ecosystem when NEP is positive when NEP isnegative there is a net carbon loss The different lines represent qualitative differences in response (and predicted long-termresponses) to experimental warming by different ecosystems as discussed in the text Harvard Forest lines A B Toolik Lakelines C D Great Dun Fell lines D E Colorado subalpine meadow line C Abisko line C

Figure 4 Harvard Forest Soil Warming ExperimentMassachusetts showing early spring greening of forestfloor In this experiment electric heating cables areburied 10 cm beneath the soil surface and 10 cm apartalong lines indicated by the string laid out on thesurface Photo Kathy Newkirk

Stocks and initial turnover rates of labile carbon inthe soil

Stocks and initial turnover rates of labile nitrogen inthe soil

Relative size of the carbon pools of plants and soil

Dominant form of available nitrogen in the soil (egorganic nitrogen ammonium nitrate)

Soil water and precipitation regime

Chemical composition and turnover rates of liveplant tissues and litter

Longevity of individuals and population turnoverrates of dominant plant species

The effects of these initial conditions on the response ofecosystems to warming can be illustrated by comparingchanges in NPP NEP or R

h over time in contrastingecosystems that have been manipulated experimentallyThis approach has already led to useful insights as shownby the following examples and illustrated in Figure 3Nitrogen redistribution model (Harvard ForestMassachusetts United States) At Harvard Forest incentral Massachusetts (a temperate deciduous forestecosystem) a long-term soil warming experiment usedelectric heating cables buried in the soil to elevate soil tem-peratures at a depth of 10 cm by 5 degC above ambient (Fig-ure 4 Peterjohn et al 1994 Melillo et al 1995) At this sitethe important initial conditions include

Relatively small labile soil carbon pool

Large slow-turnover vegetation carbon pool

Relatively large soil nitrogen pool

Ammonification-dominated nitrogen mineralizationsequence

These initial conditions constrain how the Harvard Forestecosystem responds to soil warming over time such that athree-phase transient pattern of changes in carbon storageresults (Figure 3 line A Melillo et al in press) DuringPhase I the labile carbon pool is rapidly oxidized bydecomposing organisms leading to a short period of netcarbon loss In Phase II nitrogen is transferred from thesoil which has a low CN ratio to the vegetation whichhas a high CN ratio This redistribution of nitrogenresults in net carbon storage at the ecosystem levelalthough there is a loss of soil carbon (Rastetter et al 19911997) During Phase II there is no significant loss of nitro-gen from the ecosystem because the system has an ammo-nium economy If the system had a nitrate economy or amixed ammoniumndashnitrate economy nitrogen could belost from the system by leaching or denitrification and soreduce the nitrogen available to promote carbon storageEventually the labile and metastable nitrogen pools in thesoil decrease to the point where very little nitrogen can beredistributed from the soil to the plants At this point theecosystem may begin to lose carbon overall (ie NEPbecomes negative) because of increased respiratory lossesfrom plants and soils (Figure 3 line A) This decline may

be reversed if external nitrogen inputs (deposition or fixa-tion) are available or if plant biomass is lost due to distur-bance such as a hurricane or insect attack and is decom-posed releasing nitrogen once again (Figure 3 line B)Nitrogen redistribution with variable moistureregime (Toolik Lake Alaska United States) In amoist tundra ecosystem at Toolik Lake in northern Alaskaresponses to warming were studied using small green-houses to increase air temperature during the growing sea-son by 3ndash5 degC (Figure 5) The soil organic matter pool ofthis ecosystem is larger and contains even more labile car-bon and nitrogen than the soils of many northern tem-perate forests (Giblin et al 1991) Decomposition andnitrogen mineralization in tundra is strongly limited bylow soil temperatures and high soil moisture (Nadelhofferet al 1992) Thus if soil organic matter turnover isincreased due to warming there is a high potential forredistribution of nitrogen from soils (with low CN ratio)to vegetation (with high CN ratio) resulting in net car-bon storage with little or no net change in ecosystemnitrogen stocks (Shaver et al 1992) This response howev-er is strongly interactive with soil moisture with a rela-tively small temperature response if the soil is saturated(Nadelhoffer et al 1992)

Experimental results (Chapin et al 1995 Shaver et al1998) and modeling (McKane et al 1997a 1997b Rastet-ter et al 1997) indicate a multiphase response of NEP towarming in this tundra ecosystem that is quite similar tothat at Harvard Forest as long as soil moisture does notchange greatly (Figure 3 line C) Similar results have alsobeen obtained in the long-term monitoring and modeling

October 2000 Vol 50 No 10 BioScience 875

Articles

Figure 5 The Toolik Lake Field Greenhouse Experimentin moist tussock tundra vegetation Alaska showingmoderately increased abundance of deciduous shrubsand total vegetation mass The photograph was taken in1999 after 11 summers of warming Photo Gus Shaver

studies by Oechel and coworkers (eg Vourlitis andOechel 1997 1999 Waelbroeck et al 1997) The initialresponse to warming is a large increase in R

h leading to adecrease in NEP for the first 1ndash3 years (Phase I) Increasedsoil respiration however is accompanied by increased

nitrogen mineralization leading to increased plant nitro-gen uptake and increased NPP after 3ndash10 years (Phase II)Further increases in NPP are limited by other factors (eglight) Increased NPP leads to increased litter productionwhich eventually leads to higher R

h and NEP returns tonear zero (Phase III) after approximately 50ndash100 years

Changes in NEP in response to warming of moist tun-dra may be very different if there is an interaction between

warming and soil moisture (Billings 1987 McKane et al1997b Vourlitis and Oechel 1999) Warming combinedwith decreased soil moisture will cause a long-term loss ofboth carbon and nitrogen because of large increases in Rh

combined with losses of mineralized nitrogen by drainagefrom the system especially in areas where thaw depthincreases greatly (In such areas nitrogen losses mayinclude large amounts of dissolved organic nitrogen theproduct of incomplete mineralization) In this scenarioincreases in nitrogen uptake and NPP in Phase II areinsufficient to compensate for nitrogen and carbon lossesin leaching and respiration so that even though theecosystem eventually returns to an equilibrium whereNPP equals R

h (NEP = 0) over 50ndash100 years there is a netloss of carbon to the atmosphere (Figure 3 line D)Partial decay of large carbon pool (Great DunFell United Kingdom) In an upland grassland systemat Great Dun Fell in the United Kingdom a combinationof approaches was used to assess the impact of globalwarming on ecosystem carbon stores and soil water quali-ty In one intact vegetationsoil monoliths were moveddown an altitudinal gradient (Ineson et al 1998a) inanother in situ warming using electric heating cables (Ine-son et al 1998b) was applied Both approaches were testedin several different vegetationsoil systems including amineral acidic brown earth and a peaty stagnohumic gleyResults after 3 years of the heating treatments were quali-tatively similar to those predicted for the arctic tundraunder warming combined with reduced soil moisture(Figure 3 line D) but the net carbon losses were even larg-er

All systems at Great Dun Fell showed an increase inplant biomass under the warmer conditions but majorreductions in the total carbon content of the peaty gley(approximately 10 reduction in 3 years) demonstratedthe vulnerability of the carbon stored in this particularsoil The increases in temperature which stimulated themineralization of organic matter and release of nutrientssignificantly improved plant growth but these increases incarbon fixation were insufficient to replace the carbon lostthrough increased decomposition These increases indecomposition were caused by generally greater soil bio-logical activity and in particular by population increasesand changes in vertical distribution of the enchytraeidworm Cognettia sphagnetorum (Briones et al 1998) Theselarge losses in total soil carbon stocks clearly cannot con-tinue eventually the soils are expected to stabilize at alower carbon content analogous to the pattern shown inFigure 3 (line E) and by the native soils at the lower end ofthe same altitudinal gradientSpecies changes affect ecosystem carbon storage(subalpine meadow Colorado United States) Ina subalpine meadow ecosystem in Colorado Harte and col-leagues have used overhead infrared heaters to warm bothsoil and vegetation year-round since 1991 (Figure 6) caus-ing significant changes in NPP and NEP (Harte et al 1995

876 BioScience October 2000 Vol 50 No 10

Articles

Figure 6 Infrared heaters in place above subalpinemeadow vegetation Colorado Photo Taylor Ricketts

Figure 7 Smaller plastic greenhouses and burlap-coveredshade frames at the subarctic heath site Abisko SwedenPhoto Sven Jonasson

Saleska et al 1999) In contrast to the previous examples aprincipal mechanism driving the changes in Colorado is theeffect of the vegetation on species growth rates and func-tional type composition (Harte and Shaw 1995 deValpineand Harte in press) A 12 loss of soil carbon in the top 15cm of the heated plot soils was observed after 4 years oftreatment (Scott Saleska and John Harte University of Cal-iforniandashBerkeley unpublished data) This initial loss of soilorganic matter was not driven by a warming-inducedincrease in heterotrophic respiration because the positiveeffect of a temperature increase on R

h was cancelled by thenegative effect of soil drying Instead this initial loss of soilorganic matter was driven mainly by the effect of warmingon the rate of litter input to the soil In particular warminginduced a shift from more productive forbs to less produc-tive shrubs However litter produced by forbs is more easi-ly decomposed than litter produced by the shrubs whichhas a higher ligninnitrogen ratio Thus the loss of soilorganic matter is likely to be a transient effect with therecalcitrant litter input from shrubs causing an increase insoil organic matter over the long term and ultimately in anoverall pattern of change in NEP similar to that in Figure 3(line C)

By comparing responses observed in the warmingexperiment to changes in microclimate soil organic mat-ter and vegetation species composition along a naturalelevation gradient Harte and colleagues have shown thatthe combined use of manipulations and gradient studiescan provide a unified picture partitioned into short andlong-term components of how climate change influencessoil organic matter The research shows how the impacts ofwarming on carbon cycling can be dominated by the indi-rect effect of a warming-induced shift in relative speciescomposition on the quantity (short term) and quality(longer term) of litter input to the soilExploitation of available space (Abisko Sweden)In low-biomass ecosystems where the surface of the soil ispartially bare warming brought about by increasing plantcover may cause increases in NEP In tundra heath and fellfield ecosystems at Abisko Sweden warming with smallplastic-covered greenhouses caused such an increase inplant cover (Figure 7) Plant biomass driven in part bytemperature-related increases in plant nitrogen uptakeincreased almost twofold within 5 years (Jonasson et al1999)

The response of biomass to warming at Abisko was sim-ilar to that in the tussock tundra at Toolik Lake althoughthe response at Abisko was greater and NPP increasedwithout any pronounced changes in species compositionOverall the response followed the curve represented inFigure 3 (line C) initial declines in carbon stocks were notobserved because the experiment was not sampled imme-diately The longer term (more than 10 years) response isexpected to level off and reach an equilibrium at a net car-bon accumulation rate close to zero when the availablespace has been occupied by plants

Initial increases in biomass in response to warming neednot be a continuous process over time particularly not inthe northernmost ecosystems where the plants operate closeto the limit of their cold tolerance For instance an increasein biomass similar to that at Abisko took place over severalyears in a poorly vegetated high arctic semi-desert howev-er after an unusually mild and rainy growing season muchof the increase was lost in warmed and watered plots Theloss in biomass presumably occurred because winter hard-ening in the experimental plots was delayed and the plantssuffered frost damage when the winter set in (Robinson etal 1998) reflecting a possible effect of stochastic ldquoextremeeventsrdquo Such events are likely to be of great importance inclimatic regimes close to the tolerance limits of plantsUnder such regimes carbon pools in the living plant bio-mass may fluctuate between phases of accumulation andabrupt decline never stabilizing at zero net accumulationwhile soil organic matter is built up stepwise during theperiods of plant die-back

Experiments data and synthesis needsPast warming experiments have used a variety of heatingmethods including electric resistance heating infraredirradiation reciprocal transplants and open- and closed-top field greenhouses (Table 1) Some experiments havewarmed only portions of the ecosystem Others havewarmed the ecosystem for only part of each day or year Inmost cases the choice of warming method is constrainedby logistical and engineering considerations (eg avail-ability of electric power difficulty of enclosing an entireforest and the initial interests of the investigators) No onemethod perfectly simulates the expected climatic changesbut all have yielded useful results that with cautious inter-pretation have begun to reveal similarities differencesand patterns of response among the ecosystems studiedAlthough there is considerable intellectual appeal todesigning future experiments that use a common method-ology and realistic simulation of expected change it also istrue that different warming methods illuminate differentaspects of the overall warming response To optimizeintercomparisons future research should include a mix ofcommon manipulations in diverse ecosystem typesFuture research should also include manipulationsdesigned to provide focused insights into individualecosystems or key processes

Several networks and consortia are already working tomeet the need for a global understanding of ecosystemresponses to climate change (Ingram et al 1999) TheInternational Tundra Experiment (ITEX) uses passivelywarmed open-top chambers in 26 different arctic andalpine tundras to compare the effects of warming on plantgrowth and flowering (Henry and Molau 1997 Arft et al1999) A broader-based Network of Ecosystem WarmingStudies (NEWS) is being developed as part of the GlobalChange and Terrestrial Ecosystems (GCTE) Core Projectof the International GeospherendashBiosphere Program

October 2000 Vol 50 No 10 BioScience 877

Articles

(IGBP) Most warming experiments to date however havetaken place in north temperate boreal and arctic ecosys-tems which are cool and moist and where NPP is strong-ly nutrient-limited Although the processes that dominatethe warming responses of these ecosystems are importanteverywhere they may not dominate in other ecosystemtypes In warm and dry regions including Mediterraneanor desert ecosystems and in tropical humid ecosystemsindirect warming effects acting through changes in evapo-transpiration and soil dryness will play a major regulatoryrole We also might expect the direct effects of warming onNPP to be relatively smaller in these warm systemsbecause temperatures are already near optimum for pho-tosynthesis NPP might even decrease due to overheatingin the warmest ecosystems or microenvironments Addi-tional experiments and networks are needed to expand therange of ecosystems studied including studies of vegeta-tion with contrasting plant life histories (eg annual ver-sus perennial grasslands woodlands and evergreen versusdeciduous vegetation)

New methods are also needed for the quantitative syn-thesis of data from different studies of temperature effectson ecosystems Comparisons among different simulationmodels have also yielded important insights (Schimel et al1997) Some relatively new simulation models are

designed to facilitate comparisons among ecosystems andto predict their responses to temperature change within acommon model structure (eg Rastetter et al 1991Williams et al 1997) New statistical methods can also beused to search for and explain general patterns in diversedata sets For example we might want to know the averageeffects of experimental warming treatments on NPPacross a group of studies and the extent to which thoseeffects differ in biomes at different latitudes or with differ-ent moisture regimes Ecologists have begun to use met-analysis statistics to synthesize experimental data frommany independent studies to answer these kinds of ques-tions (Arnqvist and Wooster 1995 Gurevitch and Hedges1999) including plant responses to warming in ITEX (Arftet al 1999)

Finally there is a need to extrapolate experimentalresults over longer time scales and over larger landscapesand regions An excellent way to do so is to combineexperimentation with measurement of ecosystem proper-ties and processes along environmental transects such asthe IGBPndashGCTE Terrestrial Transects (Steffen et al1999) The transects are assumed to mirror longer-term(decades to centuries) responses of vegetation and soils toenvironmental conditions whereas warming experimentsreflect responses over the shorter-term (years to decades)

878 BioScience October 2000 Vol 50 No 10

Articles

Table 1 Comparison of currently used methods for experimental warming of terrestrial ecosystems

Method Mechanism of warming Advantages Disadvantages Examples

Field greenhouse ldquoGreenhouse warmingrdquo (ie Simple and inexpensive Little or no temperature Chapin et al 1995reflection of reradiated infrared requires no electrical power control and large tempera- Shaver et al 1998energy and reduced advective ture variability other arti- Jonasson et al 1999energy loss) facts include altered light

wind humidity and precipi- tation regimes

Passive Open-top Same as above Same as above Little or no temperature Marion et al 1997Chamber (OTC) control altered wind and Arft et al 1999

humidity only small areas can be manipulated uniformly

Active Open-top Same as above Precise control of air temp- Altered wind regime Norby et al 1997Chamber (OTC) plus warming by advection of erature or temperature humidity and

electrically heated forced difference may be combined evapotranspirationwarm air with CO2 control

Active Soil Warming Warming by conduction from Precise control of soil Altered soil moisture Peterjohn et al 1994buried electrical resistance temperature or temperature regime no effect on Ineson et al 1998bcables difference may be combined aboveground temperatures Hartley et al 1999

with greenhouses or OTCs

Electric infrared heat Warming by increased infrared Precise control of energy Warming depends entirely Harte et al 1995radiation input direct simulation of on radiation no change in Harte and Shaw 1995

global change in energy advective energy inputsbalance

Reciprocal or one-way Transplantation of plants Comparison of relatively Disturbance effects Ineson et al 1998atransplantation soils or whole plantndashsoil natural temperature gradients multiple environmental Grogan and Chapin1999 systems to warmer or cooler changes make it difficult

locations to assign specific causes to responses observed

Combined information from experiments and transectsmay thus be useful in distinguishing relatively short-termtransient changes in the experiments from longer-termnear-equilibrium changes For example global analyses ofprimary production (NPP Figure 1) and litter (the majorcontributor to R

h Figure 1) have shown that both are rea-sonably well-predicted by actual evapotranspiration(Lieth and Whittaker 1975 Meentemeyer 1984)Although short-term responses of NPP and R

h to experi-mental warming may not have the same relationshipswith actual evapotranspiration as do undisturbed ecosys-tems convergence toward the predicted relationshipmight be a useful measure of expected future change inthe experiments

Ultimately experimental and analytical approachesshould be combined with large-scale monitoring of vari-ables and processes that experimental studies have identi-fied as critical indicators of ecosystem change The combi-nation of global monitoring with ongoing experimentaldata acquisition metanalysis modeling and other datasynthesis efforts would provide an early warning system todetect and potentially mitigate these ecosystem responses

Field stations long-term ecological research sites andnature reserves already exist in many relatively unman-aged landscapes Collecting simple baseline data and mon-itoring annually or even less frequently would not be veryexpensive although data quality may be highly variable(eg Stohlgren et al 1995) Efforts to coordinate data andestablish protocols for data sharing might be daunting butthe resulting information would be immensely valuableresponses of species communities and biogeochemicalprocesses to climate change could potentially be detecteddecades earlier than they would be otherwise

Interactions with other drivers of globalchangeChanges in temperature are now occurring simultaneous-ly with other types of global change It will not be possibleto fully understand and predict ecosystem responses totemperature change without taking into account the inter-actions with the other components of global environmen-tal change (Koch and Mooney 1996 Sellers et al 1997Vitousek et al 1997 Luo and Mooney 1999) For exampleincreases in atmospheric CO

2 may modify plant tissuechemistry and ecosystem water balance both of whichmay in turn modify responses to warming Length of thegrowing season may increase with warming or it maydecrease as drought duration increases altering NPP NEPand other processes Many regions of the globe are experi-encing increased levels of nitrogen deposition and thenature of the nitrogen economy of an ecosystem is animportant factor in determining responses to other envi-ronmental factors (Aber et al 1998)

Changes in land use and disturbance regime may alsoinfluence responses to warming For instance land clear-ing over very large areas may modify local temperature or

precipitation regimes beyond what would happen withclimate change alone Similarly changes in fire regimescould be caused by and affect ecosystem responses towarming The spread of invasive species will also be affect-ed by both warming and land use change

Some interactions between warming and other globalchanges are predicted more readily than others For exam-ple outbreaks of insects or pathogens that kill the dominantplants and increase fire probability would lead to dramati-cally different results than the phased changes in response towarming alone We must expect some surprises

Clearly it will not be possible to incorporate all impor-tant dimensions of global change into experiments It willrarely be feasible to include more than one or two factorsin addition to temperature (eg CO2 concentration ornitrogen level) as experimental factors (Schulze et al1999) However through the use of combinations ofexperiments gradient studies remote sensing and simu-lation modeling it should be possible to make substantialprogress toward developing a mechanistic understandingof the most important interactions (Canadell et al 2000)

Although much current global change research focuseson feedbacks from terrestrial ecosystems to climatethrough greenhouse gases important feedbacks also existthrough changes in vegetation structure surface energybalance and water balance Leaves in the canopy act assurfaces for energy and mass exchanges with the atmos-phere and changes in canopy density and transpiration inresponse to warming will change the energy and massfluxes (McFadden et al 1998) One of the most importantand obvious effects of changes in the canopy is its albedo(ie the fraction of incoming solar radiation that isreflected) Decreasing albedo through increased leaf den-sity will increase net radiation (solar radiation absorbedby the system) The increased net radiation must be dissi-pated through one or a combination of three major ener-gy pathways or fluxes sensible heat (actual warming of theair) latent heat (evapotranspiration) and ground heat(warming the soil) If increased evapotranspirationdecreases soil moisture and species changes result in bet-ter stomatal control of water loss the result will beincreased sensible heat flux to the atmosphere Foley et al(1994) illustrated the importance of such a change duringthe last temperature optimum in the Holocene approxi-mately 6000 years ago Latitudinal treeline advanced sig-nificantly greatly reducing albedo The increased soil tem-perature and sensible heat flux from this vegetationchange was as great as the solar forcing believed to beresponsible for initiating the climate change Similarlyfurther loss of soil moisture in dry systems will decreaselatent heat flux with more energy dissipated as sensibleheat Current understanding of these types of feedbackson climate change (ie those operating through changesin surface energy and water balance) is less advanced thanthat of feedbacks through greenhouse gas fluxes andshould be a particularly high priority for future research

October 2000 Vol 50 No 10 BioScience 879

Articles

ConclusionsBecause temperature has a direct effect on virtually allecosystem processes responses of terrestrial ecosystems toglobal warming will be even more complex and difficult topredict than responses to increased atmospheric CO

2con-

centrations Nonetheless a useful conceptual frameworkfor analysis comparison and prediction of responses towarming is beginning to emerge (see box this page) Fornow the principal applications of this framework are as astatement of the major dimensions of complexity of theecosystem warming response and as a reminder of themajor constraints on long-term predictions based on rela-tively small-area short-term experimental studies A keyneed is to further develop our understanding of how ini-tial climate biogeochemical states and species character-istics might be used to generate long-term predictions ofrelative sensitivity to climate change Such a frameworkwould serve as a heuristic guide to new research

We also need improved models of the temporalsequence of ecosystem responses to temperature changebecause it is increasingly clear that long-term responses towarming and increased CO

2may differ greatly in both

magnitude and direction from initial responses Ecosys-tem-level experiments are a useful tool for developing thisunderstanding a broader array of experiments in con-trasting ecosystem types is needed including both whole-system manipulations and more focused experimentaltreatments Also needed are the transect studies long-term monitoring and the development of research net-works that will allow spatial extrapolation and validationof predictions based on intensive experimental studies

AcknowledgmentsThis paper contributes to the Global Change and Terres-trial Ecosystems (GCTE) core project of InternationalGeospherendashBiosphere Program (IGBP) and to the activi-ties of the Network of Experimental Warming Studies(GCTE-NEWS) The work was initiated with the supportof the National Center for Ecological Analysis and Synthe-sis a center funded by NSF (Grant DEB-94-21535) theUniversity of CaliforniandashSanta Barbara the CaliforniaResources Agency and the California Environmental Pro-tection Agency Additional support was provided by NSFgrant DEB-9730110 by NASA through a grant to GCTEand by USDA Forest Service

References citedAber J McDowell W Nadelhoffer K McGill A Berntson G Kamaka M

McNulty S Currie W Rustad L Fernandez I 1998 Nitrogen saturation

in forest ecosystems Hypotheses revisited BioScience 48 921ndash934Arft AM et al 1999 Response patterns of tundra plant species to experi-

mental warming A meta-analysis of the International Tundra Experi-

ment Ecological Monographs 69 491ndash512

Arnqvist FR Wooster D 1995 Meta-analysismdashsynthesizing research find-

ings in ecology and evolution Trends in Ecology and Evolution 10236ndash240

880 BioScience October 2000 Vol 50 No 10

Articles

Conceptual framework for analysis of ecosystem response to changing temperature

A Complexity Spatial and temporal variation Global tempera-

ture changes will not be uniform with mostregions warming by varying amounts but otherscooling These changes will occur over a range ofnatural temperature regimes and annual temper-ature variations that are greater than the expect-ed average temperature change

Direct and indirect effects Because temperatureaffects virtually all ecosystem processes indirecttemperature effects resulting from interactionsamong processes are likely to be very importantespecially in the long term

Variable time constants of response Diverseecosystem processes have different time constantsfor response to temperature Thus initial overallecosystem responses to change are not reliableindicators of long-term responses because theyrepresent only part of the total response

B Responses Initial response a function of initial conditions The

same temperature change applied to differentecosystems will elicit different responses depend-ing on initial position on the temperatureresponse surface (direct effects) and on initialbiogeochemical conditions and composition(indirect effects)

Changing sequence of controls The dominant con-trols over ecosystem response to temperature willchange over time as different processes change atdifferent rates This changing sequence will notnecessarily be the same in all ecosystems

Changing direction as well as magnitude ofresponse Initial losses or gains of carbon by theecosystem may be reversed over time as indirecteffects and slowly changing components of theecosystem come into play

Species changes can affect biogeochemistryChanges in species composition and abundancewill affect ecosystem functions (net primary pro-duction carbon and nutrient cycling energyfluxes) through changes in tissue chemistrydemography (both parts and individuals)turnover rates and vegetation structure (physiog-nomy leaf area index)

Response is predictable (at least qualitatively) frominitial conditions Key variables include initialtemperature regime initial carbon and nutrientpool sizes and turnover rates and initial popula-tion dynamics of vegetation

CO2 and other inorganic carbon products (eg CH4)Both NPP and R

h are affected directly by temperaturechange Both are usually increased by warming althoughR

h often increases more rapidly in the short term (Wood-well 1995) In addition warming can affect NPP and R

h

indirectly by altering the ecosystemrsquos moisture regimenitrogen availability length of its growing season orspecies composition (Figure 1) Warming-driven changesin moisture nitrogen or species composition may alsohave intermediate effects on other ecosystem processes orstates (eg litter quality and quantity which affect both R

h

and nitrogen mineralization) leading to multistep indi-rect effects including losses of carbon through fire orleaching and changes in the balance of NPP and R

h (Melil-lo et al 1990 Chapin et al 1997)

For many ecosystems the indirect effects of a tempera-ture increase on carbon balance are likely to be moreimportant than the direct effects Nutrient-limited tundraand northern forest ecosystems for example are muchmore responsive in the short term (1ndash10 years) to changesin nutrient availability which is likely to increase with soilwarming than to the direct effect of increased tempera-ture (Clark and Rosswall 1981 Chapin 1983 Jonasson andShaver 1999) In many dry ecosystems increased evapora-tive water loss at higher temperatures resulting in dryersoils may strongly limit soil and plant processes so thatpotential temperature-driven increases in process rates arenot achieved (Mooney et al 1999 Saleska et al 1999)Finally changes in temperature may alter speciesrsquo compet-itive interactions and the activity of herbivores andpathogens (Mooney et al 1991 Smith et al 1997deValpine and Harte in press) leading to changes in lightwater and nutrient limitations with complex long-termeffects on carbon turnover NPP and Rh (Herbert et al1999)

Variable time scales of responses towarmingWarming will affect essentially all ecosystem processes andorganic matter pools but at different rates Because theprocesses and pools are linked to each other through bio-geochemical cycles the magnitude and even the directionof the net change in the ecosystem may vary over time Forexample NPP is affected by processes operating over awide range of time scales from the very short-termresponses of leaf-level photosynthesis to the long-termchanges in storage and turnover of soil nitrogen stocks(Figure 2) Although in most ecosystems the immediateeffect of warming may be to increase NPP throughincreased photosynthesis in the longer-term the uptakeand accumulation of nitrogen in plant biomass (particu-larly in forests) might reduce the possibility for furtherincreases in NPP by reducing nitrogen availability Even-tually changes in species composition or litter qualitymight lead to further increases or decreases in NPPthrough changes in plant biomass (and nutrient) turnover

or changes in litter decomposition and nitrogen mineral-ization In some ecosystems (eg boreal forests) theincreasing fuel load associated with biomass increase mayincrease the likelihood of fire and therefore the proportionof stands with low carbon storage on the landscape Thusthe principal long-term mechanisms of regulation of NPPand overall carbon balance with respect to temperaturemay be very different from the short-term mechanisms(McKane et al 1995 Braswell et al 1997 Rastetter et al1997)

The time scales of response of individual processesshould also vary among ecosystems adding further com-plexity to long-term prediction of changes caused bywarming (Figure 2) For example in ecosystems where thedominant species in the vegetation are long-lived treeschanges in species composition might have relativelyminor effects on NPP in the first few decades of warmingbut changes in biomass allocation within the existingdominants might have major effects Principal controlsover nitrogen availability in such forests might lie in con-trols over turnover of existing large soil nitrogen pools Onthe other hand the changes in species composition in anannual grassland with a relatively small soil organic matter

October 2000 Vol 50 No 10 BioScience 873

Articles

Figure 2 Time scales of response to temperature changeby various ecosystem processes and components Each ofthe processes or components shown in the figure affect netecosystem production either directly or indirectly Forconvenience they are grouped into categories vegetationsoils and other The intent is to show how differentprocesses and components respond to temperature changeat different rates hence the overall ecosystem response(the result of the individual responses and theirinteractions) may be very different in the long-termversus the short-term Many other processes andcomponents could be added to this figure (Psphotosynthesis Rs respiration SOM soil organicmatter)

pool could be more important to NPP and changes in soilnitrogen turnover less important within the first decades

Ecosystem warming experiments if continued over sev-eral years can capture much but not all of this changingsequence of responses to temperature change (Figure 2)Changes in basic metabolism (such as photosyntheticacclimation to temperature) may take place over less than1 year and will be missed by annual sampling schedulesbut these short-term responses may have less net impacton long-term change in ecosystem properties than thewarming-induced rearrangement of ecosystem carbonand nutrient stocks and species composition which takesplace on a time scale of 1ndash100 years Some changes espe-cially those occurring over very long time scales (eg soilprofile development and organic matter accumulation) orlarge spatial scales (changes in fire regime or long-distancemovements of large herbivores or timberlines) will not bepicked up by experiments and must be investigated in oth-er ways

Examples of warming responses CasestudiesDespite the complexities of ecosystem responses to warm-ing an important goal of ecosystem science should be topredict which kinds of ecosystems are more or less respon-sive to warming and to identify the characteristics ofecosystems that cause them to be more or less responsiveOur present ability to do so is limited but evidence indi-cates that ecosystem responses to warming are stronglyaffected by initial conditions including

874 BioScience October 2000 Vol 50 No 10

Articles

Figure 3 Changes in net ecosystem production (NEP the net exchange of carbon between an ecosystem and the atmosphere)over time in response to warming There is a net carbon uptake by the ecosystem when NEP is positive when NEP isnegative there is a net carbon loss The different lines represent qualitative differences in response (and predicted long-termresponses) to experimental warming by different ecosystems as discussed in the text Harvard Forest lines A B Toolik Lakelines C D Great Dun Fell lines D E Colorado subalpine meadow line C Abisko line C

Figure 4 Harvard Forest Soil Warming ExperimentMassachusetts showing early spring greening of forestfloor In this experiment electric heating cables areburied 10 cm beneath the soil surface and 10 cm apartalong lines indicated by the string laid out on thesurface Photo Kathy Newkirk

Stocks and initial turnover rates of labile carbon inthe soil

Stocks and initial turnover rates of labile nitrogen inthe soil

Relative size of the carbon pools of plants and soil

Dominant form of available nitrogen in the soil (egorganic nitrogen ammonium nitrate)

Soil water and precipitation regime

Chemical composition and turnover rates of liveplant tissues and litter

Longevity of individuals and population turnoverrates of dominant plant species

The effects of these initial conditions on the response ofecosystems to warming can be illustrated by comparingchanges in NPP NEP or R

h over time in contrastingecosystems that have been manipulated experimentallyThis approach has already led to useful insights as shownby the following examples and illustrated in Figure 3Nitrogen redistribution model (Harvard ForestMassachusetts United States) At Harvard Forest incentral Massachusetts (a temperate deciduous forestecosystem) a long-term soil warming experiment usedelectric heating cables buried in the soil to elevate soil tem-peratures at a depth of 10 cm by 5 degC above ambient (Fig-ure 4 Peterjohn et al 1994 Melillo et al 1995) At this sitethe important initial conditions include

Relatively small labile soil carbon pool

Large slow-turnover vegetation carbon pool

Relatively large soil nitrogen pool

Ammonification-dominated nitrogen mineralizationsequence

These initial conditions constrain how the Harvard Forestecosystem responds to soil warming over time such that athree-phase transient pattern of changes in carbon storageresults (Figure 3 line A Melillo et al in press) DuringPhase I the labile carbon pool is rapidly oxidized bydecomposing organisms leading to a short period of netcarbon loss In Phase II nitrogen is transferred from thesoil which has a low CN ratio to the vegetation whichhas a high CN ratio This redistribution of nitrogenresults in net carbon storage at the ecosystem levelalthough there is a loss of soil carbon (Rastetter et al 19911997) During Phase II there is no significant loss of nitro-gen from the ecosystem because the system has an ammo-nium economy If the system had a nitrate economy or amixed ammoniumndashnitrate economy nitrogen could belost from the system by leaching or denitrification and soreduce the nitrogen available to promote carbon storageEventually the labile and metastable nitrogen pools in thesoil decrease to the point where very little nitrogen can beredistributed from the soil to the plants At this point theecosystem may begin to lose carbon overall (ie NEPbecomes negative) because of increased respiratory lossesfrom plants and soils (Figure 3 line A) This decline may

be reversed if external nitrogen inputs (deposition or fixa-tion) are available or if plant biomass is lost due to distur-bance such as a hurricane or insect attack and is decom-posed releasing nitrogen once again (Figure 3 line B)Nitrogen redistribution with variable moistureregime (Toolik Lake Alaska United States) In amoist tundra ecosystem at Toolik Lake in northern Alaskaresponses to warming were studied using small green-houses to increase air temperature during the growing sea-son by 3ndash5 degC (Figure 5) The soil organic matter pool ofthis ecosystem is larger and contains even more labile car-bon and nitrogen than the soils of many northern tem-perate forests (Giblin et al 1991) Decomposition andnitrogen mineralization in tundra is strongly limited bylow soil temperatures and high soil moisture (Nadelhofferet al 1992) Thus if soil organic matter turnover isincreased due to warming there is a high potential forredistribution of nitrogen from soils (with low CN ratio)to vegetation (with high CN ratio) resulting in net car-bon storage with little or no net change in ecosystemnitrogen stocks (Shaver et al 1992) This response howev-er is strongly interactive with soil moisture with a rela-tively small temperature response if the soil is saturated(Nadelhoffer et al 1992)

Experimental results (Chapin et al 1995 Shaver et al1998) and modeling (McKane et al 1997a 1997b Rastet-ter et al 1997) indicate a multiphase response of NEP towarming in this tundra ecosystem that is quite similar tothat at Harvard Forest as long as soil moisture does notchange greatly (Figure 3 line C) Similar results have alsobeen obtained in the long-term monitoring and modeling

October 2000 Vol 50 No 10 BioScience 875

Articles

Figure 5 The Toolik Lake Field Greenhouse Experimentin moist tussock tundra vegetation Alaska showingmoderately increased abundance of deciduous shrubsand total vegetation mass The photograph was taken in1999 after 11 summers of warming Photo Gus Shaver

studies by Oechel and coworkers (eg Vourlitis andOechel 1997 1999 Waelbroeck et al 1997) The initialresponse to warming is a large increase in R

h leading to adecrease in NEP for the first 1ndash3 years (Phase I) Increasedsoil respiration however is accompanied by increased

nitrogen mineralization leading to increased plant nitro-gen uptake and increased NPP after 3ndash10 years (Phase II)Further increases in NPP are limited by other factors (eglight) Increased NPP leads to increased litter productionwhich eventually leads to higher R

h and NEP returns tonear zero (Phase III) after approximately 50ndash100 years

Changes in NEP in response to warming of moist tun-dra may be very different if there is an interaction between

warming and soil moisture (Billings 1987 McKane et al1997b Vourlitis and Oechel 1999) Warming combinedwith decreased soil moisture will cause a long-term loss ofboth carbon and nitrogen because of large increases in Rh

combined with losses of mineralized nitrogen by drainagefrom the system especially in areas where thaw depthincreases greatly (In such areas nitrogen losses mayinclude large amounts of dissolved organic nitrogen theproduct of incomplete mineralization) In this scenarioincreases in nitrogen uptake and NPP in Phase II areinsufficient to compensate for nitrogen and carbon lossesin leaching and respiration so that even though theecosystem eventually returns to an equilibrium whereNPP equals R

h (NEP = 0) over 50ndash100 years there is a netloss of carbon to the atmosphere (Figure 3 line D)Partial decay of large carbon pool (Great DunFell United Kingdom) In an upland grassland systemat Great Dun Fell in the United Kingdom a combinationof approaches was used to assess the impact of globalwarming on ecosystem carbon stores and soil water quali-ty In one intact vegetationsoil monoliths were moveddown an altitudinal gradient (Ineson et al 1998a) inanother in situ warming using electric heating cables (Ine-son et al 1998b) was applied Both approaches were testedin several different vegetationsoil systems including amineral acidic brown earth and a peaty stagnohumic gleyResults after 3 years of the heating treatments were quali-tatively similar to those predicted for the arctic tundraunder warming combined with reduced soil moisture(Figure 3 line D) but the net carbon losses were even larg-er

All systems at Great Dun Fell showed an increase inplant biomass under the warmer conditions but majorreductions in the total carbon content of the peaty gley(approximately 10 reduction in 3 years) demonstratedthe vulnerability of the carbon stored in this particularsoil The increases in temperature which stimulated themineralization of organic matter and release of nutrientssignificantly improved plant growth but these increases incarbon fixation were insufficient to replace the carbon lostthrough increased decomposition These increases indecomposition were caused by generally greater soil bio-logical activity and in particular by population increasesand changes in vertical distribution of the enchytraeidworm Cognettia sphagnetorum (Briones et al 1998) Theselarge losses in total soil carbon stocks clearly cannot con-tinue eventually the soils are expected to stabilize at alower carbon content analogous to the pattern shown inFigure 3 (line E) and by the native soils at the lower end ofthe same altitudinal gradientSpecies changes affect ecosystem carbon storage(subalpine meadow Colorado United States) Ina subalpine meadow ecosystem in Colorado Harte and col-leagues have used overhead infrared heaters to warm bothsoil and vegetation year-round since 1991 (Figure 6) caus-ing significant changes in NPP and NEP (Harte et al 1995

876 BioScience October 2000 Vol 50 No 10

Articles

Figure 6 Infrared heaters in place above subalpinemeadow vegetation Colorado Photo Taylor Ricketts

Figure 7 Smaller plastic greenhouses and burlap-coveredshade frames at the subarctic heath site Abisko SwedenPhoto Sven Jonasson

Saleska et al 1999) In contrast to the previous examples aprincipal mechanism driving the changes in Colorado is theeffect of the vegetation on species growth rates and func-tional type composition (Harte and Shaw 1995 deValpineand Harte in press) A 12 loss of soil carbon in the top 15cm of the heated plot soils was observed after 4 years oftreatment (Scott Saleska and John Harte University of Cal-iforniandashBerkeley unpublished data) This initial loss of soilorganic matter was not driven by a warming-inducedincrease in heterotrophic respiration because the positiveeffect of a temperature increase on R

h was cancelled by thenegative effect of soil drying Instead this initial loss of soilorganic matter was driven mainly by the effect of warmingon the rate of litter input to the soil In particular warminginduced a shift from more productive forbs to less produc-tive shrubs However litter produced by forbs is more easi-ly decomposed than litter produced by the shrubs whichhas a higher ligninnitrogen ratio Thus the loss of soilorganic matter is likely to be a transient effect with therecalcitrant litter input from shrubs causing an increase insoil organic matter over the long term and ultimately in anoverall pattern of change in NEP similar to that in Figure 3(line C)

By comparing responses observed in the warmingexperiment to changes in microclimate soil organic mat-ter and vegetation species composition along a naturalelevation gradient Harte and colleagues have shown thatthe combined use of manipulations and gradient studiescan provide a unified picture partitioned into short andlong-term components of how climate change influencessoil organic matter The research shows how the impacts ofwarming on carbon cycling can be dominated by the indi-rect effect of a warming-induced shift in relative speciescomposition on the quantity (short term) and quality(longer term) of litter input to the soilExploitation of available space (Abisko Sweden)In low-biomass ecosystems where the surface of the soil ispartially bare warming brought about by increasing plantcover may cause increases in NEP In tundra heath and fellfield ecosystems at Abisko Sweden warming with smallplastic-covered greenhouses caused such an increase inplant cover (Figure 7) Plant biomass driven in part bytemperature-related increases in plant nitrogen uptakeincreased almost twofold within 5 years (Jonasson et al1999)

The response of biomass to warming at Abisko was sim-ilar to that in the tussock tundra at Toolik Lake althoughthe response at Abisko was greater and NPP increasedwithout any pronounced changes in species compositionOverall the response followed the curve represented inFigure 3 (line C) initial declines in carbon stocks were notobserved because the experiment was not sampled imme-diately The longer term (more than 10 years) response isexpected to level off and reach an equilibrium at a net car-bon accumulation rate close to zero when the availablespace has been occupied by plants

Initial increases in biomass in response to warming neednot be a continuous process over time particularly not inthe northernmost ecosystems where the plants operate closeto the limit of their cold tolerance For instance an increasein biomass similar to that at Abisko took place over severalyears in a poorly vegetated high arctic semi-desert howev-er after an unusually mild and rainy growing season muchof the increase was lost in warmed and watered plots Theloss in biomass presumably occurred because winter hard-ening in the experimental plots was delayed and the plantssuffered frost damage when the winter set in (Robinson etal 1998) reflecting a possible effect of stochastic ldquoextremeeventsrdquo Such events are likely to be of great importance inclimatic regimes close to the tolerance limits of plantsUnder such regimes carbon pools in the living plant bio-mass may fluctuate between phases of accumulation andabrupt decline never stabilizing at zero net accumulationwhile soil organic matter is built up stepwise during theperiods of plant die-back

Experiments data and synthesis needsPast warming experiments have used a variety of heatingmethods including electric resistance heating infraredirradiation reciprocal transplants and open- and closed-top field greenhouses (Table 1) Some experiments havewarmed only portions of the ecosystem Others havewarmed the ecosystem for only part of each day or year Inmost cases the choice of warming method is constrainedby logistical and engineering considerations (eg avail-ability of electric power difficulty of enclosing an entireforest and the initial interests of the investigators) No onemethod perfectly simulates the expected climatic changesbut all have yielded useful results that with cautious inter-pretation have begun to reveal similarities differencesand patterns of response among the ecosystems studiedAlthough there is considerable intellectual appeal todesigning future experiments that use a common method-ology and realistic simulation of expected change it also istrue that different warming methods illuminate differentaspects of the overall warming response To optimizeintercomparisons future research should include a mix ofcommon manipulations in diverse ecosystem typesFuture research should also include manipulationsdesigned to provide focused insights into individualecosystems or key processes

Several networks and consortia are already working tomeet the need for a global understanding of ecosystemresponses to climate change (Ingram et al 1999) TheInternational Tundra Experiment (ITEX) uses passivelywarmed open-top chambers in 26 different arctic andalpine tundras to compare the effects of warming on plantgrowth and flowering (Henry and Molau 1997 Arft et al1999) A broader-based Network of Ecosystem WarmingStudies (NEWS) is being developed as part of the GlobalChange and Terrestrial Ecosystems (GCTE) Core Projectof the International GeospherendashBiosphere Program

October 2000 Vol 50 No 10 BioScience 877

Articles

(IGBP) Most warming experiments to date however havetaken place in north temperate boreal and arctic ecosys-tems which are cool and moist and where NPP is strong-ly nutrient-limited Although the processes that dominatethe warming responses of these ecosystems are importanteverywhere they may not dominate in other ecosystemtypes In warm and dry regions including Mediterraneanor desert ecosystems and in tropical humid ecosystemsindirect warming effects acting through changes in evapo-transpiration and soil dryness will play a major regulatoryrole We also might expect the direct effects of warming onNPP to be relatively smaller in these warm systemsbecause temperatures are already near optimum for pho-tosynthesis NPP might even decrease due to overheatingin the warmest ecosystems or microenvironments Addi-tional experiments and networks are needed to expand therange of ecosystems studied including studies of vegeta-tion with contrasting plant life histories (eg annual ver-sus perennial grasslands woodlands and evergreen versusdeciduous vegetation)

New methods are also needed for the quantitative syn-thesis of data from different studies of temperature effectson ecosystems Comparisons among different simulationmodels have also yielded important insights (Schimel et al1997) Some relatively new simulation models are

designed to facilitate comparisons among ecosystems andto predict their responses to temperature change within acommon model structure (eg Rastetter et al 1991Williams et al 1997) New statistical methods can also beused to search for and explain general patterns in diversedata sets For example we might want to know the averageeffects of experimental warming treatments on NPPacross a group of studies and the extent to which thoseeffects differ in biomes at different latitudes or with differ-ent moisture regimes Ecologists have begun to use met-analysis statistics to synthesize experimental data frommany independent studies to answer these kinds of ques-tions (Arnqvist and Wooster 1995 Gurevitch and Hedges1999) including plant responses to warming in ITEX (Arftet al 1999)

Finally there is a need to extrapolate experimentalresults over longer time scales and over larger landscapesand regions An excellent way to do so is to combineexperimentation with measurement of ecosystem proper-ties and processes along environmental transects such asthe IGBPndashGCTE Terrestrial Transects (Steffen et al1999) The transects are assumed to mirror longer-term(decades to centuries) responses of vegetation and soils toenvironmental conditions whereas warming experimentsreflect responses over the shorter-term (years to decades)

878 BioScience October 2000 Vol 50 No 10

Articles

Table 1 Comparison of currently used methods for experimental warming of terrestrial ecosystems

Method Mechanism of warming Advantages Disadvantages Examples

Field greenhouse ldquoGreenhouse warmingrdquo (ie Simple and inexpensive Little or no temperature Chapin et al 1995reflection of reradiated infrared requires no electrical power control and large tempera- Shaver et al 1998energy and reduced advective ture variability other arti- Jonasson et al 1999energy loss) facts include altered light

wind humidity and precipi- tation regimes

Passive Open-top Same as above Same as above Little or no temperature Marion et al 1997Chamber (OTC) control altered wind and Arft et al 1999

humidity only small areas can be manipulated uniformly

Active Open-top Same as above Precise control of air temp- Altered wind regime Norby et al 1997Chamber (OTC) plus warming by advection of erature or temperature humidity and

electrically heated forced difference may be combined evapotranspirationwarm air with CO2 control

Active Soil Warming Warming by conduction from Precise control of soil Altered soil moisture Peterjohn et al 1994buried electrical resistance temperature or temperature regime no effect on Ineson et al 1998bcables difference may be combined aboveground temperatures Hartley et al 1999

with greenhouses or OTCs

Electric infrared heat Warming by increased infrared Precise control of energy Warming depends entirely Harte et al 1995radiation input direct simulation of on radiation no change in Harte and Shaw 1995

global change in energy advective energy inputsbalance

Reciprocal or one-way Transplantation of plants Comparison of relatively Disturbance effects Ineson et al 1998atransplantation soils or whole plantndashsoil natural temperature gradients multiple environmental Grogan and Chapin1999 systems to warmer or cooler changes make it difficult

locations to assign specific causes to responses observed

Combined information from experiments and transectsmay thus be useful in distinguishing relatively short-termtransient changes in the experiments from longer-termnear-equilibrium changes For example global analyses ofprimary production (NPP Figure 1) and litter (the majorcontributor to R

h Figure 1) have shown that both are rea-sonably well-predicted by actual evapotranspiration(Lieth and Whittaker 1975 Meentemeyer 1984)Although short-term responses of NPP and R

h to experi-mental warming may not have the same relationshipswith actual evapotranspiration as do undisturbed ecosys-tems convergence toward the predicted relationshipmight be a useful measure of expected future change inthe experiments

Ultimately experimental and analytical approachesshould be combined with large-scale monitoring of vari-ables and processes that experimental studies have identi-fied as critical indicators of ecosystem change The combi-nation of global monitoring with ongoing experimentaldata acquisition metanalysis modeling and other datasynthesis efforts would provide an early warning system todetect and potentially mitigate these ecosystem responses

Field stations long-term ecological research sites andnature reserves already exist in many relatively unman-aged landscapes Collecting simple baseline data and mon-itoring annually or even less frequently would not be veryexpensive although data quality may be highly variable(eg Stohlgren et al 1995) Efforts to coordinate data andestablish protocols for data sharing might be daunting butthe resulting information would be immensely valuableresponses of species communities and biogeochemicalprocesses to climate change could potentially be detecteddecades earlier than they would be otherwise

Interactions with other drivers of globalchangeChanges in temperature are now occurring simultaneous-ly with other types of global change It will not be possibleto fully understand and predict ecosystem responses totemperature change without taking into account the inter-actions with the other components of global environmen-tal change (Koch and Mooney 1996 Sellers et al 1997Vitousek et al 1997 Luo and Mooney 1999) For exampleincreases in atmospheric CO

2 may modify plant tissuechemistry and ecosystem water balance both of whichmay in turn modify responses to warming Length of thegrowing season may increase with warming or it maydecrease as drought duration increases altering NPP NEPand other processes Many regions of the globe are experi-encing increased levels of nitrogen deposition and thenature of the nitrogen economy of an ecosystem is animportant factor in determining responses to other envi-ronmental factors (Aber et al 1998)

Changes in land use and disturbance regime may alsoinfluence responses to warming For instance land clear-ing over very large areas may modify local temperature or

precipitation regimes beyond what would happen withclimate change alone Similarly changes in fire regimescould be caused by and affect ecosystem responses towarming The spread of invasive species will also be affect-ed by both warming and land use change

Some interactions between warming and other globalchanges are predicted more readily than others For exam-ple outbreaks of insects or pathogens that kill the dominantplants and increase fire probability would lead to dramati-cally different results than the phased changes in response towarming alone We must expect some surprises

Clearly it will not be possible to incorporate all impor-tant dimensions of global change into experiments It willrarely be feasible to include more than one or two factorsin addition to temperature (eg CO2 concentration ornitrogen level) as experimental factors (Schulze et al1999) However through the use of combinations ofexperiments gradient studies remote sensing and simu-lation modeling it should be possible to make substantialprogress toward developing a mechanistic understandingof the most important interactions (Canadell et al 2000)

Although much current global change research focuseson feedbacks from terrestrial ecosystems to climatethrough greenhouse gases important feedbacks also existthrough changes in vegetation structure surface energybalance and water balance Leaves in the canopy act assurfaces for energy and mass exchanges with the atmos-phere and changes in canopy density and transpiration inresponse to warming will change the energy and massfluxes (McFadden et al 1998) One of the most importantand obvious effects of changes in the canopy is its albedo(ie the fraction of incoming solar radiation that isreflected) Decreasing albedo through increased leaf den-sity will increase net radiation (solar radiation absorbedby the system) The increased net radiation must be dissi-pated through one or a combination of three major ener-gy pathways or fluxes sensible heat (actual warming of theair) latent heat (evapotranspiration) and ground heat(warming the soil) If increased evapotranspirationdecreases soil moisture and species changes result in bet-ter stomatal control of water loss the result will beincreased sensible heat flux to the atmosphere Foley et al(1994) illustrated the importance of such a change duringthe last temperature optimum in the Holocene approxi-mately 6000 years ago Latitudinal treeline advanced sig-nificantly greatly reducing albedo The increased soil tem-perature and sensible heat flux from this vegetationchange was as great as the solar forcing believed to beresponsible for initiating the climate change Similarlyfurther loss of soil moisture in dry systems will decreaselatent heat flux with more energy dissipated as sensibleheat Current understanding of these types of feedbackson climate change (ie those operating through changesin surface energy and water balance) is less advanced thanthat of feedbacks through greenhouse gas fluxes andshould be a particularly high priority for future research

October 2000 Vol 50 No 10 BioScience 879

Articles

ConclusionsBecause temperature has a direct effect on virtually allecosystem processes responses of terrestrial ecosystems toglobal warming will be even more complex and difficult topredict than responses to increased atmospheric CO

2con-

centrations Nonetheless a useful conceptual frameworkfor analysis comparison and prediction of responses towarming is beginning to emerge (see box this page) Fornow the principal applications of this framework are as astatement of the major dimensions of complexity of theecosystem warming response and as a reminder of themajor constraints on long-term predictions based on rela-tively small-area short-term experimental studies A keyneed is to further develop our understanding of how ini-tial climate biogeochemical states and species character-istics might be used to generate long-term predictions ofrelative sensitivity to climate change Such a frameworkwould serve as a heuristic guide to new research

We also need improved models of the temporalsequence of ecosystem responses to temperature changebecause it is increasingly clear that long-term responses towarming and increased CO

2may differ greatly in both

magnitude and direction from initial responses Ecosys-tem-level experiments are a useful tool for developing thisunderstanding a broader array of experiments in con-trasting ecosystem types is needed including both whole-system manipulations and more focused experimentaltreatments Also needed are the transect studies long-term monitoring and the development of research net-works that will allow spatial extrapolation and validationof predictions based on intensive experimental studies

AcknowledgmentsThis paper contributes to the Global Change and Terres-trial Ecosystems (GCTE) core project of InternationalGeospherendashBiosphere Program (IGBP) and to the activi-ties of the Network of Experimental Warming Studies(GCTE-NEWS) The work was initiated with the supportof the National Center for Ecological Analysis and Synthe-sis a center funded by NSF (Grant DEB-94-21535) theUniversity of CaliforniandashSanta Barbara the CaliforniaResources Agency and the California Environmental Pro-tection Agency Additional support was provided by NSFgrant DEB-9730110 by NASA through a grant to GCTEand by USDA Forest Service

References citedAber J McDowell W Nadelhoffer K McGill A Berntson G Kamaka M

McNulty S Currie W Rustad L Fernandez I 1998 Nitrogen saturation

in forest ecosystems Hypotheses revisited BioScience 48 921ndash934Arft AM et al 1999 Response patterns of tundra plant species to experi-

mental warming A meta-analysis of the International Tundra Experi-

ment Ecological Monographs 69 491ndash512

Arnqvist FR Wooster D 1995 Meta-analysismdashsynthesizing research find-

ings in ecology and evolution Trends in Ecology and Evolution 10236ndash240

880 BioScience October 2000 Vol 50 No 10

Articles

Conceptual framework for analysis of ecosystem response to changing temperature

A Complexity Spatial and temporal variation Global tempera-

ture changes will not be uniform with mostregions warming by varying amounts but otherscooling These changes will occur over a range ofnatural temperature regimes and annual temper-ature variations that are greater than the expect-ed average temperature change

Direct and indirect effects Because temperatureaffects virtually all ecosystem processes indirecttemperature effects resulting from interactionsamong processes are likely to be very importantespecially in the long term

Variable time constants of response Diverseecosystem processes have different time constantsfor response to temperature Thus initial overallecosystem responses to change are not reliableindicators of long-term responses because theyrepresent only part of the total response

B Responses Initial response a function of initial conditions The

same temperature change applied to differentecosystems will elicit different responses depend-ing on initial position on the temperatureresponse surface (direct effects) and on initialbiogeochemical conditions and composition(indirect effects)

Changing sequence of controls The dominant con-trols over ecosystem response to temperature willchange over time as different processes change atdifferent rates This changing sequence will notnecessarily be the same in all ecosystems

Changing direction as well as magnitude ofresponse Initial losses or gains of carbon by theecosystem may be reversed over time as indirecteffects and slowly changing components of theecosystem come into play

Species changes can affect biogeochemistryChanges in species composition and abundancewill affect ecosystem functions (net primary pro-duction carbon and nutrient cycling energyfluxes) through changes in tissue chemistrydemography (both parts and individuals)turnover rates and vegetation structure (physiog-nomy leaf area index)

Response is predictable (at least qualitatively) frominitial conditions Key variables include initialtemperature regime initial carbon and nutrientpool sizes and turnover rates and initial popula-tion dynamics of vegetation

pool could be more important to NPP and changes in soilnitrogen turnover less important within the first decades

Ecosystem warming experiments if continued over sev-eral years can capture much but not all of this changingsequence of responses to temperature change (Figure 2)Changes in basic metabolism (such as photosyntheticacclimation to temperature) may take place over less than1 year and will be missed by annual sampling schedulesbut these short-term responses may have less net impacton long-term change in ecosystem properties than thewarming-induced rearrangement of ecosystem carbonand nutrient stocks and species composition which takesplace on a time scale of 1ndash100 years Some changes espe-cially those occurring over very long time scales (eg soilprofile development and organic matter accumulation) orlarge spatial scales (changes in fire regime or long-distancemovements of large herbivores or timberlines) will not bepicked up by experiments and must be investigated in oth-er ways

Examples of warming responses CasestudiesDespite the complexities of ecosystem responses to warm-ing an important goal of ecosystem science should be topredict which kinds of ecosystems are more or less respon-sive to warming and to identify the characteristics ofecosystems that cause them to be more or less responsiveOur present ability to do so is limited but evidence indi-cates that ecosystem responses to warming are stronglyaffected by initial conditions including

874 BioScience October 2000 Vol 50 No 10

Articles

Figure 3 Changes in net ecosystem production (NEP the net exchange of carbon between an ecosystem and the atmosphere)over time in response to warming There is a net carbon uptake by the ecosystem when NEP is positive when NEP isnegative there is a net carbon loss The different lines represent qualitative differences in response (and predicted long-termresponses) to experimental warming by different ecosystems as discussed in the text Harvard Forest lines A B Toolik Lakelines C D Great Dun Fell lines D E Colorado subalpine meadow line C Abisko line C

Figure 4 Harvard Forest Soil Warming ExperimentMassachusetts showing early spring greening of forestfloor In this experiment electric heating cables areburied 10 cm beneath the soil surface and 10 cm apartalong lines indicated by the string laid out on thesurface Photo Kathy Newkirk

Stocks and initial turnover rates of labile carbon inthe soil

Stocks and initial turnover rates of labile nitrogen inthe soil

Relative size of the carbon pools of plants and soil

Dominant form of available nitrogen in the soil (egorganic nitrogen ammonium nitrate)

Soil water and precipitation regime

Chemical composition and turnover rates of liveplant tissues and litter

Longevity of individuals and population turnoverrates of dominant plant species

The effects of these initial conditions on the response ofecosystems to warming can be illustrated by comparingchanges in NPP NEP or R

h over time in contrastingecosystems that have been manipulated experimentallyThis approach has already led to useful insights as shownby the following examples and illustrated in Figure 3Nitrogen redistribution model (Harvard ForestMassachusetts United States) At Harvard Forest incentral Massachusetts (a temperate deciduous forestecosystem) a long-term soil warming experiment usedelectric heating cables buried in the soil to elevate soil tem-peratures at a depth of 10 cm by 5 degC above ambient (Fig-ure 4 Peterjohn et al 1994 Melillo et al 1995) At this sitethe important initial conditions include

Relatively small labile soil carbon pool

Large slow-turnover vegetation carbon pool

Relatively large soil nitrogen pool

Ammonification-dominated nitrogen mineralizationsequence

These initial conditions constrain how the Harvard Forestecosystem responds to soil warming over time such that athree-phase transient pattern of changes in carbon storageresults (Figure 3 line A Melillo et al in press) DuringPhase I the labile carbon pool is rapidly oxidized bydecomposing organisms leading to a short period of netcarbon loss In Phase II nitrogen is transferred from thesoil which has a low CN ratio to the vegetation whichhas a high CN ratio This redistribution of nitrogenresults in net carbon storage at the ecosystem levelalthough there is a loss of soil carbon (Rastetter et al 19911997) During Phase II there is no significant loss of nitro-gen from the ecosystem because the system has an ammo-nium economy If the system had a nitrate economy or amixed ammoniumndashnitrate economy nitrogen could belost from the system by leaching or denitrification and soreduce the nitrogen available to promote carbon storageEventually the labile and metastable nitrogen pools in thesoil decrease to the point where very little nitrogen can beredistributed from the soil to the plants At this point theecosystem may begin to lose carbon overall (ie NEPbecomes negative) because of increased respiratory lossesfrom plants and soils (Figure 3 line A) This decline may

be reversed if external nitrogen inputs (deposition or fixa-tion) are available or if plant biomass is lost due to distur-bance such as a hurricane or insect attack and is decom-posed releasing nitrogen once again (Figure 3 line B)Nitrogen redistribution with variable moistureregime (Toolik Lake Alaska United States) In amoist tundra ecosystem at Toolik Lake in northern Alaskaresponses to warming were studied using small green-houses to increase air temperature during the growing sea-son by 3ndash5 degC (Figure 5) The soil organic matter pool ofthis ecosystem is larger and contains even more labile car-bon and nitrogen than the soils of many northern tem-perate forests (Giblin et al 1991) Decomposition andnitrogen mineralization in tundra is strongly limited bylow soil temperatures and high soil moisture (Nadelhofferet al 1992) Thus if soil organic matter turnover isincreased due to warming there is a high potential forredistribution of nitrogen from soils (with low CN ratio)to vegetation (with high CN ratio) resulting in net car-bon storage with little or no net change in ecosystemnitrogen stocks (Shaver et al 1992) This response howev-er is strongly interactive with soil moisture with a rela-tively small temperature response if the soil is saturated(Nadelhoffer et al 1992)

Experimental results (Chapin et al 1995 Shaver et al1998) and modeling (McKane et al 1997a 1997b Rastet-ter et al 1997) indicate a multiphase response of NEP towarming in this tundra ecosystem that is quite similar tothat at Harvard Forest as long as soil moisture does notchange greatly (Figure 3 line C) Similar results have alsobeen obtained in the long-term monitoring and modeling

October 2000 Vol 50 No 10 BioScience 875

Articles

Figure 5 The Toolik Lake Field Greenhouse Experimentin moist tussock tundra vegetation Alaska showingmoderately increased abundance of deciduous shrubsand total vegetation mass The photograph was taken in1999 after 11 summers of warming Photo Gus Shaver

studies by Oechel and coworkers (eg Vourlitis andOechel 1997 1999 Waelbroeck et al 1997) The initialresponse to warming is a large increase in R

h leading to adecrease in NEP for the first 1ndash3 years (Phase I) Increasedsoil respiration however is accompanied by increased

nitrogen mineralization leading to increased plant nitro-gen uptake and increased NPP after 3ndash10 years (Phase II)Further increases in NPP are limited by other factors (eglight) Increased NPP leads to increased litter productionwhich eventually leads to higher R

h and NEP returns tonear zero (Phase III) after approximately 50ndash100 years

Changes in NEP in response to warming of moist tun-dra may be very different if there is an interaction between

warming and soil moisture (Billings 1987 McKane et al1997b Vourlitis and Oechel 1999) Warming combinedwith decreased soil moisture will cause a long-term loss ofboth carbon and nitrogen because of large increases in Rh

combined with losses of mineralized nitrogen by drainagefrom the system especially in areas where thaw depthincreases greatly (In such areas nitrogen losses mayinclude large amounts of dissolved organic nitrogen theproduct of incomplete mineralization) In this scenarioincreases in nitrogen uptake and NPP in Phase II areinsufficient to compensate for nitrogen and carbon lossesin leaching and respiration so that even though theecosystem eventually returns to an equilibrium whereNPP equals R

h (NEP = 0) over 50ndash100 years there is a netloss of carbon to the atmosphere (Figure 3 line D)Partial decay of large carbon pool (Great DunFell United Kingdom) In an upland grassland systemat Great Dun Fell in the United Kingdom a combinationof approaches was used to assess the impact of globalwarming on ecosystem carbon stores and soil water quali-ty In one intact vegetationsoil monoliths were moveddown an altitudinal gradient (Ineson et al 1998a) inanother in situ warming using electric heating cables (Ine-son et al 1998b) was applied Both approaches were testedin several different vegetationsoil systems including amineral acidic brown earth and a peaty stagnohumic gleyResults after 3 years of the heating treatments were quali-tatively similar to those predicted for the arctic tundraunder warming combined with reduced soil moisture(Figure 3 line D) but the net carbon losses were even larg-er

All systems at Great Dun Fell showed an increase inplant biomass under the warmer conditions but majorreductions in the total carbon content of the peaty gley(approximately 10 reduction in 3 years) demonstratedthe vulnerability of the carbon stored in this particularsoil The increases in temperature which stimulated themineralization of organic matter and release of nutrientssignificantly improved plant growth but these increases incarbon fixation were insufficient to replace the carbon lostthrough increased decomposition These increases indecomposition were caused by generally greater soil bio-logical activity and in particular by population increasesand changes in vertical distribution of the enchytraeidworm Cognettia sphagnetorum (Briones et al 1998) Theselarge losses in total soil carbon stocks clearly cannot con-tinue eventually the soils are expected to stabilize at alower carbon content analogous to the pattern shown inFigure 3 (line E) and by the native soils at the lower end ofthe same altitudinal gradientSpecies changes affect ecosystem carbon storage(subalpine meadow Colorado United States) Ina subalpine meadow ecosystem in Colorado Harte and col-leagues have used overhead infrared heaters to warm bothsoil and vegetation year-round since 1991 (Figure 6) caus-ing significant changes in NPP and NEP (Harte et al 1995

876 BioScience October 2000 Vol 50 No 10

Articles

Figure 6 Infrared heaters in place above subalpinemeadow vegetation Colorado Photo Taylor Ricketts

Figure 7 Smaller plastic greenhouses and burlap-coveredshade frames at the subarctic heath site Abisko SwedenPhoto Sven Jonasson

Saleska et al 1999) In contrast to the previous examples aprincipal mechanism driving the changes in Colorado is theeffect of the vegetation on species growth rates and func-tional type composition (Harte and Shaw 1995 deValpineand Harte in press) A 12 loss of soil carbon in the top 15cm of the heated plot soils was observed after 4 years oftreatment (Scott Saleska and John Harte University of Cal-iforniandashBerkeley unpublished data) This initial loss of soilorganic matter was not driven by a warming-inducedincrease in heterotrophic respiration because the positiveeffect of a temperature increase on R

h was cancelled by thenegative effect of soil drying Instead this initial loss of soilorganic matter was driven mainly by the effect of warmingon the rate of litter input to the soil In particular warminginduced a shift from more productive forbs to less produc-tive shrubs However litter produced by forbs is more easi-ly decomposed than litter produced by the shrubs whichhas a higher ligninnitrogen ratio Thus the loss of soilorganic matter is likely to be a transient effect with therecalcitrant litter input from shrubs causing an increase insoil organic matter over the long term and ultimately in anoverall pattern of change in NEP similar to that in Figure 3(line C)

By comparing responses observed in the warmingexperiment to changes in microclimate soil organic mat-ter and vegetation species composition along a naturalelevation gradient Harte and colleagues have shown thatthe combined use of manipulations and gradient studiescan provide a unified picture partitioned into short andlong-term components of how climate change influencessoil organic matter The research shows how the impacts ofwarming on carbon cycling can be dominated by the indi-rect effect of a warming-induced shift in relative speciescomposition on the quantity (short term) and quality(longer term) of litter input to the soilExploitation of available space (Abisko Sweden)In low-biomass ecosystems where the surface of the soil ispartially bare warming brought about by increasing plantcover may cause increases in NEP In tundra heath and fellfield ecosystems at Abisko Sweden warming with smallplastic-covered greenhouses caused such an increase inplant cover (Figure 7) Plant biomass driven in part bytemperature-related increases in plant nitrogen uptakeincreased almost twofold within 5 years (Jonasson et al1999)

The response of biomass to warming at Abisko was sim-ilar to that in the tussock tundra at Toolik Lake althoughthe response at Abisko was greater and NPP increasedwithout any pronounced changes in species compositionOverall the response followed the curve represented inFigure 3 (line C) initial declines in carbon stocks were notobserved because the experiment was not sampled imme-diately The longer term (more than 10 years) response isexpected to level off and reach an equilibrium at a net car-bon accumulation rate close to zero when the availablespace has been occupied by plants

Initial increases in biomass in response to warming neednot be a continuous process over time particularly not inthe northernmost ecosystems where the plants operate closeto the limit of their cold tolerance For instance an increasein biomass similar to that at Abisko took place over severalyears in a poorly vegetated high arctic semi-desert howev-er after an unusually mild and rainy growing season muchof the increase was lost in warmed and watered plots Theloss in biomass presumably occurred because winter hard-ening in the experimental plots was delayed and the plantssuffered frost damage when the winter set in (Robinson etal 1998) reflecting a possible effect of stochastic ldquoextremeeventsrdquo Such events are likely to be of great importance inclimatic regimes close to the tolerance limits of plantsUnder such regimes carbon pools in the living plant bio-mass may fluctuate between phases of accumulation andabrupt decline never stabilizing at zero net accumulationwhile soil organic matter is built up stepwise during theperiods of plant die-back

Experiments data and synthesis needsPast warming experiments have used a variety of heatingmethods including electric resistance heating infraredirradiation reciprocal transplants and open- and closed-top field greenhouses (Table 1) Some experiments havewarmed only portions of the ecosystem Others havewarmed the ecosystem for only part of each day or year Inmost cases the choice of warming method is constrainedby logistical and engineering considerations (eg avail-ability of electric power difficulty of enclosing an entireforest and the initial interests of the investigators) No onemethod perfectly simulates the expected climatic changesbut all have yielded useful results that with cautious inter-pretation have begun to reveal similarities differencesand patterns of response among the ecosystems studiedAlthough there is considerable intellectual appeal todesigning future experiments that use a common method-ology and realistic simulation of expected change it also istrue that different warming methods illuminate differentaspects of the overall warming response To optimizeintercomparisons future research should include a mix ofcommon manipulations in diverse ecosystem typesFuture research should also include manipulationsdesigned to provide focused insights into individualecosystems or key processes

Several networks and consortia are already working tomeet the need for a global understanding of ecosystemresponses to climate change (Ingram et al 1999) TheInternational Tundra Experiment (ITEX) uses passivelywarmed open-top chambers in 26 different arctic andalpine tundras to compare the effects of warming on plantgrowth and flowering (Henry and Molau 1997 Arft et al1999) A broader-based Network of Ecosystem WarmingStudies (NEWS) is being developed as part of the GlobalChange and Terrestrial Ecosystems (GCTE) Core Projectof the International GeospherendashBiosphere Program

October 2000 Vol 50 No 10 BioScience 877

Articles

(IGBP) Most warming experiments to date however havetaken place in north temperate boreal and arctic ecosys-tems which are cool and moist and where NPP is strong-ly nutrient-limited Although the processes that dominatethe warming responses of these ecosystems are importanteverywhere they may not dominate in other ecosystemtypes In warm and dry regions including Mediterraneanor desert ecosystems and in tropical humid ecosystemsindirect warming effects acting through changes in evapo-transpiration and soil dryness will play a major regulatoryrole We also might expect the direct effects of warming onNPP to be relatively smaller in these warm systemsbecause temperatures are already near optimum for pho-tosynthesis NPP might even decrease due to overheatingin the warmest ecosystems or microenvironments Addi-tional experiments and networks are needed to expand therange of ecosystems studied including studies of vegeta-tion with contrasting plant life histories (eg annual ver-sus perennial grasslands woodlands and evergreen versusdeciduous vegetation)

New methods are also needed for the quantitative syn-thesis of data from different studies of temperature effectson ecosystems Comparisons among different simulationmodels have also yielded important insights (Schimel et al1997) Some relatively new simulation models are

designed to facilitate comparisons among ecosystems andto predict their responses to temperature change within acommon model structure (eg Rastetter et al 1991Williams et al 1997) New statistical methods can also beused to search for and explain general patterns in diversedata sets For example we might want to know the averageeffects of experimental warming treatments on NPPacross a group of studies and the extent to which thoseeffects differ in biomes at different latitudes or with differ-ent moisture regimes Ecologists have begun to use met-analysis statistics to synthesize experimental data frommany independent studies to answer these kinds of ques-tions (Arnqvist and Wooster 1995 Gurevitch and Hedges1999) including plant responses to warming in ITEX (Arftet al 1999)

Finally there is a need to extrapolate experimentalresults over longer time scales and over larger landscapesand regions An excellent way to do so is to combineexperimentation with measurement of ecosystem proper-ties and processes along environmental transects such asthe IGBPndashGCTE Terrestrial Transects (Steffen et al1999) The transects are assumed to mirror longer-term(decades to centuries) responses of vegetation and soils toenvironmental conditions whereas warming experimentsreflect responses over the shorter-term (years to decades)

878 BioScience October 2000 Vol 50 No 10

Articles

Table 1 Comparison of currently used methods for experimental warming of terrestrial ecosystems

Method Mechanism of warming Advantages Disadvantages Examples

Field greenhouse ldquoGreenhouse warmingrdquo (ie Simple and inexpensive Little or no temperature Chapin et al 1995reflection of reradiated infrared requires no electrical power control and large tempera- Shaver et al 1998energy and reduced advective ture variability other arti- Jonasson et al 1999energy loss) facts include altered light

wind humidity and precipi- tation regimes

Passive Open-top Same as above Same as above Little or no temperature Marion et al 1997Chamber (OTC) control altered wind and Arft et al 1999

humidity only small areas can be manipulated uniformly

Active Open-top Same as above Precise control of air temp- Altered wind regime Norby et al 1997Chamber (OTC) plus warming by advection of erature or temperature humidity and

electrically heated forced difference may be combined evapotranspirationwarm air with CO2 control

Active Soil Warming Warming by conduction from Precise control of soil Altered soil moisture Peterjohn et al 1994buried electrical resistance temperature or temperature regime no effect on Ineson et al 1998bcables difference may be combined aboveground temperatures Hartley et al 1999

with greenhouses or OTCs

Electric infrared heat Warming by increased infrared Precise control of energy Warming depends entirely Harte et al 1995radiation input direct simulation of on radiation no change in Harte and Shaw 1995

global change in energy advective energy inputsbalance

Reciprocal or one-way Transplantation of plants Comparison of relatively Disturbance effects Ineson et al 1998atransplantation soils or whole plantndashsoil natural temperature gradients multiple environmental Grogan and Chapin1999 systems to warmer or cooler changes make it difficult

locations to assign specific causes to responses observed

Combined information from experiments and transectsmay thus be useful in distinguishing relatively short-termtransient changes in the experiments from longer-termnear-equilibrium changes For example global analyses ofprimary production (NPP Figure 1) and litter (the majorcontributor to R

h Figure 1) have shown that both are rea-sonably well-predicted by actual evapotranspiration(Lieth and Whittaker 1975 Meentemeyer 1984)Although short-term responses of NPP and R

h to experi-mental warming may not have the same relationshipswith actual evapotranspiration as do undisturbed ecosys-tems convergence toward the predicted relationshipmight be a useful measure of expected future change inthe experiments

Ultimately experimental and analytical approachesshould be combined with large-scale monitoring of vari-ables and processes that experimental studies have identi-fied as critical indicators of ecosystem change The combi-nation of global monitoring with ongoing experimentaldata acquisition metanalysis modeling and other datasynthesis efforts would provide an early warning system todetect and potentially mitigate these ecosystem responses

Field stations long-term ecological research sites andnature reserves already exist in many relatively unman-aged landscapes Collecting simple baseline data and mon-itoring annually or even less frequently would not be veryexpensive although data quality may be highly variable(eg Stohlgren et al 1995) Efforts to coordinate data andestablish protocols for data sharing might be daunting butthe resulting information would be immensely valuableresponses of species communities and biogeochemicalprocesses to climate change could potentially be detecteddecades earlier than they would be otherwise

Interactions with other drivers of globalchangeChanges in temperature are now occurring simultaneous-ly with other types of global change It will not be possibleto fully understand and predict ecosystem responses totemperature change without taking into account the inter-actions with the other components of global environmen-tal change (Koch and Mooney 1996 Sellers et al 1997Vitousek et al 1997 Luo and Mooney 1999) For exampleincreases in atmospheric CO

2 may modify plant tissuechemistry and ecosystem water balance both of whichmay in turn modify responses to warming Length of thegrowing season may increase with warming or it maydecrease as drought duration increases altering NPP NEPand other processes Many regions of the globe are experi-encing increased levels of nitrogen deposition and thenature of the nitrogen economy of an ecosystem is animportant factor in determining responses to other envi-ronmental factors (Aber et al 1998)

Changes in land use and disturbance regime may alsoinfluence responses to warming For instance land clear-ing over very large areas may modify local temperature or

precipitation regimes beyond what would happen withclimate change alone Similarly changes in fire regimescould be caused by and affect ecosystem responses towarming The spread of invasive species will also be affect-ed by both warming and land use change

Some interactions between warming and other globalchanges are predicted more readily than others For exam-ple outbreaks of insects or pathogens that kill the dominantplants and increase fire probability would lead to dramati-cally different results than the phased changes in response towarming alone We must expect some surprises

Clearly it will not be possible to incorporate all impor-tant dimensions of global change into experiments It willrarely be feasible to include more than one or two factorsin addition to temperature (eg CO2 concentration ornitrogen level) as experimental factors (Schulze et al1999) However through the use of combinations ofexperiments gradient studies remote sensing and simu-lation modeling it should be possible to make substantialprogress toward developing a mechanistic understandingof the most important interactions (Canadell et al 2000)

Although much current global change research focuseson feedbacks from terrestrial ecosystems to climatethrough greenhouse gases important feedbacks also existthrough changes in vegetation structure surface energybalance and water balance Leaves in the canopy act assurfaces for energy and mass exchanges with the atmos-phere and changes in canopy density and transpiration inresponse to warming will change the energy and massfluxes (McFadden et al 1998) One of the most importantand obvious effects of changes in the canopy is its albedo(ie the fraction of incoming solar radiation that isreflected) Decreasing albedo through increased leaf den-sity will increase net radiation (solar radiation absorbedby the system) The increased net radiation must be dissi-pated through one or a combination of three major ener-gy pathways or fluxes sensible heat (actual warming of theair) latent heat (evapotranspiration) and ground heat(warming the soil) If increased evapotranspirationdecreases soil moisture and species changes result in bet-ter stomatal control of water loss the result will beincreased sensible heat flux to the atmosphere Foley et al(1994) illustrated the importance of such a change duringthe last temperature optimum in the Holocene approxi-mately 6000 years ago Latitudinal treeline advanced sig-nificantly greatly reducing albedo The increased soil tem-perature and sensible heat flux from this vegetationchange was as great as the solar forcing believed to beresponsible for initiating the climate change Similarlyfurther loss of soil moisture in dry systems will decreaselatent heat flux with more energy dissipated as sensibleheat Current understanding of these types of feedbackson climate change (ie those operating through changesin surface energy and water balance) is less advanced thanthat of feedbacks through greenhouse gas fluxes andshould be a particularly high priority for future research

October 2000 Vol 50 No 10 BioScience 879

Articles

ConclusionsBecause temperature has a direct effect on virtually allecosystem processes responses of terrestrial ecosystems toglobal warming will be even more complex and difficult topredict than responses to increased atmospheric CO

2con-

centrations Nonetheless a useful conceptual frameworkfor analysis comparison and prediction of responses towarming is beginning to emerge (see box this page) Fornow the principal applications of this framework are as astatement of the major dimensions of complexity of theecosystem warming response and as a reminder of themajor constraints on long-term predictions based on rela-tively small-area short-term experimental studies A keyneed is to further develop our understanding of how ini-tial climate biogeochemical states and species character-istics might be used to generate long-term predictions ofrelative sensitivity to climate change Such a frameworkwould serve as a heuristic guide to new research

We also need improved models of the temporalsequence of ecosystem responses to temperature changebecause it is increasingly clear that long-term responses towarming and increased CO

2may differ greatly in both

magnitude and direction from initial responses Ecosys-tem-level experiments are a useful tool for developing thisunderstanding a broader array of experiments in con-trasting ecosystem types is needed including both whole-system manipulations and more focused experimentaltreatments Also needed are the transect studies long-term monitoring and the development of research net-works that will allow spatial extrapolation and validationof predictions based on intensive experimental studies

AcknowledgmentsThis paper contributes to the Global Change and Terres-trial Ecosystems (GCTE) core project of InternationalGeospherendashBiosphere Program (IGBP) and to the activi-ties of the Network of Experimental Warming Studies(GCTE-NEWS) The work was initiated with the supportof the National Center for Ecological Analysis and Synthe-sis a center funded by NSF (Grant DEB-94-21535) theUniversity of CaliforniandashSanta Barbara the CaliforniaResources Agency and the California Environmental Pro-tection Agency Additional support was provided by NSFgrant DEB-9730110 by NASA through a grant to GCTEand by USDA Forest Service

References citedAber J McDowell W Nadelhoffer K McGill A Berntson G Kamaka M

McNulty S Currie W Rustad L Fernandez I 1998 Nitrogen saturation

in forest ecosystems Hypotheses revisited BioScience 48 921ndash934Arft AM et al 1999 Response patterns of tundra plant species to experi-

mental warming A meta-analysis of the International Tundra Experi-

ment Ecological Monographs 69 491ndash512

Arnqvist FR Wooster D 1995 Meta-analysismdashsynthesizing research find-

ings in ecology and evolution Trends in Ecology and Evolution 10236ndash240

880 BioScience October 2000 Vol 50 No 10

Articles

Conceptual framework for analysis of ecosystem response to changing temperature

A Complexity Spatial and temporal variation Global tempera-

ture changes will not be uniform with mostregions warming by varying amounts but otherscooling These changes will occur over a range ofnatural temperature regimes and annual temper-ature variations that are greater than the expect-ed average temperature change

Direct and indirect effects Because temperatureaffects virtually all ecosystem processes indirecttemperature effects resulting from interactionsamong processes are likely to be very importantespecially in the long term

Variable time constants of response Diverseecosystem processes have different time constantsfor response to temperature Thus initial overallecosystem responses to change are not reliableindicators of long-term responses because theyrepresent only part of the total response

B Responses Initial response a function of initial conditions The

same temperature change applied to differentecosystems will elicit different responses depend-ing on initial position on the temperatureresponse surface (direct effects) and on initialbiogeochemical conditions and composition(indirect effects)

Changing sequence of controls The dominant con-trols over ecosystem response to temperature willchange over time as different processes change atdifferent rates This changing sequence will notnecessarily be the same in all ecosystems

Changing direction as well as magnitude ofresponse Initial losses or gains of carbon by theecosystem may be reversed over time as indirecteffects and slowly changing components of theecosystem come into play

Species changes can affect biogeochemistryChanges in species composition and abundancewill affect ecosystem functions (net primary pro-duction carbon and nutrient cycling energyfluxes) through changes in tissue chemistrydemography (both parts and individuals)turnover rates and vegetation structure (physiog-nomy leaf area index)

Response is predictable (at least qualitatively) frominitial conditions Key variables include initialtemperature regime initial carbon and nutrientpool sizes and turnover rates and initial popula-tion dynamics of vegetation

Stocks and initial turnover rates of labile carbon inthe soil

Stocks and initial turnover rates of labile nitrogen inthe soil

Relative size of the carbon pools of plants and soil

Dominant form of available nitrogen in the soil (egorganic nitrogen ammonium nitrate)

Soil water and precipitation regime

Chemical composition and turnover rates of liveplant tissues and litter

Longevity of individuals and population turnoverrates of dominant plant species

The effects of these initial conditions on the response ofecosystems to warming can be illustrated by comparingchanges in NPP NEP or R

h over time in contrastingecosystems that have been manipulated experimentallyThis approach has already led to useful insights as shownby the following examples and illustrated in Figure 3Nitrogen redistribution model (Harvard ForestMassachusetts United States) At Harvard Forest incentral Massachusetts (a temperate deciduous forestecosystem) a long-term soil warming experiment usedelectric heating cables buried in the soil to elevate soil tem-peratures at a depth of 10 cm by 5 degC above ambient (Fig-ure 4 Peterjohn et al 1994 Melillo et al 1995) At this sitethe important initial conditions include

Relatively small labile soil carbon pool

Large slow-turnover vegetation carbon pool

Relatively large soil nitrogen pool

Ammonification-dominated nitrogen mineralizationsequence

These initial conditions constrain how the Harvard Forestecosystem responds to soil warming over time such that athree-phase transient pattern of changes in carbon storageresults (Figure 3 line A Melillo et al in press) DuringPhase I the labile carbon pool is rapidly oxidized bydecomposing organisms leading to a short period of netcarbon loss In Phase II nitrogen is transferred from thesoil which has a low CN ratio to the vegetation whichhas a high CN ratio This redistribution of nitrogenresults in net carbon storage at the ecosystem levelalthough there is a loss of soil carbon (Rastetter et al 19911997) During Phase II there is no significant loss of nitro-gen from the ecosystem because the system has an ammo-nium economy If the system had a nitrate economy or amixed ammoniumndashnitrate economy nitrogen could belost from the system by leaching or denitrification and soreduce the nitrogen available to promote carbon storageEventually the labile and metastable nitrogen pools in thesoil decrease to the point where very little nitrogen can beredistributed from the soil to the plants At this point theecosystem may begin to lose carbon overall (ie NEPbecomes negative) because of increased respiratory lossesfrom plants and soils (Figure 3 line A) This decline may

be reversed if external nitrogen inputs (deposition or fixa-tion) are available or if plant biomass is lost due to distur-bance such as a hurricane or insect attack and is decom-posed releasing nitrogen once again (Figure 3 line B)Nitrogen redistribution with variable moistureregime (Toolik Lake Alaska United States) In amoist tundra ecosystem at Toolik Lake in northern Alaskaresponses to warming were studied using small green-houses to increase air temperature during the growing sea-son by 3ndash5 degC (Figure 5) The soil organic matter pool ofthis ecosystem is larger and contains even more labile car-bon and nitrogen than the soils of many northern tem-perate forests (Giblin et al 1991) Decomposition andnitrogen mineralization in tundra is strongly limited bylow soil temperatures and high soil moisture (Nadelhofferet al 1992) Thus if soil organic matter turnover isincreased due to warming there is a high potential forredistribution of nitrogen from soils (with low CN ratio)to vegetation (with high CN ratio) resulting in net car-bon storage with little or no net change in ecosystemnitrogen stocks (Shaver et al 1992) This response howev-er is strongly interactive with soil moisture with a rela-tively small temperature response if the soil is saturated(Nadelhoffer et al 1992)

Experimental results (Chapin et al 1995 Shaver et al1998) and modeling (McKane et al 1997a 1997b Rastet-ter et al 1997) indicate a multiphase response of NEP towarming in this tundra ecosystem that is quite similar tothat at Harvard Forest as long as soil moisture does notchange greatly (Figure 3 line C) Similar results have alsobeen obtained in the long-term monitoring and modeling

October 2000 Vol 50 No 10 BioScience 875

Articles

Figure 5 The Toolik Lake Field Greenhouse Experimentin moist tussock tundra vegetation Alaska showingmoderately increased abundance of deciduous shrubsand total vegetation mass The photograph was taken in1999 after 11 summers of warming Photo Gus Shaver

studies by Oechel and coworkers (eg Vourlitis andOechel 1997 1999 Waelbroeck et al 1997) The initialresponse to warming is a large increase in R

h leading to adecrease in NEP for the first 1ndash3 years (Phase I) Increasedsoil respiration however is accompanied by increased

nitrogen mineralization leading to increased plant nitro-gen uptake and increased NPP after 3ndash10 years (Phase II)Further increases in NPP are limited by other factors (eglight) Increased NPP leads to increased litter productionwhich eventually leads to higher R

h and NEP returns tonear zero (Phase III) after approximately 50ndash100 years

Changes in NEP in response to warming of moist tun-dra may be very different if there is an interaction between

warming and soil moisture (Billings 1987 McKane et al1997b Vourlitis and Oechel 1999) Warming combinedwith decreased soil moisture will cause a long-term loss ofboth carbon and nitrogen because of large increases in Rh

combined with losses of mineralized nitrogen by drainagefrom the system especially in areas where thaw depthincreases greatly (In such areas nitrogen losses mayinclude large amounts of dissolved organic nitrogen theproduct of incomplete mineralization) In this scenarioincreases in nitrogen uptake and NPP in Phase II areinsufficient to compensate for nitrogen and carbon lossesin leaching and respiration so that even though theecosystem eventually returns to an equilibrium whereNPP equals R

h (NEP = 0) over 50ndash100 years there is a netloss of carbon to the atmosphere (Figure 3 line D)Partial decay of large carbon pool (Great DunFell United Kingdom) In an upland grassland systemat Great Dun Fell in the United Kingdom a combinationof approaches was used to assess the impact of globalwarming on ecosystem carbon stores and soil water quali-ty In one intact vegetationsoil monoliths were moveddown an altitudinal gradient (Ineson et al 1998a) inanother in situ warming using electric heating cables (Ine-son et al 1998b) was applied Both approaches were testedin several different vegetationsoil systems including amineral acidic brown earth and a peaty stagnohumic gleyResults after 3 years of the heating treatments were quali-tatively similar to those predicted for the arctic tundraunder warming combined with reduced soil moisture(Figure 3 line D) but the net carbon losses were even larg-er

All systems at Great Dun Fell showed an increase inplant biomass under the warmer conditions but majorreductions in the total carbon content of the peaty gley(approximately 10 reduction in 3 years) demonstratedthe vulnerability of the carbon stored in this particularsoil The increases in temperature which stimulated themineralization of organic matter and release of nutrientssignificantly improved plant growth but these increases incarbon fixation were insufficient to replace the carbon lostthrough increased decomposition These increases indecomposition were caused by generally greater soil bio-logical activity and in particular by population increasesand changes in vertical distribution of the enchytraeidworm Cognettia sphagnetorum (Briones et al 1998) Theselarge losses in total soil carbon stocks clearly cannot con-tinue eventually the soils are expected to stabilize at alower carbon content analogous to the pattern shown inFigure 3 (line E) and by the native soils at the lower end ofthe same altitudinal gradientSpecies changes affect ecosystem carbon storage(subalpine meadow Colorado United States) Ina subalpine meadow ecosystem in Colorado Harte and col-leagues have used overhead infrared heaters to warm bothsoil and vegetation year-round since 1991 (Figure 6) caus-ing significant changes in NPP and NEP (Harte et al 1995

876 BioScience October 2000 Vol 50 No 10

Articles

Figure 6 Infrared heaters in place above subalpinemeadow vegetation Colorado Photo Taylor Ricketts

Figure 7 Smaller plastic greenhouses and burlap-coveredshade frames at the subarctic heath site Abisko SwedenPhoto Sven Jonasson

Saleska et al 1999) In contrast to the previous examples aprincipal mechanism driving the changes in Colorado is theeffect of the vegetation on species growth rates and func-tional type composition (Harte and Shaw 1995 deValpineand Harte in press) A 12 loss of soil carbon in the top 15cm of the heated plot soils was observed after 4 years oftreatment (Scott Saleska and John Harte University of Cal-iforniandashBerkeley unpublished data) This initial loss of soilorganic matter was not driven by a warming-inducedincrease in heterotrophic respiration because the positiveeffect of a temperature increase on R

h was cancelled by thenegative effect of soil drying Instead this initial loss of soilorganic matter was driven mainly by the effect of warmingon the rate of litter input to the soil In particular warminginduced a shift from more productive forbs to less produc-tive shrubs However litter produced by forbs is more easi-ly decomposed than litter produced by the shrubs whichhas a higher ligninnitrogen ratio Thus the loss of soilorganic matter is likely to be a transient effect with therecalcitrant litter input from shrubs causing an increase insoil organic matter over the long term and ultimately in anoverall pattern of change in NEP similar to that in Figure 3(line C)

By comparing responses observed in the warmingexperiment to changes in microclimate soil organic mat-ter and vegetation species composition along a naturalelevation gradient Harte and colleagues have shown thatthe combined use of manipulations and gradient studiescan provide a unified picture partitioned into short andlong-term components of how climate change influencessoil organic matter The research shows how the impacts ofwarming on carbon cycling can be dominated by the indi-rect effect of a warming-induced shift in relative speciescomposition on the quantity (short term) and quality(longer term) of litter input to the soilExploitation of available space (Abisko Sweden)In low-biomass ecosystems where the surface of the soil ispartially bare warming brought about by increasing plantcover may cause increases in NEP In tundra heath and fellfield ecosystems at Abisko Sweden warming with smallplastic-covered greenhouses caused such an increase inplant cover (Figure 7) Plant biomass driven in part bytemperature-related increases in plant nitrogen uptakeincreased almost twofold within 5 years (Jonasson et al1999)

The response of biomass to warming at Abisko was sim-ilar to that in the tussock tundra at Toolik Lake althoughthe response at Abisko was greater and NPP increasedwithout any pronounced changes in species compositionOverall the response followed the curve represented inFigure 3 (line C) initial declines in carbon stocks were notobserved because the experiment was not sampled imme-diately The longer term (more than 10 years) response isexpected to level off and reach an equilibrium at a net car-bon accumulation rate close to zero when the availablespace has been occupied by plants

Initial increases in biomass in response to warming neednot be a continuous process over time particularly not inthe northernmost ecosystems where the plants operate closeto the limit of their cold tolerance For instance an increasein biomass similar to that at Abisko took place over severalyears in a poorly vegetated high arctic semi-desert howev-er after an unusually mild and rainy growing season muchof the increase was lost in warmed and watered plots Theloss in biomass presumably occurred because winter hard-ening in the experimental plots was delayed and the plantssuffered frost damage when the winter set in (Robinson etal 1998) reflecting a possible effect of stochastic ldquoextremeeventsrdquo Such events are likely to be of great importance inclimatic regimes close to the tolerance limits of plantsUnder such regimes carbon pools in the living plant bio-mass may fluctuate between phases of accumulation andabrupt decline never stabilizing at zero net accumulationwhile soil organic matter is built up stepwise during theperiods of plant die-back

Experiments data and synthesis needsPast warming experiments have used a variety of heatingmethods including electric resistance heating infraredirradiation reciprocal transplants and open- and closed-top field greenhouses (Table 1) Some experiments havewarmed only portions of the ecosystem Others havewarmed the ecosystem for only part of each day or year Inmost cases the choice of warming method is constrainedby logistical and engineering considerations (eg avail-ability of electric power difficulty of enclosing an entireforest and the initial interests of the investigators) No onemethod perfectly simulates the expected climatic changesbut all have yielded useful results that with cautious inter-pretation have begun to reveal similarities differencesand patterns of response among the ecosystems studiedAlthough there is considerable intellectual appeal todesigning future experiments that use a common method-ology and realistic simulation of expected change it also istrue that different warming methods illuminate differentaspects of the overall warming response To optimizeintercomparisons future research should include a mix ofcommon manipulations in diverse ecosystem typesFuture research should also include manipulationsdesigned to provide focused insights into individualecosystems or key processes

Several networks and consortia are already working tomeet the need for a global understanding of ecosystemresponses to climate change (Ingram et al 1999) TheInternational Tundra Experiment (ITEX) uses passivelywarmed open-top chambers in 26 different arctic andalpine tundras to compare the effects of warming on plantgrowth and flowering (Henry and Molau 1997 Arft et al1999) A broader-based Network of Ecosystem WarmingStudies (NEWS) is being developed as part of the GlobalChange and Terrestrial Ecosystems (GCTE) Core Projectof the International GeospherendashBiosphere Program

October 2000 Vol 50 No 10 BioScience 877

Articles

(IGBP) Most warming experiments to date however havetaken place in north temperate boreal and arctic ecosys-tems which are cool and moist and where NPP is strong-ly nutrient-limited Although the processes that dominatethe warming responses of these ecosystems are importanteverywhere they may not dominate in other ecosystemtypes In warm and dry regions including Mediterraneanor desert ecosystems and in tropical humid ecosystemsindirect warming effects acting through changes in evapo-transpiration and soil dryness will play a major regulatoryrole We also might expect the direct effects of warming onNPP to be relatively smaller in these warm systemsbecause temperatures are already near optimum for pho-tosynthesis NPP might even decrease due to overheatingin the warmest ecosystems or microenvironments Addi-tional experiments and networks are needed to expand therange of ecosystems studied including studies of vegeta-tion with contrasting plant life histories (eg annual ver-sus perennial grasslands woodlands and evergreen versusdeciduous vegetation)

New methods are also needed for the quantitative syn-thesis of data from different studies of temperature effectson ecosystems Comparisons among different simulationmodels have also yielded important insights (Schimel et al1997) Some relatively new simulation models are

designed to facilitate comparisons among ecosystems andto predict their responses to temperature change within acommon model structure (eg Rastetter et al 1991Williams et al 1997) New statistical methods can also beused to search for and explain general patterns in diversedata sets For example we might want to know the averageeffects of experimental warming treatments on NPPacross a group of studies and the extent to which thoseeffects differ in biomes at different latitudes or with differ-ent moisture regimes Ecologists have begun to use met-analysis statistics to synthesize experimental data frommany independent studies to answer these kinds of ques-tions (Arnqvist and Wooster 1995 Gurevitch and Hedges1999) including plant responses to warming in ITEX (Arftet al 1999)

Finally there is a need to extrapolate experimentalresults over longer time scales and over larger landscapesand regions An excellent way to do so is to combineexperimentation with measurement of ecosystem proper-ties and processes along environmental transects such asthe IGBPndashGCTE Terrestrial Transects (Steffen et al1999) The transects are assumed to mirror longer-term(decades to centuries) responses of vegetation and soils toenvironmental conditions whereas warming experimentsreflect responses over the shorter-term (years to decades)

878 BioScience October 2000 Vol 50 No 10

Articles

Table 1 Comparison of currently used methods for experimental warming of terrestrial ecosystems

Method Mechanism of warming Advantages Disadvantages Examples

Field greenhouse ldquoGreenhouse warmingrdquo (ie Simple and inexpensive Little or no temperature Chapin et al 1995reflection of reradiated infrared requires no electrical power control and large tempera- Shaver et al 1998energy and reduced advective ture variability other arti- Jonasson et al 1999energy loss) facts include altered light

wind humidity and precipi- tation regimes

Passive Open-top Same as above Same as above Little or no temperature Marion et al 1997Chamber (OTC) control altered wind and Arft et al 1999

humidity only small areas can be manipulated uniformly

Active Open-top Same as above Precise control of air temp- Altered wind regime Norby et al 1997Chamber (OTC) plus warming by advection of erature or temperature humidity and

electrically heated forced difference may be combined evapotranspirationwarm air with CO2 control

Active Soil Warming Warming by conduction from Precise control of soil Altered soil moisture Peterjohn et al 1994buried electrical resistance temperature or temperature regime no effect on Ineson et al 1998bcables difference may be combined aboveground temperatures Hartley et al 1999

with greenhouses or OTCs

Electric infrared heat Warming by increased infrared Precise control of energy Warming depends entirely Harte et al 1995radiation input direct simulation of on radiation no change in Harte and Shaw 1995

global change in energy advective energy inputsbalance

Reciprocal or one-way Transplantation of plants Comparison of relatively Disturbance effects Ineson et al 1998atransplantation soils or whole plantndashsoil natural temperature gradients multiple environmental Grogan and Chapin1999 systems to warmer or cooler changes make it difficult

locations to assign specific causes to responses observed

Combined information from experiments and transectsmay thus be useful in distinguishing relatively short-termtransient changes in the experiments from longer-termnear-equilibrium changes For example global analyses ofprimary production (NPP Figure 1) and litter (the majorcontributor to R

h Figure 1) have shown that both are rea-sonably well-predicted by actual evapotranspiration(Lieth and Whittaker 1975 Meentemeyer 1984)Although short-term responses of NPP and R

h to experi-mental warming may not have the same relationshipswith actual evapotranspiration as do undisturbed ecosys-tems convergence toward the predicted relationshipmight be a useful measure of expected future change inthe experiments

Ultimately experimental and analytical approachesshould be combined with large-scale monitoring of vari-ables and processes that experimental studies have identi-fied as critical indicators of ecosystem change The combi-nation of global monitoring with ongoing experimentaldata acquisition metanalysis modeling and other datasynthesis efforts would provide an early warning system todetect and potentially mitigate these ecosystem responses

Field stations long-term ecological research sites andnature reserves already exist in many relatively unman-aged landscapes Collecting simple baseline data and mon-itoring annually or even less frequently would not be veryexpensive although data quality may be highly variable(eg Stohlgren et al 1995) Efforts to coordinate data andestablish protocols for data sharing might be daunting butthe resulting information would be immensely valuableresponses of species communities and biogeochemicalprocesses to climate change could potentially be detecteddecades earlier than they would be otherwise

Interactions with other drivers of globalchangeChanges in temperature are now occurring simultaneous-ly with other types of global change It will not be possibleto fully understand and predict ecosystem responses totemperature change without taking into account the inter-actions with the other components of global environmen-tal change (Koch and Mooney 1996 Sellers et al 1997Vitousek et al 1997 Luo and Mooney 1999) For exampleincreases in atmospheric CO

2 may modify plant tissuechemistry and ecosystem water balance both of whichmay in turn modify responses to warming Length of thegrowing season may increase with warming or it maydecrease as drought duration increases altering NPP NEPand other processes Many regions of the globe are experi-encing increased levels of nitrogen deposition and thenature of the nitrogen economy of an ecosystem is animportant factor in determining responses to other envi-ronmental factors (Aber et al 1998)

Changes in land use and disturbance regime may alsoinfluence responses to warming For instance land clear-ing over very large areas may modify local temperature or

precipitation regimes beyond what would happen withclimate change alone Similarly changes in fire regimescould be caused by and affect ecosystem responses towarming The spread of invasive species will also be affect-ed by both warming and land use change

Some interactions between warming and other globalchanges are predicted more readily than others For exam-ple outbreaks of insects or pathogens that kill the dominantplants and increase fire probability would lead to dramati-cally different results than the phased changes in response towarming alone We must expect some surprises

Clearly it will not be possible to incorporate all impor-tant dimensions of global change into experiments It willrarely be feasible to include more than one or two factorsin addition to temperature (eg CO2 concentration ornitrogen level) as experimental factors (Schulze et al1999) However through the use of combinations ofexperiments gradient studies remote sensing and simu-lation modeling it should be possible to make substantialprogress toward developing a mechanistic understandingof the most important interactions (Canadell et al 2000)

Although much current global change research focuseson feedbacks from terrestrial ecosystems to climatethrough greenhouse gases important feedbacks also existthrough changes in vegetation structure surface energybalance and water balance Leaves in the canopy act assurfaces for energy and mass exchanges with the atmos-phere and changes in canopy density and transpiration inresponse to warming will change the energy and massfluxes (McFadden et al 1998) One of the most importantand obvious effects of changes in the canopy is its albedo(ie the fraction of incoming solar radiation that isreflected) Decreasing albedo through increased leaf den-sity will increase net radiation (solar radiation absorbedby the system) The increased net radiation must be dissi-pated through one or a combination of three major ener-gy pathways or fluxes sensible heat (actual warming of theair) latent heat (evapotranspiration) and ground heat(warming the soil) If increased evapotranspirationdecreases soil moisture and species changes result in bet-ter stomatal control of water loss the result will beincreased sensible heat flux to the atmosphere Foley et al(1994) illustrated the importance of such a change duringthe last temperature optimum in the Holocene approxi-mately 6000 years ago Latitudinal treeline advanced sig-nificantly greatly reducing albedo The increased soil tem-perature and sensible heat flux from this vegetationchange was as great as the solar forcing believed to beresponsible for initiating the climate change Similarlyfurther loss of soil moisture in dry systems will decreaselatent heat flux with more energy dissipated as sensibleheat Current understanding of these types of feedbackson climate change (ie those operating through changesin surface energy and water balance) is less advanced thanthat of feedbacks through greenhouse gas fluxes andshould be a particularly high priority for future research

October 2000 Vol 50 No 10 BioScience 879

Articles

ConclusionsBecause temperature has a direct effect on virtually allecosystem processes responses of terrestrial ecosystems toglobal warming will be even more complex and difficult topredict than responses to increased atmospheric CO

2con-

centrations Nonetheless a useful conceptual frameworkfor analysis comparison and prediction of responses towarming is beginning to emerge (see box this page) Fornow the principal applications of this framework are as astatement of the major dimensions of complexity of theecosystem warming response and as a reminder of themajor constraints on long-term predictions based on rela-tively small-area short-term experimental studies A keyneed is to further develop our understanding of how ini-tial climate biogeochemical states and species character-istics might be used to generate long-term predictions ofrelative sensitivity to climate change Such a frameworkwould serve as a heuristic guide to new research

We also need improved models of the temporalsequence of ecosystem responses to temperature changebecause it is increasingly clear that long-term responses towarming and increased CO

2may differ greatly in both

magnitude and direction from initial responses Ecosys-tem-level experiments are a useful tool for developing thisunderstanding a broader array of experiments in con-trasting ecosystem types is needed including both whole-system manipulations and more focused experimentaltreatments Also needed are the transect studies long-term monitoring and the development of research net-works that will allow spatial extrapolation and validationof predictions based on intensive experimental studies

AcknowledgmentsThis paper contributes to the Global Change and Terres-trial Ecosystems (GCTE) core project of InternationalGeospherendashBiosphere Program (IGBP) and to the activi-ties of the Network of Experimental Warming Studies(GCTE-NEWS) The work was initiated with the supportof the National Center for Ecological Analysis and Synthe-sis a center funded by NSF (Grant DEB-94-21535) theUniversity of CaliforniandashSanta Barbara the CaliforniaResources Agency and the California Environmental Pro-tection Agency Additional support was provided by NSFgrant DEB-9730110 by NASA through a grant to GCTEand by USDA Forest Service

References citedAber J McDowell W Nadelhoffer K McGill A Berntson G Kamaka M

McNulty S Currie W Rustad L Fernandez I 1998 Nitrogen saturation

in forest ecosystems Hypotheses revisited BioScience 48 921ndash934Arft AM et al 1999 Response patterns of tundra plant species to experi-

mental warming A meta-analysis of the International Tundra Experi-

ment Ecological Monographs 69 491ndash512

Arnqvist FR Wooster D 1995 Meta-analysismdashsynthesizing research find-

ings in ecology and evolution Trends in Ecology and Evolution 10236ndash240

880 BioScience October 2000 Vol 50 No 10

Articles

Conceptual framework for analysis of ecosystem response to changing temperature

A Complexity Spatial and temporal variation Global tempera-

ture changes will not be uniform with mostregions warming by varying amounts but otherscooling These changes will occur over a range ofnatural temperature regimes and annual temper-ature variations that are greater than the expect-ed average temperature change

Direct and indirect effects Because temperatureaffects virtually all ecosystem processes indirecttemperature effects resulting from interactionsamong processes are likely to be very importantespecially in the long term

Variable time constants of response Diverseecosystem processes have different time constantsfor response to temperature Thus initial overallecosystem responses to change are not reliableindicators of long-term responses because theyrepresent only part of the total response

B Responses Initial response a function of initial conditions The

same temperature change applied to differentecosystems will elicit different responses depend-ing on initial position on the temperatureresponse surface (direct effects) and on initialbiogeochemical conditions and composition(indirect effects)

Changing sequence of controls The dominant con-trols over ecosystem response to temperature willchange over time as different processes change atdifferent rates This changing sequence will notnecessarily be the same in all ecosystems

Changing direction as well as magnitude ofresponse Initial losses or gains of carbon by theecosystem may be reversed over time as indirecteffects and slowly changing components of theecosystem come into play

Species changes can affect biogeochemistryChanges in species composition and abundancewill affect ecosystem functions (net primary pro-duction carbon and nutrient cycling energyfluxes) through changes in tissue chemistrydemography (both parts and individuals)turnover rates and vegetation structure (physiog-nomy leaf area index)

Response is predictable (at least qualitatively) frominitial conditions Key variables include initialtemperature regime initial carbon and nutrientpool sizes and turnover rates and initial popula-tion dynamics of vegetation

studies by Oechel and coworkers (eg Vourlitis andOechel 1997 1999 Waelbroeck et al 1997) The initialresponse to warming is a large increase in R

h leading to adecrease in NEP for the first 1ndash3 years (Phase I) Increasedsoil respiration however is accompanied by increased

nitrogen mineralization leading to increased plant nitro-gen uptake and increased NPP after 3ndash10 years (Phase II)Further increases in NPP are limited by other factors (eglight) Increased NPP leads to increased litter productionwhich eventually leads to higher R

h and NEP returns tonear zero (Phase III) after approximately 50ndash100 years

Changes in NEP in response to warming of moist tun-dra may be very different if there is an interaction between

warming and soil moisture (Billings 1987 McKane et al1997b Vourlitis and Oechel 1999) Warming combinedwith decreased soil moisture will cause a long-term loss ofboth carbon and nitrogen because of large increases in Rh

combined with losses of mineralized nitrogen by drainagefrom the system especially in areas where thaw depthincreases greatly (In such areas nitrogen losses mayinclude large amounts of dissolved organic nitrogen theproduct of incomplete mineralization) In this scenarioincreases in nitrogen uptake and NPP in Phase II areinsufficient to compensate for nitrogen and carbon lossesin leaching and respiration so that even though theecosystem eventually returns to an equilibrium whereNPP equals R

h (NEP = 0) over 50ndash100 years there is a netloss of carbon to the atmosphere (Figure 3 line D)Partial decay of large carbon pool (Great DunFell United Kingdom) In an upland grassland systemat Great Dun Fell in the United Kingdom a combinationof approaches was used to assess the impact of globalwarming on ecosystem carbon stores and soil water quali-ty In one intact vegetationsoil monoliths were moveddown an altitudinal gradient (Ineson et al 1998a) inanother in situ warming using electric heating cables (Ine-son et al 1998b) was applied Both approaches were testedin several different vegetationsoil systems including amineral acidic brown earth and a peaty stagnohumic gleyResults after 3 years of the heating treatments were quali-tatively similar to those predicted for the arctic tundraunder warming combined with reduced soil moisture(Figure 3 line D) but the net carbon losses were even larg-er

All systems at Great Dun Fell showed an increase inplant biomass under the warmer conditions but majorreductions in the total carbon content of the peaty gley(approximately 10 reduction in 3 years) demonstratedthe vulnerability of the carbon stored in this particularsoil The increases in temperature which stimulated themineralization of organic matter and release of nutrientssignificantly improved plant growth but these increases incarbon fixation were insufficient to replace the carbon lostthrough increased decomposition These increases indecomposition were caused by generally greater soil bio-logical activity and in particular by population increasesand changes in vertical distribution of the enchytraeidworm Cognettia sphagnetorum (Briones et al 1998) Theselarge losses in total soil carbon stocks clearly cannot con-tinue eventually the soils are expected to stabilize at alower carbon content analogous to the pattern shown inFigure 3 (line E) and by the native soils at the lower end ofthe same altitudinal gradientSpecies changes affect ecosystem carbon storage(subalpine meadow Colorado United States) Ina subalpine meadow ecosystem in Colorado Harte and col-leagues have used overhead infrared heaters to warm bothsoil and vegetation year-round since 1991 (Figure 6) caus-ing significant changes in NPP and NEP (Harte et al 1995

876 BioScience October 2000 Vol 50 No 10

Articles

Figure 6 Infrared heaters in place above subalpinemeadow vegetation Colorado Photo Taylor Ricketts

Figure 7 Smaller plastic greenhouses and burlap-coveredshade frames at the subarctic heath site Abisko SwedenPhoto Sven Jonasson

Saleska et al 1999) In contrast to the previous examples aprincipal mechanism driving the changes in Colorado is theeffect of the vegetation on species growth rates and func-tional type composition (Harte and Shaw 1995 deValpineand Harte in press) A 12 loss of soil carbon in the top 15cm of the heated plot soils was observed after 4 years oftreatment (Scott Saleska and John Harte University of Cal-iforniandashBerkeley unpublished data) This initial loss of soilorganic matter was not driven by a warming-inducedincrease in heterotrophic respiration because the positiveeffect of a temperature increase on R

h was cancelled by thenegative effect of soil drying Instead this initial loss of soilorganic matter was driven mainly by the effect of warmingon the rate of litter input to the soil In particular warminginduced a shift from more productive forbs to less produc-tive shrubs However litter produced by forbs is more easi-ly decomposed than litter produced by the shrubs whichhas a higher ligninnitrogen ratio Thus the loss of soilorganic matter is likely to be a transient effect with therecalcitrant litter input from shrubs causing an increase insoil organic matter over the long term and ultimately in anoverall pattern of change in NEP similar to that in Figure 3(line C)

By comparing responses observed in the warmingexperiment to changes in microclimate soil organic mat-ter and vegetation species composition along a naturalelevation gradient Harte and colleagues have shown thatthe combined use of manipulations and gradient studiescan provide a unified picture partitioned into short andlong-term components of how climate change influencessoil organic matter The research shows how the impacts ofwarming on carbon cycling can be dominated by the indi-rect effect of a warming-induced shift in relative speciescomposition on the quantity (short term) and quality(longer term) of litter input to the soilExploitation of available space (Abisko Sweden)In low-biomass ecosystems where the surface of the soil ispartially bare warming brought about by increasing plantcover may cause increases in NEP In tundra heath and fellfield ecosystems at Abisko Sweden warming with smallplastic-covered greenhouses caused such an increase inplant cover (Figure 7) Plant biomass driven in part bytemperature-related increases in plant nitrogen uptakeincreased almost twofold within 5 years (Jonasson et al1999)

The response of biomass to warming at Abisko was sim-ilar to that in the tussock tundra at Toolik Lake althoughthe response at Abisko was greater and NPP increasedwithout any pronounced changes in species compositionOverall the response followed the curve represented inFigure 3 (line C) initial declines in carbon stocks were notobserved because the experiment was not sampled imme-diately The longer term (more than 10 years) response isexpected to level off and reach an equilibrium at a net car-bon accumulation rate close to zero when the availablespace has been occupied by plants

Initial increases in biomass in response to warming neednot be a continuous process over time particularly not inthe northernmost ecosystems where the plants operate closeto the limit of their cold tolerance For instance an increasein biomass similar to that at Abisko took place over severalyears in a poorly vegetated high arctic semi-desert howev-er after an unusually mild and rainy growing season muchof the increase was lost in warmed and watered plots Theloss in biomass presumably occurred because winter hard-ening in the experimental plots was delayed and the plantssuffered frost damage when the winter set in (Robinson etal 1998) reflecting a possible effect of stochastic ldquoextremeeventsrdquo Such events are likely to be of great importance inclimatic regimes close to the tolerance limits of plantsUnder such regimes carbon pools in the living plant bio-mass may fluctuate between phases of accumulation andabrupt decline never stabilizing at zero net accumulationwhile soil organic matter is built up stepwise during theperiods of plant die-back

Experiments data and synthesis needsPast warming experiments have used a variety of heatingmethods including electric resistance heating infraredirradiation reciprocal transplants and open- and closed-top field greenhouses (Table 1) Some experiments havewarmed only portions of the ecosystem Others havewarmed the ecosystem for only part of each day or year Inmost cases the choice of warming method is constrainedby logistical and engineering considerations (eg avail-ability of electric power difficulty of enclosing an entireforest and the initial interests of the investigators) No onemethod perfectly simulates the expected climatic changesbut all have yielded useful results that with cautious inter-pretation have begun to reveal similarities differencesand patterns of response among the ecosystems studiedAlthough there is considerable intellectual appeal todesigning future experiments that use a common method-ology and realistic simulation of expected change it also istrue that different warming methods illuminate differentaspects of the overall warming response To optimizeintercomparisons future research should include a mix ofcommon manipulations in diverse ecosystem typesFuture research should also include manipulationsdesigned to provide focused insights into individualecosystems or key processes

Several networks and consortia are already working tomeet the need for a global understanding of ecosystemresponses to climate change (Ingram et al 1999) TheInternational Tundra Experiment (ITEX) uses passivelywarmed open-top chambers in 26 different arctic andalpine tundras to compare the effects of warming on plantgrowth and flowering (Henry and Molau 1997 Arft et al1999) A broader-based Network of Ecosystem WarmingStudies (NEWS) is being developed as part of the GlobalChange and Terrestrial Ecosystems (GCTE) Core Projectof the International GeospherendashBiosphere Program

October 2000 Vol 50 No 10 BioScience 877

Articles

(IGBP) Most warming experiments to date however havetaken place in north temperate boreal and arctic ecosys-tems which are cool and moist and where NPP is strong-ly nutrient-limited Although the processes that dominatethe warming responses of these ecosystems are importanteverywhere they may not dominate in other ecosystemtypes In warm and dry regions including Mediterraneanor desert ecosystems and in tropical humid ecosystemsindirect warming effects acting through changes in evapo-transpiration and soil dryness will play a major regulatoryrole We also might expect the direct effects of warming onNPP to be relatively smaller in these warm systemsbecause temperatures are already near optimum for pho-tosynthesis NPP might even decrease due to overheatingin the warmest ecosystems or microenvironments Addi-tional experiments and networks are needed to expand therange of ecosystems studied including studies of vegeta-tion with contrasting plant life histories (eg annual ver-sus perennial grasslands woodlands and evergreen versusdeciduous vegetation)

New methods are also needed for the quantitative syn-thesis of data from different studies of temperature effectson ecosystems Comparisons among different simulationmodels have also yielded important insights (Schimel et al1997) Some relatively new simulation models are

designed to facilitate comparisons among ecosystems andto predict their responses to temperature change within acommon model structure (eg Rastetter et al 1991Williams et al 1997) New statistical methods can also beused to search for and explain general patterns in diversedata sets For example we might want to know the averageeffects of experimental warming treatments on NPPacross a group of studies and the extent to which thoseeffects differ in biomes at different latitudes or with differ-ent moisture regimes Ecologists have begun to use met-analysis statistics to synthesize experimental data frommany independent studies to answer these kinds of ques-tions (Arnqvist and Wooster 1995 Gurevitch and Hedges1999) including plant responses to warming in ITEX (Arftet al 1999)

Finally there is a need to extrapolate experimentalresults over longer time scales and over larger landscapesand regions An excellent way to do so is to combineexperimentation with measurement of ecosystem proper-ties and processes along environmental transects such asthe IGBPndashGCTE Terrestrial Transects (Steffen et al1999) The transects are assumed to mirror longer-term(decades to centuries) responses of vegetation and soils toenvironmental conditions whereas warming experimentsreflect responses over the shorter-term (years to decades)

878 BioScience October 2000 Vol 50 No 10

Articles

Table 1 Comparison of currently used methods for experimental warming of terrestrial ecosystems

Method Mechanism of warming Advantages Disadvantages Examples

Field greenhouse ldquoGreenhouse warmingrdquo (ie Simple and inexpensive Little or no temperature Chapin et al 1995reflection of reradiated infrared requires no electrical power control and large tempera- Shaver et al 1998energy and reduced advective ture variability other arti- Jonasson et al 1999energy loss) facts include altered light

wind humidity and precipi- tation regimes

Passive Open-top Same as above Same as above Little or no temperature Marion et al 1997Chamber (OTC) control altered wind and Arft et al 1999

humidity only small areas can be manipulated uniformly

Active Open-top Same as above Precise control of air temp- Altered wind regime Norby et al 1997Chamber (OTC) plus warming by advection of erature or temperature humidity and

electrically heated forced difference may be combined evapotranspirationwarm air with CO2 control

Active Soil Warming Warming by conduction from Precise control of soil Altered soil moisture Peterjohn et al 1994buried electrical resistance temperature or temperature regime no effect on Ineson et al 1998bcables difference may be combined aboveground temperatures Hartley et al 1999

with greenhouses or OTCs

Electric infrared heat Warming by increased infrared Precise control of energy Warming depends entirely Harte et al 1995radiation input direct simulation of on radiation no change in Harte and Shaw 1995

global change in energy advective energy inputsbalance

Reciprocal or one-way Transplantation of plants Comparison of relatively Disturbance effects Ineson et al 1998atransplantation soils or whole plantndashsoil natural temperature gradients multiple environmental Grogan and Chapin1999 systems to warmer or cooler changes make it difficult

locations to assign specific causes to responses observed

Combined information from experiments and transectsmay thus be useful in distinguishing relatively short-termtransient changes in the experiments from longer-termnear-equilibrium changes For example global analyses ofprimary production (NPP Figure 1) and litter (the majorcontributor to R

h Figure 1) have shown that both are rea-sonably well-predicted by actual evapotranspiration(Lieth and Whittaker 1975 Meentemeyer 1984)Although short-term responses of NPP and R

h to experi-mental warming may not have the same relationshipswith actual evapotranspiration as do undisturbed ecosys-tems convergence toward the predicted relationshipmight be a useful measure of expected future change inthe experiments

Ultimately experimental and analytical approachesshould be combined with large-scale monitoring of vari-ables and processes that experimental studies have identi-fied as critical indicators of ecosystem change The combi-nation of global monitoring with ongoing experimentaldata acquisition metanalysis modeling and other datasynthesis efforts would provide an early warning system todetect and potentially mitigate these ecosystem responses

Field stations long-term ecological research sites andnature reserves already exist in many relatively unman-aged landscapes Collecting simple baseline data and mon-itoring annually or even less frequently would not be veryexpensive although data quality may be highly variable(eg Stohlgren et al 1995) Efforts to coordinate data andestablish protocols for data sharing might be daunting butthe resulting information would be immensely valuableresponses of species communities and biogeochemicalprocesses to climate change could potentially be detecteddecades earlier than they would be otherwise

Interactions with other drivers of globalchangeChanges in temperature are now occurring simultaneous-ly with other types of global change It will not be possibleto fully understand and predict ecosystem responses totemperature change without taking into account the inter-actions with the other components of global environmen-tal change (Koch and Mooney 1996 Sellers et al 1997Vitousek et al 1997 Luo and Mooney 1999) For exampleincreases in atmospheric CO

2 may modify plant tissuechemistry and ecosystem water balance both of whichmay in turn modify responses to warming Length of thegrowing season may increase with warming or it maydecrease as drought duration increases altering NPP NEPand other processes Many regions of the globe are experi-encing increased levels of nitrogen deposition and thenature of the nitrogen economy of an ecosystem is animportant factor in determining responses to other envi-ronmental factors (Aber et al 1998)

Changes in land use and disturbance regime may alsoinfluence responses to warming For instance land clear-ing over very large areas may modify local temperature or

precipitation regimes beyond what would happen withclimate change alone Similarly changes in fire regimescould be caused by and affect ecosystem responses towarming The spread of invasive species will also be affect-ed by both warming and land use change

Some interactions between warming and other globalchanges are predicted more readily than others For exam-ple outbreaks of insects or pathogens that kill the dominantplants and increase fire probability would lead to dramati-cally different results than the phased changes in response towarming alone We must expect some surprises

Clearly it will not be possible to incorporate all impor-tant dimensions of global change into experiments It willrarely be feasible to include more than one or two factorsin addition to temperature (eg CO2 concentration ornitrogen level) as experimental factors (Schulze et al1999) However through the use of combinations ofexperiments gradient studies remote sensing and simu-lation modeling it should be possible to make substantialprogress toward developing a mechanistic understandingof the most important interactions (Canadell et al 2000)

Although much current global change research focuseson feedbacks from terrestrial ecosystems to climatethrough greenhouse gases important feedbacks also existthrough changes in vegetation structure surface energybalance and water balance Leaves in the canopy act assurfaces for energy and mass exchanges with the atmos-phere and changes in canopy density and transpiration inresponse to warming will change the energy and massfluxes (McFadden et al 1998) One of the most importantand obvious effects of changes in the canopy is its albedo(ie the fraction of incoming solar radiation that isreflected) Decreasing albedo through increased leaf den-sity will increase net radiation (solar radiation absorbedby the system) The increased net radiation must be dissi-pated through one or a combination of three major ener-gy pathways or fluxes sensible heat (actual warming of theair) latent heat (evapotranspiration) and ground heat(warming the soil) If increased evapotranspirationdecreases soil moisture and species changes result in bet-ter stomatal control of water loss the result will beincreased sensible heat flux to the atmosphere Foley et al(1994) illustrated the importance of such a change duringthe last temperature optimum in the Holocene approxi-mately 6000 years ago Latitudinal treeline advanced sig-nificantly greatly reducing albedo The increased soil tem-perature and sensible heat flux from this vegetationchange was as great as the solar forcing believed to beresponsible for initiating the climate change Similarlyfurther loss of soil moisture in dry systems will decreaselatent heat flux with more energy dissipated as sensibleheat Current understanding of these types of feedbackson climate change (ie those operating through changesin surface energy and water balance) is less advanced thanthat of feedbacks through greenhouse gas fluxes andshould be a particularly high priority for future research

October 2000 Vol 50 No 10 BioScience 879

Articles

ConclusionsBecause temperature has a direct effect on virtually allecosystem processes responses of terrestrial ecosystems toglobal warming will be even more complex and difficult topredict than responses to increased atmospheric CO

2con-

centrations Nonetheless a useful conceptual frameworkfor analysis comparison and prediction of responses towarming is beginning to emerge (see box this page) Fornow the principal applications of this framework are as astatement of the major dimensions of complexity of theecosystem warming response and as a reminder of themajor constraints on long-term predictions based on rela-tively small-area short-term experimental studies A keyneed is to further develop our understanding of how ini-tial climate biogeochemical states and species character-istics might be used to generate long-term predictions ofrelative sensitivity to climate change Such a frameworkwould serve as a heuristic guide to new research

We also need improved models of the temporalsequence of ecosystem responses to temperature changebecause it is increasingly clear that long-term responses towarming and increased CO

2may differ greatly in both

magnitude and direction from initial responses Ecosys-tem-level experiments are a useful tool for developing thisunderstanding a broader array of experiments in con-trasting ecosystem types is needed including both whole-system manipulations and more focused experimentaltreatments Also needed are the transect studies long-term monitoring and the development of research net-works that will allow spatial extrapolation and validationof predictions based on intensive experimental studies

AcknowledgmentsThis paper contributes to the Global Change and Terres-trial Ecosystems (GCTE) core project of InternationalGeospherendashBiosphere Program (IGBP) and to the activi-ties of the Network of Experimental Warming Studies(GCTE-NEWS) The work was initiated with the supportof the National Center for Ecological Analysis and Synthe-sis a center funded by NSF (Grant DEB-94-21535) theUniversity of CaliforniandashSanta Barbara the CaliforniaResources Agency and the California Environmental Pro-tection Agency Additional support was provided by NSFgrant DEB-9730110 by NASA through a grant to GCTEand by USDA Forest Service

References citedAber J McDowell W Nadelhoffer K McGill A Berntson G Kamaka M

McNulty S Currie W Rustad L Fernandez I 1998 Nitrogen saturation

in forest ecosystems Hypotheses revisited BioScience 48 921ndash934Arft AM et al 1999 Response patterns of tundra plant species to experi-

mental warming A meta-analysis of the International Tundra Experi-

ment Ecological Monographs 69 491ndash512

Arnqvist FR Wooster D 1995 Meta-analysismdashsynthesizing research find-

ings in ecology and evolution Trends in Ecology and Evolution 10236ndash240

880 BioScience October 2000 Vol 50 No 10

Articles

Conceptual framework for analysis of ecosystem response to changing temperature

A Complexity Spatial and temporal variation Global tempera-

ture changes will not be uniform with mostregions warming by varying amounts but otherscooling These changes will occur over a range ofnatural temperature regimes and annual temper-ature variations that are greater than the expect-ed average temperature change

Direct and indirect effects Because temperatureaffects virtually all ecosystem processes indirecttemperature effects resulting from interactionsamong processes are likely to be very importantespecially in the long term

Variable time constants of response Diverseecosystem processes have different time constantsfor response to temperature Thus initial overallecosystem responses to change are not reliableindicators of long-term responses because theyrepresent only part of the total response

B Responses Initial response a function of initial conditions The

same temperature change applied to differentecosystems will elicit different responses depend-ing on initial position on the temperatureresponse surface (direct effects) and on initialbiogeochemical conditions and composition(indirect effects)

Changing sequence of controls The dominant con-trols over ecosystem response to temperature willchange over time as different processes change atdifferent rates This changing sequence will notnecessarily be the same in all ecosystems

Changing direction as well as magnitude ofresponse Initial losses or gains of carbon by theecosystem may be reversed over time as indirecteffects and slowly changing components of theecosystem come into play

Species changes can affect biogeochemistryChanges in species composition and abundancewill affect ecosystem functions (net primary pro-duction carbon and nutrient cycling energyfluxes) through changes in tissue chemistrydemography (both parts and individuals)turnover rates and vegetation structure (physiog-nomy leaf area index)

Response is predictable (at least qualitatively) frominitial conditions Key variables include initialtemperature regime initial carbon and nutrientpool sizes and turnover rates and initial popula-tion dynamics of vegetation

Saleska et al 1999) In contrast to the previous examples aprincipal mechanism driving the changes in Colorado is theeffect of the vegetation on species growth rates and func-tional type composition (Harte and Shaw 1995 deValpineand Harte in press) A 12 loss of soil carbon in the top 15cm of the heated plot soils was observed after 4 years oftreatment (Scott Saleska and John Harte University of Cal-iforniandashBerkeley unpublished data) This initial loss of soilorganic matter was not driven by a warming-inducedincrease in heterotrophic respiration because the positiveeffect of a temperature increase on R

h was cancelled by thenegative effect of soil drying Instead this initial loss of soilorganic matter was driven mainly by the effect of warmingon the rate of litter input to the soil In particular warminginduced a shift from more productive forbs to less produc-tive shrubs However litter produced by forbs is more easi-ly decomposed than litter produced by the shrubs whichhas a higher ligninnitrogen ratio Thus the loss of soilorganic matter is likely to be a transient effect with therecalcitrant litter input from shrubs causing an increase insoil organic matter over the long term and ultimately in anoverall pattern of change in NEP similar to that in Figure 3(line C)

By comparing responses observed in the warmingexperiment to changes in microclimate soil organic mat-ter and vegetation species composition along a naturalelevation gradient Harte and colleagues have shown thatthe combined use of manipulations and gradient studiescan provide a unified picture partitioned into short andlong-term components of how climate change influencessoil organic matter The research shows how the impacts ofwarming on carbon cycling can be dominated by the indi-rect effect of a warming-induced shift in relative speciescomposition on the quantity (short term) and quality(longer term) of litter input to the soilExploitation of available space (Abisko Sweden)In low-biomass ecosystems where the surface of the soil ispartially bare warming brought about by increasing plantcover may cause increases in NEP In tundra heath and fellfield ecosystems at Abisko Sweden warming with smallplastic-covered greenhouses caused such an increase inplant cover (Figure 7) Plant biomass driven in part bytemperature-related increases in plant nitrogen uptakeincreased almost twofold within 5 years (Jonasson et al1999)

The response of biomass to warming at Abisko was sim-ilar to that in the tussock tundra at Toolik Lake althoughthe response at Abisko was greater and NPP increasedwithout any pronounced changes in species compositionOverall the response followed the curve represented inFigure 3 (line C) initial declines in carbon stocks were notobserved because the experiment was not sampled imme-diately The longer term (more than 10 years) response isexpected to level off and reach an equilibrium at a net car-bon accumulation rate close to zero when the availablespace has been occupied by plants

Initial increases in biomass in response to warming neednot be a continuous process over time particularly not inthe northernmost ecosystems where the plants operate closeto the limit of their cold tolerance For instance an increasein biomass similar to that at Abisko took place over severalyears in a poorly vegetated high arctic semi-desert howev-er after an unusually mild and rainy growing season muchof the increase was lost in warmed and watered plots Theloss in biomass presumably occurred because winter hard-ening in the experimental plots was delayed and the plantssuffered frost damage when the winter set in (Robinson etal 1998) reflecting a possible effect of stochastic ldquoextremeeventsrdquo Such events are likely to be of great importance inclimatic regimes close to the tolerance limits of plantsUnder such regimes carbon pools in the living plant bio-mass may fluctuate between phases of accumulation andabrupt decline never stabilizing at zero net accumulationwhile soil organic matter is built up stepwise during theperiods of plant die-back

Experiments data and synthesis needsPast warming experiments have used a variety of heatingmethods including electric resistance heating infraredirradiation reciprocal transplants and open- and closed-top field greenhouses (Table 1) Some experiments havewarmed only portions of the ecosystem Others havewarmed the ecosystem for only part of each day or year Inmost cases the choice of warming method is constrainedby logistical and engineering considerations (eg avail-ability of electric power difficulty of enclosing an entireforest and the initial interests of the investigators) No onemethod perfectly simulates the expected climatic changesbut all have yielded useful results that with cautious inter-pretation have begun to reveal similarities differencesand patterns of response among the ecosystems studiedAlthough there is considerable intellectual appeal todesigning future experiments that use a common method-ology and realistic simulation of expected change it also istrue that different warming methods illuminate differentaspects of the overall warming response To optimizeintercomparisons future research should include a mix ofcommon manipulations in diverse ecosystem typesFuture research should also include manipulationsdesigned to provide focused insights into individualecosystems or key processes

Several networks and consortia are already working tomeet the need for a global understanding of ecosystemresponses to climate change (Ingram et al 1999) TheInternational Tundra Experiment (ITEX) uses passivelywarmed open-top chambers in 26 different arctic andalpine tundras to compare the effects of warming on plantgrowth and flowering (Henry and Molau 1997 Arft et al1999) A broader-based Network of Ecosystem WarmingStudies (NEWS) is being developed as part of the GlobalChange and Terrestrial Ecosystems (GCTE) Core Projectof the International GeospherendashBiosphere Program

October 2000 Vol 50 No 10 BioScience 877

Articles

(IGBP) Most warming experiments to date however havetaken place in north temperate boreal and arctic ecosys-tems which are cool and moist and where NPP is strong-ly nutrient-limited Although the processes that dominatethe warming responses of these ecosystems are importanteverywhere they may not dominate in other ecosystemtypes In warm and dry regions including Mediterraneanor desert ecosystems and in tropical humid ecosystemsindirect warming effects acting through changes in evapo-transpiration and soil dryness will play a major regulatoryrole We also might expect the direct effects of warming onNPP to be relatively smaller in these warm systemsbecause temperatures are already near optimum for pho-tosynthesis NPP might even decrease due to overheatingin the warmest ecosystems or microenvironments Addi-tional experiments and networks are needed to expand therange of ecosystems studied including studies of vegeta-tion with contrasting plant life histories (eg annual ver-sus perennial grasslands woodlands and evergreen versusdeciduous vegetation)

New methods are also needed for the quantitative syn-thesis of data from different studies of temperature effectson ecosystems Comparisons among different simulationmodels have also yielded important insights (Schimel et al1997) Some relatively new simulation models are

designed to facilitate comparisons among ecosystems andto predict their responses to temperature change within acommon model structure (eg Rastetter et al 1991Williams et al 1997) New statistical methods can also beused to search for and explain general patterns in diversedata sets For example we might want to know the averageeffects of experimental warming treatments on NPPacross a group of studies and the extent to which thoseeffects differ in biomes at different latitudes or with differ-ent moisture regimes Ecologists have begun to use met-analysis statistics to synthesize experimental data frommany independent studies to answer these kinds of ques-tions (Arnqvist and Wooster 1995 Gurevitch and Hedges1999) including plant responses to warming in ITEX (Arftet al 1999)

Finally there is a need to extrapolate experimentalresults over longer time scales and over larger landscapesand regions An excellent way to do so is to combineexperimentation with measurement of ecosystem proper-ties and processes along environmental transects such asthe IGBPndashGCTE Terrestrial Transects (Steffen et al1999) The transects are assumed to mirror longer-term(decades to centuries) responses of vegetation and soils toenvironmental conditions whereas warming experimentsreflect responses over the shorter-term (years to decades)

878 BioScience October 2000 Vol 50 No 10

Articles

Table 1 Comparison of currently used methods for experimental warming of terrestrial ecosystems

Method Mechanism of warming Advantages Disadvantages Examples

Field greenhouse ldquoGreenhouse warmingrdquo (ie Simple and inexpensive Little or no temperature Chapin et al 1995reflection of reradiated infrared requires no electrical power control and large tempera- Shaver et al 1998energy and reduced advective ture variability other arti- Jonasson et al 1999energy loss) facts include altered light

wind humidity and precipi- tation regimes

Passive Open-top Same as above Same as above Little or no temperature Marion et al 1997Chamber (OTC) control altered wind and Arft et al 1999

humidity only small areas can be manipulated uniformly

Active Open-top Same as above Precise control of air temp- Altered wind regime Norby et al 1997Chamber (OTC) plus warming by advection of erature or temperature humidity and

electrically heated forced difference may be combined evapotranspirationwarm air with CO2 control

Active Soil Warming Warming by conduction from Precise control of soil Altered soil moisture Peterjohn et al 1994buried electrical resistance temperature or temperature regime no effect on Ineson et al 1998bcables difference may be combined aboveground temperatures Hartley et al 1999

with greenhouses or OTCs

Electric infrared heat Warming by increased infrared Precise control of energy Warming depends entirely Harte et al 1995radiation input direct simulation of on radiation no change in Harte and Shaw 1995

global change in energy advective energy inputsbalance

Reciprocal or one-way Transplantation of plants Comparison of relatively Disturbance effects Ineson et al 1998atransplantation soils or whole plantndashsoil natural temperature gradients multiple environmental Grogan and Chapin1999 systems to warmer or cooler changes make it difficult

locations to assign specific causes to responses observed

Combined information from experiments and transectsmay thus be useful in distinguishing relatively short-termtransient changes in the experiments from longer-termnear-equilibrium changes For example global analyses ofprimary production (NPP Figure 1) and litter (the majorcontributor to R

h Figure 1) have shown that both are rea-sonably well-predicted by actual evapotranspiration(Lieth and Whittaker 1975 Meentemeyer 1984)Although short-term responses of NPP and R

h to experi-mental warming may not have the same relationshipswith actual evapotranspiration as do undisturbed ecosys-tems convergence toward the predicted relationshipmight be a useful measure of expected future change inthe experiments

Ultimately experimental and analytical approachesshould be combined with large-scale monitoring of vari-ables and processes that experimental studies have identi-fied as critical indicators of ecosystem change The combi-nation of global monitoring with ongoing experimentaldata acquisition metanalysis modeling and other datasynthesis efforts would provide an early warning system todetect and potentially mitigate these ecosystem responses

Field stations long-term ecological research sites andnature reserves already exist in many relatively unman-aged landscapes Collecting simple baseline data and mon-itoring annually or even less frequently would not be veryexpensive although data quality may be highly variable(eg Stohlgren et al 1995) Efforts to coordinate data andestablish protocols for data sharing might be daunting butthe resulting information would be immensely valuableresponses of species communities and biogeochemicalprocesses to climate change could potentially be detecteddecades earlier than they would be otherwise

Interactions with other drivers of globalchangeChanges in temperature are now occurring simultaneous-ly with other types of global change It will not be possibleto fully understand and predict ecosystem responses totemperature change without taking into account the inter-actions with the other components of global environmen-tal change (Koch and Mooney 1996 Sellers et al 1997Vitousek et al 1997 Luo and Mooney 1999) For exampleincreases in atmospheric CO

2 may modify plant tissuechemistry and ecosystem water balance both of whichmay in turn modify responses to warming Length of thegrowing season may increase with warming or it maydecrease as drought duration increases altering NPP NEPand other processes Many regions of the globe are experi-encing increased levels of nitrogen deposition and thenature of the nitrogen economy of an ecosystem is animportant factor in determining responses to other envi-ronmental factors (Aber et al 1998)

Changes in land use and disturbance regime may alsoinfluence responses to warming For instance land clear-ing over very large areas may modify local temperature or

precipitation regimes beyond what would happen withclimate change alone Similarly changes in fire regimescould be caused by and affect ecosystem responses towarming The spread of invasive species will also be affect-ed by both warming and land use change

Some interactions between warming and other globalchanges are predicted more readily than others For exam-ple outbreaks of insects or pathogens that kill the dominantplants and increase fire probability would lead to dramati-cally different results than the phased changes in response towarming alone We must expect some surprises

Clearly it will not be possible to incorporate all impor-tant dimensions of global change into experiments It willrarely be feasible to include more than one or two factorsin addition to temperature (eg CO2 concentration ornitrogen level) as experimental factors (Schulze et al1999) However through the use of combinations ofexperiments gradient studies remote sensing and simu-lation modeling it should be possible to make substantialprogress toward developing a mechanistic understandingof the most important interactions (Canadell et al 2000)

Although much current global change research focuseson feedbacks from terrestrial ecosystems to climatethrough greenhouse gases important feedbacks also existthrough changes in vegetation structure surface energybalance and water balance Leaves in the canopy act assurfaces for energy and mass exchanges with the atmos-phere and changes in canopy density and transpiration inresponse to warming will change the energy and massfluxes (McFadden et al 1998) One of the most importantand obvious effects of changes in the canopy is its albedo(ie the fraction of incoming solar radiation that isreflected) Decreasing albedo through increased leaf den-sity will increase net radiation (solar radiation absorbedby the system) The increased net radiation must be dissi-pated through one or a combination of three major ener-gy pathways or fluxes sensible heat (actual warming of theair) latent heat (evapotranspiration) and ground heat(warming the soil) If increased evapotranspirationdecreases soil moisture and species changes result in bet-ter stomatal control of water loss the result will beincreased sensible heat flux to the atmosphere Foley et al(1994) illustrated the importance of such a change duringthe last temperature optimum in the Holocene approxi-mately 6000 years ago Latitudinal treeline advanced sig-nificantly greatly reducing albedo The increased soil tem-perature and sensible heat flux from this vegetationchange was as great as the solar forcing believed to beresponsible for initiating the climate change Similarlyfurther loss of soil moisture in dry systems will decreaselatent heat flux with more energy dissipated as sensibleheat Current understanding of these types of feedbackson climate change (ie those operating through changesin surface energy and water balance) is less advanced thanthat of feedbacks through greenhouse gas fluxes andshould be a particularly high priority for future research

October 2000 Vol 50 No 10 BioScience 879

Articles

ConclusionsBecause temperature has a direct effect on virtually allecosystem processes responses of terrestrial ecosystems toglobal warming will be even more complex and difficult topredict than responses to increased atmospheric CO

2con-

centrations Nonetheless a useful conceptual frameworkfor analysis comparison and prediction of responses towarming is beginning to emerge (see box this page) Fornow the principal applications of this framework are as astatement of the major dimensions of complexity of theecosystem warming response and as a reminder of themajor constraints on long-term predictions based on rela-tively small-area short-term experimental studies A keyneed is to further develop our understanding of how ini-tial climate biogeochemical states and species character-istics might be used to generate long-term predictions ofrelative sensitivity to climate change Such a frameworkwould serve as a heuristic guide to new research

We also need improved models of the temporalsequence of ecosystem responses to temperature changebecause it is increasingly clear that long-term responses towarming and increased CO

2may differ greatly in both

magnitude and direction from initial responses Ecosys-tem-level experiments are a useful tool for developing thisunderstanding a broader array of experiments in con-trasting ecosystem types is needed including both whole-system manipulations and more focused experimentaltreatments Also needed are the transect studies long-term monitoring and the development of research net-works that will allow spatial extrapolation and validationof predictions based on intensive experimental studies

AcknowledgmentsThis paper contributes to the Global Change and Terres-trial Ecosystems (GCTE) core project of InternationalGeospherendashBiosphere Program (IGBP) and to the activi-ties of the Network of Experimental Warming Studies(GCTE-NEWS) The work was initiated with the supportof the National Center for Ecological Analysis and Synthe-sis a center funded by NSF (Grant DEB-94-21535) theUniversity of CaliforniandashSanta Barbara the CaliforniaResources Agency and the California Environmental Pro-tection Agency Additional support was provided by NSFgrant DEB-9730110 by NASA through a grant to GCTEand by USDA Forest Service

References citedAber J McDowell W Nadelhoffer K McGill A Berntson G Kamaka M

McNulty S Currie W Rustad L Fernandez I 1998 Nitrogen saturation

in forest ecosystems Hypotheses revisited BioScience 48 921ndash934Arft AM et al 1999 Response patterns of tundra plant species to experi-

mental warming A meta-analysis of the International Tundra Experi-

ment Ecological Monographs 69 491ndash512

Arnqvist FR Wooster D 1995 Meta-analysismdashsynthesizing research find-

ings in ecology and evolution Trends in Ecology and Evolution 10236ndash240

880 BioScience October 2000 Vol 50 No 10

Articles

Conceptual framework for analysis of ecosystem response to changing temperature

A Complexity Spatial and temporal variation Global tempera-

ture changes will not be uniform with mostregions warming by varying amounts but otherscooling These changes will occur over a range ofnatural temperature regimes and annual temper-ature variations that are greater than the expect-ed average temperature change

Direct and indirect effects Because temperatureaffects virtually all ecosystem processes indirecttemperature effects resulting from interactionsamong processes are likely to be very importantespecially in the long term

Variable time constants of response Diverseecosystem processes have different time constantsfor response to temperature Thus initial overallecosystem responses to change are not reliableindicators of long-term responses because theyrepresent only part of the total response

B Responses Initial response a function of initial conditions The

same temperature change applied to differentecosystems will elicit different responses depend-ing on initial position on the temperatureresponse surface (direct effects) and on initialbiogeochemical conditions and composition(indirect effects)

Changing sequence of controls The dominant con-trols over ecosystem response to temperature willchange over time as different processes change atdifferent rates This changing sequence will notnecessarily be the same in all ecosystems

Changing direction as well as magnitude ofresponse Initial losses or gains of carbon by theecosystem may be reversed over time as indirecteffects and slowly changing components of theecosystem come into play

Species changes can affect biogeochemistryChanges in species composition and abundancewill affect ecosystem functions (net primary pro-duction carbon and nutrient cycling energyfluxes) through changes in tissue chemistrydemography (both parts and individuals)turnover rates and vegetation structure (physiog-nomy leaf area index)

Response is predictable (at least qualitatively) frominitial conditions Key variables include initialtemperature regime initial carbon and nutrientpool sizes and turnover rates and initial popula-tion dynamics of vegetation

(IGBP) Most warming experiments to date however havetaken place in north temperate boreal and arctic ecosys-tems which are cool and moist and where NPP is strong-ly nutrient-limited Although the processes that dominatethe warming responses of these ecosystems are importanteverywhere they may not dominate in other ecosystemtypes In warm and dry regions including Mediterraneanor desert ecosystems and in tropical humid ecosystemsindirect warming effects acting through changes in evapo-transpiration and soil dryness will play a major regulatoryrole We also might expect the direct effects of warming onNPP to be relatively smaller in these warm systemsbecause temperatures are already near optimum for pho-tosynthesis NPP might even decrease due to overheatingin the warmest ecosystems or microenvironments Addi-tional experiments and networks are needed to expand therange of ecosystems studied including studies of vegeta-tion with contrasting plant life histories (eg annual ver-sus perennial grasslands woodlands and evergreen versusdeciduous vegetation)

New methods are also needed for the quantitative syn-thesis of data from different studies of temperature effectson ecosystems Comparisons among different simulationmodels have also yielded important insights (Schimel et al1997) Some relatively new simulation models are

designed to facilitate comparisons among ecosystems andto predict their responses to temperature change within acommon model structure (eg Rastetter et al 1991Williams et al 1997) New statistical methods can also beused to search for and explain general patterns in diversedata sets For example we might want to know the averageeffects of experimental warming treatments on NPPacross a group of studies and the extent to which thoseeffects differ in biomes at different latitudes or with differ-ent moisture regimes Ecologists have begun to use met-analysis statistics to synthesize experimental data frommany independent studies to answer these kinds of ques-tions (Arnqvist and Wooster 1995 Gurevitch and Hedges1999) including plant responses to warming in ITEX (Arftet al 1999)

Finally there is a need to extrapolate experimentalresults over longer time scales and over larger landscapesand regions An excellent way to do so is to combineexperimentation with measurement of ecosystem proper-ties and processes along environmental transects such asthe IGBPndashGCTE Terrestrial Transects (Steffen et al1999) The transects are assumed to mirror longer-term(decades to centuries) responses of vegetation and soils toenvironmental conditions whereas warming experimentsreflect responses over the shorter-term (years to decades)

878 BioScience October 2000 Vol 50 No 10

Articles

Table 1 Comparison of currently used methods for experimental warming of terrestrial ecosystems

Method Mechanism of warming Advantages Disadvantages Examples

Field greenhouse ldquoGreenhouse warmingrdquo (ie Simple and inexpensive Little or no temperature Chapin et al 1995reflection of reradiated infrared requires no electrical power control and large tempera- Shaver et al 1998energy and reduced advective ture variability other arti- Jonasson et al 1999energy loss) facts include altered light

wind humidity and precipi- tation regimes

Passive Open-top Same as above Same as above Little or no temperature Marion et al 1997Chamber (OTC) control altered wind and Arft et al 1999

humidity only small areas can be manipulated uniformly

Active Open-top Same as above Precise control of air temp- Altered wind regime Norby et al 1997Chamber (OTC) plus warming by advection of erature or temperature humidity and

electrically heated forced difference may be combined evapotranspirationwarm air with CO2 control

Active Soil Warming Warming by conduction from Precise control of soil Altered soil moisture Peterjohn et al 1994buried electrical resistance temperature or temperature regime no effect on Ineson et al 1998bcables difference may be combined aboveground temperatures Hartley et al 1999

with greenhouses or OTCs

Electric infrared heat Warming by increased infrared Precise control of energy Warming depends entirely Harte et al 1995radiation input direct simulation of on radiation no change in Harte and Shaw 1995

global change in energy advective energy inputsbalance

Reciprocal or one-way Transplantation of plants Comparison of relatively Disturbance effects Ineson et al 1998atransplantation soils or whole plantndashsoil natural temperature gradients multiple environmental Grogan and Chapin1999 systems to warmer or cooler changes make it difficult

locations to assign specific causes to responses observed

Combined information from experiments and transectsmay thus be useful in distinguishing relatively short-termtransient changes in the experiments from longer-termnear-equilibrium changes For example global analyses ofprimary production (NPP Figure 1) and litter (the majorcontributor to R

h Figure 1) have shown that both are rea-sonably well-predicted by actual evapotranspiration(Lieth and Whittaker 1975 Meentemeyer 1984)Although short-term responses of NPP and R

h to experi-mental warming may not have the same relationshipswith actual evapotranspiration as do undisturbed ecosys-tems convergence toward the predicted relationshipmight be a useful measure of expected future change inthe experiments

Ultimately experimental and analytical approachesshould be combined with large-scale monitoring of vari-ables and processes that experimental studies have identi-fied as critical indicators of ecosystem change The combi-nation of global monitoring with ongoing experimentaldata acquisition metanalysis modeling and other datasynthesis efforts would provide an early warning system todetect and potentially mitigate these ecosystem responses

Field stations long-term ecological research sites andnature reserves already exist in many relatively unman-aged landscapes Collecting simple baseline data and mon-itoring annually or even less frequently would not be veryexpensive although data quality may be highly variable(eg Stohlgren et al 1995) Efforts to coordinate data andestablish protocols for data sharing might be daunting butthe resulting information would be immensely valuableresponses of species communities and biogeochemicalprocesses to climate change could potentially be detecteddecades earlier than they would be otherwise

Interactions with other drivers of globalchangeChanges in temperature are now occurring simultaneous-ly with other types of global change It will not be possibleto fully understand and predict ecosystem responses totemperature change without taking into account the inter-actions with the other components of global environmen-tal change (Koch and Mooney 1996 Sellers et al 1997Vitousek et al 1997 Luo and Mooney 1999) For exampleincreases in atmospheric CO

2 may modify plant tissuechemistry and ecosystem water balance both of whichmay in turn modify responses to warming Length of thegrowing season may increase with warming or it maydecrease as drought duration increases altering NPP NEPand other processes Many regions of the globe are experi-encing increased levels of nitrogen deposition and thenature of the nitrogen economy of an ecosystem is animportant factor in determining responses to other envi-ronmental factors (Aber et al 1998)

Changes in land use and disturbance regime may alsoinfluence responses to warming For instance land clear-ing over very large areas may modify local temperature or

precipitation regimes beyond what would happen withclimate change alone Similarly changes in fire regimescould be caused by and affect ecosystem responses towarming The spread of invasive species will also be affect-ed by both warming and land use change

Some interactions between warming and other globalchanges are predicted more readily than others For exam-ple outbreaks of insects or pathogens that kill the dominantplants and increase fire probability would lead to dramati-cally different results than the phased changes in response towarming alone We must expect some surprises

Clearly it will not be possible to incorporate all impor-tant dimensions of global change into experiments It willrarely be feasible to include more than one or two factorsin addition to temperature (eg CO2 concentration ornitrogen level) as experimental factors (Schulze et al1999) However through the use of combinations ofexperiments gradient studies remote sensing and simu-lation modeling it should be possible to make substantialprogress toward developing a mechanistic understandingof the most important interactions (Canadell et al 2000)

Although much current global change research focuseson feedbacks from terrestrial ecosystems to climatethrough greenhouse gases important feedbacks also existthrough changes in vegetation structure surface energybalance and water balance Leaves in the canopy act assurfaces for energy and mass exchanges with the atmos-phere and changes in canopy density and transpiration inresponse to warming will change the energy and massfluxes (McFadden et al 1998) One of the most importantand obvious effects of changes in the canopy is its albedo(ie the fraction of incoming solar radiation that isreflected) Decreasing albedo through increased leaf den-sity will increase net radiation (solar radiation absorbedby the system) The increased net radiation must be dissi-pated through one or a combination of three major ener-gy pathways or fluxes sensible heat (actual warming of theair) latent heat (evapotranspiration) and ground heat(warming the soil) If increased evapotranspirationdecreases soil moisture and species changes result in bet-ter stomatal control of water loss the result will beincreased sensible heat flux to the atmosphere Foley et al(1994) illustrated the importance of such a change duringthe last temperature optimum in the Holocene approxi-mately 6000 years ago Latitudinal treeline advanced sig-nificantly greatly reducing albedo The increased soil tem-perature and sensible heat flux from this vegetationchange was as great as the solar forcing believed to beresponsible for initiating the climate change Similarlyfurther loss of soil moisture in dry systems will decreaselatent heat flux with more energy dissipated as sensibleheat Current understanding of these types of feedbackson climate change (ie those operating through changesin surface energy and water balance) is less advanced thanthat of feedbacks through greenhouse gas fluxes andshould be a particularly high priority for future research

October 2000 Vol 50 No 10 BioScience 879

Articles

ConclusionsBecause temperature has a direct effect on virtually allecosystem processes responses of terrestrial ecosystems toglobal warming will be even more complex and difficult topredict than responses to increased atmospheric CO

2con-

centrations Nonetheless a useful conceptual frameworkfor analysis comparison and prediction of responses towarming is beginning to emerge (see box this page) Fornow the principal applications of this framework are as astatement of the major dimensions of complexity of theecosystem warming response and as a reminder of themajor constraints on long-term predictions based on rela-tively small-area short-term experimental studies A keyneed is to further develop our understanding of how ini-tial climate biogeochemical states and species character-istics might be used to generate long-term predictions ofrelative sensitivity to climate change Such a frameworkwould serve as a heuristic guide to new research

We also need improved models of the temporalsequence of ecosystem responses to temperature changebecause it is increasingly clear that long-term responses towarming and increased CO

2may differ greatly in both

magnitude and direction from initial responses Ecosys-tem-level experiments are a useful tool for developing thisunderstanding a broader array of experiments in con-trasting ecosystem types is needed including both whole-system manipulations and more focused experimentaltreatments Also needed are the transect studies long-term monitoring and the development of research net-works that will allow spatial extrapolation and validationof predictions based on intensive experimental studies

AcknowledgmentsThis paper contributes to the Global Change and Terres-trial Ecosystems (GCTE) core project of InternationalGeospherendashBiosphere Program (IGBP) and to the activi-ties of the Network of Experimental Warming Studies(GCTE-NEWS) The work was initiated with the supportof the National Center for Ecological Analysis and Synthe-sis a center funded by NSF (Grant DEB-94-21535) theUniversity of CaliforniandashSanta Barbara the CaliforniaResources Agency and the California Environmental Pro-tection Agency Additional support was provided by NSFgrant DEB-9730110 by NASA through a grant to GCTEand by USDA Forest Service

References citedAber J McDowell W Nadelhoffer K McGill A Berntson G Kamaka M

McNulty S Currie W Rustad L Fernandez I 1998 Nitrogen saturation

in forest ecosystems Hypotheses revisited BioScience 48 921ndash934Arft AM et al 1999 Response patterns of tundra plant species to experi-

mental warming A meta-analysis of the International Tundra Experi-

ment Ecological Monographs 69 491ndash512

Arnqvist FR Wooster D 1995 Meta-analysismdashsynthesizing research find-

ings in ecology and evolution Trends in Ecology and Evolution 10236ndash240

880 BioScience October 2000 Vol 50 No 10

Articles

Conceptual framework for analysis of ecosystem response to changing temperature

A Complexity Spatial and temporal variation Global tempera-

ture changes will not be uniform with mostregions warming by varying amounts but otherscooling These changes will occur over a range ofnatural temperature regimes and annual temper-ature variations that are greater than the expect-ed average temperature change

Direct and indirect effects Because temperatureaffects virtually all ecosystem processes indirecttemperature effects resulting from interactionsamong processes are likely to be very importantespecially in the long term

Variable time constants of response Diverseecosystem processes have different time constantsfor response to temperature Thus initial overallecosystem responses to change are not reliableindicators of long-term responses because theyrepresent only part of the total response

B Responses Initial response a function of initial conditions The

same temperature change applied to differentecosystems will elicit different responses depend-ing on initial position on the temperatureresponse surface (direct effects) and on initialbiogeochemical conditions and composition(indirect effects)

Changing sequence of controls The dominant con-trols over ecosystem response to temperature willchange over time as different processes change atdifferent rates This changing sequence will notnecessarily be the same in all ecosystems

Changing direction as well as magnitude ofresponse Initial losses or gains of carbon by theecosystem may be reversed over time as indirecteffects and slowly changing components of theecosystem come into play

Species changes can affect biogeochemistryChanges in species composition and abundancewill affect ecosystem functions (net primary pro-duction carbon and nutrient cycling energyfluxes) through changes in tissue chemistrydemography (both parts and individuals)turnover rates and vegetation structure (physiog-nomy leaf area index)

Response is predictable (at least qualitatively) frominitial conditions Key variables include initialtemperature regime initial carbon and nutrientpool sizes and turnover rates and initial popula-tion dynamics of vegetation

Combined information from experiments and transectsmay thus be useful in distinguishing relatively short-termtransient changes in the experiments from longer-termnear-equilibrium changes For example global analyses ofprimary production (NPP Figure 1) and litter (the majorcontributor to R

h Figure 1) have shown that both are rea-sonably well-predicted by actual evapotranspiration(Lieth and Whittaker 1975 Meentemeyer 1984)Although short-term responses of NPP and R

h to experi-mental warming may not have the same relationshipswith actual evapotranspiration as do undisturbed ecosys-tems convergence toward the predicted relationshipmight be a useful measure of expected future change inthe experiments

Ultimately experimental and analytical approachesshould be combined with large-scale monitoring of vari-ables and processes that experimental studies have identi-fied as critical indicators of ecosystem change The combi-nation of global monitoring with ongoing experimentaldata acquisition metanalysis modeling and other datasynthesis efforts would provide an early warning system todetect and potentially mitigate these ecosystem responses

Field stations long-term ecological research sites andnature reserves already exist in many relatively unman-aged landscapes Collecting simple baseline data and mon-itoring annually or even less frequently would not be veryexpensive although data quality may be highly variable(eg Stohlgren et al 1995) Efforts to coordinate data andestablish protocols for data sharing might be daunting butthe resulting information would be immensely valuableresponses of species communities and biogeochemicalprocesses to climate change could potentially be detecteddecades earlier than they would be otherwise

Interactions with other drivers of globalchangeChanges in temperature are now occurring simultaneous-ly with other types of global change It will not be possibleto fully understand and predict ecosystem responses totemperature change without taking into account the inter-actions with the other components of global environmen-tal change (Koch and Mooney 1996 Sellers et al 1997Vitousek et al 1997 Luo and Mooney 1999) For exampleincreases in atmospheric CO

2 may modify plant tissuechemistry and ecosystem water balance both of whichmay in turn modify responses to warming Length of thegrowing season may increase with warming or it maydecrease as drought duration increases altering NPP NEPand other processes Many regions of the globe are experi-encing increased levels of nitrogen deposition and thenature of the nitrogen economy of an ecosystem is animportant factor in determining responses to other envi-ronmental factors (Aber et al 1998)

Changes in land use and disturbance regime may alsoinfluence responses to warming For instance land clear-ing over very large areas may modify local temperature or

precipitation regimes beyond what would happen withclimate change alone Similarly changes in fire regimescould be caused by and affect ecosystem responses towarming The spread of invasive species will also be affect-ed by both warming and land use change

Some interactions between warming and other globalchanges are predicted more readily than others For exam-ple outbreaks of insects or pathogens that kill the dominantplants and increase fire probability would lead to dramati-cally different results than the phased changes in response towarming alone We must expect some surprises

Clearly it will not be possible to incorporate all impor-tant dimensions of global change into experiments It willrarely be feasible to include more than one or two factorsin addition to temperature (eg CO2 concentration ornitrogen level) as experimental factors (Schulze et al1999) However through the use of combinations ofexperiments gradient studies remote sensing and simu-lation modeling it should be possible to make substantialprogress toward developing a mechanistic understandingof the most important interactions (Canadell et al 2000)

Although much current global change research focuseson feedbacks from terrestrial ecosystems to climatethrough greenhouse gases important feedbacks also existthrough changes in vegetation structure surface energybalance and water balance Leaves in the canopy act assurfaces for energy and mass exchanges with the atmos-phere and changes in canopy density and transpiration inresponse to warming will change the energy and massfluxes (McFadden et al 1998) One of the most importantand obvious effects of changes in the canopy is its albedo(ie the fraction of incoming solar radiation that isreflected) Decreasing albedo through increased leaf den-sity will increase net radiation (solar radiation absorbedby the system) The increased net radiation must be dissi-pated through one or a combination of three major ener-gy pathways or fluxes sensible heat (actual warming of theair) latent heat (evapotranspiration) and ground heat(warming the soil) If increased evapotranspirationdecreases soil moisture and species changes result in bet-ter stomatal control of water loss the result will beincreased sensible heat flux to the atmosphere Foley et al(1994) illustrated the importance of such a change duringthe last temperature optimum in the Holocene approxi-mately 6000 years ago Latitudinal treeline advanced sig-nificantly greatly reducing albedo The increased soil tem-perature and sensible heat flux from this vegetationchange was as great as the solar forcing believed to beresponsible for initiating the climate change Similarlyfurther loss of soil moisture in dry systems will decreaselatent heat flux with more energy dissipated as sensibleheat Current understanding of these types of feedbackson climate change (ie those operating through changesin surface energy and water balance) is less advanced thanthat of feedbacks through greenhouse gas fluxes andshould be a particularly high priority for future research

October 2000 Vol 50 No 10 BioScience 879

Articles

ConclusionsBecause temperature has a direct effect on virtually allecosystem processes responses of terrestrial ecosystems toglobal warming will be even more complex and difficult topredict than responses to increased atmospheric CO

2con-

centrations Nonetheless a useful conceptual frameworkfor analysis comparison and prediction of responses towarming is beginning to emerge (see box this page) Fornow the principal applications of this framework are as astatement of the major dimensions of complexity of theecosystem warming response and as a reminder of themajor constraints on long-term predictions based on rela-tively small-area short-term experimental studies A keyneed is to further develop our understanding of how ini-tial climate biogeochemical states and species character-istics might be used to generate long-term predictions ofrelative sensitivity to climate change Such a frameworkwould serve as a heuristic guide to new research

We also need improved models of the temporalsequence of ecosystem responses to temperature changebecause it is increasingly clear that long-term responses towarming and increased CO

2may differ greatly in both

magnitude and direction from initial responses Ecosys-tem-level experiments are a useful tool for developing thisunderstanding a broader array of experiments in con-trasting ecosystem types is needed including both whole-system manipulations and more focused experimentaltreatments Also needed are the transect studies long-term monitoring and the development of research net-works that will allow spatial extrapolation and validationof predictions based on intensive experimental studies

AcknowledgmentsThis paper contributes to the Global Change and Terres-trial Ecosystems (GCTE) core project of InternationalGeospherendashBiosphere Program (IGBP) and to the activi-ties of the Network of Experimental Warming Studies(GCTE-NEWS) The work was initiated with the supportof the National Center for Ecological Analysis and Synthe-sis a center funded by NSF (Grant DEB-94-21535) theUniversity of CaliforniandashSanta Barbara the CaliforniaResources Agency and the California Environmental Pro-tection Agency Additional support was provided by NSFgrant DEB-9730110 by NASA through a grant to GCTEand by USDA Forest Service

References citedAber J McDowell W Nadelhoffer K McGill A Berntson G Kamaka M

McNulty S Currie W Rustad L Fernandez I 1998 Nitrogen saturation

in forest ecosystems Hypotheses revisited BioScience 48 921ndash934Arft AM et al 1999 Response patterns of tundra plant species to experi-

mental warming A meta-analysis of the International Tundra Experi-

ment Ecological Monographs 69 491ndash512

Arnqvist FR Wooster D 1995 Meta-analysismdashsynthesizing research find-

ings in ecology and evolution Trends in Ecology and Evolution 10236ndash240

880 BioScience October 2000 Vol 50 No 10

Articles

Conceptual framework for analysis of ecosystem response to changing temperature

A Complexity Spatial and temporal variation Global tempera-

ture changes will not be uniform with mostregions warming by varying amounts but otherscooling These changes will occur over a range ofnatural temperature regimes and annual temper-ature variations that are greater than the expect-ed average temperature change

Direct and indirect effects Because temperatureaffects virtually all ecosystem processes indirecttemperature effects resulting from interactionsamong processes are likely to be very importantespecially in the long term

Variable time constants of response Diverseecosystem processes have different time constantsfor response to temperature Thus initial overallecosystem responses to change are not reliableindicators of long-term responses because theyrepresent only part of the total response

B Responses Initial response a function of initial conditions The

same temperature change applied to differentecosystems will elicit different responses depend-ing on initial position on the temperatureresponse surface (direct effects) and on initialbiogeochemical conditions and composition(indirect effects)

Changing sequence of controls The dominant con-trols over ecosystem response to temperature willchange over time as different processes change atdifferent rates This changing sequence will notnecessarily be the same in all ecosystems

Changing direction as well as magnitude ofresponse Initial losses or gains of carbon by theecosystem may be reversed over time as indirecteffects and slowly changing components of theecosystem come into play

Species changes can affect biogeochemistryChanges in species composition and abundancewill affect ecosystem functions (net primary pro-duction carbon and nutrient cycling energyfluxes) through changes in tissue chemistrydemography (both parts and individuals)turnover rates and vegetation structure (physiog-nomy leaf area index)

Response is predictable (at least qualitatively) frominitial conditions Key variables include initialtemperature regime initial carbon and nutrientpool sizes and turnover rates and initial popula-tion dynamics of vegetation

ConclusionsBecause temperature has a direct effect on virtually allecosystem processes responses of terrestrial ecosystems toglobal warming will be even more complex and difficult topredict than responses to increased atmospheric CO

2con-

centrations Nonetheless a useful conceptual frameworkfor analysis comparison and prediction of responses towarming is beginning to emerge (see box this page) Fornow the principal applications of this framework are as astatement of the major dimensions of complexity of theecosystem warming response and as a reminder of themajor constraints on long-term predictions based on rela-tively small-area short-term experimental studies A keyneed is to further develop our understanding of how ini-tial climate biogeochemical states and species character-istics might be used to generate long-term predictions ofrelative sensitivity to climate change Such a frameworkwould serve as a heuristic guide to new research

We also need improved models of the temporalsequence of ecosystem responses to temperature changebecause it is increasingly clear that long-term responses towarming and increased CO

2may differ greatly in both

magnitude and direction from initial responses Ecosys-tem-level experiments are a useful tool for developing thisunderstanding a broader array of experiments in con-trasting ecosystem types is needed including both whole-system manipulations and more focused experimentaltreatments Also needed are the transect studies long-term monitoring and the development of research net-works that will allow spatial extrapolation and validationof predictions based on intensive experimental studies

AcknowledgmentsThis paper contributes to the Global Change and Terres-trial Ecosystems (GCTE) core project of InternationalGeospherendashBiosphere Program (IGBP) and to the activi-ties of the Network of Experimental Warming Studies(GCTE-NEWS) The work was initiated with the supportof the National Center for Ecological Analysis and Synthe-sis a center funded by NSF (Grant DEB-94-21535) theUniversity of CaliforniandashSanta Barbara the CaliforniaResources Agency and the California Environmental Pro-tection Agency Additional support was provided by NSFgrant DEB-9730110 by NASA through a grant to GCTEand by USDA Forest Service

References citedAber J McDowell W Nadelhoffer K McGill A Berntson G Kamaka M

McNulty S Currie W Rustad L Fernandez I 1998 Nitrogen saturation

in forest ecosystems Hypotheses revisited BioScience 48 921ndash934Arft AM et al 1999 Response patterns of tundra plant species to experi-

mental warming A meta-analysis of the International Tundra Experi-

ment Ecological Monographs 69 491ndash512

Arnqvist FR Wooster D 1995 Meta-analysismdashsynthesizing research find-

ings in ecology and evolution Trends in Ecology and Evolution 10236ndash240

880 BioScience October 2000 Vol 50 No 10

Articles

Conceptual framework for analysis of ecosystem response to changing temperature

A Complexity Spatial and temporal variation Global tempera-

ture changes will not be uniform with mostregions warming by varying amounts but otherscooling These changes will occur over a range ofnatural temperature regimes and annual temper-ature variations that are greater than the expect-ed average temperature change

Direct and indirect effects Because temperatureaffects virtually all ecosystem processes indirecttemperature effects resulting from interactionsamong processes are likely to be very importantespecially in the long term

Variable time constants of response Diverseecosystem processes have different time constantsfor response to temperature Thus initial overallecosystem responses to change are not reliableindicators of long-term responses because theyrepresent only part of the total response

B Responses Initial response a function of initial conditions The

same temperature change applied to differentecosystems will elicit different responses depend-ing on initial position on the temperatureresponse surface (direct effects) and on initialbiogeochemical conditions and composition(indirect effects)

Changing sequence of controls The dominant con-trols over ecosystem response to temperature willchange over time as different processes change atdifferent rates This changing sequence will notnecessarily be the same in all ecosystems

Changing direction as well as magnitude ofresponse Initial losses or gains of carbon by theecosystem may be reversed over time as indirecteffects and slowly changing components of theecosystem come into play

Species changes can affect biogeochemistryChanges in species composition and abundancewill affect ecosystem functions (net primary pro-duction carbon and nutrient cycling energyfluxes) through changes in tissue chemistrydemography (both parts and individuals)turnover rates and vegetation structure (physiog-nomy leaf area index)

Response is predictable (at least qualitatively) frominitial conditions Key variables include initialtemperature regime initial carbon and nutrientpool sizes and turnover rates and initial popula-tion dynamics of vegetation

Billings WD 1987 Carbon balance of Alaskan tundra and taiga ecosys-tems Past present and future Quaternary Science Reviews 6165ndash177

Braswell BH Schimel DS Linder E Moore B III 1997 The response ofglobal terrestrial ecosystems to interannual temperature variabilityScience 278 870ndash872

Briones MJI Ineson P Poskitt J 1998 Climate change and Cognettia sphag-netorum Effects on carbon leaching in organic soils Functional Ecol-ogy 12 528ndash535

Canadell JG et al 2000 Carbon metabolism of the terrestrial biosphere Amulti-technique approach for improved understanding Ecosystems 3115ndash130

Chapin FS III 1983 Direct and indirect effects of temperature on arcticplants Polar Biology 2 47ndash52

Chapin FS III Shaver GR Giblin AE Nadelhoffer KJ Laundre JA 1995Responses of arctic tundra to experimental and observed changes inclimate Ecology 76 694ndash711

Chapin FS III Walker BH Hobbs RJ Hooper DU Lawton JH Sala OETilman D 1997 Biotic control over the functioning of ecosystems Sci-ence 277 500ndash504

Clark FE Rosswall T eds 1981 Terrestrial Nitrogen Cycles ProcessesEcosystem Strategies and Management Impacts Ecological BulletinsNo 33 Stockholm Swedish Natural Science Research Council

deValpine P Harte J In press Effects of warming on a montane meadowecosystem How species responses comprise the ecosystem responseEcology

Foley JA Kutzbach JE Coe MT Levis S 1994 Feedbacks between climateand boreal forest during the Holocene epoch Nature 371 52ndash54

Giblin AE Nadelhoffer KJ Shaver GR Laundre JA McKerrow AJ 1991Biogeochemical diversity along a riverside toposequence in arctic Alas-ka Ecological Monographs 61 415ndash436

Grogan P Chapin FS III 1999 Arctic soil respiration Effects of climate andvegetation depend on season Ecosystems 2 451ndash459

Gurevitch J Hedges LV 1999 Statistical issues in conducting ecologicalmeta-analyses Ecology 80 1142ndash1149

Harte J Shaw R 1995 Shifting dominance within a montane vegetationcommunity Results of a climate-warming experiment Science 267876ndash880

Harte J Torn MS Chang F-R Feifarek B Kinzig AP Shaw R Shen K 1995Global warming and soil microclimate Results from a meadow-warm-ing experiment Ecological Applications 5 132ndash150

Hartley AE Neill C Melillo JM Crabtree R Bowles FP 1999 Plant perfor-mance and soil nitrogen mineralization in response to simulated cli-mate change in subarctic dwarf shrub heath Oikos 86 331ndash343

Henry GHR Molau U 1997 Tundra plants and climate change The Inter-national Tundra Experiment (ITEX) Global Change Biology 3 (Sup-plement 1) 1ndash9

Herbert DA Rastetter EB Shaver GR Aringgren G 1999 Effects of plantgrowth characteristics on biogeochemistry and community composi-tion in a changing climate Ecosystems 2 367ndash382

Houghton JT Meira Filho LG Bruce J Lee H Callander BA Haites E Har-ris N Maskell K 1995 Climate Change 1994 Radiative Forcing of Cli-mate Change and An Evaluation of the IPCC IS92 Emission ScenariosReports of Working Groups I and III of the Intergovernmental Panelon Climate Change Cambridge (UK) Cambridge University Press

Houghton JT Meira Filho LG Callander BA Harris N Kattenberg AMaskell K 1996 Climate Change 1995 The Science of ClimateChange Contribution of WG1 to the Second Assessment Report of theIntergovernmental Panel on Climate Change Cambridge (UK) Cam-bridge University Press

Ineson P Taylor K Harrison AF Poskitt J Benham DG Tipping E Woof C1998a Effects of climate change on nitrogen dynamics in upland soilsI A transplant approach Global Change Biology 4 143ndash152

Ineson P Benham DG Poskitt J Harrison AF Taylor K Woods C 1998bEffects of climate change on nitrogen dynamics in upland soils II Asoil warming study Global Change Biology 4 153ndash162

Ingram JSI Canadell J Elliott T Hunt LA Linder S Murdiyarso D StaffordSmith M Valentin C 1999 Networks and consortia Pages 45ndash65 in

Walker B Steffen W Canadell J Ingram J eds The Terrestrial Bios-phere and Global Change Implications for Natural and ManagedEcosystems Cambridge (UK) Cambridge University Press

Jonasson S Shaver GR 1999 Within-stand nutrient cycling in arctic andboreal herbaceous and forested wetlands Ecology 80 2139ndash2150

Jonasson S Michelsen A Schmidt IK Nielsen EV 1999 Responses inmicrobes and plants to changed temperature nutrient and lightregimes in the arctic Ecology 80 1828ndash1843

Koch GW Mooney HA eds 1996 Carbon Dioxide and Terrestrial Ecosys-tems San Diego Academic Press

Larcher W 1995 Physiological Plant Ecology 3rd ed New York Springer-Verlag

Lieth H Whittaker RH eds 1975 Primary Productivity of the BiosphereSpringer-Verlag Ecological Studies Series Vol 14 New York Springer-Verlag

Luo Y Mooney HA eds 1999 Carbon Dioxide and Environmental StressSan Diego Academic Press

Marion GM et al 1997 Open-top designs for manipulating field temper-atures in high-latitude ecosystems Global Change Biology 3 (Supple-ment 1) 20ndash32

McFadden JP Chapin FS III Hollinger DY 1998 Subgrid-scale variabilityin the surface energy balance of arctic tundra Journal of GeophysicalResearch 103 28 947-28

McKane RB Rastetter EB Melillo JM Shaver GR Hopkinson CS Fernan-dez DN 1995 Effects of global change on carbon storage in tropicalforests of South America Global Biogeochemical Cycles 9 329ndash350

McKane RB Rastetter EB Shaver GR Nadelhoffer KJ Giblin AE LaundreJA 1997a Climatic effects on tundra carbon storage inferred fromexperimental data and a model Ecology 78 1170ndash1187

______ 1997b Reconstruction and analysis of historic changes in carbonstorage in arctic tundra Ecology 78 1188ndash1198

Meentemeyer V 1984 The geography of organic decomposition ratesAnnals of the Association of American Geographers 74 551ndash560

Melillo JM Callaghan TV Woodward FI Salati E Sinha SK 1990 Climatechangemdasheffects on ecosystems Pages 282ndash310 in Houghton JT Jenk-ins GJ Ephraums JJ eds Climate ChangemdashThe IPCC ScientificAssessment Cambridge (UK) Cambridge University Press

Melillo JM Kicklighter DW McGuire AD Peterjohn WT Newkirk K 1995Global change and its effects on soil organic carbon stocks Pages175ndash189 in Zepp R Sonntag Ch eds Report of the Dahlem Workshopon the Role of Nonliving Organic Matter in the Earthrsquos Carbon CycleNew York John Wiley amp Sons

Melillo JM Newkirk KM Catricala CE Steudler PA Aber JP Bowles FP Inpress Chapter 6 Soil warming In Foster D Aber J eds Forest Land-scape Dynamics in Southern New England Ecosystem Structure andFunction as a Consequence of 5000 Years of Change Synthesis Volumeof the Harvard Forest LTER Program Oxford Oxford University Press

Mooney HA Winner WE Pell EJ eds 1991 Response of Plants to Multi-ple Stresses San Diego Academic Press

Mooney H Canadell J Chapin FS III Ehleringer J Koumlrner Ch McMurtrieR Parton W Pitelka L Schulze E-D 1999 Ecosystem physiologyresponses to global change Pages 141ndash149 in Walker BH WL SteffenCanadell J Ingram JSI eds The Terrestrial Biosphere and GlobalChange Implications for Natural and Managed Ecosystems Cam-bridge (UK) Cambridge University Press

Nadelhoffer KJ Linkins AE Giblin AE Shaver GR 1992 Microbialprocesses and plant nutrient availability in arctic soils Pages 281ndash300in Chapin FS III Jefferies R Reynolds J Shaver GR Svoboda J eds Arc-tic Ecosystems in a Changing Climate An Ecophysiological Perspec-tive New York Academic Press

Norby RJ Edwards NT Riggs JS Abner CH Wullschleger SD GundersonCA OrsquoNeill EG 1997 Temperature-controlled open-top chambers forglobal change research Global Change Biology 3 359ndash367

Peterjohn WT Melillo JM Steudler PA Newkirk KM Bowles FP Aber JD1994 The response of trace gas fluxes and N availability to elevated soiltemperatures Ecological Applications 4 617ndash625

Rastetter EB Ryan MG Shaver GR Melillo JM Nadelhoffer KJ Hobbie JEAber JD 1991 A general biogeochemical model describing the

October 2000 Vol 50 No 10 BioScience 881

Articles

responses of the C and N cycles in terrestrial ecosystems to changes inCO2 climate and N deposition Tree Physiology 9 101ndash126

Rastetter EB Aringgren GI Shaver GR 1997 Responses of N-limited ecosys-tems to increased CO2 Application of a balanced-nutrition coupled-element-cycles model Ecological Applications 7 444ndash460

Robinson CH Wookey PA Lee JA Callaghan TV Press MC 1998 Plantcommunity responses to simulated environmental change at a higharctic polar semi-desert Ecology 79 856ndash866

Saleska S Harte J Torn M 1999 The effect of experimental ecosystemwarming on CO2 fluxes in a montane meadow Global Change Biolo-gy 5 125ndash141

Schimel DS VEMAP Participants Braswell BH 1997 Continental scalevariability in ecosystem processes Models data and the role of distur-bance Ecological Monographs 76 251ndash271

Schulze E-D Canadell J Scholes R Ehleringer J Hunt T Sutherst B ChapinFS III Steffen W 1999 The study of ecosystems in the context of glob-al change Pages 19ndash44 in Walker BH WL Steffen Canadell J IngramJSI eds The Terrestrial Biosphere and Global Change Implications forNatural and Managed Ecosystems Cambridge (UK) Cambridge Uni-versity Press

Sellers PJ et al 1997 Modeling the exchanges of energy water and carbonbetween continents and the atmosphere Science 275 502ndash509

Shaver GR Billings WD Chapin FS III Giblin AE Nadelhoffer KJ OechelWC Rastetter EB 1992 Global change and the carbon balance of arc-tic ecosystems BioScience 42 433ndash441

Shaver GR Johnson LC Cades DH Murray G Laundre JA Rastetter EBNadelhoffer KJ Giblin AE 1998 Biomass accumulation and CO2 fluxin three Alaskan wet sedge tundras Responses to nutrients tempera-ture and light Ecological Monographs 68 75ndash99

Smith TM Shugart HH Woodward FI eds 1997 Plant Functional TypesTheir Relevance to Ecosystem Properties and Global Change Cam-bridge (UK) Cambridge University Press

Steffen WL Scholes RJ Valentin C Hang XZ Menaut J-C Schulze E-D1999 The IGBP terrestrial transects Pages 66ndash87 in Walker B Steffen

W Canadell J Ingram J eds The Terrestrial Biospehere and GlobalChange Implications for Natural and Managed Ecosystems Cam-bridge (UK) Cambridge University Press

Stohlgren TJ Quinn JF Ruggiero M Waggoner GS 1995 Status of bioticinventories in US national parks Biological Conservation 71 97ndash106

Vitousek PM Mooney HA Lubchenco J Melillo JM 1997 Human domi-nation of Earthrsquos ecosystems Science 277 494ndash499

Vourlitis GL Oechel WC 1997 Landscape-scale CO2 water vapor andenergy flux of moist-wet coastal tundra ecosystems over two growingseasons Journal of Ecology 85 575ndash590

Vourlitis GL Oechel WC 1999 Eddy covariance measurements of CO2

and energy fluxes of an Alaskan tussock tundra ecosystem Ecology 80686ndash701

Waelbroeck C Moufrey P Oechel WC Hastings S Vourlitis G 1997 Theimpact of permafrost thawing on the carbon dynamics of tundra Geo-physical Research Letters 24 229ndash232

Walker B Steffen W 1999 The nature of global change Pages 1ndash19 inWalker BH Steffen WL Canadell J Ingram JSI eds The TerrestrialBiosphere and Global Change Implications for Natural and ManagedEcosystems Cambridge (UK) Cambridge University Press

Williams M Rastetter EB Fernandez DM Goulden ML Shaver GR John-son LC 1997 Predicting gross primary productivity in terrestrialecosystems Ecological Applications 7 882ndash894

Woodwell GM 1995 Biotic feedbacks from the warming of the earthPages 3ndash21 in Woodwell GM McKenzie FT eds Biotic Feedbacks inthe Global Climatic System Will the Warming Feed the WarmingOxford Oxford University Press

Woodwell GM McKenzie FT eds 1995 Biotic Feedbacks in the Global Cli-matic System Will the Warming Feed the Warming Oxford OxfordUniversity Press

Woodwell GM Whittaker RW 1968 Primary production in terrestrialecosystems American Zoologist 8 19ndash30

882 BioScience October 2000 Vol 50 No 10

Articles