Predicting pan-tropical climate change induced forest stock gains and losses—implications for REDD

This content has been downloaded from IOPscience Please scroll down to see the full text

Download details

IP Address 95154199100

This content was downloaded on 08102013 at 2332

Please note that terms and conditions apply

Predicting pan-tropical climate change induced forest stock gains and lossesmdashimplications for

REDD

View the table of contents for this issue or go to the journal homepage for more

2010 Environ Res Lett 5 014013

(httpiopscienceioporg1748-932651014013)

Home Search Collections Journals About Contact us My IOPscience

IOP PUBLISHING ENVIRONMENTAL RESEARCH LETTERS

Environ Res Lett 5 (2010) 014013 (15pp) doi1010881748-932651014013

Predicting pan-tropical climate changeinduced forest stock gains andlossesmdashimplications for REDDMarlies Gumpenberger1 Katrin Vohland13 Ursula Heyder1Benjamin Poulter14 Kirsten Macey2 Anja Rammig1Alexander Popp1 and Wolfgang Cramer1

1 Potsdam Institute for Climate Impact Research (PIK) Telegraphenberg A 62 D-14473Potsdam Germany2 Climate Analytics Telegraphenberg 14412 Potsdam Germany

E-mail MarliesGumpenbergerpik-potsdamde

Received 14 November 2009Accepted for publication 3 February 2010Published 16 February 2010Online at stacksioporgERL5014013

AbstractDeforestation is a major threat to tropical forests worldwide contributing up to one-fifth of globalcarbon emissions into the atmosphere Despite protection efforts deforestation of tropical forests hascontinued in recent years Providing incentives to reducing deforestation has been proposed in theUnited Nations Framework Convention on Climate Change (UNFCCC) Bali negotiations in 2007 todecelerate emissions from deforestation (REDDmdashreduced emissions from deforestation and forestdegradation) A number of methodological issues such as ensuring permanence establishing referenceemissions levels that do not reward business-as-usual and having a measuring reporting andverification system in place are essential elements in implementing successful REDD schemes Toassess the combined impacts of climate and land-use change on tropical forest carbon stocks in the 21stcentury we use a dynamic global vegetation model (LPJ DGVM) driven by five different climatechange projections under a given greenhouse gas emission scenario (SRES A2) and two contrastingland-use change scenarios We find that even under a complete stop of deforestation after the period ofthe Kyoto Protocol (post-2012) some countries may continue to lose carbon stocks due to climatechange Especially at risk is tropical Latin America although the presence and magnitude of the riskdepends on the climate change scenario By contrast strong protection of forests could increase carbonuptake in many tropical countries due to CO2 fertilization effects even under altered climate regimes

Keywords REDD (reduced emissions from deforestation and forest degradation) modelling carboncycle tropical forests climate change climate policy

1 Introduction

11 Deforestation and climate change

Deforestation is the largest source of emissions fromthe LULUCF (land use land-use change and forestry)

3 Present address Museum fur Naturkunde Leibniz Institute for Research onEvolution and Biodiversity at the Humboldt University Berlin Invalidenstraszlige43 10115 Berlin Germany4 Present address Swiss Federal Institute for Forest Snow and LandscapeResearch (WSL) Zurcherstrasse 111 CH-8903 Birmensdorf Switzerland

greenhouse gas inventory sector within the UNFCCC (UnitedNations Framework Convention on Climate Change) andaccounts for 12ndash20 of global anthropogenic greenhousegas emissions [1ndash4] Land-use change related fluxes to theatmosphere from the tropics have been estimated to be as highas 22 plusmn 06 Pg C yrminus1 for the 1990s [5] Recent estimatesfor carbon emissions from deforestation and forest degradationshow lower rates of 12 Pg C yrminus1 over the period 1997ndash2006 with additional 03 Pg C yrminus1 from tropical peatlandoxidation [3] Forest loss in Latin America accounts for 60

1748-932610014013+15$3000 copy 2010 IOP Publishing Ltd Printed in the UK1

Environ Res Lett 5 (2010) 014013 M Gumpenberger et al

of total tropical biome clearing (Brazil 48) Over one-third ofclearing occurs in Asia (Indonesia 13) and Africa contributes5 to the estimated loss of humid tropical forest cover [6]Agriculture logging and mining are the direct drivers oftropical deforestation and result from or are amplified bypopulation growth agricultural subsidies and infrastructureinvestment [7 8]

12 Policy incentives to reduce deforestation

Proposals to finance deforestation reduction have been debatedfor some years [9] More recently opportunities have arisento provide incentives for developing countries to reduceemissions from deforestation and forest degradation Whilethe severity of the expected impacts of climate changehas increased as described by the IPCC Fourth AssessmentReport [10 11] reducing emissions from deforestationis a cost-effective option for mitigating climate change(although over time marginal costs would rise) [12ndash14]The Bali Action Plan provided a mandate to consider thepolicy incentives to reduce emissions from deforestation andforest degradation (REDD) as part of the post-2012 climateregime

Full success of REDD would mean halting deforestationimmediately However even a reduction in deforestationrates is considered as progress [15] Without successfulimplementation of forest protection tropical deforestation islikely to continue throughout this century According to a studyby Kindermann et al [16] todayrsquos forest cover would shrinkby around 500 million hectares until 2100 without carbonprice incentive schemes on deforestation However thereare various methodological challenges in the implementationof an effective regime on REDD This includes establishingreference emission levels which do not reward business-as-usual address leakage or emissions displacement ensuringpolicies resulting in permanent emission reductions anddeveloping an effective measuring reporting and verificationsystem (MRV) [17ndash22]

13 Predicting future forest carbon stocks

While losses due to ongoing deforestation prevail in theinternational discussion on policy schemes climate changeincreasingly is acknowledged as a possible risk for forestcarbon stocks [23] The aim of this study is to givea first assessment of risks arising from climate change incombination with a successful REDD scheme Since futurechanges in forest integrity and carbon storage cannot beextrapolated linearly from current observations we use theadvanced dynamic global vegetation model LPJmL [24ndash26]to disentangle the success of REDD in terms of reduceddeforestation against the background of different climatechange scenarios on a country scale The different projectionsof reducing deforestation success are assessed by applying twoextreme land-use change scenarios In the first scenario forestsare completely protected in every country from 2012 onwardsIn the second scenario half of the forest area existing in 2012is deforested by the end of the twenty-first century with aconstant area deforested every year We set the year 2012

as earliest possible start point to stop deforestation becauseREDD mechanisms will not be implemented beyond pilotstudies before the expiration of the Kyoto Protocol We run theLPJmL model with IPCC AR4 climate change projections offive different general circulation models (GCMs) under forcingfrom SRES A2 emissions The results from this study could beof use for policy makers who need to evaluate climate changeinduced risks for REDD schemes

2 Data and methods

In this study we investigate the role of climate change anddeforestation on the development of future tropical forestcarbon stocks We applied the dynamic global vegetationmodel LPJmL (described in section 21) with two contrastingland-use change scenarios (section 22) and five climate changescenarios under SRES A2 emission trajectories (section 23)Simulations were conducted for the historic period and the21st century (section 24) The analysis was performed with afocus on tropical countries (more details on selected countriesin section 25)

21 LPJmL model

Process-based dynamic global vegetation models providean important perspective for understanding the combinedeffects of increasing levels of atmospheric CO2 watercycling and global warming on plant productivity andtheir component fluxes of water and carbon at spatiallydifferentiated scale The process-based LPJmL DGVM isa global grid-based biogeographyndashbiogeochemistry modelwhich has been comprehensively validated for a broad rangeof conditions and quantities [24ndash30] LPJmL realisticallyreproduces terrestrial carbon pool sizes and fluxes and thebiogeographical distribution of vegetation [26] The waterbalance computed by the model performs on the level of state-of-the-art global hydrological models [25] The representationof agricultural land allows for the quantification of the impactsof land use on water and carbon cycles [24]

The simulation in any grid cell is driven by input ofmonthly climatology soil type atmospheric CO2 concentra-tion and agricultural land use No ecosystem features areprescribed plant type presence and the associated carbonstocks arise as a function of the environment In ourcalculations LPJmL is run off-line therefore no feedbackmechanisms from vegetation to the atmosphere are consideredNatural vegetation is represented by nine different plantfunctional types (PFTs) of which two are herbaceous andseven woody Different PFTs coexist within each grid cellbut their abundance is constrained by climatic conditionsand competition Vegetation structure responds dynamicallyto changes in climate including invasion of new habitatsand dieback For the tropics the prevailing PFTs arelsquotropical broad-leaved evergreenrsquo trees lsquotropical broad-leavedraingreenrsquo trees and the C4 photosynthetic grasses LPJmLsimulates processes as photosynthesis and transpiration main-tenance and growth respiration and reproduction cost Netprimary production (NPP) is allocated to the different plant

2

Environ Res Lett 5 (2010) 014013 M Gumpenberger et al

compartments (vegetation carbon pool) and enters the litterand soil carbon pools due to litter-fall and mortality Firedisturbance is driven by a threshold litter load and a soilmoisture function [31]

As this study focuses on forests carbon stocks we donot simulate the 11 different crop functional types (CFTs)contained in LPJmL instead we use only one type ofagricultural land which is rain-fed managed grassland Naturalvegetation and managed grasslands are simulated as separatestands in each grid cell each having its own soil carbon andwater pools The annual fractional coverage of agriculturalland in each grid cell is provided by the land-use input toLPJmL If deforestation occurs natural vegetation is reducedand the deforested carbon is allocated to the litter pooleventually entering the soil carbon pool from where it isrespired back to the atmosphere The occurrence of fireleads to an alternative pathway allowing carbon to return tothe atmosphere directly from standing biomass or litter Ifagricultural land is abandoned forest regrowth occurs

22 Land-use change

Several global gridded datasets for historic land use have beendeveloped in recent years [32ndash35] The HYDEv30 historicland-use dataset [33 36] comprises cropland and pasture areasfrom the years 1700 to 2000 with decadal time-steps andwas used in this study to determine the fractions of naturalvegetation and agricultural land in each grid cell of LPJmLfor the historic period The land-use dataset is based onsatellite data and agricultural statistics from the United NationsFood and Agriculture Organization (FAO) and other sub-national land-use data Distribution of population densityland suitability distance to major rivers and natural land coverare used as weighting maps to allocate historical cropland(The HYDE dataset is available at ftpftpmnpnlhyde) Weaggregated the 5prime times 5prime (longitudelatitude) resolution data to30prime (05) which is the spatial resolution of the LPJmL inputdrivers Between the time-slices of each decade land-usechange was linearly interpolated for each grid cell to providea quasi-continuous yearly historical dataset We retaineddeforestation rates from 1990 to 2000 for the period from2001 to 2012 as for example Hansen et al [6] showed thatrates of clearing from 2000 to 2005 in the humid tropicalbiome remained comparable with those observed in the 1990sPost-2012 we applied two extreme land-use scenarios a forestprotection and a deforestation scenario In the protectionscenario we assume full forest protection where the share ofnatural vegetation in each grid cell is kept constant from 2012onwards In the deforestation scenario every year an equalfraction of natural vegetated land is converted to managedgrassland until only 50 of the natural coverage in 2012 isleft at the end of the 21st century which corresponds to a pan-tropical forest loss of 555 million hectares by 2100 (definingforest with a minimum tree canopy cover of 30) [37] Thedeforestation scenario after 2012 does not include regionallydifferentiated deforestation rates and land abandonment wasnot taken into account

23 Climate change and C O2 projections

Climate projections from five general circulation mod-els (GCMs) ECHAM5MPI-OM CONSECHO-G UKMO-HadCM3 GFDL-CM21 and NCARCCSM30 under forcingfrom the SRES A2 emission scenario were used [38] Themodels have been used in the World Climate ResearchProgrammersquos (WCRPrsquos) Coupled Model IntercomparisonProject phase 3 (CMIP3) (available from httpsesgllnlgov8443) carried out for the IPCC Fourth AssessmentReport [39] A documentation of all GCMs can befound at www-pcmdillnlgovipccmodel documentationipcc model documentationphp Predicted climate anomaliesof monthly fields of precipitation and surface air temperaturefor the years 1860ndash2100 are calculated for each of the fiveclimate models with respect to the reference period (1960ndash1990) Those anomalies are interpolated to 05 resolutionand are combined with the mean climatology for the referenceperiod of an extended CRU TS21 climate dataset [40 41]Table 1 gives an overview of the GCMs used in thisstudy including bias-corrected projections for temperature andprecipitation in the tropical zone For the SRES A2 scenarioall models simulate a temperature increase over land surfacesand broad spatial patterns of increase are similar betweenGCMs In contrast there are major differences between GCMsin projected changes in precipitation in which the regionalpatterns vary greatly (figure 1)

We ran the LPJmL model with CO2 concentrationsincreasing as they did for the IPCC SRES A2 emissionscenario which is 395 ppm in 2012 rising to 532 ppmin 2050 and reaching 847 ppm in 2099 The SRES A2scenario includes anthropogenic CO2 emissions from fossil-fuel consumption and land-use change projections for the21st century with a relative contribution from each sourceof about 95 and 5 respectively [38] The SRES A2 isone of the highest emission scenarios of the IPCC range ofprojections with increasing growth rates of greenhouse gasemissions during the course of the 21st century Howeverrecent observations show that growth rates of greenhouse gasemissions are extending beyond the upper boundary of theenvelope of IPCC emissions scenarios [42]

24 Simulation protocol

In most ecosystems carbon pools in soil and vegetation reachequilibrium only after a long time Therefore a 1000 year spin-up simulation with natural vegetation was carried out Thefirst spin-up was followed by a second spin-up for 398 yearswith natural vegetation and managed grassland using land-usepatterns from 1860 In the spin-ups LPJmL was driven withclimate data from the University of East Angliarsquos ClimaticResearch Unit (CRU) [40] with repeating cycles from 1901to 1930 and with pre-industrial CO2 concentrations After thespin-ups the simulations from 1871 to 2099 were conductedwith five IPCC AR4 climate change projections SRES A2 CO2

concentrations and the two land-use scenarios described above

3

Environ Res Lett 5 (2010) 014013 M Gumpenberger et al

Figure 1 Precipitation anomalies (bias-corrected mmmonths) for midcentury (2041ndash2050) and the end of the 21st century (2090ndash2099) incomparison to the reference period (1991ndash2000) for five different climate scenarios used in this study

Table 1 Overview of five different general circulation models (GCMs) Projections from these models (bias-corrected) where used asclimate inputs in simulations with the LPJmL dynamic global vegetation model Projected changes in temperature (dT ) and precipitation(dPrec) between the reference period (1991ndash2000) and the end of this century (2089ndash2098) are shown for the SRES A2 emission scenario asaverage values for land surfaces (zone between the tropic of Cancer and Capricorn)

Centre Model name Referencesa dT (K) dPrec (mmmonth)

Max Planck Institutefor MeteorologyGermany

ECHAM5MPI-OM Jungclaus et al (2005) 45 16

MeteorologicalInstitute of theUniversity of Bonn(Germany) Institute ofKMA (Korea) andModel and Data Group

ECHO-G wwwmadzmawde Grotzneret al (1996)

36 115

Hadley Centre forClimate Prediction andResearch Met OfficeUnited Kingdom

UKMO-HadCM3 Gordon et al (2000) Pope et al(2000) Johns et al (2003)

46 minus70

Geophysical FluidDynamics LaboratoryNOAA USA

GFDL-CM21 Delworth et al (2004)Gnanadesikan et al (2004)Wittenberg et al (2004)

38 15

National Center forAtmospheric Research(NCAR) NSF DOENASA NOAA USA

CCSM3 wwwccsmucaredu Collinset al (2006)

38 123

a A full list of references is found at the model documentation site www-pcmdillnlgovipccmodel documentationipcc model documentationphp

25 Analysis of model output

The countries selected for this study are the same as listedin the study by Gibbs et al [43] (see table A1) Weadded Argentina Pakistan and Sudan because these countrieshad requested participation in the Forest Carbon PartnershipFacility (FCPF wwwcarbonfinanceorgfcpf whereas onlyArgentina has been selected as a REDD country) Except

for Bhutan Nepal and Pakistan all countries are at leastpartially located within the tropics of Cancer and CapricornAll countries except French Guiana are listed as non-AnnexI parties to the UNFCCC convention The countries Bruneiand Gambia contained less than eight grid cells and wereexcluded from the analysis (grid cell at 05 times 05 resolutioncorresponding sim50 km times 50 km) because of inaccuracies inarea calculation

4

Environ Res Lett 5 (2010) 014013 M Gumpenberger et al

We evaluated LPJmL outputs for vegetation carbon ofnatural vegetation by comparing with forest carbon estimationsgiven in [43] They synthesized mapped and updatedprominent forest biomass carbon databases to create a set ofnational-level forest carbon stock estimates for the year 2000In addition we compared the coverage of tree PFTs simulatedby LPJmL with country-based forest area referenced in theForest Resources Assessment (FRA) of the FAO [44] Avalidation of soil pools simulated by LPJmL is more difficultLiterature data on tropical soil depths and carbon contentsare limited and differ strongly Some datasets include carboncontents for a soil depth of one metre eg the Soil OrganicCarbon Map of NRCS (httpsoilsusdagovuseworldsoils)The LPJmL version we used has a uniform soil depth of2 m However tropical soils can be much deeper even if itis difficult to estimate the real extent Nevertheless soil carbonis an important component in the ecological system and for theBrazilian Amazon estimates are as high as 27ndash32 Pg C [45]Milne et al [45] used detailed geo-referenced datasets ofsoils climate land use and management information and amodelling system to produce soil organic carbon stocks Wecompare LPJmL output for the Brazilian Amazon region andfor Kenya with these estimates

We analysed future changes in carbon stocks by summingup simulated carbon pools for each country and comparing theoutput of the LPJmL model for the mid (2041ndash2050) and theend of the 21st century (2090ndash2099) with a reference period(1991ndash2000) We also looked at trends over the simulatedperiod and for different carbon pools spanning the tropicalcountries we selected We include all carbon pools simulatedby LPJmL ie vegetation litter and soil pools of naturalvegetation and managed land if not specified otherwise Giventhe uncertainty of tropical soil carbon pools and in order toallow comparison with other data we present results of thisstudy in part for above-ground carbon stocks only

3 Results

31 Impact of climate and land-use change on pan-tropicalcarbon balances

In total vegetation carbon stocks in the pan-tropics areranging between 154 and 291 Pg C during the historicalperiod from 1901 to early 21st century (figure 2) Underthe GFDL-CM21 climate scenario the lowest carbon poolsare projected while the other four models are in the samerange Overall tropical carbon stocks decreased during the20th century reaching a minimum around 1990 increasingthereafter until 2012 From 2012 on the effects of thetwo contrasting land-use change scenarios become evidentGenerally under the forest protection scenario carbon stocksin the tropics are increasing in our simulations due to theeffects of CO2 fertilization Simulations with CONSECHO-G GFDL-CM21 and NCARCCSM30 climate projectionshowed higher gains in carbon stocks with forest protectionin comparison to simulations with ECHAM5MPI-OM orUKMO-HadCM3 climate change projections Under thedeforestation scenario carbon stocks generally decrease

Figure 2 Trends of pan-tropical vegetation carbon stocks asprojected by LPJmL for five climate scenarios under the SRES A2emission trajectory and for the applied protection (solid line) and thedeforestation scenario (dashed line) The climate models applied aredescribed in more detail in the methods section

Stronger decreases in carbon stocks can be observed forthe ECHAM5MPI-OM UKMO-HadCM3 CONSECHO-Gand NCARCCSM30 climate the scenarios for which LPJmLprojects higher carbon stocks under current conditionsFor the low carbon stock GFDL-CM21 scenario pan-tropical vegetation carbon stocks show almost no decrease(minus24 Pg C)

The simulated tropical vegetation carbon pool (as shownin figure 2) was higher than the soil carbon pool which heldbetween 204 and 236 Pg C during the historical period from1901 to early 21st century Soil and litter pool combinedcontained about one half of all carbon stocks simulated byLPJmL The high variability in changes of carbon stocksbetween different climate projection and land-use scenarioswas mainly due to the high variability in the simulatedvegetation carbon pool soil and litter carbon pools were muchless affected When simulated vegetation soil and littercarbon pools are accounted for deforestation was reflectedby diminishing carbon pools in tropical countries betweenminus35 Pg C (GFDL-CM21) to minus134 Pg C (UKMO-HadCM3)until the end of the 21st century Without deforestationtropical carbon pools stabilized to even higher levels than todaywith an increase ranging from +7 Pg C (UKMO-HadCM3) to+121 Pg C (NCARCCSM30)

The sensibility of LPJmL for CO2 fertilization was testedin order to estimate its effect on simulated carbon stocks Wefound that without an increase in CO2 concentration duringthe course of the 21st century rising temperatures under theSRES A2 climate projection trigger high tree mortality ratesfrom heat stress in LPJmL causing drastic break downs of pan-tropical carbon stocks (minus54 Pg C GFDL-CM21 to minus172 Pg CUKMO-HadCM3) without deforestation (see section 42 fordiscussion on the CO2 fertilization effect)

32 Regional differentiation of carbon stocks projections

The changes in carbon stocks were regionally differentiated(figures 3 and 4 table A1) In Africa and in Asia and when the

5

Environ Res Lett 5 (2010) 014013 M Gumpenberger et al

Figure 3 Relative changes of vegetation carbon stocks (in kg C mminus2) in tropical regions between the reference period (1991ndash2000) and(a) midcentury (2041ndash2050) as well as (b) the end of the 21st century (2090ndash2099) Differences are shown for the forest protection and thedeforestation scenario and for climate anomalies of five different GCMs under SRES A2 emissions

forest protection scenario was applied carbon stocks mainlyincreased whereas in Latin America carbon stocks decreasedor increased according to the different climate projectionUnder the UKMO-HadCM3 climate projection the LPJmLmodel simulated a strong reduction of carbon stocks in theAmazon region

The Asian countries Bangladesh Cambodia Sri Lankaand Thailand showed the largest relative increase of theircarbon stocks under the forest protection scenario witha high agreement between the different climate scenariosIn Bangladesh carbon stocks increased even under the

deforestation scenario (up to +103) Malaysia was oneof the countries with the highest relative loss under thedeforestation scenario (up to minus326) For Indonesia thecountry with the highest carbon stock resources in this regionthe model simulated carbon uptakes with forest protection (upto +248) and carbon stock decreases under the deforestationscenario (up to minus280) under all climate projections

On the African continent Cameroon Central AfricanRepublic DR Congo Ethiopia Gabon and Kenya showedthe largest relative increase of carbon stocks under the forestprotection scenario On the other hand Madagascar and

6

Environ Res Lett 5 (2010) 014013 M Gumpenberger et al

Figure 4 Relative changes of carbon stocks (inclusive soil) for the end of the twenty-first century (2090ndash2099) compared to 1991ndash2000 forcarbon-rich tropical countries

Sudan lost carbon stocks even under the protection scenario(up to minus130 minus156 respectively) Burundi showed acarbon loss under the forest protection scenario in simulationswith four out of five climate scenarios (minus88 to +108)In contrast Ethiopian carbon stocks increased even underthe deforestation scenario (+142 to +405) likewise inKenya carbon stocks increased in simulations with four climatechange scenarios (minus42 to +307) In DR Congo thecountry with the largest carbon stocks in Africa carbon stocksincreased ranging from +219 to +586 under the forestprotection scenario and decreased under the deforestationscenario with four climate scenarios (minus229 to +63)In Senegal and with forest protection the highest variabilitybetween the different climate change scenarios was found(minus337 to +371)

In Latin America the variability in carbon stockschanges resulting from different climate scenarios was higherespecially in Costa Rica El Salvador French Guiana GuyanaHonduras Nicaragua Suriname and Venezuela Despiteforest protection and under the UKMO-HadCM3 climateprojection the LPJmL simulated a vegetation dieback (morethan minus45 carbon loss) in Costa Rica El Salvador GuyanaNicaragua and Suriname However in the same countries

and under different climate scenarios carbon uptakes werepossible for example in Suriname and Guyana with morethan +50 under the GFDL-CM21 climate projection InBrazil and with forest protection simulated gains in carbonstocks increased under the CONSECHO-G NCARCCSM30and GFDL-CM21 climate projections (up to +381) anddecreased under UKMO-HadCM3 and ECHAM5MPI-OM(up to minus248) Under the deforestation scenario and theUKMO-HadCM3 climate projection there was a simulated lossof minus451 in carbon stocks

33 Comparison with other estimates of carbon stocks andemissions

To evaluate how well simulated carbon stocks compare withliterature values we used the country-based estimates for forestbiomass carbon stocks for the year 2000 given by Gibbs et al[43] Simulated vegetation carbon stocks were well within theranges for most of the tropical countries (figure 5 table A1)For soil carbon stocks we compared LPJmL output with valuesgiven in [45] for the Brazilian Amazon and for Kenya forthe year 2000 LPJmL simulated soil carbon stocks wereunderestimated for the Brazilian Amazon and overestimated

7

Environ Res Lett 5 (2010) 014013 M Gumpenberger et al

Figure 5 Vegetation carbon stocks (including trunk branches leaves roots) simulated by LPJmL for natural vegetation for the period1991ndash2000 (dark grey bars) compared to forest carbon stocks estimates for the year 2000 referenced in [43] (light grey bars) for carbon-richtropical countries The bars give the average vegetation carbon stocks the error bars indicate the minimum and maximum values

for Kenya but within the same order of magnitude For theBrazilian Amazon the simulated soil carbon stocks withoutcoarse roots were 17 Pg C (21 Pg C including litter) comparedto 27ndash32 Pg C given in [45] For Kenya simulated carbonstocks were 24 Pg C (27 Pg C including litter) compared to14ndash20 Pg C In addition we analysed how well the LPJmLsimulated coverage of tree PFTs constrained by land usecompares with country-based forest inventory data for 2005 bythe FAO [44] and found a positive correlation (R2 = 052p lt 00001)

We show a range of deforestation losses for the tropicsfrom minus35 to minus134 Pg C and gains from forest protection from7 to 121 Pg C by the end of the 21st century for all carbon poolssimulated by LPJmL (forested and not forested land aboveand belowground carbon stocks) In a study by Gullison et al

[46] estimated losses from tropical deforestation ranged fromminus87 to minus130 Pg C by 2100 Estimates by Cramer et al [47]using an earlier version of the LPJ model ranged from minus101 tominus367 Pg C for the tropics by 2100 For the SRES A2 scenariosthe cumulative emissions from land-use from 1990 to 2100range from 49 to 181 Pg C For comparison the emissions fromfossil fuels range from 1303 to 1860 Pg C [38]

4 Discussion

Generally we found a high interregional variability betweencarbon losses and gains for the different scenarios Inconsequence countries may benefit differentially from forestprotection which can be attributed to changing of regionalclimate regimes In our simulations forest protection strongly

8

Environ Res Lett 5 (2010) 014013 M Gumpenberger et al

increased carbon stocks in many regions which is mainly dueto growth enhancing effects of CO2 Deforestation on theother hand leads to strong carbon stock reduction in mostregions Below we discuss (1) the potential future impactson tropical carbon stocks under contrasting climate and land-use change scenarios (2) the uncertainties in the estimationof future tropical carbon stocks and (3) the implications for asuccessful REDD mechanism

41 Carbon winners and losers under contrasting climate andland-use change scenarios

During recent decades old-growth and intact forests inthe tropics were carbon sinks accumulating approximately08ndash16 Pg C yrminus1 [48] In Africa the increasing carbonstorage of intact tropical forests has been attributed toan increase in resource availability including fertilizationby atmospheric CO2 changes in solar radiation at theEarthrsquos surface increases in nutrient deposition and changesin rainfall [48] How the carbon storage potential oftropical forests will change under future climate conditionsis nevertheless highly uncertain Changes in precipitationpatterns and temperature increase among other factors couldstrongly alter vegetation dynamics Over the past two decadesair temperatures in the tropical forest biome have increasedon average by 026 Cdecade [49] There has been a strongand significant decline in rainfall in the northern Africantropics but no significant trend in other tropical regionsSimilarly strength and intensity of the dry season havesignificantly increased in Africa but not in Latin America orAsia [49] Despite some recent progress in global climatemodel development [50] climate scenarios continue to containsubstantial uncertainties In terms of their ability to forecastlong-term trends there are important differences betweenclimate models especially on a regional scale [51 52]Most climate models project increasing temperatures withsimilar spatial patterns More pronounced differences exist forprojected changes in precipitation (table 1 figure 1)

For tropical Asia most GCMs simulate a general increasein precipitation until the end of the century although theseasonal distribution remains uncertain In Africa theprediction for changes in precipitation patterns is not uniformFor central Africa four out of five climate models predict anincrease in precipitation (figure 1) In Asia and Africa climatechange in combination with increasing CO2 concentrationshad an overall positive effect on carbon storage potentials insimulations with LPJmL For some regions eg parts of theAfrican highlands (Ethiopia Kenya) gains in carbon stockwere simulated despite a reduction of 50 of the countriesnaturally vegetated area under the deforestation scenarioCarbon losses from deforestation were overcompensated bythe combined effects of CO2 fertilization and climate changeHowever simulated carbon stocks in the reference periodare overestimated for these countries which might be dueto missing disturbance processes in the LPJmL modelNevertheless the simulated abundance of tree PFTs was stillvery low in this region Climatic change increased treecover (replacing C4 grasses) and there was vegetation growth

in previously non-vegetated areas In addition the CO2

fertilization effect increased NPP and both effects were leadingto the relatively strong carbon sink

In Latin America GCMs vary greatly in their projectionsof future climate change [53ndash55] accordingly the congruencein simulated changes of carbon stocks between differentclimate scenarios was particularly low for this region(figure 3) A high inter-annual variability in precipitation inthe GFDL-CM21 climate projection caused an underestimatednet primary production (NPP) in tropical Latin Americaconsequently reducing pan-tropical vegetation carbon stockswith relatively little changes in the 21st century under thedeforestation scenario (figure 2) This demonstrates the relativeimportance of tropical rainforests in Latin America for pan-tropical carbon stocks In simulations with UKMO-HadCM3climate projection where a strong decrease in precipitationis projected for the Amazon region the LPJmL modelsimulated a vegetation dieback even without the additionalpressure of increasing land use (figure 3) This result isin accordance with findings of other studies in which forparts of the Amazon basin a tipping for the rain forest intosavannah is shown [56ndash58] Other recent studies on theAmazonian rainforest emphasize the high vulnerability ofthis region due to climate change in combination with land-use change [54 59 60] Land-use change including large-scale deforestation and fragmentation might trigger or stronglyenhance climatic change effects For carbon stocks and the netcarbon exchange land-use change may well be more importantthan climatic change [30 47] Tropical Latin America has ahigher risk to lose large amounts of its carbon stocks duringthe course of this century

42 Uncertainties in the estimation of future tropical carbonstocks

Generally our simulated carbon stocks are in the range ofother studies (figure 5 table A1) In the model land useconstrains the area of natural vegetation which is forested ifclimate conditions allow it Thus the size of the forestedarea determines the natural vegetation carbon balances Weused the HYDE30 gridded dataset to constrain historic andcurrent land use in LPJmL However different land-usedatasets are not consistent and can differ especially regionallybecause of the differences in the methods applied the use ofdifferent input data and definitions (eg for pasture land) [61]One of the most important reference dataset for forests anddeforestation trends is the Forest Resources Assessment (FRA)of the FAO [44] But changing classification schemes overtime adjustments in the presentation of trends as well as inaggregating algorithms make the data an inconsistent source ofglobal deforestation rates and trends [62] The inconsistenciesin different datasets may explain that the correlation we foundbetween simulated forest areas and country-based forest areasgiven by the FAO was not high (R2 = 052) As it is difficultto determine current land use and land-use change rates largeuncertainties exist over the changing rate of deforestation inthe future The IMAGE model has been used to projectfuture land-use changes under different SRES scenarios [63]

9

Environ Res Lett 5 (2010) 014013 M Gumpenberger et al

IMAGE land-use projections have been applied to study theeffects of climate and land-use change on the global terrestrialcarbon cycle for the 21st century using the LPJmL model [64]The current study mainly focuses on changes in tropical forestcarbon stocks by comparing hypothetical land-use scenarioswith climate scenarios temporal and regional differentiatedland-use scenarios were not used or developed

Our study shows that under the protection scenario insome countries the carbon gain is large (figures 3 and 4table A1) This is due to the modelrsquos assumption of enhancedwater use efficiency by CO2 fertilization There is noconsensus in the scientific community about the magnitudeof the CO2 fertilization effect with rising CO2 concentrationsunder climate change The sensibility towards CO2 in LPJmight be rather over-than underestimated [47] Hickleret al [28] showed that the LPJ-GUESS dynamic vegetationmodel reproduces the magnitude of the NPP enhancement attemperate forest FACE experiments but in tropical forestspredicted NPP enhancement was more than twice as high as inboreal forests suggesting that currently available FACE resultsare not applicable to tropical ecosystems It has been arguedthat the availability of nutrients will constrain NPP responses toCO2 enhancement [28] However in LPJmL CO2 fertilizationis limited only by the availability of water and processesfor nitrogen and phosphorus limitation which are especiallyimportant in the tropics [65 66] are not represented

Other factors influencing the estimation of changes infuture carbon stocks are selective logging fire forest grazingand edge effects in fragmented landscapes [54] Forestdegradation is difficult to detect at large scale and is notnecessarily stopped with deforestation [62 67] Fire inthe tropics is primarily associated with human activity andinfluence on land cover lightning strikes rarely lead toforest fires as these events are usually associated withheavy rainfall [68] Fire as a disturbance factor is causingbiomass loss and modified site conditions might delay orprevent regeneration of the vegetation In the LPJmL modelfire disturbance is included by a process-based fire-modulewhich allows for fires in natural vegetation ignited only bylightning [31] Deforestation and forest degradation frequentlylead to nutrient depletion soil degradation or erosionmdashprocesses that reduce a regionrsquos growth potential irreversiblyon a timescale of centuries Most processes of forest orsoil degradation are not represented in LPJmL so that futurecarbon gains might be overestimated

43 Implications for REDD

Our results show that tropical forests have the potential toincrease their carbon stocks substantially if they are protectedIn contrast climate change possesses risks for forest carbonstocks to decrease without any direct human influence Thechallenge in a policy context lies in determining how incentiveswill be given to countries for reducing emissions and protectingforests In providing incentives to countries for increases incarbon stocks natural and indirect human induced effects suchas CO2 fertilization as well as the risks of climate changeimpacts must also be taken into account Thus it will be

important to understand the processes that govern currentgreenhouse gas emissions and future projections [69] As withdeveloped countries in the Kyoto Protocol it will be necessaryto improve how to factor out the impacts of CO2 fertilizationeffects and the impacts of climate change [69 70] Incentivesshould be restricted to direct human induced increases incarbon stocks and reductions in deforestation emissions belowbusiness-as-usual Therefore it must be considered toinclude not only carbon stocks alone but also other criteriathat refer to policy implementation combating the drivers ofdeforestation as a calculation basis to pay for successful forestprotection [71]

5 Conclusions

Climate change will have regionally differentiated impactson tropical carbon stocks Countries in tropical South EastAsia and Africa could profit from higher carbon densitiesmainly due to changes in precipitation patterns increase intemperature and CO2 fertilization effects Also positive effectsdue to CO2 fertilization might prevail in the coming decadeslatest at the end of the century severe losses due to climatechange induced forest degradation could be expected at leastfor some parts of the tropics eg for Latin America There is ahigher risk that large parts of the tropical Amazonian rainforestcould degrade due to a strong reduction in rainfall Limitingdeforestation and the spread of fires may be successful tools tomaintain Amazonian forest resilience under the risk of futureclimate change [54 72]

Based on the findings of this study we suggest that factorssuch as future changes of climate water availability as well asCO2 fertilization effects must be taken into account in order toachieve an effective and fair REDD mechanism Continuingto gain an understanding of the different interactions affectingcarbon stocks and related emissions from the land-use sectorwill become increasingly important in identifying the directhuman induced reductions from deforestation

Acknowledgments

This study was financially supported by the EU Marie CurieResearch Training Network GREENCYCLES (MRTN-CT-2004-512464) and by the German BMBF (Bundesministeriumfur Bildung und Forschung) Results benefitted fromdiscussions within the context of the Klima-und-GerechtigkeitProject (wwwklima-und-gerechtigkeitde) We thank twoanonymous referees for valuable comments on the manuscriptWe acknowledge the modelling groups the Program forClimate Model Diagnosis and Intercomparison (PCMDI) andthe WCRPrsquos Working Group on Coupled Modelling (WGCM)for their roles in making available the WCRP CMIP3 multi-model dataset Support of this dataset is provided by the Officeof Science US Department of Energy

10

Environ Res Lett 5 (2010) 014013 M Gumpenberger et al

Appendix

Table A1 Countries as listed in the study of Gibbs et al [43] to which we additionally added Argentina Pakistan and Sudan(a) Above-ground forest carbon stocks (Tg C) as estimated from [43] and as projected by LPJmL (including trunk branches leaves and roots)for natural vegetation The simulated values are displayed for the reference period (1991ndash2000) For 2041ndash2050 and 2090ndash2099 the absolutedifferences to the reference period are given showing the range of the two land-use scenarios (deforestation protection) based on fivedifferent climate scenarios (min max) (b) Carbon stocks including all carbon pools simulated by LPJmL ie vegetation soil and littercarbon pools for natural vegetation and managed grassland

(a) Vegetation carbon (Tg C) of natural vegetation as projected by LPJmL

Absolute difference2041ndash2050 to 1991ndash2000

Absolute difference2090ndash2099 to 1991ndash2000

Gibbs et al (2007) 1991ndash2000 Protection Deforestation Protection Deforestation

Country MIN MAX MIN MAX MIN MAX MIN MAX MIN MAX MIN MAX

Angola 3 557 11 767 3 672 5 291 minus92 1 730 minus1 018 394 minus162 3 844 minus2 417 minus436Argentina nv nv 1 200 3 060 255 1 552 minus21 735 563 4 233 minus263 964Bangladesh 65 158 264 320 182 247 87 145 291 475 4 107Belize 198 318 148 363 minus7 121 minus45 55 minus6 197 minus107 0Benin 260 792 446 574 71 102 minus41 minus9 230 296 minus116 minus75Bhutan 1 121 185 263 minus50 55 minus90 3 minus56 19 minus135 minus92Bolivia 2 469 9 189 2 521 5 541 minus265 1 780 minus1 269 427 minus702 4 545 minus2 970 minus77Brazil 54 697 82 699 39 622 60 759 minus3617 13 539 minus13 830 1268 minus21 306 31 402 minus38 228 minus1864Burundi 9 69 35 102 minus15 19 minus31 3 22 47 minus32 8Cambodia 957 1 914 989 1 275 249 384 minus7 69 460 785 minus263 minus95Cameroon 3 454 6 138 2 615 4 506 740 1 189 minus193 182 1 455 2 686 minus1 201 minus130CentrAfrRep 3 176 7 405 3 452 5 652 1004 1 787 156 477 2 686 4 092 minus769 minus190Colombia 2 529 11 467 7 250 12 429 800 2 026 minus1 499 268 minus1 630 5 251 minus6 159 minus604Congo 3 458 5 472 1 214 4 136 535 717 minus274 201 935 1 568 minus1 379 264Costa Rica 471 704 262 592 minus80 99 minus177 26 minus284 235 minus314 3DR Congo 20 416 36 672 12 149 30 039 4800 6 397 minus1 804 1570 9 401 14 963 minus9 076 1233Ecuador 351 2 071 1 738 2 687 152 446 minus388 30 455 1 015 minus1 017 minus275El Salvador 105 153 76 125 minus27 5 minus36 minus12 minus73 19 minus75 minus30Eq Guinea 268 474 176 440 22 79 minus64 30 62 184 minus171 16Ethiopia 153 867 1 415 2 171 1218 2 187 718 1454 2 737 5 382 791 2150French Guiana 403 1 683 586 1 390 minus219 396 minus373 108 minus428 732 minus710 22Gabon 3 063 4 742 1 041 3 635 389 502 minus337 164 828 1 373 minus1 230 242Ghana 609 2 172 709 878 minus30 0 minus187 minus141 41 261 minus325 minus236Guatemala 787 1 147 502 1 024 82 243 minus127 66 minus192 345 minus409 minus95Guinea 598 2 051 830 1 221 11 338 minus196 116 minus234 719 minus629 minus69Guinea Bissau 78 381 28 57 minus3 24 minus13 9 5 72 minus14 17Guyana 923 3 354 1 679 3 243 minus604 1 043 minus809 481 minus1 286 1 517 minus1 478 minus21Honduras 852 1 268 568 1 017 50 325 minus92 73 minus289 767 minus478 minus61India 5 085 8 997 3 250 4 034 770 1 639 minus39 647 1 830 4 201 minus664 519Indonesia 10 252 25 547 13 654 29 542 3148 4 189 minus2 644 minus37 3 460 8 255 minus9 864 minus3065Ivory Coast 750 3 355 1167 1 432 117 235 minus140 minus52 minus131 713 minus640 minus280Kenya 163 618 276 1 018 222 644 37 329 785 1 727 79 441Laos 718 1 870 1 574 2 107 320 743 minus127 278 367 1 167 minus690 minus368Liberia 506 1 302 660 788 77 277 minus88 86 minus176 603 minus461 minus12Madagascar 1 043 2 114 2 310 2 918 minus412 minus24 minus846 minus465 minus550 177 minus1 375 minus1024Malawi 152 391 257 447 minus84 129 minus135 20 minus22 312 minus177 minus43Malaysia 2 405 4 821 2 838 5 677 403 533 minus738 minus108 606 1 098 minus2 208 minus749Mexico 4 361 5 924 1 899 3 507 57 815 minus314 minus5 477 2 259 minus1 100 minus443Mozambique 1 894 5 148 1 345 2 157 minus31 575 minus390 72 41 1 398 minus876 minus208Myanmar 2 377 5 182 3 764 4 517 736 1 400 minus264 355 1 063 2 343 minus1 551 minus650Nepal 246 393 178 364 11 104 minus58 36 minus2 293 minus129 38Nicaragua 930 1 395 629 1 384 minus139 91 minus310 minus118 minus486 338 minus578 minus141Nigeria 1 278 3 952 992 1 289 535 681 246 311 1 145 1 492 28 206Pakistan nv nv 255 292 minus53 110 minus91 33 minus158 222 minus203 minus19Panama 509 763 544 1 069 minus337 167 minus476 32 minus305 462 minus663 minus9Papua N Guinea 4 154 8 037 5 885 8 820 165 1 890 minus1 119 409 571 2 317 minus3 023 minus2458Paraguay 1 087 3 659 171 1 678 minus63 536 minus133 228 minus39 1 253 minus674 minus45Peru 2 782 13 241 6 358 12 302 1288 1 940 minus1 097 9 minus2 628 4 886 minus7 154 minus840Philippines 765 2 503 2 062 3 065 377 618 minus277 59 666 1 546 minus771 minus528Rwanda 6 48 40 183 2 44 minus28 28 101 132 minus32 51Senegal 86 228 52 76 minus7 46 minus19 23 minus28 178 minus39 66Sierra Leone 114 683 373 485 46 136 minus46 39 minus53 291 minus234 minus21

11

Environ Res Lett 5 (2010) 014013 M Gumpenberger et al

Table A1 (Continued)

(a) Vegetation carbon (Tg C) of natural vegetation as projected by LPJmL

Absolute difference2041ndash2050 to 1991ndash2000

Absolute difference2090ndash2099 to 1991ndash2000

Gibbs et al (2007) 1991ndash2000 Protection Deforestation Protection Deforestation

Country MIN MAX MIN MAX MIN MAX MIN MAX MIN MAX MIN MAX

Sri Lanka 138 509 271 356 67 171 2 86 189 386 minus35 46Sudan nv nv 457 740 minus308 minus182 minus390 minus234 minus141 minus49 minus421 minus240Suriname 663 2753 1299 2337 minus590 674 minus725 205 minus1186 1254 minus1342 minus1Tanzania 1281 3400 2803 5402 817 1340 minus149 308 1506 3350 minus1221 470Thailand 1346 2489 2023 2617 511 1021 minus15 348 1486 1901 minus385 minus139Togo 145 510 148 187 minus15 0 minus48 minus28 19 44 minus66 minus46Uganda 429 1237 314 1379 117 384 minus124 144 531 852 minus260 162Venezuela 2326 9202 6347 7968 minus1322 2402 minus2277 497 minus3202 4027 minus4675 minus959Vietnam 774 1642 2236 2838 70 616 minus441 73 234 1411 minus924 minus573Zambia 1455 6378 2115 3312 245 1019 minus313 304 603 2491 minus764 42

(b)Above and belowground carbon (Tg C) including litter and soil fornatural vegetation and managed grassland as projected from LPJmL

Absolute difference2041ndash2050 to 1991ndash2000

Absolute difference2090ndash2099 to 1991ndash2000

1991ndash2000 Protection Deforestation Protection Deforestation

Country MIN MAX MIN MAX MIN MAX MIN MAX MIN MAX

Angola 11 083 13 092 minus253 1 750 minus1 227 418 minus656 4 038 minus3 554 minus873Argentina 17 836 21 594 minus387 1 689 minus860 648 minus525 4 294 minus2 143 minus843Bangladesh 855 915 197 281 103 179 320 500 minus2 91Belize 319 525 19 125 minus22 60 5 217 minus112 minus3Benin 1 022 1 144 37 67 minus60 minus41 192 287 minus184 minus134Bhutan 683 722 minus19 48 minus59 2 minus26 35 minus141 minus111Bolivia 9 804 14 122 minus901 1 864 minus2 000 426 minus1 924 4 563 minus4 851 minus782Brazil 85 852 109 762 minus6020 11 441 minus16 715 6 minus26 248 32 689 minus47 746 minus5925Burundi 247 335 minus57 minus12 minus74 minus22 minus29 27 minus95 minus21Cambodia 1 952 2 234 247 362 minus4 56 456 797 minus361 minus155Cameroon 5 349 7 278 785 1 265 minus128 255 1 510 2 821 minus1 466 minus231CentrAfrRep 7 521 9 926 1056 2 046 195 643 3 109 4 778 minus1 060 minus212Colombia 14 393 19 443 939 1 948 minus1 404 123 minus1 483 5 325 minus6 847 minus1284Congo 3 018 6 184 569 791 minus198 222 1 052 1 845 minus1 577 315Costa Rica 701 1 029 minus79 90 minus180 minus9 minus382 229 minus428 minus52DR Congo 26 086 45 423 5250 6 867 minus1 309 2037 9 934 15 953 minus10 409 1642Ecuador 3 883 4 736 79 297 minus457 minus136 300 897 minus1 300 minus573El Salvador 241 295 minus41 minus16 minus51 minus32 minus139 6 minus143 minus48Eq Guinea 322 599 34 85 minus54 36 66 216 minus189 21Ethiopia 7 100 8 183 1665 2 686 1 099 1945 3 652 7 083 1 101 3109French Guiana 1 131 1 991 minus157 405 minus323 115 minus376 778 minus742 minus20Gabon 2 370 5 021 428 587 minus244 177 903 1 584 minus1 319 264Ghana 1 806 1 988 minus125 minus87 minus273 minus226 minus54 162 minus482 minus383Guatemala 1 447 1 934 25 176 minus175 minus5 minus292 303 minus586 minus218Guinea 2 164 2 540 44 287 minus166 68 minus245 683 minus738 minus174Guinea Bissau 190 213 minus9 14 minus20 minus1 minus23 54 minus46 minus5Guyana 3 002 4 678 minus421 1 015 minus672 440 minus1 539 1 587 minus1 858 minus164Honduras 1 462 1 873 79 359 minus64 102 minus393 854 minus669 minus92India 16 669 18 882 623 2 266 minus272 1185 1 778 5 339 minus1 463 876Indonesia 26 103 42 123 2977 4 540 minus2 571 minus381 4 702 9 019 minus11 060 minus4472Ivory Coast 2 778 3 055 23 107 minus224 minus180 minus150 577 minus810 minus477Kenya 2 309 3 948 220 726 minus41 378 947 1 974 minus163 710Laos 3 097 3 563 335 804 minus104 281 517 1 289 minus736 minus386Liberia 1 174 1 287 122 301 minus47 97 minus77 601 minus460 minus72Madagascar 6 070 6 708 minus610 minus281 minus1 105 minus736 minus788 minus11 minus1 838 minus1441Malawi 958 1 193 minus120 94 minus180 minus14 minus106 259 minus314 minus134Malaysia 5 107 7 907 350 434 minus795 minus295 549 1 092 minus2 512 minus1151Mexico 9 083 11 148 minus528 933 minus925 94 226 3 203 minus1 807 24Mozambique 5 337 6 416 minus135 533 minus538 80 minus312 1 374 minus1 516 minus393Myanmar 7 769 8 513 786 1 510 minus194 454 1 449 2 641 minus1 568 minus684Nepal 1 559 1 793 50 173 minus2 84 84 415 minus97 27

12

Environ Res Lett 5 (2010) 014013 M Gumpenberger et al

Table A1 (Continued)

(b)Above and belowground carbon (Tg C) including litter and soil fornatural vegetation and managed grassland as projected from LPJmL

Absolute difference2041ndash2050 to 1991ndash2000

Absolute difference2090ndash2099 to 1991ndash2000

1991ndash2000 Protection Deforestation Protection Deforestation

Country MIN MAX MIN MAX MIN MAX MIN MAX MIN MAX

Nicaragua 1 568 2 339 minus188 21 minus353 minus185 minus713 277 minus867 minus286Nigeria 4 456 4 789 527 674 228 312 1025 1498 minus213 128Pakistan 2 420 2 568 minus16 228 minus78 141 minus324 564 minus498 182Panama 1 180 1 681 minus249 200 minus404 minus3 minus250 487 minus694 minus73Papua N Guinea 9 118 12 024 529 1943 minus829 420 954 2657 minus3222 minus2710Paraguay 2 417 4 078 minus255 363 minus350 21 minus549 1120 minus1196 minus402Peru 17 380 23 962 1388 2168 minus1072 55 minus2014 5277 minus7955 minus1413Philippines 4 000 5 064 319 543 minus341 14 700 1515 minus997 minus755Rwanda 297 452 minus23 36 minus56 18 51 139 minus96 41Senegal 646 784 minus128 94 minus145 65 minus217 268 minus242 112Sierra Leone 728 836 63 137 minus28 42 minus5 299 minus232 minus40Sri Lanka 639 732 61 166 minus8 77 210 390 minus62 5Sudan 5 040 5 769 minus622 minus455 minus695 minus506 minus868 minus182 minus1141 minus391Suriname 2 255 3 359 minus427 669 minus604 193 minus1256 1299 minus1538 minus93Tanzania 8 280 11 751 682 1286 minus315 506 1152 4192 minus2073 704Thailand 4 627 5 145 484 1095 minus46 424 1418 1906 minus541 minus361Togo 404 445 minus39 minus25 minus68 minus53 minus16 13 minus111 minus87Uganda 1 687 3 067 minus22 242 minus271 61 341 745 minus563 222Venezuela 12 083 13 820 minus1139 2393 minus2214 467 minus4233 4514 minus6193 minus1255Vietnam 4 433 5 042 22 533 minus506 minus8 251 1370 minus1150 minus745Zambia 7 274 8 721 minus69 679 minus663 minus43 412 2048 minus1738 minus737

References

[1] Achard F Eva H D Mayaux P Stibig H-J and Belward A 2004Improved estimates of net carbon emissions from land coverchange in the tropics for the 1990s Glob BiogeochemCycles 18 GB2008

[2] Schimel D S et al 2001 Recent patterns and mechanisms ofcarbon exchange by terrestrial ecosystems Nature414 169ndash72

[3] van der Werf G R Morton D C DeFries R S Olivier J G JKasibhatla P S Jackson R B Collatz G J and Randerson J T2009 CO2 emissions from forest loss Nat Geosci 2 737ndash8

[4] IPCC 2000 Special Report on Land Use Land-Use Change andForestry ed R T Watson I R Noble B BolinN H Ravindranath D J Verardo and D J Dokken(Cambridge Cambridge University Press) p 377 available atwwwipccchipccreportssresland useindexphpidp=0

[5] Houghton R A 2003 Revised estimates of the annual net flux ofcarbon to the atmosphere from changes in land use and landmanagement 1850ndash2000 Tellus B 55 378ndash90

[6] Hansen M C et al 2008 Humid tropical forest clearing from2000 to 2005 quantified by using multitemporal andmultiresolution remotely sensed data Proc Natl Acad SciUSA 105 9439ndash44

[7] Lambin E F Geist H J and Lepers E 2003 Dynamics ofland-use and land-cover change in tropical regions Ann RevEnviron Resour 28 205ndash41

[8] Santilli M Moutinho P Schwartzman S Nepstad DCurran L and Nobre C 2005 Tropical deforestation and theKyoto Protocol Clim Change 71 267ndash76

[9] Fearnside P M 2001 Saving tropical forests as a globalwarming countermeasure an issue that divides theenvironmental movement Ecol Econ 39 167ndash84

[10] IPCC 2007 Contribution of Working Group II to the FourthAssessment Report of the Intergovernmental Panel onClimate Change ed M L Parry O F Canziani J P Palutikof

P J van der Linden and C E Hanson (Cambridge CambridgeUniversity Press) p 976 available at wwwipccchpublications and dataar4wg2encontentshtml

[11] Smith J B et al 2009 Assessing dangerous climate changethrough an update of the Intergovernmental Panel onClimate Change (IPCC) lsquoreasons for concernrsquo Proc NatlAcad Sci USA 106 4133ndash7

[12] Kindermann G Obersteiner M Sohngen B Sathaye JAndrasko K Rametsteiner E Schlamadinger BWunder S and Beach R 2008 Global cost estimates ofreducing carbon emissions through avoided deforestationProc Natl Acad Sci USA 105 10302ndash7

[13] H M Treasury 2006 Stern Review on the Economics of ClimateChange (London H M Treasury) available at wwwhm-treasurygovukstern review reporthtm

[14] Strassburg B Turner R K Fisher B Schaeffer R andLovett A 2009 Reducing emissions from deforestationmdashthelsquocombined incentivesrsquo mechanism and empirical simulationsGlob Environ Change 19 265ndash78

[15] Gurney K R and Raymond L 2008 Targeting deforestation ratesin climate change policy a lsquoPreservation Pathwayrsquo approachCarbon Balance Manag 3 doi101186750-0680-3-2

[16] Kindermann G Obersteiner M Rametsteiner E andMcCallum I 2006 Predicting the deforestation-trend underdifferent carbon-prices Carbon Balance Manag 1doi101186750-0680-1-15

[17] Fry I 2008 Reducing emissions from deforestation and forestdegradation opportunities and pitfalls in developing a newlegal regime Rev European Community Int Environ Law17 166ndash82

[18] Karsenty A 2008 The architecture of proposed REDD schemesafter Bali facing critical choices Int Forest Rev 10 443ndash57

[19] Angelsen A 2008 REDD models and baselines Int Forest Rev10 465ndash75

[20] Dutschke M and Wolf R 2007 Reducing emissions fromdeforestation in developing countries the way forward

13

Environ Res Lett 5 (2010) 014013 M Gumpenberger et al

Deutsche Gesellschaft fur Technische Zusammenarbeit(GTZ) Eschborn Germany available at wwwgtzdededokumenteen-climate-reducing-emissionspdf

[21] Murray B C 2008 Leakage from an avoided deforestationcompensation policy concepts empirical evidence andcorrective policy options Working Paper ed C Palmer andS Engel Nicholas Institute for Environmental PolicySolutions Duke University available at httpnicholasdukeeduinstitutewp-leakagepdf

[22] UNFCCC 2008 Results of the work on scientific andmethodological aspects of the proposal by Brazil TheTwenty-Eighth Session of the Subsidiary Body for Scientificand Technological Advice United Nations FrameworkConvention on Climate Change Bonn Germany available athttpunfcccintresourcedocs2008sbstaengmisc01pdf

[23] Ebeling J and Yasue M 2008 Generating carbon financethrough avoided deforestation and its potential to createclimatic conservation and human development benefits PhilTrans R Soc B 363 1917ndash24

[24] Bondeau A et al 2007 Modelling the role of agriculture for the20th century global terrestrial carbon balance Glob ChangeBiol 13 679ndash706

[25] Gerten D Schaphoff S Haberlandt U Lucht W andSitch S 2004 Terrestrial vegetation and waterbalancemdashhydrological evaluation of a dynamic globalvegetation model J Hydrol 286 249ndash70

[26] Sitch S Smith B and Prentice I C 2003 Evaluation ofecosystem dynamics plant geography and terrestrial carboncycling in the LPJ dynamic global vegetation model GlobChange Biol 9 161ndash85

[27] Cowling S A and Shin Y 2006 Simulated ecosystem thresholdresponses to co-varying temperature precipitation andatmospheric CO2 within a region of Amazonia Glob EcolBiogeogr 15 553ndash66

[28] Hickler T Smith B Prentice I C Mjofors K Miller PArneth A and Sykes M T 2008 CO2 fertilization in temperateFACE experiments not representative of boreal and tropicalforests Glob Change Biol 14 1531ndash42

[29] Lucht W Prentice I C Myneni R B Sitch S Friedlingstein PCramer W Bousquet P Buermann W and Smith B 2002Climatic control of the high-latitude vegetation greeningtrend and Pinatubo effect Science 296 1687ndash9

[30] Poulter B Aragao L Heyder U Gumpenberger M Heinke JLangerwisch F Rammig A Thonicke K andCramer W 2009 Net biome production of the Amazon Basinin the 21st century Glob Change Bioldoi101111j365-248600902064x

[31] Thonicke K Venevsky S Sitch S and Cramer W 2001 The roleof fire disturbance for global vegetation dynamics couplingfire into a dynamic global vegetation model Glob EcolBiogeogr 10 661ndash77

[32] Erb K-H Gaube V Krausmann F Plutzar C Bondeau A andHaberl H 2007 A comprehensive global 5 min resolutionland-use data set for the year 2000 consistent with nationalcensus data J Land Use Sci 2 191ndash224

[33] Goldewijk K K van Drecht G and Bouwman A F 2007Mapping contemporary global cropland and grasslanddistributions on a 5 times 5 minute resolution J Land Use Sci2 167ndash90

[34] Ramankutty N Evan A T Monfreda C and Foley J A 2008Farming the planet 1 Geographic distribution of globalagricultural lands in the year 2000 Glob BiogeochemCycles 22 GB1003

[35] Ramankutty N and Foley J A 1999 Estimating historicalchanges in global land cover croplands from 1700 to 1992Glob Biogeochem Cycles 13 997ndash1027

[36] Goldewijk K K and van Drecht G 2006 HYDE 3 current andhistorical population and land cover Integrated Modelling ofGlobal Environmental Change An Overview of IMAGE 24ed A F Bouwman T Kram and K K Goldewijk (BilthovenNetherlands Environmental Assessment Agency) available atwwwrivmnlbibliotheekrapporten500110002pdf

[37] FAO 2006 Choosing a forest definition for the CleanDevelopment Mechanism Forests and Climate ChangeWorking Paper 4 ed T Neeff H von Luepke andD Schoene (Rome Food and Agriculture Organization ofthe United Nations) available at wwwfaoorgforestry11280-1-0pdf

[38] IPCC 2000 Special Report on Emissions Scenarios A SpecialReport of Working Group III of the Intergovernmental Panelon Climate Change ed N Nakicenovic and R Swart(Cambridge Cambridge University Press) p 599 available atwwwipccchipccreportssresemissionindexphpidp=0

[39] IPCC 2007 Contribution of Working Group I to the FourthAssessment Report of the Intergovernmental Panel onClimate Change ed S Solomon D Qin M Manning Z ChenM Marquis K B Averyt M Tignor and H L Miller(Cambridge Cambridge University Press) p 996 available atwwwipccchpublications and dataar4wg1encontentshtml

[40] Mitchell T D and Jones P D 2005 An improved method ofconstructing a database of monthly climate observations andassociated high-resolution grids Int J Climatol 25 693ndash712

[41] Osterle H Gerstengarbe F W and Werner P C 2003Homogenisierung und Aktualisierung des Klimadatensatzesder Climate Research Unit der University of East AngliaNorwich Terra Nostra 6 326ndash9

[42] Richardson K et al 2009 Climate Change Global RisksChallenges and Decisions University of Copenhagenhttpclimatecongresskudk

[43] Gibbs H K Brown S Niles J O and Foley J A 2007 Monitoringand estimating tropical forest carbon stocks making REDDa reality Environ Res Lett 2 045023

[44] FAO 2006 Global Forest Resources Assessment 2005 ProgressTowards Sustainable Forest Management (Rome Food andAgriculture Organization of the United Nations) available atwwwfaoorgDOCREP008a0400ea0400e00htm

[45] Milne E et al 2007 An increased understanding of soilorganic carbon stocks and changes in non-temperate areasnational and global implications Agric Ecosyst Environ122 125ndash36

[46] Gullison R E et al 2007 Tropical forests and climate policiesScience 316 985ndash6

[47] Cramer W Bondeau A Schaphoff S Lucht W Smith B andSitch S 2004 Tropical forests and the global carbon cycleimpacts of atmospheric carbon dioxide climate change andrate of deforestation Phil Trans R Soc B 359 331ndash43

[48] Lewis S L et al 2009 Increasing carbon storage in intactAfrican tropical forests Nature 457 1003ndash6

[49] Lewis S L Malhi Y and Phillips O L 2004 Fingerprinting theimpacts of global change on tropical forests Phil Trans RSoc B 359 437ndash62

[50] Reichler T and Kim J 2008 How well do coupled modelssimulate todayrsquos climate Bull Am Meteorol Soc89 303ndash11

[51] Giorgi F 2006 Climate change hot-spots Geophys Res Lett33 L08707

[52] Gleckler P J Taylor K E and Doutriaux C 2008 Performancemetrics for climate models J Geophys Res 113 D06104

[53] Cook K H and Vizy E K 2008 Effects of twenty-first-centuryclimate change on the Amazon rain forest J Clim21 542ndash60

14

Environ Res Lett 5 (2010) 014013 M Gumpenberger et al

[54] Malhi Y Aragao L E O C Galbraith D Huntingford CFisher R Zelazowski P Sitch S McSweeney C andMeir P 2009 Exploring the likelihood and mechanism of aclimate-change-induces dieback of the Amazon rainforestProc Natl Acad Sci USA 106 20610ndash5

[55] Vera C and Silvestri G 2009 Precipitation interannualvariability in South America from the WCRP-CMIP3multi-model dataset Clim Dyn 32 1003ndash14

[56] Cowling S A Betts R A Cox P M Ettwein V J Jones C DMaslin M A and Spall S A 2004 Contrasting simulated pastand future responses of the Amazon forest to atmosphericchange Phil Trans R Soc B 359 539ndash47

[57] Cox P M Betts R A Collins M Harris P P Huntingford C andJones C D 2004 Amazonian forest dieback underclimate-carbon cycle projections for the 21st century TheorAppl Climatol 78 137ndash56

[58] Phillips O L et al 2009 Drought sensitivity of the Amazonrainforest Science 323 1344ndash7

[59] Senna M C A Costa M H and Pires G F 2009Vegetation-atmosphere-soil nutrient feedbacks in theAmazon for different deforestation scenarios J GeophysRes 114 D04104

[60] Nepstad D C Stickler C M Soares-Filho B and Merry F 2008Interactions among Amazon land use forests and climateprospects for a near-term forest tipping point Phil Trans RSoc B 363 1737ndash46

[61] Goldewijk K K and Ramankutty N 2004 Land cover changeover the last three centuries due to human activities theavailability of new global data sets GeoJournal61 335ndash44

[62] Grainger A 2008 Difficulties in tracking the long-term globaltrend in tropical forest areas Proc Natl Acad Sci USA105 818ndash23

[63] Strengers B Leemans R Eickhout B de Vries B andBouwman L 2004 The land-use projections and resultingemissions in the IPCC SRES scenarios as simulated by theIMAGE 22 model GeoJournal 61 381ndash93

[64] Muller C Eickhout B Zaehle S Bondeau A Cramer W andLucht W 2007 Effects of changes in CO2 climate and landuse on the carbon balance of the land biosphere during the21st century J Geophys Res 112 G02032

[65] Sanchez P 2002 Soil fertility and hunger in Africa Science295 2019ndash20

[66] Zougmore R Zida Z and Kamboua N F 2003 Role of nutrientamendments in the success of half-moon soil and waterconservation practice in semiarid Burkina Faso Soil TillageRes 71 143ndash9

[67] Foley J et al 2007 Amazonia revealed forest degradation andloss of ecosystem goods and services in the Amazon BasinFront Ecol Environ 5 25ndash32

[68] Cochrane M A 2003 Fire science for rainforests Nature421 913ndash9

[69] Canadell J G Kirschbaum M Kurz W Sanz M-JSchlamadinger B and Yamagata Y 2007 Factoring outnatural and indirect human effects on terrestrial carbonsources and sinks Environ Sci Policy 10 370ndash84

[70] IPCC 2003 IPCC meeting on current scientific understandingof the processes affecting terrestrial carbon stocks andhuman influences upon them Expert Meeting Report(Geneva July 2003) available at wwwipccchpdfsupporting-materialipcc-meeting-2003-07pdf

[71] Motel P C Pirard R and Combes J-L 2009 A methodology toestimate impacts of domestic policies on deforestationCompensated Successful Efforts for lsquoavoided deforestationrsquo(REDD) Ecol Econ 68 680ndash91

[72] Cochrane M A and Laurance W F 2008 Synergisms among fireland use and climate change in the Amazon Ambio37 522ndash7

15