Mortality as a key driver of the spatial distribution of aboveground biomass in Amazonian forest

14 Pages·2017·0.82 MB·English

View metadata, citation and similar papers at core.ac.uk brought to you by CORE
provided by HAL-CEA
Mortality as a key driver of the spatial distribution of
aboveground biomass in Amazonian forest: Results from
a dynamic vegetation model
N. Delbart, P. Ciais, J. Chave, N. Viovy, Y. Malhi, T. Le Toan
To cite this version:
N. Delbart, P. Ciais, J. Chave, N. Viovy, Y. Malhi, et al.. Mortality as a key driver of the
spatial distribution of aboveground biomass in Amazonian forest: Results from a dynamic
vegetation model. Biogeosciences, European Geosciences Union, 2010, 7 (10), pp.3027-3039.
<10.5194/bg-7-3027-2010>. <hal-01050836>
HAL Id: hal-01050836
https://hal.archives-ouvertes.fr/hal-01050836
Submitted on 25 Jul 2014
HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est
archive for the deposit and dissemination of sci- destin´ee au d´epˆot et `a la diffusion de documents
entific research documents, whether they are pub- scientifiques de niveau recherche, publi´es ou non,
lished or not. The documents may come from ´emanant des ´etablissements d’enseignement et de
teaching and research institutions in France or recherche fran¸cais ou ´etrangers, des laboratoires
abroad, or from public or private research centers. publics ou priv´es.
Biogeosciences,7,3027–3039,2010
Biogeosciences
www.biogeosciences.net/7/3027/2010/
doi:10.5194/bg-7-3027-2010
©Author(s)2010. CCAttribution3.0License.
Mortality as a key driver of the spatial distribution of aboveground
biomass in Amazonian forest: results from a dynamic vegetation
model
N.Delbart1,*,P.Ciais1,J.Chave2,N.Viovy1,Y.Malhi3,andT.LeToan4
1LaboratoiredesSciencesduClimatetdel’Environnement,UMR8212(CEA/CNRS/UVSQ),CEA-ormedesMerisiers,
91191Gif-sur-Yvette,France
2LaboratoireEvolutionetDiversite´ Biologique,UMR5174(CNRS/UniversitePaulSabatierToulouse3),
31062Toulouse,France
3EnvironmentalChangeInstitute,SchoolofGeographyandtheEnvironment,OxfordUniversity,
SouthParksRoad,Oxford,UK
4Centred’EtudesSpatialesdelaBiosphe`re,UMR5126(CNRS/CNES/IRD/UPS),Toulouse,France
*nowat: Universite´ ParisDiderot-Paris7,Paris,France
Received: 7April2010–PublishedinBiogeosciencesDiscuss.: 29April2010
Revised: 31August2010–Accepted: 17September2010–Published: 6October2010
Abstract.DynamicVegetationModels(DVMs)simulateen- PreviousstudiesshowedthatAmazonianforestswithhigh
ergy, water and carbon fluxes between the ecosystem and productivity have a higher mortality rate than forests with
the atmosphere, between the vegetation and the soil, and lowerproductivity. Weintroducethisrelationship,whichre-
between plant organs. They also estimate the potential sults in strongly improved modelling of biomass and of its
biomass of a forest in equilibrium having grown under a spatial variations. We discuss the possibility of modifying
given climate and atmospheric CO level. In this study, themortalitymodellinginORCHIDEE,andtheopportunity
2
we evaluate the Above Ground Woody Biomass (AGWB) to improve forest productivity modelling through the inte-
and the above ground woody Net Primary Productivity grationofbiomassmeasurements,inparticularfromremote
(NPP )simulatedbytheDVMORCHIDEEacrossAma- sensing.
AGW
zonianforests,bycomparingthesimulationresultstoalarge
setofgroundmeasurements(220sitesforbiomass,104sites
for NPP ). We found that the NPP is on average
AGW AGW
overestimated by 63%. We also found that the fraction of
1 Introduction
biomassthatislostthroughmortalityis85%toohigh.These
model biases nearly compensate each other to give an aver-
Tropical rainforests play a crucial but poorly known role in
agesimulatedAGWBclosetothegroundmeasurementaver-
theglobalcarboncycle(MalhiandGrace,2000). Deforesta-
age. Nevertheless,thesimulatedAGWBspatialdistribution
tion and forest degradation contribute significantly to CO
differssignificantlyfromtheobservations. Then,weanalyse 2
emissionstotheatmosphereandareequivalenttoabout15–
thediscrepanciesinbiomasswithregardstodiscrepanciesin
25% of fossil fuel emissions (IPCC, 2007; Le Que´re´ et al.,
NPP and those in the rate of mortality. When we cor-
AGW
2009). ThereisalsostrongevidencethatundisturbedAma-
rectfortheerrorinNPP ,theerrorsonthespatialvaria-
AGW
zonian and African tropical forests are currently a sink of
tionsinAGWBareexacerbated,showingclearlythatalarge
carbonastheyareaccumulatingmorebiomass(Stephenset
partofthemisrepresentationofbiomasscomesfromawrong
al.,2007;Lewisetal.,2009a;Phillipsetal.,2004;Lewiset
modellingofmortalityprocesses.
al.,2009b;butseeJacobson,2007). Themechanismsbehind
this carbon sink are subject to debate (Lewis et al., 2009;
Correspondenceto: N.Delbart Ko¨rner,2009): thecarbonsinkcouldbeduetophotosynthe-
([email protected]) sisstimulationbyincreasingatmosphericCO rate(Phillips
2
PublishedbyCopernicusPublicationsonbehalfoftheEuropeanGeosciencesUnion.
3028 N.Delbartetal.: Amazonianforest: resultsfromadynamicvegetationmodel
etal.,2008;Glooretal.,2009;Ciaisetal.,2007)ortoregen- based measurements of above ground biomass and net pri-
erationfromearlierdisturbance(Chaveetal.,2008). maryproductivityallocatedtotheabovegroundpartsofthe
Diagnosing the carbon balance of tropical forests should plants,KeelingandPhillips(2007)showedthatthispremise
relyonrobustandrepeatedbiomassestimationsatlargescale is unrealistic. Specifically, they found that biomass stock
(Houghton, 2005; DeFries et al., 2001), but until recently is not linearly related to the productivity of tropical forests.
maps of biomass stock in Amazonia retrieved from remote Consideringarangeofsites,above-groundbiomassshowsa
sensing,groundmeasurementsormodellingshowedstriking hump-shaped variation with productivity, as sites with high
differences(Houghtonetal.,2001).Recently,advancescom- productivity (high carbon input) also have high turnover
bining field data, remote sensing techniques and innovative (high carbon output). To faithfully depict future trends in
statisticalmethodshaveprovidednewinsightsonthedistri- tropicalforestbiomassstocks,aDVMshouldcorrectlysim-
butionofAboveGroundWoodyBiomass(AGWB)inAma- ulate both woody productivity and mortality fluxes. Some
zonia(Malhietal.,2006;Saatchietal.,2007). Inthefuture, DVMs (e.g. Moorcroft et al., 2001) simulate temporal vari-
dedicated remote sensing mission such as the BIOMASS ationsinmortalitywithforestregeneration(pioneerspecies
mission(ESA,2008;LeToanetal.,2010)willprovidecon- have a high turnover rate) or drought that increases mortal-
sistent,global,andgriddedAGWBdatasetatspatialresolu- ity. Still, such DVMs are unable to reproduce the regional
tionof100m,andwillallowmonitoringforestbiomasslost variationofbiomassinAmazonia,thehighestbiomasslevels
due to disturbances and ensuing biomass increment. Con- beingobtainedinNorthWestAmazonia(Huntingfordetal.,
sistency among recent maps suggests that the important re- 2008) whereas ground measurements show higher biomass
gional variation in AGWB across Amazonia is not an arte- level in central Amazonia or Guyana (Malhi et al., 2006).
fact of measurement methods. Biomass distribution results Thus,theinitialconditionsofthesimulationspredictingfor-
fromvariationsbothinthetreeallometry(Chaveetal.,2005) estdiebackwithlowerprecipitationdonotmatchtheobser-
and in wood density (Baker et al., 2004). It was also found vations.
that wood density, stem turnover, and productivity co-vary: In this study, we evaluate the spatial variations in
the species with higher productivity have lower wood den- the biomass across the Amazon simulated by the OR-
sity and higher turnover rate. As a result, highest biomass CHIDEEDVM(Krinneretal.,2005). InORCHIDEE,con-
storesarefoundforsiteswithlowerturnoverrates(Malhiet trarily to other DVM (Sato et al., 2007), individuals are not
al.,2006). represented and forests are considered as one average tree.
Hypotheses about future carbon balance of Amazonian Photosynthesised carbon remaining after plant maintenance
forests may be formulated using ecosystem models. One and growth respiration computation is distributed to leaf,
General Circulation Model coupled with a Dynamic Vege- above ground wood, below ground wood, fine roots, fruits
tation Models (DVM) predicted massive biomass loss dur- and reserves. Primary production and allocation of carbon
ing the 21st century induced by reduced precipitation (Cox areruledbyprocess-orientedschemes(Friedlingsteinetal.,
et al., 2001; Huntingford et al., 2008), in line with results 1999; Krinner et al., 2005). However, the loss of biomass
from ecosystem manipulation experiments (Nepstad et al., throughmortalityismodelledverysimply:everyyearafixed
2007), although most climate models showed a more mod- fractionofthetotalcarboninwoodislosttolitter. Themor-
erateintensificationofseasonaldrought(Malhietal.,2009). tality rate is defined as the inverse of the time of residence
DVMscomputeenergy,waterandcarbonfluxesbetweenthe of carbon in wood, which is constant and prescribed. As
ecosystem and the atmosphere, between the vegetation and there is no spatially explicit individual-based representation
the soil, and among plant organs. Most DVMs employ the andasthereisonecarbonreservoirfortrunksandbranches,
conceptofan“averageplant”(butseee.g.Satoetal.,2007). thetimeofresidenceofcarboninwoodisequivalenttothe
For each average plant organ, a DVM simulates the carbon averagetreelifespanwithinthismodelframework.
input (primary production allocated to the organ, transloca- OurfirstobjectiveistoexplorewhyDVMsfailtoprovide
tionfromanotherorgan)andoutput(mortalityorsenescence, spatial gradients in AGWB under the current climatic con-
translocationtoanotherorgan).Underaconstantclimateand ditions. Using a large set of AGWB observations, we show
atmosphericCO level, aDVMreachesasteady-stateequi- that the DVM ORCHIDEE outputs do not match empirical
2
libriumatwhichcarboninputsandoutputsforeachcompart- observations of regional variation in biomass over Amazo-
mentbalanceeachotheronthelongterm. Dependingonthe nian forests. We analyse the discrepancies between model
studyecosystem,thisequilibriummaybereachedaftersev- outputsanddatawithregardtodiscrepanciesincarboninput
eraldecadesorcenturiesofsimulation. Thus, DVMssimu- toabovegroundwoodandintherateofmortality. Wethen
latethepotentialclimaxbiomassofamatureforestshaving testtheideaofrelatingmortalitytoproductivityasitissug-
grownunderagivenclimateandatmosphericCO level. gested by empirical evidence from the field data (Malhi et
2
By construction, any increase in primary productivity al.,2004). Wediscusstheconsequencesofthismodification
driven by climatic gradients or temporal changes (nitrogen in predicting the spatial distribution of Amazonian biomass
deposition or CO increase) should result in an increase in stores, and discuss possible improvements in the mortality
2
biomass stock in a DVM. In a recent synthesis of ground moduleofDVMs.
Biogeosciences,7,3027–3039,2010 www.biogeosciences.net/7/3027/2010/
N.Delbartetal.: Amazonianforest: resultsfromadynamicvegetationmodel 3029
Our second objective is to explore the opportunity to im- Then,
prove forest productivity modelling through the integration
1
ofmoreaccurateandspatiallyexplicitobservationsonforest AGWB(n) = AGWB(n − 1) × 1 − (4)
(cid:18) t (cid:19)
biomass stocks from satellite remote sensing. Here we dis- residence
+NPP (n − 1).
cuss the use of future BIOMASS satellite data to constrain AGW
thecarbonfluxessimulatedbyaDVMformatureforests.
Further details about the calculation of GPP, R and
a
falloc−organ areprovidedinKrinneretal.(2005). Thevalue
of t is prescribed and constant, set equal to 30 years
2 Materialandmethods residence
forthetropicalforestbiome.
2.1 ORCHIDEE Ourobjectiveistotesttheseassumptionsforundisturbed
tropicalforests. Thuswefixedthelengthofthesimulations
ORCHIDEE (Krinner et al., 2005) is a dynamic vegetation (Nyears) to 206 years (from 1801 to 2006), after checking
model. ItistheresultofthecouplingoftheSECHIBAland- thatbiomassstoresequilibrateafter100years.
surfacescheme(Ducoudre´etal.,1993),whichisdedicatedto
surfaceenergyandwaterbalances, andthecarbonandveg- 2.2 Fielddata
etation model STOMATE that calculates the carbon fluxes
LocationsofgroundobservationsareshowninFig.1.
betweentheatmosphere,thevegetationandthesoil. Athird
componentisthevegetationdynamicsmodulethatcalculates
2.2.1 Biomass
thespatialdistributionofvegetationtypes,butitisnotacti-
vatedinoursimulationsasweprescribetheevergreentropi-
Ground-based above ground woody biomass measurements
calvegetationtype.
across220AmazonianforestsitesweretakenfromMalhiet
In this study we concentrate on the carbon dynamics
al.(2006). Someofthesevaluesweredirectlyderivedfrom
sub-model STOMATE. In this model, Net Primary Produc-
individual tree diameter data using the allometric relation-
tion (NPP) is modelled as Gross Primary Production (GPP
ship of Baker et al. (2004) that incorporates wood density
the amount of atmospheric carbon assimilated by photo-
information. At other sites where only plot-level basal area
synthesis) minus autrotrophic respiration (R ). Both her-
a information was available, AGWB estimates were derived
bivore grazing of NPP (Keeling and Phillips, 2007), and
from an allometric model that relates AGWB to basal area
Volatile Organic Compounds emission of C to the atmo-
measurementsandcorrectingforvariationsinwooddensity.
sphere (Kesselmeier et al., 1999) are ignored. NPP is allo-
cated to several vegetation organs or compartments: above 2.2.2 WoodyNPPandresidencetime
ground wood (AGW), below ground wood (BGW), fine
roots,leaves,fruitsandnonstructuralreservecarbonstores. Net primary production allocated to aboveground wood
Organs lose carbon through mortality or senescence. The (NPP )wasestimatedfor104Amazonianforestsitesby
AGW
amountofbiomassallocatedtoeachorganiscalculatedfrom Malhi et al. (2004), based on dendrometric measurements
thefollowingequation: conducted at the same site during two or more censuses.
Onlytreesthatreachedtheminimalcensussizeof10cmdbh
NPPorgan = falloc−organ × NPP (1) were considered. Specifically, NPP is the sum of two
AGW
components: 1/ the individual tree biomass increment mea-
with NPP and NPP expressed in tonsC/ha/year, and
organ
sured during the interval (inferred from the increment in
falloc−organ being a dimensionless fraction ranging from 0
basalarea),plus2/themassoftreesrecruitedduringtheinter-
to1.
val. Acensus-intervalcorrectionwasintroducedtoaccount
Atyearn,AGWBisgiven(intonsC/ha)by:
for trees that recruited, grew and died between censuses.
AGWB(n) = AGWB(n − 1) (2) Theconversionfromthebasalareaincrementtobiomassin-
+NPP (n − 1) − mortality(n − 1) crement incorporates wood density estimates (Baker et al.,
AGW
2004). As all sites did not have all necessary information,
wheremortality(intonsC/ha/year)equals the correction due to wood density was determined empiri-
callyandappliedtotheavailabledata(seedetailsinMalhiet
AGWB(n − 1)
mortality(n − 1) = , (3) al.,2004). Malhietal.(2004)alsoprovidedestimatesofthe
t
residence residencetimeofcarbonintreewoodybiomass,definedby
withtresidence(inyears)beingthetimeofresidenceofcarbon theratioofAGWBtoNPPAGW.
in wood. Note that the inverse of t is equal to the
residence
rateofmortality,i.e.thefractionofAGWBlostannuallyvia 2.2.3 Leafandfruitallocation
mortality.
Assuming that the forest canopy biomass is in equilib-
rium, NPPallocationtoleafandfruitwasinferredfromthe
www.biogeosciences.net/7/3027/2010/ Biogeosciences,7,3027–3039,2010
3030 N.Delbartetal.: Amazonianforest: resultsfromadynamicvegetationmodel
15
10
5
0
-5
-10
-15
-20
-25
-90 -85 -80 -75 -70 -65 -60 -55 -50 -45 -40 -35 -30
Fig.1. Locationsofgroundobservations. Blacktriangles: abovegroundwoodybiomass(Malhietal.,2006). Greendots: aboveground
woodyNPP(Malhietal.,2004). Reddiamonds:allocationfractions(Araga˜oetal.,2009). Bluesquares:leafandfruitallocation(Chaveet
al.,2010).
correspondinglitterfallfor62(leaf)and51(fruit)sitesfrom variablesAGWB,NPP,NPP separatelyinordertoeval-
organs
Chaveetal.(2010). uate their mean values, their range and their spatial distri-
bution. We used the statistical indicators as formulated in
2.2.4 Completeallocationpattern Willmott(1982).
For ten sites ranging across lowland Amazonia, Araga˜o et 2.5 ImpactofcorrectingNPP whilekeepinga
AGW
al.(2009)estimatedthetotalNPPanditspartitioningamong prescribedt
residence
organs,includingcoarseandfineroots.
AGWB reaches equilibrium when NPP and mortality
AGW
2.3 Climaticdata compensateeachotheronaverageoverseveralyears. Equi-
libriumAGWB isthusdefinedas:
max
The climate dataset CRU-NCEP used for driving OR-
CHIDEE in this study is a combination of two existing AGWB = NPP × t (5)
max AGW residence
datasets: the CRU TS.2.1 0.5◦×0.5◦, (Mitchell and Jones,
2005)monthlyclimatologycoveringtheperiod1901to2002 Here the objective is to determine the potential impact on
and the NCEP reanalysis 2.5◦×2.5◦ 6h time step begin- the equilibrium AGWB of improving the modelling of
max
ning in 1948 and available in near real time (Kalnay et al., NPP while preserving a prescribed mortality rate. For
AGW
1996).Theprocessing(seeappendixA)ofCRU-NCEPaims thispurpose,weapplyamultiplicativecorrectionont
residence
atbuildingaclimaticdatasetat0.5◦×0.5◦spatialresolution, to make it equal to the average value from the ground
which keeps the diurnal and daily variability of NCEP and measurements and we report this correction on AGWB
max
the monthly averages from CRU and which covers the pe- through Eq. (5). Then we test two types of correction on
riod from 1901 until now. We used the climate data from NPP ,andevaluatetheirimpactsimilarlythroughEq.(5).
AGW
years 1901–1950 repeatedly to simulate the forest growth
over1801–1900. – First, we test the effect on AGWBmax of removing the
average bias on NPP that was found from the sta-
AGW
2.4 StatisticalanalysisofORCHIDEEresults tistical analysis. This allows testing of what would be
thesimulatedAGWB byadjustingthemodelparam-
max
We ran ORCHIDEE at each site where ground measure- eterswhilekeepingthesameformulations.
ments were available. We averaged the model outputs over
the last 50 years of the simulations. We tested a shorter – Second, we correct the site specific bias on NPP .
AGW
period (10 years) and showed that this had no impact on This allows testing what would be the modelled
the interpretation of results. We analyzed the model output AGWB if NPP and therefore all processes
max AGW
Biogeosciences,7,3027–3039,2010 www.biogeosciences.net/7/3027/2010/
N.Delbartetal.: Amazonianforest: resultsfromadynamicvegetationmodel 3031
(a) (b)
220 ar) 7
ha) 200 a/ye 6
onsC/ 180 nsC/h 5
WB (t 114600 (toW 4
G G
E A 120 PPA 3
DE 100 E N 2
HI E
C 80 D
R HI 1
O 60 C
R
40 O 0
40 60 80 100 120 140 160 180 200 220 0 1 2 3 4 5 6 7
In situ AGWB (tonsC/ha) In situ NPPAGW (tonsC/ha/year)
(c) (d)
220 220
200 200
a) 180 a) 180
h h
C/ 160 C/ 160
s s
on 140 on 140
B (t 120 B (t 120
W W
G 100 G 100
A A
80 80
60 Data Model 60 Data Model
40 40
1 2 3 4 5 6 7 -80 -75 -70 -65 -60 -55 -50 -45
NPPAGW (tonsC/ha/year) Longitude (degrees)
Fig.2. ComparisonofORCHIDEEoutputswithgroundmeasurements. (a)Abovegroundwoodybiomass(AGWB);(b)NetPrimaryPro-
ductionallocatedtoabovegroundwood(NPPAGW).(c)AGWBagainstNPPAGWfromORCHIDEEandgroundmeasurements.(d)AGWB
◦ ◦
againstlongitudeforlatitudebetween2 northand10 south,fromORCHIDEEandgroundmeasurements.Groundmeasurementsarefrom
Malhietal.(2004,2006). Foreachx-axisclass:theverticalboxrepresentsthe25and75percentiles;medianinrepresentedbyalightline;
averageisrepresentedbyaboldline;whiskersextremitiesshowtheminimumandmaximumvalues. In(c)and(d),redisforORCHIDEE
outputsandblackforthegroundmeasurements.Thewidthoftheboxisreducedbyhalfiflessthan5sitesbelongtoaspecificx-axisclass.
involved in photosynthesis, respiration and alloca- α would be equal to 0. If AGWB were proportional to
max
tion were perfectly modelled. Remaining errors on NPP , the residence time would be fixed and α would
AGW
AGWB comefromthemortalityonly,whichinthis beequalto1asincurrentORCHIDEEparameterisation. If
max
caseisdefinedfromaconstantt . AGWB decreasedwithNPP ,αwouldbenegative.
residence max AGW
We evaluate K and α by minimizing the average abso-
2.6 Introducinganewmortalitymodel
lute difference between Eq. (7) and t from Malhi et
residence
al. (2004) for the NPP values provided. We then apply
t is given by Eq. (5) where AGWB is the equi- AGW
residence max Eq.(6)firsttoretrieveAGWBfromgroundmeasurementsof
librium AGWB. Across Amazonia, AGWB may change
max NPP ,secondtoretrieveNPP fromgroundmeasure-
withvariationsinspeciesdistribution,climateandsoilprop- AGW AGW
mentsofAGWB.
erties. Consequently, AGWB could also spatially vary
max
withecosystemproductivitysuchthat:
AGWBmax = K × NPPαAGW (6) 3 Results
Then t is obtained combining Eqs. (5) and (6) such
residence
that: 3.1 EvaluationofAGWBandNPPAGWsimulatedby
ORCHIDEE
t = K × NPPα × NPP−1 (7)
residence AGW AGW
αistheparameterthatquantifiesthevariationsinAGWB Using ORCHIDEE’s original parameters, we compared the
max
with NPP . If AGWB were not related to NPP , model outputs with ground measurements. On average,
AGW max AGW
www.biogeosciences.net/7/3027/2010/ Biogeosciences,7,3027–3039,2010
3032 N.Delbartetal.: Amazonianforest: resultsfromadynamicvegetationmodel
Table1.comparisonofORCHIDEEoutputs(AGWBintonsC/ha,NPPandNPPorgansintonsC/ha/year)withgroundmeasurements.
NPP
AGWB NPPAGW NPPLeaf NPPFruit NPPFruit+Leaf NPPAGW NPPBGW NPPFineroot TotalNPP
Sourceof Malhiet Malhiet Chaveet Chaveet Araga˜oet Araga˜oet Araga˜oet Araga˜oet Araga˜oet
grounddata al.,2006 al.,2004 al.,2010 al.,2010 al.,2009 al.,2009 al.,2009 al.,2009 al.,2009
N 220 100 62 51 10 10 10 10 10
ObsMean 146.62 3.04 2.72 0.31 4.33 3.71 0.61 4.07 12.83
ModelMean 138.83 4.96 1.79 1.00 2.71 5.05 1.28 0.64 9.68
RMSEs 29.87 2.01 1.25 0.72 1.96 1.51 0.70 3.83 3.98
RMSEu 16.71 0.6 0.18 0.12 0.28 0.52 0.13 0.17 1.02
RMSE 34.23 2.1 1.26 0.73 1.98 1.60 0.71 3.84 4.11
Slope 0.07 0.25 −0.01 −0.01 −0.01 0.10 0 0.04 0.01
Intercept 129.14 4.21 1.81 1.01 2.75 4.68 1.28 0.46 9.54
R 0.12 0.32 −0.04 −0.01 −0.04 0.14 0 0.42 0.03
the simulated AGWB was close to the ground measure- 3.2 ImpactonAGWBofcorrectingNPP while
AGW
ment (Fig. 2a), but the simulated NPP was too high keepingaconstantmortalityrate
AGW
(Fig. 2b, Table 1). We found that to set the AGWB to real-
isticvalues,theoverestimationofNPPAGWwasbalancedby First,wecorrectedthemodeloutputsbyde-biasingNPPAGW
anoverestimationofthemortalityrate(3.33%year−1 given and fixing tresidence as the average of the ground measure-
t =30years). ThisrateassumedinORCHIDEEwas ments. Thiscorrectioncorrespondstowhatcouldbeimple-
residence
higherthanobservedonaverageatthegroundmeasurement mentedinORCHIDEEbyonlyadjustingmodelparameters
sites(1.8%year−1asaveraget =55years). while keeping the logic of a constant rate of mortality. The
residence
DespiteoverestimationofNPP (Fig.2,Table1),total averageAGWBwasstillclosetotheaverageofgroundmea-
AGW
NPP (above and below ground) was found to be underesti- surements.However,asthemortalitywasassumedtobecon-
mated by 25% (Fig. 3a). This is explained by the fact that stant,therewasnoimprovementinthespatialdistributionof
the allocation fraction to above ground wood was overesti- AGWBcomparedtoresultsinFig.2.
mated(Fig.3b,Table1)inthemodelcomparedtoempirical Second, we corrected the model outputs by forcing
data (Araga˜o et al., 2009; Chave et al., 2010). Allocation NPPAGW with the data while keeping a constant tresidence
to below ground wood and to fruits was also overestimated (55years). Thistestallowedassessmentoftheerrorcoming
(Fig.3b–c,Table1).Bycontrast,allocationtoleaveswasun- from tresidence alone avoiding all errors on NPPAGW coming
derestimatedby34%,andallocationtofinerootsby84%in from the modelling of GPP, autotrophic respiration, alloca-
ORCHIDEE.Fornoneofthetestedparameterswasthereei- tion and uncertainties in the climatic data. The model-data
therasignificantcorrelationoralinearregressionslopethat discrepanciesinspatialdistributionofAGWBarethenexac-
iscloseto1(Table1),showingthemodelcannotreproduce erbated(Fig.4). Thisindicatesthatimprovingthemodelling
the observed spatial patterns. The simulated NPPleaf+fruit ofGPP,respirationandallocationwouldleadtoaworsedis-
isequalto0.54NPP ,whereasthegroundmeasurements tribution of AGWB in the absence of improvement in the
AGW
indicatethatNPPleaf+fruit=1.67NPPAGWonaverage. mortalitymodel.
ThespatialdistributionofAGWBacrosstheAmazonfor-
3.3 ImpactonAGWBofaresidencetimeinwood
est, as modelled by ORCHIDEE’s default parameters, dif-
governedbyNPP
fered from empirical observations, with a peak in AGWB AGW
in western Amazonia while ground measurements showed
We found that the best empirical fit for the relationship
maximumAGWBincentralAmazonia(Fig.2d). Simulated
between t and NPP gave an α value of −0.32
residence AGW
AGWB increased with NPP in contrast to direct obser-
AGW (Figs.5and6).ThisindicatesaslightdecreaseinAGWB
max
vations(Fig.2c). Inthefollowingweexplorehowmuchof
withincreasingNPP . Thisdifferssignificantlyfromthe
AGW
thisisexplainedbythefactthatt isconstantinOR-
residence value α=1 that corresponds to the fixed mortality rate im-
CHIDEE,whereasthegroundmeasurementsdatashowthat
plementedinORCHIDEE.
itdecreaseswithincreasingNPP .
AGW We then derived AGWB from Eq. (6) applied to the ob-
servationsofNPP . TheregionaldistributionofAGWB
AGW
matchedbettertotheobservationsthanwithafixedmortality
Biogeosciences,7,3027–3039,2010 www.biogeosciences.net/7/3027/2010/
N.Delbartetal.: Amazonianforest: resultsfromadynamicvegetationmodel 3033
(a) rate(Fig.7aandc). Hence,inferringtresidencefromNPPAGW
20 resulted in a clear improvement for most of the sites in our
dataset compared to the results presented in Fig. 4. Then,
we derived NPP from Eq. (6) applied to ground obser-
AGW
ear) 15 vationsofAGWB.TheretrievedNPPAGW givenAGWBbe-
a/y came close to the observations (Fig. 7b and d). This shows
h
C/ thatdirectobservationsofAGWBmaybeusedtoconstrain
s
P (ton 10 bothNPPAGWandmortalityinaDVM.
P
N
al
ot 4 Discussionandoutlook
RC t 5
O
Here we have demonstrated that the predictions of a DVM
over the Amazon were incorrect in two major respects: the
0 spatial distribution of AGWB did not match empirical ob-
0 5 10 15 20
servations,andthemodelassumedatoohighturnover(both
In situ total NPP (tonsC/ha/year)
NPP andmortality). Onemajorfindingisthatmortality
(b) AGW
0.6 rateisasimportantasnetprimaryproductivityandallocation
in determining spatial gradients of above ground biomass,
d (f)alloc-organ 00..45 asvtnaardniadthtiinaotgncsbaiilnocumalbaaotsivsnegpgrterhoveuemnndtosrbtciaoolrimtryeacsatss.saWimceountlhastteainonnptrfoorafpcotthsioeendsopanfattaihale-l
e
at ternativestrategytoaccountforthesebiases,whichconsists
P alloc 0.3 in relating mortality to NPPAGW on the basis of empirical
NP evidence. Thiswaseffectivetoexplainandreproducepartly
n of 0.2 thevariationsinAGWBofourdataset.Weacknowledgethis
o
acti strategy should be tested over an independent dataset in the
Fr 0.1
future,forexampleincentralAfrica.
Here, we calculated the mortality rate from a long term
0
AGW Leaf + fruit Coarse roots Fine roots averagedNPP . Thisapproachisprobablyonlyvalidfor
(c) AGW
the near-equilibrium context of mature, old growth forests.
4.5
Two examples of where this equilibrium validity breaks
4
down are given below. First, immediately after disturbance
ar 3.5 pioneer species with high turnover rate are favoured over
e
a/y 3 late-successionalones. Thus,tresidenceshouldbesmallerdur-
h
C/ 2.5 ingthefirstyearsofsimulation(Moorcroftetal.,2001). We
s
on expect that this should not affect steady-state biomass in a
P (torgan 1.25 breingtlleyarfemcoovdeerlinsgucfhroamsOaRreCcHenItDdEisEtubrubtaanncyefwooreusldttnhoattimsactucrh-
P
N
1 perfectly with ORCHIDEE’s predictions. Second, mortal-
0.5 ityincreasesasaconsequencetodroughtstress(Nepstadet
al., 2007). ORCHIDEEwouldsimulaterealisticallythede-
0
Leaf Fruit crease in primary productivity in case of drought, but then
Eq.(7)wouldleadtoadecreaseofthemortalityrate,which
Fig. 3. Comparison of ORCHIDEE allocation with ground mea-
is at odds with observations during the 2005 drought in the
surements. (a)TotalNetPrimaryproduction;(b)Fractionofallo-
Amazon (Phillips et al., 2009). Then, the background mor-
cationoftotalNPPbetweenorgans(falloc−organs): AboveGround talityrateasmodelledbyEq.(7),whichappearsfromourre-
Wood (AGW), Leaf and fruit, Coarse Roots, Fine Roots (c) NPP
sultsnecessarytoreproducedregionalvariationsinAGWB,
allocated to leaf and fruit. Ground measurements for (a) and (b)
shouldbemodulatedbyshorttermvariationswheremortal-
arefor10sites(Araga˜oetal.,2009).Groundmeasurementsfor(c)
arebasedonlitterfallmeasurementsmadefor62sites(Chaveetal., ity increases in case of adverse climate conditions in order
2010). Foreachx-axisclass:theverticalboxrepresentsthe25and to simulate temporal variations in AGWB. However, mod-
75percentiles;medianinrepresentedbyalightline;averageisrep- erate and progressive decrease in precipitation may favour
resented by a bold line; whiskers extremities show the minimum slow-growing species, with a low turnover rate and a high
andmaximumvalues.In(b)and(c),redisforORCHIDEEoutputs biomass. ThispointwasignoredinpreviousAmazonianfor-
andblackforthegroundmeasurements. estdiebacksimulations(Coxetal.,2004;Huntingfordetal.,
www.biogeosciences.net/7/3027/2010/ Biogeosciences,7,3027–3039,2010
3034 N.Delbartetal.: Amazonianforest: resultsfromadynamicvegetationmodel
(a) (b)
ar) 6
ha) 250 a/ye 5
sC/ C/h
on 200 ns 4
WB (t (toW
G G 3
A 150 PA
E P
E N 2
D E
HI 100 E
C D
R HI 1
O C
50 R
O 0
50 100 150 200 250 0 1 2 3 4 5 6
In situ AGWB (tonsC/ha) In situ NPPAGW (tonsC/ha/year)
(c) (d)
250 250
C/ha) 200 C/ha) 200
s s
n n
B (to 150 B (to 150
W W
G G
A A 100
100
Data Model 50 Data Model
50
1 2 3 4 5 6 -80 -75 -70 -65 -60 -55 -50 -45
NPPAGW (tonsC/ha/year) Longitude (degrees)
Fig.4.ImpactofforcingNPPAGWwhilekeepingaconstantmortalityrate.SamelegendthanFig.2exceptthatweusetheonly72sitesfor
whichwehavebothgroundmeasurementsofNPPAGWandAGWB.
2008)andcouldbemodelledthroughEq.(7). Nevertheless, withbettersoilphysicalpropertiesandlowerfertilityinduc-
asindicatedbytheobservationsofcurrentbiomespatialdis- ingrespectivelyalowermortalityrateandalowerNPP .
AGW
tribution (Malhi et al., 2009b), forest might be replaced by Thesetwofactorsfavourhighwooddensitylate-successional
savannah if a large decrease in precipitation is experienced species, which ends up in higher AGWB. Equation (7) is
inthefutureintheAmazonianregion. in line with this explanation, as long as physical properties
Soil type is an important factor influencing NPP and andfertilitypropertiesco-vary,whichappearstobethecase
AGW
t , as shown in Fig. 8. For example, forests with low from the soil properties measurements reported in Quesada
residence
NPP and long t are favoured on older oxisol, etal.(2010).
AGW residence
whereas forests with high NPP and short t are Thisstronginfluenceofsoilpropertiescouldbeakeyissue
AGW residence
favoured on entisol. Based on ground measurements, Que- when modelling the future evolution of Amazonian forests
sada et al. (2009) analysed the impact of soil properties on under a climate change scenario, as soil type may limit
the mortality rate and on NPP . The mortality rate was the floristic composition change that we suggest to model
AGW
found essentially influenced by the soil physical properties throughEq.(7).However,thismayalsoallowderivingmaps
(topography, soil depth, structure), whereas NPPAGW was of average tresidence, NPPAGW and thus AGWB from a soil
found primarily driven by fertility parameters, essentially typemap.
phosphorus availability. The authors proposed that AGWB Anegativeα valueisconsistentwiththeobservationthat
gradients can be explained by the ecosystem dynamics that slow growing forests as in central Amazonia have higher
is essentially driven by these soil properties. In Western biomassthanthefastgrowingforestasinWesternAmazonia.
Amazonia, poor soil physical properties (steep slope, shal- However we cannot exclude that this apparent trend is af-
low soils) favour high mortality rate, which favours early- fectedbyAGWBorNPP measurementerrors. Ifαwere
AGW
successional species with low wood density, whereas the equal to zero, observed spatial variations of biomass would
high phosphorus availability induces higher NPP . In be independent of NPP . Then, this would mean that a
AGW AGW
contrast,incentralAmazonia,ecosystemsarelessdynamic, DVM that would perfectly simulate carbon fluxes could at
Biogeosciences,7,3027–3039,2010 www.biogeosciences.net/7/3027/2010/
N.Delbartetal.: Amazonianforest: resultsfromadynamicvegetationmodel 3035
160 -1 10
140 -0.8 9.5
s)
ar
e
n (y 120 -0.6 9
o
b
car 100
e of α -0.4 8.5
c
en 80
d
esi -0.2 8
of r 60
e
m
Ti 0 7.5
40
0.2 7
20 50 100 150 200 250 300 350 400 450 500
1 2 3 4 5 6
K
Above ground woody NPP (tonsC/ha/year)
Fig. 5. time of residence of carbon in wood plotted versus Fig.6. Averageabsolutedifference(years)betweenthecomputed
NPPAGW.Squares:groundmeasurementsfromMalhietal.2004. tresidence (fromEq.7)andthegroundmeasurements(Malhietal.,
Boldlinerepresentsthebestoverallfit:tresidence=217×NPP−AG1.W32 2004)inasystematicexplorationofthevaluesofαandK.
(averageabsoluteerror: 7.30years). Thinlinerepresentsthebest
fit for α=0: tresidence=154×NPP−AG1W (average absolute error :
8.6years). mustbeexpectedasreallocatingcarbontoleafandfineroots
would stimulate photosynthesis and increasing water con-
sumptionbyplants.
best give an AGWB that is spatially constant. In this case, Thanks to the new formulation of mortality and because
biomass observations such as those from remote sensing αwasfounddifferentfromzero,observationsofAGWBcan
couldbeusedtoestimatetheKvalue(thatwouldvaryspa- constrainNPP andt .Whiletheretrievedt
AGW residence residence
tiallyindependentlyfromproductivity)fromEq.(6)andthen could be directly ingested by the model, correcting a DVM
establish the relationship between mortality and NPPAGW in order to reach the NPPAGW value retrieved from AGWB
fromEq.(7). wouldnotbetrivial. Itmayinvolveacombinationofpossi-
Our results show clearly that an α value of 1, which is blecorrectionsonparameterisationofphotosynthesis,respi-
equivalent to the mortality calculation as it is done in OR- ration and allocation. Then, adjusting the model should re-
CHIDEE, cannot explain the patterns in the data for the spectsomeconstraints. First, CarbonUseEfficiency(CUE,
Amazonianforests. However,wefoundthatkeepingαequal whichistheratioofNPPtoGPP)shouldremainclosetothe
to 1 does not prevent from reproducing spatial variations in 0.30–0.35 values derived from carbon cycling studies made
AGWBfortemperateandborealforestsbiomes(notshown). inAmazonianmatureforests(Malhietal.,2009)asitisthe
In fact, as AGWB displays a hump-shaped variation with case in current ORCHIDEE simulation (CUE=0.36). Sec-
productivity when analysed over a range of biome types ond, carbon allocation should fall in the intervals given by
(Keeling and Phillips, 2007), it is unlikely that Eq. (7) ap- the ground measurements. Then, because photosynthesis
pliestomanyotherbiomes,ifany. increases with nutrient availability (Davidson et al., 2004;
Implementingthenewmortalitycomputationrequiresthat Quesadaetal.,2009),andbecauseallocationdoesnotseem
NPP was modelled correctly. In ORCHIDEE, the pri- to vary with it (Araga˜o et al., 2009), model photosynthesis
AGW
ority in order to improve NPP is to reduce the fraction parameters could be adjusted to reach the NPP value
AGW AGW
of allocation to wood and distribute more carbon to leaves derived from AGWB and Eq. (7). Forcing the model with
and fine roots. At high leaf area values, the DVM tends to biomass observations may thus help accounting implicitly
allocatemorecarbontowoodinordertosimulatethecom- forphosphoruslimitationandotherNPPcontrollingfactors
petitionforlight. However,underthecurrentformulationof in a model like ORCHIDEE, or constrain the modelling of
the allocation pattern, limitation by water or nutrients can- the nitrogen cycle in DVMs (Zaelhe et al., 2010). The use
not be larger than the limitation by light for high leaf area ofbiomassdatatoconstrainttheprocessesmodelledinOR-
index forests. Thus excessive carbon is allocated to wood CHIDEEremainsspeculative,buttheconstraintonNPP
AGW
forourevergreentropicalforestsimulations. Modelparame- looks robust from our results (Fig. 7b and d). It is ex-
tersshouldbeadjustedtomakethemodelledallocationfrac- pected that biomass increments estimated by satellites such
tions to the different organs fall within the intervals given as the proposed BIOMASS mission (Le Toan et al., 2010)
bythegroundmeasurements. Non-lineareffectsinallfluxes at several year interval could be used to infer NPP for
AGW
www.biogeosciences.net/7/3027/2010/ Biogeosciences,7,3027–3039,2010