etude du déterminisme des variations interannuelles des échanges carbonés des écosystèmes...
Post on 04-Jan-2016
214 Views
Preview:
TRANSCRIPT
Etude du déterminisme des variations interannuelles des échanges carbonés des
écosystèmes forestiers européens: une approche basée sur la modélisation des
processusNicolas DelpierreNicolas Delpierre (dir. Eric Dufrêne)
Saclay, 14 décembre 2009
Terrestrial vegetation modulates atmospheric [CO2]
2
1960 1970 1980 1990 2000 2010
Flu
x (G
t C
y-1
)
0
2
4
6
8
10
Fossil fuels + Land Use Change
Le Quéré et al., 2009
Terrestrial vegetation modulates atmospheric [CO2]
2
1960 1970 1980 1990 2000 2010
Flu
x (G
t C
y-1
)
0
2
4
6
8
10
Atmosph.40%
Fossil fuels + Land Use ChangeAtm increase
Le Quéré et al., 2009
Terrestrial vegetation modulates atmospheric [CO2]
2
1960 1970 1980 1990 2000 2010
Flu
x (G
t C
y-1
)
0
2
4
6
8
10
Atmosph.40%
Ocean30%
Fossil fuels + Land Use ChangeAtm increaseOcean uptake
Le Quéré et al., 2009
Terrestrial vegetation modulates atmospheric [CO2]
2
1960 1970 1980 1990 2000 2010
Flu
x (G
t C
y-1
)
0
2
4
6
8
10
Atmosph.40%
Ocean30%
Vegetation30%
Vegetation C uptakeVegetation C uptake accounts for most of the IAV
Forests ~60% of vegetation uptake
Fossil fuels + Land Use ChangeAtm increaseOcean uptakeVegetation uptake
Le Quéré et al., 2009
FLUXNETMonitoring the vegetation / atmosphere C exchanges
Forest sitesNon-forest sites
8CARBOEUROPE CARBOEUROPE
networknetwork
CARBOEUROPEEcological gradient
Coniferous forests
Pinus spp.Picea spp.
Deciduous forests
Fagus sylvaticaQuercus spp.
Evergreen Broadleaves
Quercus ilex
Mixed forests
9
CARBOEUROPEAnnual NEP sums
Boreal (Pinus)
Temperate (Picea)
Temperate (Fagus)
Mediterranean (Q.ilex)
2001 2003 2005 2007
2001 2003 2005 2007
2001 2003 2005 2007
2001 2003 2005 200711
R²=0.40R²=0.40 R²=0.80R²=0.80
CARBOEUROPEExplaining Intersite variations of the C balance
Water balance Temperature
GPP(gC / m² / y)
12
One color =
One site
Southern <52°N Northern >52°N
adapted from Reichstein et al., 2007
CARBOEUROPEExplaining Intersite variations of the C balance
R²=0.40R²=0.40 R²=0.80R²=0.80
R²=0.30R²=0.30 R²=0.70R²=0.70
GPP(gC / m² / y)
Reco(gC / m² / y)
12
Water balance Temperature
Southern <52°N Northern >52°N
adapted from Reichstein et al., 2007
CARBOEUROPEExplaining Intersite variations of the C balance
R²=0.40R²=0.40 R²=0.80R²=0.80
GPP(gC / m² / y)
R²=0.30R²=0.30 R²=0.70R²=0.70
Reco(gC / m² / y)
R²<0.10R²<0.10 R²=0.20R²=0.20
NEP(gC / m² / y)
12
What about interannual variations
???
Southern <52°N Northern >52°N
Water balance Temperatureadapted from Reichstein et al., 2007
CARBOEUROPEExplaining Interannual variations of the C balance
GPP(gC / m² / y)
Reco(gC / m² / y)
NEP(gC / m² / y)
SignificantRelationships
5 sites over 25
SignificantRelationships
3 sites over 25
SignificantRelationships
4 sites over 25
13
Southern <52°N Northern >52°N
Water balance Temperature
Info used Logical linkInterannual
Climate indexes
Annual climate correlative
Empirical vs. Process-based models
14
Statisticalmodels
Info used Logical linkInterannual
Climate indexes
Annual climate correlative
CASTANEACASTANEAClimateClimate
Biological Biological driversdrivers
?? ?? To be To be
tested ??tested ??
Empirical vs. Process-based models
14
Statisticalmodels
Process
Basedmodel
Info used Logical linkInterannual
Climate indexes
Annual climate correlative
CASTANEACASTANEAClimateClimate
Biological Biological driversdrivers
explanatory Proces
sBasedmodel
Empirical vs. Process-based models
14
Quantify the influences of ClimateClimate and Biological driversBiological driversoperating at several timescalesseveral timescales to determine the interannual variations of GPP, Reco and NEP
Statisticalmodels
3)Availability of Statistical tools for signal deconvolution
Criteria for using CASTANEA as a deconvolution tool
1) Biological realism of the simulated processes
2) Accuracy of flux simulations
15
Seasonality of photosynthesis in conifers Seasonality of photosynthesis in deciduous species
- Spring phase- Autumn phase
Evaluation of data quality Model validation at multiple time scales
SA technique revealing seasonal influences SA technique revealing influences at multiple time scales
4. Influence of climate and biological drivers
across time scales
OUTLINEOUTLINE
2. Modelling canopy senescence in deciduous
forests
16
1. Materials & methodsAn overview of the CASTANEA model
3. Model Validation
CO2
Solar radiation
temperature
Radiation interceptionGlobal PAR
PhotosynthesisStomatal Cond.
CASTANEAmodel
Dufrêne et al., 2005
Transpiration
Water vapour
GP
P
17
Solar radiation
temperature
Water vapour
Radiation interceptionGlobal PAR
Photosynthesis
Precipitations
Canopy interception
Throughfall
Stem flow
Litter
Surface
Root zone
drainage
Soil evaporation
TranspirationCanopy
evaporation
CO2
Stomatal Cond.
GP
P
CASTANEAmodel
Dufrêne et al., 2005 17
Solar radiation
temperature
Water vapour
Radiation interceptionGlobal PAR
Photosynthesis
Precipitations
Canopy interception
Throughfall
Stem flow
Litter
Surface
Root zone
Soil evaporation
Transpiration
Carbon AllocationC leaves
C coarse roots
C fine roots
Growth Respiration
C litter
C surface
C deep
HeterotrophicRespiration
CO2
Canopy evaporation
drainage
Stomatal Cond.
GP
P
Rec
o
C aerial wood
C reserves Maintenance Respiration
CASTANEAmodel
Dufrêne et al., 2005 17
CASTANEA Modelling the C balance of European forests
Coniferous forests
Hyytiälä(Boreal Pine)
Tharandt(Temperate Spruce)
Evergreen Bleaves
Puéchabon(Mediterranean Q. ilex)
Deciduous forests
SoroeHainich (Temperate Beech)Hesse
18
Canopy senescence Original modelling scheme
Hesse forestFagus sylvatica
49°N
Leaf fall
19
N resorption
Sep Oct Nov Dec
Sep Oct Nov Dec
Sep Oct Nov Dec
Modelled NEP
Davi et al. (2005)
Canopy senescence Original modelling scheme
Hesse forestFagus sylvatica
49°N
Leaf fall
19
N resorption
Sep Oct Nov Dec
Sep Oct Nov Dec
Modelled NEP
Davi et al. (2005)Sep Oct Nov Dec
Canopy senescence Original modelling scheme
Hesse forestFagus sylvatica
49°N
Leaf fall
19
N resorption
Sep Oct Nov Dec
Sep Oct Nov Dec
Modelled NEP
Davi et al. (2005)Sep Oct Nov Dec
Canopy senescence Original modelling scheme
Hesse forestFagus sylvatica
49°N
Leaf fall
19
N resorption
Sep Oct Nov Dec
Sep Oct Nov Dec
Modelled NEP
Davi et al. (2005)Sep Oct Nov Dec
OakOakBeechBeech
RENECOFOR observationsSenSen9090 = 90% x 36 trees = 90% x 36 trees
Sites n MAT (°C) Alt (m) Sen90
Beech 18 159 9.8 400 20 October
Autumn phenology The RENECOFOR dataset (1997-2006)
20
Driver Effect References
TemperaturesTemperatures
Addicott, 1968
Low temperatures ++ Schnelle, 1952
Schulze, 1970
PhotoperiodPhotoperiod
Long days ++ / / --Addicott, 1968Seyfert, 1970Chuine, 2001
Estrella & Menzel, 2006
Short days ++ / / --Other potential driversOther potential drivers
•Water balance•Mineral deficits
•atmospheric pollution•parasites…
Designing a bioclimatic modelLiterature review
21
8
10
12
14
16
Rel
ati
ve
sen
esc
ence
Tem
per
atu
reDesigning a bioclimatic model
Model formulation
22
Model parameters
Rel
ativ
e se
nes
cen
ceT
emp
erat
ure
Day
len
gth
Senescence initiation date Base temperature Critical T sum
Model formulation
non-linear T x DayLength effects
5
10
15
20
25
0.0
0.2
0.4
0.6
0.8
1.0Jul Aug Sep Oct Nov Dec
Jul Aug Sep Oct Nov Dec
Tbase
Beech
RMSE ( days )
ME (%)
Null model 16a 0
White et al. 15a 10
Jolly et al. 15a 7
This study 13b 33
BeechBeech
Fitting subsetValidation subset
Senescence model assessment (1)
23
Prediction error =13 days(Observation resolution = 7 days)
Observations
Sim
ula
tio
ns
Delpierre et al., 2009
RMSE (days)
ME (%)
Beech 2 46
Senescence model assessment (2)
24
BeechBeech
Yel
low
ing
dat
e (D
oY
)
Validation statistics
Important reduction of the prediction error
Observation uncertainty averaging Reduced contribution of extreme dates
1997 1999 2001 2003 2005
observations
simulations
Canopy senescence Original modelling scheme
Hesse forestFagus sylvatica
49°N
Leaf fall
25
N resorption
Sep Oct Nov Dec
Davi et al. (2005)Sep Oct Nov Dec
Impro
ved
Canopy senescence Original modelling scheme
Hesse forestFagus sylvatica
49°N
Leaf fall
25
N resorption
Sep Oct Nov Dec
Sep Oct Nov Dec
Modelled NEP
Davi et al. (2005)Sep Oct Nov Dec
Impro
ved
Over
estim
ation
Canopy senescence Original modelling scheme
Hesse forestFagus sylvatica
49°N
Leaf fall
25
Sep Oct Nov Dec
Sep Oct Nov Dec
Modelled NEP
Davi et al. (2005)Sep Oct Nov Dec
Impro
ved
Impro
ved
N investment
xN resorption
Model validation across time scalesDAILYDAILY
Hyytiälä(Pinus)
R²=0.92bias= +0.11
Tharandt(Picea)
R²=0.91bias= +0.10
Puéchabon(Q. ilex)
R²=0.74bias= +0.21
Hainich(Fagus)
R²=0.95bias= -0.08
2000 2001 2002 2003 2004 2005 2006 2007
2000 2001 2002 2003 2004 2005 2006 2007
26
Model validation across time scalesANNUALANNUAL
FIHyy RMSE=13, r²=0.82DETha RMSE=66, r²=0.51FRPue RMSE=59, r²=0.82
CASTANEA reproduces 36% - 82% of C flux interannual variance36% - 82% of C flux interannual variance
Model validated Model challenged
27
Defining Flux IAV across time scales
GPP Tharandt (Picea abies) 2000-2007 29
Jan Jul DecApr Oct
Mean annualpattern
Influence on Influence on GPPGPP
Influence on Influence on RecoReco
Climate drivers
Incident Radiation
Temperature
Relative Humidity
Soil water content
Biological drivers
Thermal acclimation
Canopy dynamics (LAI)
Woody biomass
Soil C stock
No effect
No effect
No effect
No effect
No effect
32
Conifers
Conifers
dailycycle
annualcycle
Climate and biological drivers interact at different time scales
Climate drivers Biological drivers
Var
ian
ce i
nd
ex
Var
ian
ce i
nd
ex
Global radiationTemperatureVPD
Leaf Area Index
annualcycle
33
hour day month year hour day month year
dailycycle
annualcycle
Climate and biological drivers interact at different time scales
Climate drivers Biological drivers
Var
ian
ce i
nd
ex
Var
ian
ce i
nd
ex
Global radiationTemperatureVPD
GPP
Leaf Area Index
GPP
annualcycle
Climate modulates short-term flux variability Climate + Biological drivers modulate flux IAV 33
hour day month year hour day month year
Constrained simulations
blue = « mean Rg » referencegrey = original flux (year 2000)
Single driver contribution to flux modulationSingle driver contribution to flux modulation
Day of Year
Day of Year
Hyytiälä, Boreal Pine
34
Proper Rg effect on GPP
Constrained simulations
8 years of dailyGPP anomalies due to radiation
8 years of dailyGPP anomalies
due to Water Stress
2000 2002 2004 2006
2000 2002 2004 2006
Hyytiälä, Boreal Pine
Hyytiälä, Boreal Pine
35
0.000
0.002
0.006
0.008
0.010
OWT variance decomposition
Orthonormal wavelet transform
(Haar basis)
Localize and compare residual signal variances
across time scales
Residual signals variance spectra
36
d w m s y >y
2000 2002 2004 2006
2000 2002 2004 2006
Hyytiälä, Boreal Pine
Hyytiälä, Boreal Pine
0.0
0.2
0.4
0.6
0.8
1.0
OWT variance decomposition
37
Residual signals relative influences
Orthonormal wavelet transform
(Haar basis)
calculate relative influencesof both drivers
d w m s y >y
2000 2002 2004 2006
2000 2002 2004 2006
Hyytiälä, Boreal Pine
Hyytiälä, Boreal Pine
decreasing influence of climate drivers at higher timescales
Deconvolution across time scales
Hyytiälä (Boreal Pine)GPP
d w m s y >y
clim
ate
clim
ate
bio
log
ica
lb
iolo
gic
alAccP
38
RglobalRglobal + LAILAI + droughtdrought control GPP annual IAV
decreasing influence of climate drivers at higher timescales
Deconvolution across time scales
Hyytiälä (Boreal Pine)GPP
clim
ate
clim
ate
bio
log
ica
lb
iolo
gic
alAccP
39
RglobalRglobal + LAILAI + droughtdrought control GPP annual IAV
RglobLAI
REWAccP
Significant contribution of biological driversbiological drivers to GPP-IAV modulation
Deconvolution across time scales
Hyytiälä (Boreal Pine)GPP
clim
ate
clim
ate
bio
log
ica
lb
iolo
gic
alAccP
39
Climatedrivers
60%
Biologicaldrivers
40%
AccP45%
AccP9%
GPP-IAV controls in conifers(2000-2007)
Hyytiälä (Boreal Pine)GPP
Tharandt (Temperate Spruce)GPP
Stronger influence of thermal acclimationthermal acclimation at the warmer site !!!
40
+9°C+9°C +4°C+4°C
RglobTair
VPDVPD
REW
AccPLAIBwood
CsoilCsoilClim
ate
Bio
log
ical
Bio
log
ical
Thermal acclimation AccPThermal acclimation AccP
Jan Jul Nov
Hyytiälä (Boreal Pine)GPP
Tharandt (Temperate Spruce)GPP
GPP-IAV controls in conifers(2000-2007)
40
+9°C+9°C +4°C+4°C
AccP45%
AccP9%
Ac
cP
0.0
0.2
0.4
0.6
0.8
1.0
AccP constraint ++++++AccP constraint ++
Acc
P
0.0
0.2
0.4
0.6
0.8
1.0Thermal acclimation AccPThermal acclimation AccP
Jan Jul Nov
AccP constraint ++++++AccP variations ++
AccP constraint ++AccP variations ++++++
Hyytiälä (Boreal Pine)GPP
Tharandt (Temperate Spruce)GPP
GPP-IAV controls in conifers(2000-2007)
40
+9°C+9°C +4°C+4°C
AccP45%
AccP9%
HyyHyyBoreal Boreal PinusPinus
ThaThaTemperateTemperate
PiceaPicea
HaiHaiTemperateTemperate
FagusFagus
PuePueMedit.Medit.Q. ilexQ. ilex
Contrast of thermal acclimation influence in conifers
Strong influence of REW• Recurrent in Puéchabon
• 2003 drought in Hainich
RglobTair
VPDVPD
REW
AccPLAIBwood
CsoilCsoilClim
ate
Bio
log
ical
Bio
log
ical
Flux-IAV controls in European forests (2000-2007)
41
GPPGPP
REW
REW
AccP
LAI Rg
REW
RecoReco
HyyHyyBoreal Boreal PinusPinus
ThaThaTemperateTemperate
PiceaPicea
HaiHaiTemperateTemperate
FagusFagus
PuePueMedit.Medit.Q. ilexQ. ilex
Temperature vs Soil Water control
Low influence of Biomass
RglobTair
VPDVPD
REW
AccPLAIBwood
CsoilCsoilClim
ate
Bio
log
ical
Bio
log
ical
Flux-IAV controls in European forests (2000-2007)
41
GPPGPP
REW
REW
AccP
LAI Rg
REW REW
REW
Temp
Temp
RecoReco
HyyHyyBoreal Boreal PinusPinus
ThaThaTemperateTemperate
PiceaPicea
HaiHaiTemperateTemperate
FagusFagus
PuePueMedit.Medit.Q. ilexQ. ilex
RglobTair
VPDVPD
REW
AccPLAIBwood
CsoilCsoilClim
ate
Bio
log
ical
Bio
log
ical
Flux-IAV controls in European forests (2000-2007)
41
GPPGPP
REW
REW
AccP
LAI Rg
REW REW
REW
Temp
Temp
NEPNEP
Rg
AccP Temp
REW
REWTemp
NEP control unpredictable from
elementary flux responses
Compensation effects
noticeable
CONCLUSION
Process-based modelsallow to address the determinism of C fluxes
from detailed processes to ecosystem scale
from hourly to decadal time scale
42
CONCLUSION
42
Increased contribution of Biological Drivers
at higher timescales
Constraint vs. Modulation
Unpredictability of NEP controls from GPP/Reco controls
Process-based modelsallow to address the determinism of C fluxes
CONCLUSION
42
Limits of the approach
Poor quality of Eddy Covariance nighttime fluxes Are our models reliable ?
(van Gorsel et al., 2009)
Model challenged at some sites (Hesse, Soroe)Are we missing something ?
Process-based modelsallow to address the determinism of C fluxes
PERSPECTIVES
43
Deconvolution methodology on longer time series
for an increased number of sites
Model developmentsSupra-decadal simulations
age effects (C allocation) acclimation (e.g. respiration, phenology)
Carbon-Water-Nitrogen coupling
Genericity
top related