Transcript Assimilating satellite observations of land surface properties into

Making C flux calculations interact with
satellite observations of land surface
properties
Shaun Quegan and friends
Global Carbon Data Assimilation System
Geo-referenced
Geo-referenced
emissions
emissionsinventories
inventories
Climate and weather
fields
Ocean time series
Biogeochemical
pCO2
Surface
observation
pCO2
nutrients
Water column
inventories
Atmospheric
Atmospheric
measurements
measurements
Remote
Remote sensing
sensing of
of
atmospheric
atmospheric CO
CO22
Atmospheric
Atmospheric Transport
Transport
Model
Model
Ocean
Ocean Carbon
Carbon
Model
Model
Coastal
Coastal
studies
studies
Optimised
Optimised
fluxes
fluxes
Terrestrial
Terrestrial
Carbon
Carbon Model
Model
rivers
Lateral fluxes
Data
assimilation
link
Optimised
Optimised
model
model
parameters
parameters
Eddy-covariance
flux towers
Biomass soil
carbon
inventories
Ecological
studies
Ocean remote sensing
Ocean colour
Altimetry
Winds
SST
SSS
Remote sensing of
vegetation properties
Growth cycle
Fires
Biomass
Radiation
Land cover/use
Ciais et al. 2003 IGOS-P Integrated Global Carbon Observing Strategy
Terrestrial Component
Remote
Remote sensing
sensing of
of
atmospheric
CO
atmospheric CO22
Climate and weather
fields
Atmospheric
Atmospheric Transport
Transport
Model
Model
Optimised
Optimised
fluxes
fluxes
Terrestrial
Terrestrial
Carbon
Carbon Model
Model
rivers
Lateral fluxes
Optimised
Optimised
model
model
parameters
parameters
Eddy-covariance
flux towers
Biomass soil
carbon
inventories
Ecological
studies
+ Water components:
SWE
soil moisture
Remote sensing of
vegetation properties
Growth cycle
Fires
Biomass
Radiation
Land cover/use
The SDGVM carbon cycle
ATMOSPHERIC
CO2
Photosynthesis
Fire
GPP
GROWTH
BIOPHYSICS
Mortality
NPP
Thinning
Litter
Soil
LEACHED
NBP
Biomass
Disturbance
The Structure of a Dynamic Vegetation Model
Parameters
Climate
Sn
Soil
texture
DVM
Processes
Sn+1
Testing
EO interactions with the DVM
Phenology
Snow water
Burnt area
Parameters
Processes
Climate
Land
cover
Forest age
Sn
Soils
DVM
Sn+1
Observable
Possible feedback
Testing:
Radiance
fAPAR
Matching of concepts
Real world
S
Primary observation
Model
Model
Derived parameter
MODIS LAI/fAPAR
biome Landcover
2000
MODIS/IGBP
Landcover 2000
MODIS/UMD
Landcover 2000
CEH LCM2000
GLC2000 (SPOT-VGT)
Scale effects on flux estimates (GLC-LCM)
GPP
+1.0%
NPP
NEP
+6.4%
+16.1%
Difference in annual predicted fluxes
for GB, 1999. GLC – LCM.
Lessons 1
1. Land cover matters.
2. ‘Subjective’ land cover may be more useful than
‘objective’ land cover.
3. Scale matters.
4. Can we do this better?
The SDGVM budburst algorithm
When
 min(0, T – T0) > Threshold, budburst occurs.
days
The sum is the red area. Optimise over the 2 parameters, Threshold
and T0 (minimum effective temperature).
T0
Start of budburst
Data
 SPOT-VEG budburst 1998, 2000-02: 0.1o
 Ground data; Komarov RAS, dates of bud-burst
at 9 sites in the region.
 Temperature data: ERA-40, 1.125o
 GTOPO-30 DEM
 Land cover: GLC2000
The Date of budburst derived from
minimum NDWI (VGT sensor, 2000) N. Delbart, CESBIO
Day of year
Variability in optimising coefficients
Application of model to entire boreal regions
Model 1985
EO 1985
Model 2002
EO 2002
Comparison of ground data with calibrated model
Impact on Carbon Calculations
1 day advance: NPP increases by 10.1 gCm-2yr-1
15 days advance: 38% bias in annual NPP
Observations
Carbon Calculation
Phenology model
Picard et al.,GCB, 2005
Dynamic Vegetation
Model
Comparison Model-EO: RMSE
Model needs to be region specific,
here include chilling requirement ?
Lessons 2
1. A simple 2-parameter spring warming model gives a
good fit between model and EO data
2. RMS differences between model, VGT data and
ground data are ~6.5 days.
3. Ground data are crucial in investigating bias.
4. Model failures are identifiable.
5. Noise errors in NPP estimates are ~8%. Bias effects
are ~2.2% per day.
6. Biophysical content of the parameters is low.
SDGVM module driven by climate data
Cloud cover
Precipitation
Evaporation
Humidity
Atmosphere
Temperature
Snowpack
Ground
Snow melt
Snow water
equivalent (SWE)
SWE estimated from SSM/I data over Siberia
CTCD: Comparison model and EO (& IIASA snow map)
Snow Water Equivalent (mm) 01/97
SDGVM using ECMWF
SSM/I
IIASA maximum snow storage
Lessons 3
1. The physical quantity inferred from the EO data is
almost certainly not what it is called.
2. The problem here is making the model and the EO
data communicate. Until communication is established,
the data cannot be used to test or calibrate the model.
Severity of disagreement – AVHRR/SDGVM
1998
r > 0.497 OR r.m.s.e < 0.2
r < 0.497 AND r.m.s.e > 0.2
r < 0.497 AND r.m.s.e > 0.3
Severity of disagreement – example
Mid Europe
Severity of disagreement – example
SW China
Lessons 5
1. The DVM as currently formulated only supports a
simple observation operator. This allows meaningful
estimates of time series of observables; absolute
values of the observables are of dubious value.
2. These time series permit the model to be interrogated
with satellite data, and model failures to be identified.
Detecting incorrect land cover
Crop class incorrectly set
Crop class correctly set
0.9
0.0
Pearson’s product moment
Temporal correlation
Final remarks
 The link between satellite measurements and
most surface parameters used by the C models
(and how they are represented) is indirect.
 In many cases, the only viable source of
information on surface properties is from
satellites.
 The art is to find the right means of
communication between the data and the
models.
Environmental effects on coherence
Coherence of Kielder Forest, July 1995
Measurements by radar
satellites are sensitive to
biomass, but:
• only for younger ages
• weather dependent through
soil and canopy moisture
Age Estimation Accuracy
 Small Spatial Scale
– Inter-stand variance
– Inter stand bias
Kielder Forest
Kielder Forest
North
South
 Large Scale
– Meteorology dominant
Raw Coherence
Time
coherence
-8
NEE tc ha-1 y-1
4
0
-4
N(age)
8
Estimating NEE with SAR
0
5
10
15
20 25
Age (y)
30 35
0
40
10
20
30
40
age
Age (y)
50
60
Sensitivity range
NEE = X N(A(x))
dx
X
70
Using SPA to model coherence
• Observations
+ Model with biomass
saturation information
Model Backscatter
SPA was used to predict canopy and soil moisture, and coupled with
a radar scattering model to predict coherence. Also needed was the
saturation level of biomass, which had to be measured from the data
Lessons 3
Here the carbon model is essential to interpret the data
and its variation.
UK Forest NEE Calculations 1995
Methods
FC GIS
NEE Total (MtC y-1)
[NEE per ha (tC ha y-1)]
-9.37
Area
(k ha)
[-3.2]
2,928
-10.87 [-3.7]
2,928
(extrap. private forest)
SAR Estimate
(measured private forest)
National Inventory
( land class only)
-2.8
[-1.75]
1,600
MODIS Burned Area
Russian Federation
500m burned areas
1 month 2002
MODIS Active Fires (& FRP)
Russian Federation
1km active fires
1 month 2002