Carbon, soil moisture and
fAPAR assimilation
Wolfgang Knorr
Max-Planck Institute of Biogeochemistry
Jena, Germany 1
Acknowledgments: Nadine Gobron 2, Marko Scholze
3, Peter Rayner 4, Thomas Kaminski 5,
Ralf Giering 5, Heinrich Widmann1
2
IES/JRC
3
QUEST /
4
LSCE
5 FastOpt
Overview
• CO2 – climate linkages
• Satellite fAPAR as soil moisture indicator
• Assimilation of fAPAR
Atmospheric CO2
Measurements
CCDAS inverse modelling period
... and more stations in CCDAS
Carbon Cycle Data Assimilation
System (CCDAS)
Assimilated
Prescribed
Assimilated
satellite fAPAR +
Uncert.
Phenology
Hydrology
CO2
+ Uncert.
CCDAS Step 1
full BETHY
Background
CO2 fluxes*
CCDAS Step 2
BETHY+TM2
only Photosynthesis,
Energy&Carbon Balance
Optimized Params
+ Uncert.
Diagnostics
+ Uncert.
*
* ocean:
Takahashi et al. (1999), LeQuere et al. (2000); emissions: Marland et al. (2001), Andres et al. (1996); land use: Houghton et al. (1990)
Terr. biosphere–atmosphere
CO2 fluxes
ENSO
… preliminary results from extended CCDAS run
ENSO and global climate
normalized anomalies
ENSO
–precipitation
temperature
… global drying and warming trend
for more information see:
http://www.CCDAS.org
Overview
• CO2 – climate linkages
• Satellite fAPAR as soil moisture indicator
• Assimilation of fAPAR
Remotely Sensed
Vegetation Activity
ITOC
I
TOC
fAPAR:
[(ITOC+IS)–(ITOC+IS)]
/ ITOC
canopy
I S
I S
soil
SeaWiFS fAPAR archive
 developed by Nadine Gobron, Bernard Pinty,
Frédéric Melin, IES/JRC, Ispra
 3-channel algorithm taylored to SeaWiFS ocean color
instrument (blue, red, near-infrared)
 cloud screening algorithm
 requires no atmospheric correction
 starts 10/1997, continuing...
 being extended by same product for MERIS
Precipitation – fAPAR
precipitation
soil moisture
gridded station data
BETHY simulations
leaf area index
fAPAR
satellite observations
BETHY simulations
precipitation vs.
fAPAR from
SeaWiFS satellite
obs.
1-month lag
percent area with 99%
significant correlation
r>0
r<0
4-month lag
0.5°x0.5°, ≥50% cloud free, ≥75% temporal coverage
precipitation vs.
fAPAR:
satellite and
model
1-month lag
percent area with 99%
significant correlation
SeaWiFS fAPAR
r>0
r<0
BETHY simulated fAPAR
precipitation vs.
fAPAR:
satellite and
model
4-month lag
percent area with 99%
significant correlation
SeaWiFS fAPAR
r>0
r<0
BETHY simulated fAPAR
precipitation vs.
satellite fAPAR
and simulated soil
moisture
1-month lag
percent area with 99%
significant correlation
SeaWiFS fAPAR
r>0
r<0
BETHY simulated soil moisture
precipitation vs.
satellite fAPAR
and simulated soil
moisture
4-month lag
percent area with 99%
significant correlation
SeaWiFS fAPAR
r>0
r<0
BETHY simulated soil moisture
ENSO – SeaWiFS fAPAR
lagged correlation
3-month lag
Overview
• CO2 – climate linkages
• Satellite fAPAR as soil moisture indicator
• Assimilation of fAPAR
fAPAR Assimilation
Prescribed
PFT distribution*
climate &
soils data**
ecosystem model
parameters
optimization
BETHY
carbon and water
fluxes
model-derived
fAPAR
mismatch
satellite
fAPAR
The Cost Function
Measure of the mismatch (cost function):
model diagnostics
measurements
1
1
-1
T
T
J(m)  [m  m0 ]Cm 0 [m  m0 ]  [y (m)  y0 ]C-1
[y
(m
)

y
]
y
0
2
2
assumed
model parameters
a priori
parameter values
a priori error covariance
matrix of parameters
aim: minimize J(m)
[for each grid point separately]
error covariance matrix
of measurements
The Parameters
parameter vector m={m1,m2,m3}:
represents:
vector of prior
parameter
values m0:
m1
Tf
shift of leaf
onset/shedding
temperature
temperature
limitation
Tf=0
m2
wmax
maximum soil water
holding capacity
water limitation
wmax,0
fraction of grid cell
covered with
vegetation
residual,
unmodelled
limitations (nitrogen,
land use)
m3
fc
(derived from
FAO soil map)
fc,0
(function of
P/PET and
Temp. of
warmest
month)
Prior Parameter 1
prior values:
Tf=5°C
Tf=12°C for crops
T^f=15°C
note: each 0.5°x0.5° has
mixture of up to 6 PFTs
map reflects presence of crops; red: unvegetated
Prior Parameter 2
bucket model:
precipitation
=input
runoff
=overflow
full
bucket:
wmax
current
bucket:
w
wmax,0 [mm]
Prior Parameter 3
^
fc,0=Pannual/PETannual*W(Twarmest month)/
Prior Parameter Errors
error covariance matrix of parameters Cm0:
Cm 0
1K 2

  0

 0
0
(2w max, 0 ) 2
0


0 
2 
0.25 
0
 off-diagonal elements assumed 0 here
= no prior correlation between errors of different parameters
The Assimilated Data
model diagnostics vector y={y1,y2,...,y12}:
yi
modelled fAPAR of month i
satellite-derived diagnostics vector y0={y0,1,y0,2,...,y0,12}:
y0,i
SeaWiFS derived fAPAR of month i
Prior Errors of Measurements
error covariance matrix of measurements Cy:
Cy i, j
0.052
 
i=j  
if valid measurement
if data gap
2
  y,i
 off-diagonal elements again 0
= no prior correlation between errors of different months
Parameter 2 (regional)
soil water-holding capacity
prior
optimized
local site
Local Simulations
precipitation [mm/month]
fAPAR
Paragominas
3°S 48°W 63 m
remote sensing data
1992
evapotranspiration [mm/month]
NPP [gC/(m2 month)]
-
no remote sens. data
fAPAR prescribed
fAPAR assimilated
Measured Soil Moisture
precipitation [mm/month]
fAPAR
Paragominas
3°S 48°W 63 m
remote sensing data
1992
0...2m depth
0...8m depth
-
1992
1992
no remote sens. data
fAPAR prescribed
fAPAR assimilated
evapotranspiration (regional)
prior
mm/year
optimized
mm/year
July soil moisture (regional,
dry season)
prior
optimized
Parag.
mm
mm
Conclusions
• The carbon cycle is highly sensitive to climate fluctuations
• Vegetation can be quantified reliably from space
• fAPAR lags precipitation by ~1–4(?) months
• seems to behave similar to soil moisture
• assimilation of fAPAR can deliver valuable information on
soil moisture status
Conclusions
• Need to improve phenology model
• Implement sequential 2-D var assimilation scheme
• Assimilate fAPAR into coupled ECHAM5-BETHY model
(hope not too distant) goal: make fAPAR what SST
is for ocean-atmosphere interactions... and
improve seasonal forecasts
Thank You For Your Attention!