1) The study assimilated SMOS soil moisture observations into a terrestrial biosphere model to quantify the added value of soil moisture data for constraining terrestrial carbon fluxes compared to assimilating only atmospheric CO2 concentrations. 2) Assimilating both soil moisture and CO2 observations led to greater reductions in uncertainties of carbon fluxes and process parameters than assimilating CO2 observations alone. 3) The results demonstrate the potential of jointly assimilating remotely sensed soil moisture with atmospheric CO2 observations to improve estimates and uncertainties of global carbon fluxes from land to the atmosphere.

1. Constraining terrestrial carbon fluxes by
assimilating the SMOS soil moisture product
into a model of the global terrestrial
biosphere
ICOS Science Conference
27-29 September 2016, Helsinki
M. Scholze1, T. Kaminski2, S. Blessing3, W. Knorr1, M. Vossbeck2,
J. Grant1* & K. Scipal4
1 Lund University, 2 Inversion Lab (previously at FastOpt),
3 FastOpt, 4 ESA, * now at Netherlands Space Office

2. Global Carbon Budget
1.0±0.5 PgC/yr 10%
2.6±0.8 PgC/yr
Land
28%
4.3±0.1 PgC/yr
Atmosphere
46%
2.5±0.5 PgC/yr
Oceans
26%
8.3±0.4 PgC/yr 90%
+

3. The case for data assimilation
 Carbon Cycle Data Assimilation System
= ecophysiological constraints from forward modelling
+ observational constraints from inverse modelling
Large uncertainty from land
to predict C-balance (GCP)
Cfluxtoland(PgC/yr)
Available Observations
Le Quéré et al. 2013

4. Previous results
Kaminski et al. (2012)
 Assimilation of FAPAR and atmospheric CO2 constrains water
fluxes
 Water and Carbon Cycles tightly coupled
Plant water
stress
LowHigh

5. Objective of this study
Assimilation of SMOS soil moisture observation together with
atmospheric CO2 concentration:
 To quantify the added value of remotely sensed soil moisture
observations (as provided by SMOS) on constraining terrestrial
C fluxes.
 To assess the potential of a SMOS-based NEE product.

6. C-cycle data assimilation system
Optimizer
J(X) and dJ(X)/X
SMOS
soil moisture
PFT composition
ecosystem parameters
initial conditions
parameters
(X) 
J(X)
M(X)
ysoil moist
Terrestrial
ecosystem
model
J(X)J(X)
Climate
J(x) 
1
2
y  M(x) 
t
Cy
1
y  M(x)  (x  xp )t
Cp
1
(x  xp )


Cost function:
 Need to define the error matrices Cy
-1, Cp
-1
In‐situ
CO2yCO2
Adjoint model
dJ(X)/X

7. CCDAS methodology
 Based on process-based terrestrial ecosystem model (BETHY)
 Optimizing parameter values (~100) based on gradient method
 Hessian (2nd deriv.) to estimate posterior parameter uncertainty
 Error propagation by using linearised model
Scholze et al. (2007)

8. Site-scale experiments
 Substantial model development to simulate surface soil moisture
 Joint assimilation of SMOS daily SM data for 5 sites
 5 member ensembles from different starting points
 All 5 converge to same minimum

9. Global Experiments
 Coarse resolution, 2 years (2010/11)
 Running 3-member ensembles from different starting points
 Baseline: in-situ atm. CO2 (10 sites) concentrations only
 Baseline + SMOS daily soil moisture with variance/mean
scaling

10. Results: process-parameters
CO2 & SMOSCO2 only
Scholze et al. (2016)
Photosynthesis Eco. Respiration
& C balance
Phenology Soil Hydrology Photosynthesis Eco. Respiration
& C balance
Phenology Soil Hydrology

11. Results: atm CO2 (also for validation)
CO2 only
CO2 & SMOS
MLO ALT

12. Results: soil moisture (RMS)
CO2 & SMOSCO2 only
1 2 3 4 5 6 1 2 3 4 5 6

13. Results: CO2 fluxes (NEP)
CO2 & SMOSCO2 only
-400 -300 -200 -100 0 100 200 300 400
g C / m2
-400 -300 -200 -100 0 100 200 300 400
g C / m2

14. Results: CO2 fluxes (NPP)
CO2 & SMOSCO2 only
-3000 -2000 -1000 0 1000 2000 3000 -3000 -2000 -1000 0 1000 2000 3000

15. Validation: soil moisture at site level
CO2 & SMOSCO2 only

16. Relative flux (NEP & NPP) uncertainty
reduction for 6 regions
Red: CO2 only Blue: CO2 & SMOS
Percentage
NEP NPP

17. Conclusions
• CCDAS combines process understanding with observations,
provides an integrated view on global carbon cycle and delivers
elaborated products based on ICOS data (among others)
• Site scale experiments to test and validate assimilation system
=> demonstrates capabilities of CCDAS to assimilate SM
• First global experiments assimilating remotely sensed SM and
atm. CO2 simultaneously
• Significant added value (unc. reduction) compared to CO2 only
• Further work using SMOS observations together with atm. CO2:
– assimilate the full 6 years of SM data at high resolution
– extend the assimilation system to additionally include
microwave Vegetation Optical Depth data as a proxy for
above ground biomass (vegetation water content) over
various land-cover types