Wu, Mousong: Optimizing ecosystem carbon and water fluxes using COS and SIF within the adjoint-based data assimilation system (NUCAS)
1. Optimizing ecosystem carbon and water fluxes using COS and SIF within
the adjoint-based data assimilation system (NUCAS)
MOUSONG WU1*, THOMAS KAMINSKI2, MICHAEL VOSSBECK2, FEI JIANG1, WEIMIN JU1, YONGGUANG
ZHANG1, MARKO SCHOLZE3, JING M. CHEN4, KUKKA-MAARIA KOHONEN5, TIMO VESALA5, HUAJIE ZHU1,
HUILIN CHEN1
1NANJING UNIVERSITY, NANJING, CHINA
2THE INVERSION LAB, HAMBURG, GERMANY
3LUND UNIVERSITY, LUND, SWEDEN
4UNIVERSITY OF TORONTO, TORONTO, CANADA
5UNIVERSITY OF HELSINKI, HELSINKI, FINLAND
2. Water and carbon fluxes in leaf
Energy control
Stomatal control
The fluxes of H2O, CO2, COS in the leaf, adopted from Berry et al. (2013)
Carbonyl sulfide
Outside-of-leaf-cells
Acquisition of ‘raw materials’
Inside-of-leaf-cells
Transformation of ‘raw materials’
[H2O]
3. COS: reflects the opening of stomata (“outside-of-leaf-cells” stomatal conductance), Transpiration
SIF: reflects the photochemistry (“inside-of-leaf-cells” photochemistry), Photosynthesis
COS+SIF: constraints on the whole processes in leaf cell, Constraining water and carbon fluxes
Wohlfahrt et al., 2011
COS, SIF and the water and carbon fluxes
fluorescence
The pathways of CO2, H2O and COS into/out of a leaf
• COS diffuses into the leaf and is hydrolyzed in the
cell, without back-diffusion
• CO2 diffuses into the leaf and fixed in the cell, with
fluorescence emitted
• H2O diffuses out of the leaf via stomata
4. COS emissions:ocean, anthropogenic
activities, biomass burning, low-oxygen soil
and wetland
COS uptake:plant, soil, atmospheric
chemistry
COS budget
46%
66%
COS in ecosystems
Ecosystem is an important sink for COS!
5. COS observations
COS concentrations:1994-,14 stations
COS fluxes:2011-,7 sites
(DBF,ENF,SAV,GRA,CRO)
From 2022, two sites in China will be
observing COS in paddy rice and maize
fields, owned by Nanjing University
Global COS concentrations (blue circle) and fluxes (orange triangle)
observations sites
6. COS fluxes simulated by scaling the GPP (COS = F*GPP)
DK-Sor COS and CO2 fluxes simulated by CoupModel
R2=0.21
R2=0.66
R2=0.03
Validity of COS for stomatal conductance and
GPP
A diagnostic test was conducted using the process-based CoupModel, which has been well calibrated at the COS site
for CO2 and water fluxes. We modelled the COS fluxes (plant uptake) by scaling either GPP or transpiration.
7. COS fluxes simulated by scaling the transpiration
(COS= F*Transpiration)
DK-Sor COS and CO2 fluxes simulated by CoupModel
R2=0.36
R2=0.66
R2=0.03
Validity of COS for stomatal conductance and
GPP
Transpiration
scaling model
GPP scaling
model
COS is more linked to the
stomatal processes than the
photochemical processes at
diurnal scales
8. COS uptake by plants
1. Conductance model—SiB4/ORCHIDEE/BEPS(with two-leaf scheme for COS)
assimCOS=cosCAS×[1.56/gb+1.94/gs+1.0/gCOS]-1
• gCOS = 1400×(1.0+5.33×C4) ×Vcmax×fs, with fs as scaling factor, C4=1 for C4, 0 for C3 plant
2. LRU (leaf relative uptake) model—CASA,SiBCASA, SiB3
assimCOS=GPP× [COS]a/[CO2]a×LRU
• LRU=1.68 for C4 and 1.21 for C3
COS fluxes in soils
1. Scaling of heterotrophic respiration—SiB3/BEPS
grndCOS=fs×Rh
• fs is scaling factors considering water, temperature, freezing, Rh is
heterotrophic respiration
2. COS diffusion model: COSSM—SiB4/ORCHIDEE
COSSM, Sun et al., 2015
COS simulation approaches
9. BEPS
ecosystem (SM,
SIF, COS,
carbon and
energy)
Meteo.
PFT and
Soil Texture
Parameters
Optimized water,
carbon fluxes
Adjoint-
based
optimization
RS SM/SIF
EC COS flux
RS
LAI
Nanjing University Carbon Assimilation System
(NUCAS)
Cost function:
• NUCAS is a newly developed data assimilation
system for assimilating multiple sources of
observations
• It uses the two-leaf BEPS model as the core
model to simulate SIF, COS
• The optimization algorithm is adjoint-based,
i.e. variational data assimilation
• The optimization goal is to minimize the cost
function
10. NUCAS preliminary results
1. Twin-experiment: using model forward simulation output as pseudo observations, perturb parameters and
then run the assimilation
2. Single variable assimilation: data assimilation using SIF from GOME-2, SM from SMOS, or COS fluxes from
EC separately
3. Multiple variables assimilation: data assimilation using SIF+SM, SIF+COS, SM+COS, or SM+SIF+COS,
respectively
1. Trace of gradient, and cost function
2. Performance of assimilation results
3. Changes in parameters among different assimilation experiments
Assimilation experiments
Evaluation of assimilation experiments
Assimilation window of 15 days, tested at site FI-Hyy, a coniferous forest in Finland
11. Twin experiment—robustness of NUCAS
Both the gradient and cost function reaches very small values, the NUCAS is robust!
13. COS/GPP fluxes under different assimilation
experiments
• The assimilation of SM or COS can improve COS fluxes simulations
• The assimilation of COS or SIF can improve GPP simulations
• Night-time COS fluxes are captured by the COS model, but day-time data shows larger variations
14. LH/SH fluxes under different assimilation
experiments
• The inclusion of COS or SIF can capture LH and SH variations
• LH is more complicated as it includes transpiration, soil/snow evaporation and
snow/water interception evaporation
15. Parameters changes under different assimilation
experiments
The assimilation of three variables can constrain plant carbon related parameters
16. Parameters changes under different assimilation
experiments
The assimilation of soil moisture can constrain soil hydraulic related parameters
17. Parameters changes under different assimilation
experiments
The assimilation of soil moisture and SIF can constrain SIF simulation related parameters
18. Take home messages
• We developed the two-leaf BEPS model considering COS, SIF and soil moisture
simulations
• We optimized model parameters using the NUCAS with the adjoint-based
algorithm
• We for the first time tested the usefulness of simultaneously assimilating COS,
SIF, and SM in constraining ecosystem water and carbon fluxes
• The assimilation window, the impacts of the model and observation uncertainties
on assimilation need further tests
• Extend the NUCAS to regional assimilation using multiple satellite observations