SIF_v1A
Optimization of SIF model v1.1 vs TROPOSIF data with 8 optimized parameters (set A)

Optimization configuration

• L-BFGS-B algorithm limited to 100 iterations
• FLUXNET2015 sites (117 in total) grouped in 12 PFTs simulated using in-situ atmospheric data for different periods across 1992-2015
• driven by the fit to SIF data (mean seasonal cycle at daily frequency calculated across all available years for CTESSEL and 2019-2020 for TROPOSIF)
• 8 optimized parameters (set A), including:
• 3 parameters from Farquhar module (ASJ, ASV, GB_REF)
• 5 parameters from Fluor module (ALA, SLA, K_F, A_PS, K_P)
• parameter ranges of variation are set individually for each parameter (see table below)
• prior parameter uncertainty is set to 40% of its range of variation
• prior model uncertainty is set to the root-mean-square difference between the prior simulations and TROPOSIF data (mean seasonal cycle)

Select PFT: PFT1 (Crops, mixed farming, C3) PFT2 (Short grass, C3) PFT3 (Evergreen needleleaf, C3) PFT5 (Deciduous broadleaf, C3) PFT6 (Evergreen broadleaf, C3) PFT7 (Tall grass, C4) PFT9 (Tundra, C3) PFT11 (Semidesert, C4) PFT13 (Bogs and marshes, C3) PFT16 (Evergreen shrubs, C4) PFT17 (Deciduous shrubs, C3) PFT18 (Mixed forest - Wood, C3)

Parameter	RALA_1	RSLA_1	RASJ_2	RASV_2	RK_F	RA_PS	RK_P	RGB_REF
prior	57.4	0.026	630	641.64	0.05	0.5	4	0.04
min	45	0.013	315	321	0.01	0.01	3	0
max	65	0.05	945	962	0.15	0.99	8	0.08
post	54.7	0.0278	630	642	0.05	0.0804	5.19	0.0397

Optimization status after 80 iterations : CONVERGENCE: REL_REDUCTION_OF_F_<=_FACTR*EPSMCH

SITE LOCATION	SIF	GPP	Simulations	Data	Optimization vs TROPOMI-SIF data (mean seasonal cycle)