Improving the E3SM Land Model Photosynthesis Using Satellite Solar-Induced Chlorophyll Fluorescence and Machine Learning Techniques
Anping Chen1, Daniel M. Ricciuto2, Jiafu Mao2* (firstname.lastname@example.org), Jiawei Wang3, Dan Lu2, Fandong Meng3, Paul J. Hanson2
1Colorado State University, Fort Collins, CO; 2Oak Ridge National Laboratory, Oak Ridge, TN; 3College of Urban and Environmental Sciences, Peking University, Beijing, China
The parameterization of key photosynthesis parameters is one of the key uncertain sources in modelling ecosystem gross primary productivity (GPP). Solar-induced chlorophyll fluorescence (SIF) offers a good proxy for GPP since it measures the actual process of photosynthesis; while machine learning (ML) provides a robust approach to model the GPP-SIF relationship. The research team trained the Boosted Regressing Tree (BRT) and the Random Forest (RF) ML models with Greenhouse Gases Observing Satellite SIF data and in situ GPP observations from 49 eddy covariance (EC) towers. These trained ML GPP-SIF models were fed into the E3SM Land Model (ELM) to generate ELM-simulated global SIF estimates, which were benchmarked against satellite SIF observations with a surrogate modelling approach. Results indicated good modeling performance of the ML-based GPP-SIF relationship. The ELM model when fed with the ML GPP-SIF models also can well predict the spatial-temporal variations in SIF. The research team also found high model accuracy for the surrogate modeling. Model parameter sensitivity analysis suggested that the fraction of leaf nitrogen in the photosynthetic enzyme ribulose-1,5-bisphosphate carboxylase-oxygenase is the most sensitive parameter to the SIF; other sensitive parameters include the Ball-Berry stomatal conductance slope (mbbopt) and the Vcmax entropy (vcmaxse). The posterior uncertainty in simulated GPP was greatly reduced after benchmarking, and the model produced improved spatial patterns of mean GPP relative to FluxCom GPP. This integrated approach provides a new avenue for improving land models and using remote-sensing SIF, which can be further improved in the future with more ground- and satellite-based observations.