Advancing Satellite Solar‐Induced Chlorophyll Fluorescence Data Assimilation in Land Surface Models: An Assessment Over Boreal Evergreen Needleleaf Forests

Abstract

Boreal forests represent about 20% of the global forest carbon sink and contribute to climate change mitigation, yet land surface models (LSMs) struggle to simulate their gross primary production (GPP). Space‐borne observations of solar‐induced fluorescence (SIF) have been used to improve LSM GPP through data assimilation. However, previous work has shown that assimilating SIF alone can degrade GPP for some vegetation types, highlighting the need for enhanced process representation and tightly constrained parameters. Uncertainties in the representation of non‐photochemical quenching (NPQ), which regulates the SIF‐GPP relationship, further complicate this challenge. Here, we develop a data assimilation strategy to assess the complementarity of satellite SIF from TROPOMI and in situ GPP in constraining GPP dynamics simulated by the ORCHIDEE LSM over boreal evergreen needleleaf forests. We evaluate this framework using two NPQ schemes, a process‐based formulation separating sustained and reversible components, and a generic model representing total NPQ. After assimilation, evaluation against independent SIF grid cells and flux sites shows improved seasonal cycles of both SIF and GPP. Regional‐scale simulations better match independent GPP products (FLUXSAT and FLUXCOM‐X‐BASE) over 2001–2020, with over 77% RMSD reduction in the mean seasonal cycle versus the prior, largely reflecting the recalibration of model parameters after implementation of the new radiative transfer scheme. SIF‐GPP co‐assimilation outperforms GPP‐only assimilation for the GPP latitudinal profile, and also for the GPP mean seasonal cycle in three out of four cases. Although the two NPQ models affect the simulated SIF seasonal cycle differently, their performances after assimilation are similar.