In the simulation studies of carbon cycling processes, accurate estimation of Gross Primary Productivity (GPP) of terrestrial ecosystems is pivotal. However, the simulation accuracy for Evergreen Broadleaf Forests (EBF) and Evergreen Shrubs (ES) requires enhancement. This study focuses on these two typical vegetation types in the southwestern China. We refined the phenology module of the Biome-BGC model and optimized its parameters to boost GPP simulation accuracy. Solar-induced fluorescence (SIF) data, which is sensitive to photosynthetic activity, was employed to replace the meteorologically-based phenology module in the original model. SIF data can more directly measure the photosynthetic activity of plants, which is closely related to GPP changes, and the calculated phenology information is more accurate than that calculated from traditional meteorological data. We then used the Embedding-Tree-Parzen-Estimator (ETPE) optimization algorithm, which is a Bayesian optimization algorithm, for parameter tuning. Calibration and validation of the model were conducted using Eddy covariance (EC) measurements from four towers. The optimized model exhibited improved accuracy in GPP simulations compared to the original. For EBF, simulation accuracy increased from R² = 0.34 and RMSE = 2.40 g/m²/day to R² = 0.42 and RMSE = 1.64 g/m²/day. For ES, accuracy improved from R² = 0.65 and RMSE = 1.20 g/m²/day to R² = 0.77 and RMSE = 0.66 g/m²/day. In conclusion, by integrating the SIF-based phenology module with the Biome-BGC model and optimizing parameters, the model exhibited better adaptability to the complex vegetation environment in the southwestern China, thereby effectively simulating GPP data. This study provides a robust method and basis for accurately simulating GPP values of the main vegetation types in the southwestern region.
 

Biome-BGC,GPP,SIF,Phenology,Southwestern China