Global-scale improvement of terrestrial gross primary productivity estimation by integrating optical remote sensing with meteorological data

• This study significantly improves global-scale gross primary productivity (GPP) estimation accuracy by integrating sunlight-induced chlorophyll fluorescence (SIF), meteorological data, and optical factors. • Sunlight-induced chlorophyll fluorescence (SIF), as the core variable, outperforms traditional vegetation indices and helps estimate gross primary productivity (GPP) more accurately. • The fusion of multi-source data enhances the model’s adaptability, especially in complex ecosystems. This study integrates sunlight-induced chlorophyll fluorescence (SIF), meteorological factors, and various optical factors (such as LAI, FAPAR, and NDVI) to estimate global-scale gross primary productivity (GPP) using a random forest model (RF). The results show that the random forest model can significantly improve the GPP estimation accuracy, and its overall coefficient of determination (R 2 ) is 0.84, the consistency index (IOA) is 0.97, and the root mean square error (RMSE) is only 1.73 g C m - 2 d -1 . As the core variable of the model, SIF showed significant contributions in all vegetation types, making up for the limitations of the traditional vegetation index in monitoring the dynamic changes of photosynthesis. The multi-source data fusion method effectively improves the adaptability and robustness of the model to the dynamic changes in GPP in complex ecosystems (such as wetlands and farmland). This study shows that the integration of SIF, meteorological factors and optical factors to construct a multi-source fusion model can not only improve the accuracy and spatiotemporal resolution ability of global-scale GPP estimation, but also provide scientific support for further research on the interaction between ecosystem carbon cycle and climate change.