Abstract: Estimating global terrestrial Gross Primary Productivity (GPP) through process-based models is challenging due to complex parameterization,while data-driven models and light-use efficiency models rely heavily on extensive datasets.The significant spatial heterogeneity of terrestrial ecosystems also limits the application of these three major GPP models in regions lacking ground-based observations.This study developed empirical GPP-SIF (Solar-Induced Chlorophyll Fluorescence) relationships specific to various vegetation types using continuous SIF observation data and land cover information.The stability of these relationships within identical vegetation types and their variation across different types were validated.Based on the established GPP-SIF relationships and the latest RTSIF data,we generated an 8-day,0.05-degree spatial resolution global GPP_SIF product for terrestrial ecosystems.Cross-validation with other GPP estimation models and site-measured GPP demonstrated that the proposed global GPP estimation method is computationally efficient with reliable results and exceptional performance in data-scarce regions.