Abstract: Accurately representing photosynthetic optimum temperature (T_opt) is essential for predicting terrestrial carbon uptake, yet conventional ecosystem-scale estimates based on peak gross primary productivity (GPP) are confounded by concurrent variations in radiation, moisture and phenology. Here, we develop a physiologically grounded approach that defines T_opt as the air temperature where light use efficiency (LUE) reaches its maximum, thereby isolating the intrinsic thermal response of photosynthesis. We derived efficiency-based T_opt (T_opt-LUE) from 131 flux observation sites. By replacing biome-based T_opt used in Vegetation Photosynthesis Model (VPM), we significantly improved GPP estimation (R² = 0.71, RMSE = 1.93 gC m^-2 d^-1) compared to the biome-based approach (R² = 0.62, RMSE = 2.84 gC m^-2 d^-1). We then used a Random Forest framework to generate global and time-varying T_opt-LUE fields (2001–2020). The results revealed that biome-based T_opt systematically overestimated T_opt-LUE across ~94% of global vegetated areas, with a mean bias of ~10°C. The global T_opt-LUE exhibit clear latitudinal gradients and biome‐specific contrasts, providing evidence for widespread thermal acclimation of ecosystem photosynthesis. This acclimation is reflected in a mean increase in T_opt-LUE of 0.021 ± 0.102 °C per year, underscoring a measurable response to long-term climate changes. When integrated into VPM, the dynamic T_opt-LUE fields reshape the spatial pattern of simulated carbon uptake, mitigating overestimation in tropical areas (~5 gC m^-2 d^-1) and enhancing underestimation in frigid areas and temperate regions including China, India and Europe. This study established a mechanistically grounded framework for quantifying ecosystem thermal acclimation, advancing the representation of temperature responses in terrestrial carbon cycle models.

Key words: Photosynthetic optimum temperature, Light use efficiency, Gross primary productivity, Thermal acclimation, Vegetation Photosynthesis Model