Photosynthetic efficiency (PE) quantifies the fraction of absorbed light used in photochemistry and is essential for understanding ecosystem productivity and the global carbon cycle1-4, particularly under conditions of vegetation stress5. However, nearly 60% of the global spatiotemporal variance in terrestrial PE remains unexplained3,5-7. Here, we integrate remote sensing and eco-evolutionary optimality theory to derive key plant traits, alongside explainable machine learning and global eddy covariance observations, to uncover the drivers of PE variations. Incorporating plant traits into our model increases the explained daily PE variance from 36% to 80% for C3 vegetation and from 54% to 84% for C4 vegetation compared to using climate data alone. Key plant traits—including chlorophyll content, leaf longevity, and leaf mass per area—consistently emerge as dominant factors across global biomes and temporal scales. Water availability and light conditions are also critical in regulating PE, underscoring the need for an integrative approach that combines plant traits with climatic factors. Overall, our findings demonstrate the potential of remote sensing and eco-evolutionary optimality theory to capture principal PE drivers, offering valuable tools for more accurately predict ecosystem productivity and improving Earth system models under climate change.