Improved Estimation of Gross Primary Productivity in Global Croplands Using a Transpiration-Based uWUE Model

Accurate estimation of gross primary productivity (GPP) in croplands is essential for quantifying terrestrial carbon uptake and understanding carbon–water coupling under increasing agricultural water stress. Conventional Light Use Efficiency (LUE) models typically rely on evaporative fraction (EF), derived from total evapotranspiration (ET), which does not distinguish between productive transpiration and non-productive evaporation. In contrast, transpiration-based framework explicitly represent the physiological coupling between carbon assimilation and water loss regulated by stomatal conductance. In this study, transpiration is estimated using a Leaf Area Index (LAI)-based approach driven by remotely sensed MODIS data and environmental variables within the underlying Water Use Efficiency (uWUE) framework.

We evaluate the transpiration-based uWUE model against an EF-based LUE model for GPP estimation using eddy covariance observations from 51 globally distributed cropland sites. The dataset includes 6 sites from India (Flux Tower and INCOMPASS networks), 3 sites from Japan (AsiaFlux), 9 sites from Europe, and 33 sites from the United States (FLUXNET), spanning a wide range of hydro-climatic and management conditions. Model performance was assessed using the coefficient of determination (R²), root mean square error (RMSE), and bias.

The transpiration-based uWUE model showed overall better agreement with observed GPP than the EF-based LUE model across the global set of crop sites. Improvements were evident in both the strength of the relationship with observations and the reduction of estimation errors. At the site level, uWUE more frequently achieved higher R² together with lower RMSE, demonstrating consistent performance across multiple evaluation metrics at a larger number of sites. Superior performance was observed at 28 sites, driven by the model’s ability to capture coupled carbon–water dynamics under varying crop types, canopy structures, and climatic conditions. In contrast, the EF-based LUE model showed advantages at a limited number of sites characterized by distinct water stress regimes or vegetation properties.

Overall, the results highlight the critical role of transpiration dynamics in GPP estimation, with higher GPP values associated with dense canopies and favorable environmental conditions. By explicitly isolating transpiration from total evapotranspiration, the uWUE framework provides a more physically meaningful representation of carbon–water interactions than ET-based approaches. These findings demonstrate that incorporating transpiration-based constraints improves GPP estimation in croplands and has important implications for large-scale agricultural carbon cycle assessments under future climate scenarios characterized by increased water stress and drought frequency.