Incorporating environmental stress improves estimation of photosynthesis from NIRvP in US Great Plains pasturelands and Midwest croplands

Near-infrared reflectance of vegetation multiplied by incoming sunlight (NIRvP) is important for gross primary production (GPP) estimation. While NIRvP is a useful indicator of canopy structure and solar radiation, its association with heat or moisture stress is not fully understood. Thus, this research aimed to explore the impact of air temperature (Ta) and vapor pressure deficit (VPD) on the NIRvP-GPP relationship. Using Moderate Resolution Imaging Spectroradiometer (MODIS) observations, eddy-covariance measurements, and the Parameter–Elevation Regressions on Independent Slopes Model (PRISM) data, we found that NIRvP cannot fully explain the response of plant photosynthesis to Ta and VPD at both seasonal and daily scales. Therefore, we incorporated a polynomial function of Ta and an exponential function of VPD to correct its seasonal response to stress and calibrated the GPP residual via a linear function of Ta and VPD time-varying derivatives to account for its daily response to stress. Leave-one-site-out cross-validation suggested that the improvements relative to its original version were especially noteworthy under stress conditions while less significant when there was no water or heat stress across grasslands and croplands. When compared to six other GPP models, the enhanced NIRvP model consistently outperformed them or performed comparably with the best model in terms of bias, RSME, and coefficient of determinant against measurements in grasslands and croplands. Moreover, we found that parameterizing the fraction of photosynthetically active radiation term using NIRv notably improved the performance of the classic MOD17 and vegetation photosynthesis model, with an average RMSE reduction of 13 % across grasslands and croplands. Overall, this study highlights the need to consider environmental stressors for improved NIRvP-based GPP and shed light on future improvements of LUE models.