Assessing the sensitivity of satellite-derived gross primary productivity to combined atmospheric dryness and soil moisture deficit

Water stress, characterized by atmospheric dryness (vapor pressure deficit, VPD) and soil moisture (SM) deficit, has a significant impact on terrestrial gross primary productivity (GPP), necessitating accurate modeling of the relative effects of VPD and SM deficit. Satellite remote sensing (RS) GPP estimations offer valuable tools for studying large-scale terrestrial GPP under water stress. However, it remains unclear how accurately they capture these relative effects compared to ground-based eddy covariance (EC) measurements. To address this gap, we quantified GPP sensitivity to VPD and SM deficit using ten widely used RS GPP products and EC measurements. By comparing GPP sensitivity patterns derived from RS and EC data across all ecosystems and within three ecosystem types (forest, grassland, and cropland), our results demonstrate that: (1) the mean of all ten RS GPP products (RSmean) captures the general directional response of GPP to VPD (i.e., mainly negative) and SM deficit (i.e., mixed positive and negative) across different VPD-SM gradients, but fails to reproduce the absolute value of GPP changes compared to EC measurements. This discrepancy could be attributed to RS products primarily capturing changes in canopy structure under water stress rather than accurately reflecting short-term plant physiological responses, while EC-derived GPP anomalies under water stress encompass both changes in canopy structure and plant physiological activities. (2) RSmean generally tracks the directional sensitivities of GPP to VPD and SM deficit within various ecosystem types, but significant magnitude differences are observed compared to EC measurements, with larger biases in forest ecosystems compared to grassland and cropland ecosystems, likely due to the lower sensitivity of deep-rooted forest ecosystems to water stress. (3) Despite the presence of biases, certain models (e.g., GOSIFGPP and BESSGPP) outperform others in terms of both GPP-VPD and GPP-SM sensitivities across all ecosystems and within different ecosystem types. Collectively, this study comprehensively assesses the ability of RS GPP estimations to capture vegetation responses to VPD and SM deficit and suggests methods for refining water stress effects in RS GPP models to enhance large-scale GPP impact assessments under water stress.