Assessment of global gridded transpiration products using the extended instrumental variable technique (EIVD)

Abstract

Transpiration (T) constitutes a significant portion of total terrestrial evapotranspiration in ecosystems and functions as the connection between the water and energy cycle. Despite its importance, large-scale quantification of T remains a formidable challenge. Global gridded T datasets are essential for investigating and predicting the regulatory mechanisms of transpiration. However, the scarcity of reliable field references data limits direct validation efforts. This study uses collocation analysis methods, including triple collocation (TC) and extended double instrumental variable algorithm (EIVD), to assess three global T products at 0.25°-daily resolutions. Signal-to-noise ratios (SNR) and absolute error variances (σε2) obtained through EIVD serve as primary indicators. Our results show that both TC and EIVD approaches yield valuable insights into the accuracy and performance of the T products. Moreover, the analysis emphasizes the necessity of considering non-zero error cross-correlation when utilizing collocation analysis for error characterization. Overall, this study presents a comprehensive examination and comparison of collocation analysis for error characterization of three T products. The findings suggested that collocation analysis methods may function as reliable alternatives to tower observations at a global scale, offering potential benefits for data assimilation and merging.