Parameter sensitivity analysis and optimization for a satellite‐based evapotranspiration model across multiple sites using Moderate Resolution Imaging Spectroradiometer and flux data. Issue 1 (11th January 2017)

• Record Type: Journal Article
• Authors: Zhang, Kun; Ma, Jinzhu; Zhu, Gaofeng; Ma, Ting; Han, Tuo; Feng, Li Li
• Is Part Of: Journal of geophysical research. Volume 122:Issue 1(2017)
• Page Start: 230
• Page End: 245
• Publication Date: 2017-01-11
• Subjects: evapotranspiration -- sensitivity analysis -- optimization -- model parameters; Atmospheric physics -- Periodicals; Geophysics -- Periodicals; 551.5
• DOI: 10.1002/2016JD025768
• Languages: English
• ISSNs: 2169-897X

Abstract

Abstract: Global and regional estimates of daily evapotranspiration are essential to our understanding of the hydrologic cycle and climate change. In this study, we selected the radiation‐based Priestly‐Taylor Jet Propulsion Laboratory (PT‐JPL) model and assessed it at a daily time scale by using 44 flux towers. These towers distributed in a wide range of ecological systems: croplands, deciduous broadleaf forest, evergreen broadleaf forest, evergreen needleleaf forest, grasslands, mixed forests, savannas, and shrublands. A regional land surface evapotranspiration model with a relatively simple structure, the PT‐JPL model largely uses ecophysiologically‐based formulation and parameters to relate potential evapotranspiration to actual evapotranspiration. The results using the original model indicate that the model always overestimates evapotranspiration in arid regions. This likely results from the misrepresentation of water limitation and energy partition in the model. By analyzing physiological processes and determining the sensitive parameters, we identified a series of parameter sets that can increase model performance. The model with optimized parameters showed better performance ( R 2 = 0.2–0.87; Nash‐Sutcliffe efficiency (NSE) = 0.1–0.87) at each site than the original model ( R 2 = 0.19–0.87; NSE = −12.14–0.85). The results of the optimization indicated that the parameter β (water control of soil evaporation) was much lower in arid regions than in relatively humid regions. Furthermore, the optimized value of parameter m 1 (plant control of canopy transpiration) was mostly between 1 to 1.3, slightly lower than the original value. Also, the optimized parameter T opt correlated well to the actual environmental temperature at each site. We suggest that using optimized parameters with the PT‐JPL model could provide an efficient way to improve the model performance. Key Points: To identify the key parameters of the PT‐JPL model specific to different biomes by using SA with forcing data To optimize the selected key parameters using the DEMC scheme for different biomes To evaluate and compare the performance of the original model and the optimized model across different biomes