Hongmei Li, Wenpeng Xie, Xiaoyang Li, Kei Yoshimura
Received 3 July, 2025
Accepted 15 September, 2025
Published online 28 November, 2025

Hongmei Li¹⁾, Wenpeng Xie²⁾, Xiaoyang Li²⁾, Kei Yoshimura²⁾

1) Natural Environmental Studies, Graduate School of Frontier Sciences, The University of Tokyo, Japan
2) Institute of Industrial Sciences, The University of Tokyo, Japan

Accurate simulation of gross primary productivity (GPP) and evapotranspiration (ET) is essential for understanding global carbon and water cycles. However, land surface models (LSMs), while widely used, still exhibit substantial uncertainties, largely due to poorly calibrated parameters. This study integrated the Morris sensitivity analysis with Differential Evolution calibration (MD framework) to improve GPP and ET simulations in the MATSIRO LSM across twenty FLUXNET sites representing diverse ecosystem types. Results demonstrated that the MD framework led to marked improvements, with ensemble mean Kling-Gupta Efficiency (KGE) for monthly GPP and ET increasing by 43% and 30%, respectively. Among all parameters, GPP and ET were most sensitive to the maximum carboxylation rate of RuBisCO (VMAX0), which was underestimated in the original model. Along with VMAX0, the MD framework identified the top eight sensitive parameters for each site and provided site-specific optimal values, enabling tailored parameterization to improve simulation performance across diverse ecosystems. In addition, our results showed that leaf area index (LAI) is a non-negligible source of uncertainty for GPP simulations in cropland sites. Multiple-cropping systems amplify the importance of accurately capturing the abrupt LAI. These findings offer valuable guidance for improving GPP and ET simulation in LSMs, particularly in managed ecosystems.

[Full Text]

Copyright (c) 2025 The Author(s) CC-BY 4.0