Wuhan Ligong Daxue Xuebao (Jiaotong Kexue Yu Gongcheng Ban)/Journal of Wuhan University of Technology (Transportation Science and Engineering)

This study assesses the performance of four monthly water balance models (MWBMs: GR2M, XM, abcd, DWBM) in arid and semi-arid catchments across Spain, Australia, and the United States, utilizing CAMELS datasets. Our large-scale assessment demonstrates that some parsimonious MWBMs yield good and robust results in these regions. A three-fold cross-validation, optimizing Nash-Sutcliffe Efficiency (NSE) or Kling-Gupta Efficiency (KGE12), revealed regional differences. The GR2M model exhibited higher performance and robustness in Australia and the US, DWBM and abcd topped in Spain, while XM consistently underperformed. Globally, KGE12 proved to be a superior optimization metric. Model performance was generally highest in Spain, followed by Australia and the USA. To predict model performance category, a logistic regression, based on mean annual runoff and precipitation as inputs, achieved from 75% to 89% success rate. The findings confirm MWBMs' capability in arid and semi-arid regions and provide empirical insights into their performance.