Answers are grounded in this paper and cite their sources.
Study region 154 unregulated catchments across Australia, representing diverse hydroclimatic conditions. Study focus This study develops and evaluates a multi-method ensemble (MME) regionalization framework for continuous streamflow prediction in ungauged catchments. Using the SIMHYD hydrological model, 11 regionalization methods were evaluated, including spatial proximity, physical similarity, principal component regression, partial least squares regression, random forests, extreme gradient boosting, and ensemble averaging strategies. The 154 catchments were classified into three relatively homogeneous clusters using principal component analysis-based clustering, and regionalization performance was assessed through leave-one-out cross-validation. New hydrological insights for the region The SIMHYD model achieved satisfactory calibration performance, with a median Nash–Sutcliffe efficiency of 0.723 during the calibration period. However, regionalization performance varied across catchments, with no single method consistently outperforming the others. Physical similarity with output averaging achieved the highest overall median square root-transformed Nash–Sutcliffe efficiency ( NSE sqrt ) of 0.695, while the MME achieved a comparable value of 0.687. Notably, Cluster 2 exhibited the poorest overall prediction performance among the three clusters, yet MME achieved the highest median NSE sqrt (0.624) within this cluster. These findings suggest that ensemble averaging provides an effective approach to continuous streamflow prediction in ungauged catchments.
No takes yet. Share an insight, caveat, or question.
Chen et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: