Thai University RankingsRESEARCH RADAR
Evidence of global relevance

Streamflow prediction with machine learning: evaluating predictability across hydroclimatic regimes

Forty-three years of daily rainfall and streamflow across 24 basins showed both models above NSE 0.90 in humid stable basins but below 0.5 in arid variable catchments.

01

Key findings

  • RF outperformed in 18 basins, especially intermittent and low-flow settings. Both achieved NSE>0.90 in humid stable basins and
02

Why this matters globally

The diagnostic framework supports regime-based model selection and warns that the most management-critical variable basins may be hardest to predict.

03

Thai researcher contribution

An Asian Institute of Technology researcher contributed to a cross-regime streamflow-model selection framework.

04

Limitations to consider

The abstract omits basin locations, final splits and physical baselines. Retrospective performance does not guarantee real-time flood forecasting, and climate nonstationarity may erode accuracy.

05

Verify the original sources

Hydrological Sciences JournalRead the original article

DOI: 10.1080/02626667.2026.2699259

KEEP EXPLORING

More Thai research to explore