Skip to content
DSML.KZ
Publications

Code · Zenodo · June 30, 2026

sbc — Snow-influenced streamflow Bias Correction

Tetyana Honcharenko, Serhii Dolhopolov, Alexandr V. Neftissov, Ilyas Kazambayev, Lalita Kirichenko, Oleksandr Kuchanskyi

Ilyas Kazambayev
Open via DOI

Abstract

An explainable deep-ensemble framework for multi-scale (decadal and daily) bias correction of the GloFAS-ERA5 global river-discharge reanalysis across the snow- and glacier-influenced transboundary headwater basins of Central Asia. Bias correction is cast as log-residual learning; a gradient-boosting trio, an entity-aware LSTM and a regime-aware probabilistic flagship are combined through stacking, evaluated with a leakage-safe temporal / leave-one-basin-out / prediction-in-ungauged-regions matrix, and explained with SHAP attributions tied to snow, glacier, temperature and precipitation drivers. Supported by grant BR24993128.

Hydrology and Watershed Management StudiesHydrological Forecasting Using AICryospheric studies and observationssnowmelt runoffglacier meltensemble deep learning

More by Ilyas Kazambayev