Answers are grounded in this paper and cite their sources.
Analysis evaluates machine learning techniques to fill daily streamflow gaps in Ethiopia, highlighting effective water management strategies.
Reliable daily streamflow observations are essential for sustainable water resource management, yet many developing countries face significant gaps in streamflow records due to conflict and financial constraints. This study evaluates machine learning (ML) and climatological mean (CM) approaches for filling long-term daily streamflow gaps in the Ribb and Gilgel Abbay watersheds of the Lake Tana sub-basin, upper Blue Nile basin. Specifically, decision tree regression (DTR), artificial neural network (ANN), linear regression (LR), seasonal autoregressive integrated moving average with exogenous variables (SARIMAX), and CM methods were compared. Input variables for ML included remote sensing-based daily rainfall (with 0–4-day lags), season index, and date features. Model performance was assessed using Nash–Sutcliffe efficiency (NSE), Bias, coefficient of correlation (CC), and Kling–Gupta efficiency (KGE). DTR outperformed all other methods, achieving NSE scores of 0.81–0.82, Bias between 6.8 and 13.9%, CC of 0.90–0.93, and KGE of 0.82–0.85. It effectively preserved temporal variability and reproduced peak-flow events more accurately than both other ML methods and the CM approach. The results demonstrate that integrating DTR with rainfall and temporal features provides a reliable framework for estimating long-term streamflow data gaps, supporting effective water resource planning and management in regions with limited continuous streamflow observations.
No takes yet. Share an insight, caveat, or question.
Lakew et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: