Shanko et al. (2026) Comparative Analysis of LSTM and Random Forest Algorithms for Streamflow Prediction: A Case Study of Diverse River Basins in the United States

⚠️ Warning: This summary was generated from the abstract only, as the full text was not available.

Identification

• Journal: Water
• Year: 2026
• Date: 2026-07-22
• Authors: Alemayehu Shanko, Assefa M. Melesse
• DOI: 10.3390/w18141768

Research Groups

Not specified in the provided text.

Short Summary

This study compares Long Short-Term Memory (LSTM) and Random Forest (RF) algorithms for daily streamflow prediction across 16 diverse US basins, finding that RF generally provides superior accuracy and flood detection performance.

Objective

• To evaluate and compare the predictive performance of LSTM and RF machine learning algorithms for daily streamflow forecasting across hydroclimatically diverse basins.

Study Configuration

• Spatial Scale: Sixteen hydroclimatically diverse basins in the contiguous United States (CAMELS dataset).
• Temporal Scale: Daily time step.

Methodology and Data

• Models used: Long Short-Term Memory (LSTM) and Random Forest (RF).
• Data sources: CAMELS dataset, utilizing five climatic features and antecedent streamflow lag features (7, 14, and 30 days).

Main Results

• Predictive Accuracy: RF outperformed LSTM with an average Nash–Sutcliffe efficiency (NSE) of 0.755 compared to 0.632.
• Hydroclimatic Performance: Both models showed high efficacy in snowmelt-dominated basins but performed poorly in flashy humid regimes.
• Flood Detection: Both models achieved accuracy rates > 88% for flood event distinction, with RF achieving a higher F1 score (0.797) than LSTM (0.734).
• Application Recommendation: RF is recommended for general operational forecasting, while LSTM is suggested for perennial snowmelt- and groundwater-influenced catchments.

Contributions

• Provides a quantitative comparison between a recurrent neural network (LSTM) and a decision-tree ensemble (RF) for streamflow prediction across diverse hydroclimatic settings, identifying the specific environmental conditions under which each model excels.

Funding

Not specified in the provided text.

Citation

@article{Shanko2026Comparative,
  author = {Shanko, Alemayehu and Melesse, Assefa M.},
  title = {Comparative Analysis of LSTM and Random Forest Algorithms for Streamflow Prediction: A Case Study of Diverse River Basins in the United States},
  journal = {Water},
  year = {2026},
  doi = {10.3390/w18141768},
  url = {https://doi.org/10.3390/w18141768}
}