Cong et al. (2026) Multi-Day Salinity Forecasting in the Vietnamese Mekong Delta: Horizon-Dependent Performance and Interpretation of XGBoost, LSTM, and a Weighted Hybrid

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

Identification

• Journal: Water
• Year: 2026
• Date: 2026-09-24
• Authors: Nguyen Phuoc Cong, Nigel K. Downes, Huỳnh Vương Thu Minh, Sudhir Kumar Singh
• DOI: 10.3390/w18192379

Research Groups

• Department of Hydrology and Water Resources, University of Arizona
• Mekong Delta Institute for Water Resources Research (MDIWR), Vietnam

Short Summary

This study compares the performance of three machine learning models (XGBoost, LSTM, and a weighted hybrid) in forecasting daily mean salinity at Vam Kenh station in the Vietnamese Mekong Delta one to five days ahead.

Objective

• Investigate the effectiveness of different machine learning models for short-range salinity forecasting in tide-influenced deltas

Study Configuration

• Spatial Scale: Local-scale (Vam Kenh station) with potential for regional applicability
• Temporal Scale: Daily, seasonal, and annual time scales

Methodology and Data

• Models used: Extreme gradient boosting (XGBoost), long short-term memory network (LSTM), weighted hybrid
• Data sources: In-situ observations from Vam Kenh station (721 daily records, 2020–2025)

Main Results

• XGBoost outperformed LSTM at one to three days ahead, while LSTM performed better at five days ahead
• The weighted hybrid model provided the best or equal-best estimates at every horizon, with modest improvement over the best constituent model
• Empirical prediction intervals achieved 0.97–1.00 coverage in the test season

Contributions

• Demonstrated horizon-dependent model complementarity and potential for improving short-range salinity forecasting in tide-influenced deltas
• Highlighted the importance of recent salinity, dry-season timing, and upstream hydrological conditions as predictive information sources

Funding

• This research was supported by the Mekong Delta Institute for Water Resources Research (MDIWR) and the University of Arizona's Department of Hydrology and Water Resources.

Citation

@article{Cong2026MultiDay,
  author = {Cong, Nguyen Phuoc and Downes, Nigel K. and Minh, Huỳnh Vương Thu and Singh, Sudhir Kumar},
  title = {Multi-Day Salinity Forecasting in the Vietnamese Mekong Delta: Horizon-Dependent Performance and Interpretation of XGBoost, LSTM, and a Weighted Hybrid},
  journal = {Water},
  year = {2026},
  doi = {10.3390/w18192379},
  url = {https://doi.org/10.3390/w18192379}
}