Shi et al. (2025) Improving runoff simulation in cold alpine regions based on VIC-glacier by combining LSTM error correction technology

Identification

• Journal: Journal of Hydrology
• Year: 2025
• Date: 2025-09-17
• Authors: Chen Shi, Qin Liu, Yungang Bai, Qiying Yu, Zhenlin Lu, Chengshuai Liu, Biao Cao, Lei Ren, Ming Li, Miao Gan, Caihong Hu
• DOI: 10.1016/j.jhydrol.2025.134251

Research Groups

• College of Water Conservancy and Transportation, Zhengzhou University, China
• Xinjiang Institute of Water Resources and Hydropower Research, China
• Yellow River Conservancy Commission Henan Hydrological and Water Resources Bureau, China

Short Summary

This study developed and optimized a coupled Variable Infiltration Capacity-Glacier (VIC-glacier) model for the upper Hotan River Basin, demonstrating that integrating Long Short-Term Memory (LSTM) error correction significantly enhances runoff simulation accuracy in cold alpine regions.

Objective

• To develop and optimize a coupled VIC-glacier model for the upper Hotan River Basin, and to improve its runoff simulation accuracy in cold alpine regions by integrating and comparing LSTM and Autoregression (AR) error correction techniques.

Study Configuration

• Spatial Scale: Upper Hotan River Basin, Xinjiang, China.
• Temporal Scale: Focused on short-term and future runoff predictions across different forecast periods. Specific historical calibration/training periods are not detailed in the provided text.

Methodology and Data

• Models used: Variable Infiltration Capacity-Glacier (VIC-glacier) model, Shuffled Complex Evolution (SCE-UA) for automatic calibration, Long Short-Term Memory (LSTM) for error correction, Autoregression (AR) for error correction.
• Data sources: Not explicitly detailed in the provided text, but implied to be hydrological and meteorological data necessary for runoff and glacier melt simulation.

Main Results

• The VIC-glacier model, when combined with LSTM error correction, significantly improved simulation accuracy across various forecast periods.
• For short-term forecasts, the Nash-Sutcliffe Efficiency (NSE) reached 0.9 during both training and testing periods.
• Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) were reduced compared to the uncorrected model.
• The LSTM error correction model consistently outperformed the AR error correction model across multiple foresight periods.

Contributions

• Development and application of a coupled VIC-glacier model specifically for the upper Hotan River Basin, a typical cold alpine region.
• Introduction and validation of LSTM error correction technology to significantly enhance the accuracy of glacier runoff simulation in cold alpine environments.
• Demonstrated the superior performance of LSTM over AR for error correction in hydrological forecasting for these regions.
• Provides a scientific basis for mitigating water resource uncertainties due to glacier changes and for more accurate future runoff predictions, supporting integrated water resource management in Xinjiang's cold alpine regions.

Funding

• No funding information is provided in the supplied text.

Citation

@article{Shi2025Improving,
  author = {Shi, Chen and Liu, Qin and Bai, Yungang and Yu, Qiying and Lu, Zhenlin and Liu, Chengshuai and Cao, Biao and Ren, Lei and Li, Ming and Gan, Miao and Hu, Caihong},
  title = {Improving runoff simulation in cold alpine regions based on VIC-glacier by combining LSTM error correction technology},
  journal = {Journal of Hydrology},
  year = {2025},
  doi = {10.1016/j.jhydrol.2025.134251},
  url = {https://doi.org/10.1016/j.jhydrol.2025.134251}
}