Stepwise decomposition-integration-prediction framework for runoff forecasting considering boundary correction.

Area of Science:

• Hydrology
• Environmental Science
• Data Science

Background:

• Accurate river runoff prediction is crucial for water management, flood control, and ecological balance.
• Existing time series decomposition methods have limitations in practical runoff forecasting applications.

Purpose of the Study:

• To develop and validate a novel stepwise decomposition-integration-prediction considering boundary correction (SDIPBC) framework for enhanced river runoff forecasting.
• To evaluate the performance of a hybrid model combining Seasonal-Trend decomposition based on Loess (STL) with Long Short-Term Memory (LSTM) networks within the SDIPBC framework.

Main Methods:

• Proposed a novel stepwise decomposition-integration-prediction considering boundary correction (SDIPBC) framework.
• Implemented a hybrid forecasting model, STL-LSTM (SDIPBC), integrating STL decomposition with LSTM neural networks.
• Validated the model using historical runoff data from Lianghekou and Jinping I Reservoirs in the Yalong River Basin.

Main Results:

• The proposed SDIPBC framework effectively avoids using future information, enhancing prediction accuracy.
• The STL-LSTM (SDIPBC) model achieved high Nash-Sutcliffe efficiency (NSE) coefficients of 0.845 and 0.862 for ten-day runoff forecasting.
• The hybrid model demonstrated improved accuracy (1.81% and 2.38% increases) compared to a single LSTM model.

Conclusions:

• The SDIPBC framework offers a practical and reliable approach for decomposition-based hybrid runoff forecasting.
• The STL-LSTM (SDIPBC) model represents a significant advancement over existing methods for mid-long term river runoff estimation.