A Time-Series Water Level Forecasting Model Based on Imputation and Variable Selection Method.

Area of Science:

• Environmental Science
• Hydrology
• Data Science

Background:

• Reservoir water levels are vital for household needs and national economic stability.
• Accurate forecasting of reservoir water levels is essential for effective water resource management.
• Existing forecasting methods may not fully leverage available atmospheric and historical reservoir data.

Purpose of the Study:

• To propose a novel time-series forecasting model for reservoir water levels.
• To enhance prediction accuracy by integrating missing value imputation and variable selection.
• To evaluate the performance of the proposed model against traditional methods.

Main Methods:

• Collected daily atmospheric and Shimen Reservoir water level data (2008-2015).
• Applied five imputation methods for handling missing data.
• Utilized factor analysis and sequential variable selection to identify key predictive variables.
• Developed a Random Forest model for water level forecasting.

Main Results:

• The Random Forest model with variable selection demonstrated superior forecasting performance compared to the listing model.
• Variable selection effectively identified crucial factors influencing water levels.
• The proposed model showed improved forecasting capability by optimizing variable inclusion.

Conclusions:

• The developed time-series forecasting model, incorporating imputation and variable selection, offers enhanced accuracy for reservoir water level prediction.
• Random Forest, combined with strategic variable selection, is a powerful tool for hydrological forecasting.
• This approach provides a valuable methodology for optimizing water resource management through improved forecasting.