Mid- and long-term runoff predictions by an improved phase-space reconstruction model

Area of Science:

• Hydrology
• Nonlinear Dynamics
• Computational Intelligence

Background:

• Traditional phase-space reconstruction methods require improvement for accurate mid- and long-term runoff predictions.
• The rainfall-runoff process is complex and influenced by multiple variables, necessitating more sophisticated modeling approaches.

Purpose of the Study:

• To develop a novel nonlinear model for monthly runoff prediction by enhancing the phase-space reconstruction method.
• To incorporate additional hydrological factors like temperature and rainfall into the runoff prediction model.
• To analyze the chaotic characteristics of the improved model and validate its performance.

Main Methods:

• Genetic algorithm optimization applied to phase-space reconstruction for monthly runoff modeling.
• Incorporation of multiple variables (temperature, rainfall) to capture the complexity of the rainfall-runoff process.
• Analysis of chaotic dynamics to confirm the model's representation of nonlinear hydrological behavior.

Main Results:

• The improved model demonstrates satisfactory medium- and long-term runoff forecasting accuracy, with a mean absolute percentage error under 15%.
• The model effectively predicts runoff trends in both wet and dry years, mitigating the traditional predictability barrier.
• Validation across six hydrological stations on the Yellow and Yangtze Rivers confirms the model's universality and reliability.

Conclusions:

• The enhanced phase-space reconstruction model offers a more reliable and stable approach to monthly runoff prediction compared to existing methods.
• The study provides a new framework for understanding the relationship between monthly runoff and other hydrological factors.
• This research presents a novel and effective method for long-term runoff forecasting.