Hydrological prediction in ungauged basins based on spatiotemporal characteristics

Area of Science:

• Hydrology
• Environmental Science
• Data Science

Background:

• Parameter transplant is a common method for hydrological prediction in ungauged basins.
• This method is labor-intensive due to its reliance on expert judgment.
• Existing methods often lack accuracy and efficiency.

Purpose of the Study:

• To develop a novel hydrological prediction model, STH-Trans, to improve accuracy and reduce labor costs.
• To leverage multiple spatiotemporal views for enhanced predictive capabilities.
• To provide a robust tool for water resource management and decision support.

Main Methods:

• Utilized geographic and topographic indicators for watershed similarity identification.
• Developed an initial regression model using the TrAdaBoost algorithm.
• Refined the model with semi-supervised learning and multiple spatiotemporal views.

Main Results:

• The STH-Trans model demonstrated high efficiency in predicting runoff for ungauged basins.
• Achieved a substantial increase in model accuracy, ranging from 7.9% to 30% compared to conventional methods.
• The model's performance was validated through experimental results.

Conclusions:

• STH-Trans offers a significant advancement in hydrological prediction for ungauged basins.
• The model provides valuable data and decision support for water resource management, flood mitigation, and disaster relief.
• This approach reduces reliance on expert input, lowering labor costs.