Short-term forecasts of streamflow in the UK based on a novel hybrid artificial intelligence algorithm

Area of Science:

• Hydrology and Water Resource Management
• Climate Change Impact Assessment
• Computational Intelligence in Environmental Science

Background:

• Climate change necessitates accurate streamflow forecasting for effective water resource management.
• Traditional forecasting methods face challenges with complex hydrological systems and varying catchment characteristics.
• Short-term streamflow prediction is critical for operational water management and flood/drought mitigation.

Purpose of the Study:

• To develop and evaluate a novel ensemble (hybrid) model for short-term streamflow forecasting.
• To assess the performance of the hybrid model against simpler ensemble and single-algorithm models.
• To investigate the model's reliability across diverse UK watercourses and forecast horizons up to 7 days.

Main Methods:

• Ensemble modeling combining Deep Learning (Nonlinear AutoRegressive network with eXogenous inputs - NARX) and Machine Learning (Multilayer Perceptron - MLP, Random Forest - RF).
• Utilized precipitation as the sole exogenous input for forecasting streamflow.
• Conducted a large-scale regional study across 18 UK watercourses with varying catchment properties.

Main Results:

• The hybrid Machine Learning-Deep Learning model significantly outperformed simpler ensemble and single Deep Learning models.
• Achieved high predictive accuracy (R² > 0.9) for numerous watercourses, particularly in complex small basins.
• Demonstrated superior performance stability with increasing forecast horizons, providing reliable 7-day predictions.

Conclusions:

• The proposed hybrid NARX-MLP-RF model offers a robust and accurate solution for short-term streamflow forecasting.
• This advanced approach enhances water resource planning and management capabilities, especially under climate change pressures.
• The model's effectiveness in diverse hydrological settings highlights its potential for widespread application in water management.