Assessing the impacts of climate change on streamflow dynamics: A machine learning perspective

Area of Science:

• Hydrology and Climate Science
• Water Resource Management
• Machine Learning Applications

Background:

• Climate change is altering river flow patterns globally, necessitating accurate streamflow predictions.
• Extreme weather events are increasing, posing risks to water resource management.
• The Hunza Basin in Pakistan is vulnerable to climate-induced hydrological changes.

Purpose of the Study:

• To assess machine learning models for streamflow prediction in the Hunza Basin under climate change scenarios.
• To compare the performance of Artificial Neural Network (ANN), Recurrent Neural Network (RNN), and Adaptive Fuzzy Neural Inference System (ANFIS) models.
• To project future streamflow using the best-performing model under different Shared Socioeconomic Pathways (SSPs).

Main Methods:

• Utilized monthly precipitation, maximum temperature, and minimum temperature as input variables.
• Employed ANN, RNN, and ANFIS models for streamflow prediction, with discharge as the output.
• Evaluated model performance using Mean Square Error (MSE), Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and Coefficient of Determination (R²).

Main Results:

• The ANN model (3-10-1 architecture) demonstrated superior performance over RNN and ANFIS.
• ANN achieved high accuracy with low MSE, RMSE, MAE, and high R² for both training and testing datasets.
• The selected ANN model successfully predicted future streamflow under SSP245 and SSP585 scenarios up to 2100.

Conclusions:

• Artificial Neural Network models show significant potential for robust streamflow prediction in data-scarce regions.
• Accurate streamflow forecasting is crucial for effective water resource management and adaptation strategies in the Hunza Basin.
• The study provides valuable data for policymakers and water managers to address climate change impacts on river systems.