Forecasting SPEI and SPI Drought Indices Using the Integrated Artificial Neural Networks

Area of Science:

• Hydrology
• Artificial Intelligence
• Climate Science

Background:

• Drought forecasting is crucial for water resource management.
• Artificial neural networks (ANNs) offer potential for improved hydrological predictions.
• Comparing different ANN architectures is essential for advancing forecasting capabilities.

Purpose of the Study:

• To compare the performance of two ANN models for drought index forecasting.
• To evaluate the efficacy of a feedforward multilayer perceptron (sANN) against an integrated neural network model (hANN).
• To assess forecast accuracy for Standardized Precipitation Index (SPI) and Standardized Precipitation Evaporation Index (SPEI).

Main Methods:

• Utilized meteorological and hydrological data from the MOPEX experiment (1948-2002).
• Trained sANN and hANN models using the adaptive differential evolution algorithm (JADE).
• Evaluated model performance using six statistical measures and analyzed forecast accuracy with four indices.

Main Results:

• The integrated neural network model (hANN) demonstrated superior performance compared to the sANN model.
• hANN provided more accurate forecasts for both SPI and SPEI drought indices.
• Model comparison was based on rigorous statistical performance metrics.

Conclusions:

• The integrated neural network model (hANN) is a more effective tool for drought index forecasting than the sANN.
• Findings suggest that advanced ANN architectures can significantly enhance hydrological drought predictions.
• This research provides valuable insights for developing robust drought management strategies.