Prediction of Fire-Induced Pollution Dispersion in Watersheds using ANN, A Case Study of Dareh-Moradbeyg River Watershed

Abstract

Post-fire runoff, driven by the loss of vegetation cover and alterations in soil characteristics, represents a significant source of pollutants in downstream ecosystems, residential, and urban areas.This study employs an artificial intelligence approach, optimized using the Artificial Bee Colony (ABC) algorithm, to predict water quality in a fire-affected watershed. pH variation data, collected following a prescribed burn in the study area, was utilized as input for developing an artificial neural network. The network’s interlayer connection weights and threshold parameters were optimized using an enhanced version of the ABC algorithm. The results indicate that, compared to conventional models, the neural network trained using the ABC algorithm demonstrates approximately 25% greater predictive accuracy. These findings emphasize the critical role of machine learning techniques in forecasting pollutant dispersion in post-fire watershed environments.