Flash Flood prediction of Panjkora River, KPK, Using Artificial Neural Networks (ANN) and Support Vector Machine (SVM)

Abstract

Globally, floods are the most damaging natural hazards, because of its severe nature and difficulty in forecasting that restrains emergency responses. Floods triggered by thunderstorms are frequent in the high mountainous area of Hindu Kush Himalaya especially in the north of Pakistan. This research delves into effective water management in the Panjkora River basin (a tributary of Swat River), combining statistical evaluations with socio-economic considerations for flood predictions. Amidst climate change uncertainties, a historical analysis of the 2010 and 2022 catastrophic events emphasises the urgency for robust predictive models. The study identifies the Artificial Neural Networks (ANN) model's proficiency in daily discharge prediction and the Support Vector Machine (SVM) model's commendable performance in monthly flood event forecasting. Focusing on daily runoff in the Panjkora River, the research leverages the R-squared (R²) metric, revealing the ANN model's superior accuracy (R² = 0.75) compared to the SVM model (R² = 0.60) in daily discharge prediction. Extending the analysis to monthly flood event prediction, the SVM model excels with an R² value of 0.60 despite limitations in capturing nuances. The significance of this research lies in the fact that a combination of multiple inputs, ANN and SVM techniques has been used to develop the flood models. This research revealed the potential of algorithm-based computational models in predicting floods and developed some useful techniques that can be used by the disaster management authorities in Pakistan.