Artificial Neural Network Technique for Annual Rainfall Generation Applied to Three Selected Sites in Kurdistan Region, Iraq

Abstract

Predicting rainfall is one of the more difficult tasks involved in weather forecasting. Due to extreme climate variations, it is now harder than ever to predict rainfall accurately. In the current study, an Artificial Neural Network (ANN) has been used to forecast the annual maximum rainfall (AMR) data from 1990 to 2021 at three chosen stations (i.e., Duhok, Erbil, and Sulaymaniya) in the Kurdistan region of Iraq. The Multilayer Perceptron (MLP) approach of ANN models was applied in generating and forecasting the AMR time series of the adopted stations. Model performance indicators such as model efficiency, correlation coefficient, root mean square error, and root mean absolute error were used to evaluate the performance of ANN for the annual rainfall prediction. The ANN models were used to forecast the AMR data for the upcoming five years (2022 to 2026). The study reveals that the MLP approach of the ANN models, which we have used is the most appropriate tool for forecasting the AMR data series in the three selected stations in the Kurdistan Region of Iraq for future time periods.