Rainwater Quality Assessment Based on Artificial Neural Network Using Mathematical Models

Abstract

One of the most crucial and difficult duties jobs performed by meteorological agencies in the world is the forecasting of weather, particularly rainfall. Furthermore, it is a difficult process that calls for knowledge from many different specialist domains. In this study, a model based on an artificial neural network (ARNN) is suggested as a method to forecast successive rainfalls according to Meteorological Station based on analyses of previous rainfall data. It is for 11 years from 2010-2021 at the Hay Al Hussian in Basrah. Based on three informative meteorological factors, the feed forward neural networks with back propagation algorithm are used for learning and predicting. The created models have been trained, validated, and tested using observations of temperature, wind speed, and relative humidity. The research discovered that the neural network ARNN organized (3-30-1) was capable of long-term rainfall forecasting at the study region with one hidden layer and layers (3-25-5-1) were capable of long-term rainfall forecasting in this region with two hidden layers. The model was able to learn the events that it had been trained to recognize, according to the results, which were backed up by strong correlation coefficient (R) values and low mean squared errors (MSE) values. The root MSE was 0.0016187, and R Value was found to be 0.997 for one hidden layer and the root MSE was 8.4731E-8, and R value was found to be 1 for two hidden layers.