Open AccessRainfall Prediction using Artificial Neural Network in Semi-Arid mountainous region, Saudi Arabia
Author(s) -
Roohul Abad Khan,
Rachida El Morabet,
Javed Mallick,
Mohammed Azam Ali,
Viola Vambol,
Sergij Vambol,
Volodymyr Sydorenko
Publication year - 2021
Publication title -
ecological questions
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.227
H-Index - 10
eISSN - 2083-5469
pISSN - 1644-7298
DOI - 10.12775/eq.2021.038
Subject(s) - arid , artificial neural network , evapotranspiration , environmental science , mean squared error , meteorology , climatology , physical geography , hydrology (agriculture) , statistics , geography , computer science , mathematics , machine learning , geology , ecology , paleontology , biology , geotechnical engineering
Rainfall prediction using Artificial Intelligence technique is gaining attention nowadays. Semi-arid region receives rainfall below potential evapotranspiration but more than arid region. However, in mountainous semi-arid region high rainfall intensity makes it highly variable. This renders rainfall prediction difficult by applying normal techniques and calls for data pre-processing. This study presents rainfall prediction in semi-arid mountainous region of Abha, KSA. The study adopted Moving Average (Method) for data pre-processing based on 2 years, 3 years, 4 years, 5 years and 10 years. The Artificial Neural Network (ANN) was trained for a period of 1978-2016 rainfall data. The neural network was validated against the existing data of period 1997-2006. The trained neural network was used to predict for period of 2017-2025. The performance of the model was evaluated against AAE, MAE, RMSE, MASE and PP. The mean absolute error was observed least in 2 years moving average model. However, the most accurate prediction models were obtained from 2 years moving average and 5 year moving average. The study concludes that ANN coupled with MA have potential of predicting rainfall in Semi-Arid mountainous region.