Open AccessApplication of wavelet artificial neural networks in forecasting seasonal rainfall time series in Queensland, Australia
Author(s) -
Meysam Ghamariadyan,
Monzur Alam Imteaz
Publication year - 2021
Publication title -
iop conference series. materials science and engineering
Language(s) - English
Resource type - Journals
eISSN - 1757-899X
pISSN - 1757-8981
DOI - 10.1088/1757-899x/1067/1/012038
Subject(s) - artificial neural network , mean squared error , wavelet , series (stratigraphy) , wavelet transform , computer science , time series , statistics , artificial intelligence , machine learning , mathematics , pattern recognition (psychology) , geology , paleontology
Predicting accurate future rainfall plays an essential role in managing the water resources as well as controlling the unforeseen phenomena like flooding or drought. In this study, an Artificial Intelligence (AI) method called wavelet artificial neural network is used to predict one-year ahead seasonal rainfall of Queensland, Australia. The technique used in this study is based on wavelet transform and artificial neural networks (ANN). Over the wavelet transforms analysis, the primary signal is split into some sub-time series called approximation and detail. The outcome of this process is some meticulous and valuable information that cannot be easily captured in the main time series. To assessment the skillfulness of the proposed method, five input sets combined of rainfall and three climate indices are defined. The comparisons showed that the wavelet-ANN (WANN) presented the lower root-mean-square-error (RMSE) with 64.8mm compared to ANN with 88.9mm, and it outperforms the ANN forecast with respect to statistical measures. Since fewer data in predicting seasonal rainfall is available compared to monthly rainfall data, the application of WANN provides promising insights into using this AI method for predicting seasonal rainfall one year in advance in Australia.