Lake Water Level Prediction Model Based on Artificial Intelligence and Classical Techniques – An Empirical Study on Lake Volta Basin, Ghana

Authors : Michael Stanley PEPRAH, Edwin Kojo LARBI

Pages : 134-150

View : 52 | Download : 8

Publication Date : 2021-03-15

Article Type : Research Paper

Abstract :Several studies in the past and recent years have suggested numerous mathematical models for Lake Water Level insert ignore into journalissuearticles values(LWL); modelling to a good precision. This study considered an empirical evaluation of Artificial Intelligence and Classical Techniques such as Wavelet Transform insert ignore into journalissuearticles values(WT);, Bayesian Regularization Backpropagation Artificial Neural Network insert ignore into journalissuearticles values(BRBPANN);, Levenberg-Marquardt Backpropagation Artificial Neural Network insert ignore into journalissuearticles values(LMBPANN);, Scaled Conjugate-Gradient Backpropagation Artificial Neural Network insert ignore into journalissuearticles values(SCGBPANN);, Radial Basis Functions Artificial Neural Network insert ignore into journalissuearticles values(RBFANN);, Generalized Regression Artificial Neural Network insert ignore into journalissuearticles values(GRANN);, Multiple Linear Regression insert ignore into journalissuearticles values(MLR);, and Autoregressive Integrated Moving Average insert ignore into journalissuearticles values(ARIMA); for LWL modelling. The motive is to apply and assess for the first time in our study area, the working efficiency of the aforementioned techniques. Satellite altimetry data provided by the United States Department of Agriculture was used in this study. The input and output variables used in this study were the decomposed LWL by the WT. Each model technique was assessed based on statistical measures such as Arithmetic Mean Error insert ignore into journalissuearticles values(AME);, Arithmetic Mean Square Error insert ignore into journalissuearticles values(AMSE);, arithmetic mean absolute percentage deviation insert ignore into journalissuearticles values(AMAPD);, minimum error value insert ignore into journalissuearticles values(rmin);, maximum error value insert ignore into journalissuearticles values(rmax);, and arithmetic standard deviation insert ignore into journalissuearticles values(ASD);. The statistical analysis of the results revealed that, all the hybridized models successfully estimate the LWL heights at a good precision for the study area. However, Discrete Wavelet Transform insert ignore into journalissuearticles values(DWT);-MLR model outperforms DWT-BRBPANN, DWT-LMBPANN, DWT-SCGBPANN, DWT-RBFANN, DWT-GRANN, and DWT-ARIMA techniques in estimating the LWL heights for the study area. In terms of AME, AMSE and ASD, DWT-MLR achieved 0.1988 m, 0.0024 m, and 0.0017 m respectively. The main conclusion drawn from this study is that, the method of using novel ensemble models is promising and can be adopted for LWL modelling in the study area. This study seeks to contribute to the existing knowledge on understanding the hydrodynamic processes in Lake Volta Basin and support water resource management.
Keywords : Artificial Intelligence, Lake Volta Basin, Stochastic Models, Time Series Analysis, Water Level Modelling

ORIGINAL ARTICLE URL