Machine learning based groundwater prediction in a data-scarce basin of Ghana
2022
Siabi, E. K. | Dile, Y. T. | Kabo-Bah, A. T. | Amo-Boateng, M. | Anornu, G. K. | Akpoti, Komlavi | Vuu, C. | Donkor, P. | Mensah, S. K. | Incoom, A. B. M. | Opoku, E. K. | Atta-Darkwa, T.
Groundwater (GW) is a key source of drinking water and irrigation to combat growing food insecurity and for improved water access in rural sub-Saharan Africa. However, there are limited studies due to data scarcity in the region. New modeling techniques such as Machine learning (ML) are found robust and promising tools to assess GW recharge with less expensive data. The study utilized ML technique in GW recharge prediction for selected locations to assess sustainability of GW resources in Ghana. Two artificial neural networks (ANN) models namely Feedforward Neural Network with Multilayer Perceptron (FNN-MLP) and Extreme Learning Machine (FNN-ELM) were used for the prediction of GW using 58 years (1960–2018) of GW data. Model evaluation between FNN-MLP and FNN-ELM showed that the former approach was better in predicting GW with R2 ranging from 0.97 to 0.99 while the latter has an R2 between 0.42 to 0.68. The overall performance of both models was acceptable and suggests that ANN is a useful forecasting tool for GW assessment. The outcomes from this study will add value to the current methods of GW assessment and development, which is one of the pillars of the sustainable development goals (SDG 6).
显示更多 [+] 显示较少 [-]