Kohonen Self Organizing Map (SOM) Aquifer Water Struck Levels in the Merti Aquifer, Northern aided Predictions of Aquifer Water Struck Levels in the Merti Aquifer, Northern Kohonen Self Organizing Map (SOM)-aided Predictions of Aquifer Water Struck Levels in the Merti Aquifer, Northern Kenya

Abstract

The aquifer water struck levels in the Merti aquifer were assessed using the Kohonen Self-Organizing Maps algorithm, which employs the Neural Networks. This algorithm mimics the biological sensory and motor neurons. The variable was inferred via predicting the aquifer ground water levels , then subtracting the same from the elevations, measured in meters above the sea levels, of the proposed well point mapped, which is one of the three variables generated using the hand-held GPS. The objective of the study is to help develop a simple prediction model for aquifer well depths in the Merti aquifer, to be used alongside the geoelectrical models generated during geophysical surveys, so as to enhance and modernize groundwater management plan for the Merti Aquifer. Data on well hydraulics for the Isiolo, Garissa and parts of Wajir (south) counties were used to generate the model predictors, employing the Kohonen R package, which clusters and predicts variables using this neural network algorithm developed by Teuvo Kohonen. The algorithm was then used to predict the expected groundwater levels of a new area that has not been developed, and this value was subsequently subtracted from the elevation levels, thereby generating water struck levels. The variables employed to achieve this task were longitudes, latitudes, elevation, aquifer depth, resistivity, and the groundwater levels, (gwl) in meters (below ground level) bgl. To predict the gwl of a newly proposed drilling point, the hydrogeological data was run on R platform and models generated inferred and interpreted. The new model was then used to predict the gwl of the new site. Subtracted from elevation of the area, wsl was derived, thus The study concludes that the neural network SOM mapping algorithm is an accurate predictor of the wsl in the Merti aquifer, as it clusters geological zones bearing the same groundwater levels and aquifer depths together. It should therefore be a useful stochastic hydrological tool for decision making on matters groundwater development in the Merti aquifer.