Groundwater Sodium Levels Estimation of Proposed Irrigation Groundwater Source for the Kumahumato Settlement of the Dadaab Subcounty, North Eastern Kenya

Abstract

The Project Area of Kumahumato is a locality located on the fringes of the Merti aquifer within a radii ranging from 5 to 10 kilometer metric units. The area is primarily inhabited Nomadic pastoralists who have limited experience with matters farming and allied agricultural techniques. Owing to the rapid change in climate patterns and with massive death toll of livestock resulting from prolonged droughts and unpredictable rains, the community leadership have deemed it fit to focus on irrigation-aided agriculture. One problem noted is that the sodium levels in the soils may be exacerbated by irrigation farming, if the groundwater sodic levels area already way above the thresholds deemed safe by the WHO, both for human usage and for soil chemistry. The sodium levels may increase with progressive usage of borehole water in the farming projects, up to a point deemed way beyond salvage-meaning the destroyed fertility may not be reclaimed or restored once the damage is done. To mitigate against the potential disastrous and irreversible consequence, the study team undertook a geophysical surveys as well as hydrochemical surveys and data analysis to understand the likely consequence of a prolonged usage of irrigation –based agriculture in the Kumahumato centre. To achieve this, eight algorithms were employed, namely, Neural Networks, Naïve Bayes, Support Vector Machines, Logistic Regression, Decision Trees, K-Nearest Neighbor and Random forests  algorithm amongst others. The final three algorithms mentioned here emerged out as the best performers, registering between 95 to hundred percent precisions levels during detailed data analysis. A point picked at random in the Kumahumato area which showed promise of good groundwater potential was analysed and found to be at suitable aquifer sodium levels, which will not be a threat to small scale agriculture envisaged in the program. Machine Learning was thus employed and proved a useful decision making tool in the Project Planning and Design Phase for the proposed food security meant to be a practical resilience response to climate change hazards.