Mapping Groundwater Potential Through an Ensemble of Big Data Methods

• DOI: 10.1111/gwat.12939
• Publication Date: 2019-09-05T02:14:48Z
• Authors: P. Martínez‐Santos; P. Renard
• Keywords: Landform; Ensemble Learning

Abstract Groundwater resources are crucial to safe drinking supplies in sub‐Saharan Africa, and will be increasingly relied upon a context of climate change. The need better understand groundwater calls for innovative approaches make the best out existing information. A methodology map potential based on an ensemble machine learning classifiers is presented. large borehole database ( n = 1848) was integrated into Geographic Information Systems (GIS) environment used train, validate test 12 algorithms. Each classifier predicts binary target (positive or negative borehole) minimum flow rate required communal domestic supplies. Classification number explanatory variables, including landforms, lineaments, soil, vegetation, geology slope, among others. Correlations between variables were then generalized develop maps. Most algorithms attained success rates 80% 96% terms score, which suggests that outcomes provide accurate picture field conditions. Statistical learners observed perform than most other algorithms, excepting random forests support vector machines. Furthermore, it concluded approach provides added value by incorporating measure uncertainty results. This technique may rapidly rural supply humanitarian emergencies areas where there sufficient historical data but comprehensive work unfeasible.