Preprocessing approaches in machine learning-based groundwater potential mapping: an application to the Koulikoro and Bamako regions, Mali

Groundwater is crucial for domestic supplies in the Sahel, where the strategic importance of aquifers can only be expected to increase in the coming years due to climate change. Groundwater potential mapping is gaining recognition as a valuable tool to underpin water management practices in the region, and hence, to improve water access. This paper presents a machine learning method to map groundwater potential and illustrates it through an application to two regions of Mali. A set of explanatory variables for the presence of groundwater is developed first. Several scaling methods (standardization, normalization, maximum absolute value and min-max scaling) are used to avoid the pitfalls associated with the reclassification of explanatory variables. A number of supervised learning classifiers is then trained and tested on a large borehole database (n = 3,345) in order to find meaningful correlations between the presence or absence of groundwater and the explanatory variables. This process identifies noisy, collinear and counterproductive variables and excludes them from the input dataset. Tree-based algorithms, including the AdaBoost, Gradient Boosting, Random Forest, Decision Tree and Extra Trees classifiers were found to outperform other algorithms on a consistent basis (accuracy > 0.85), whereas maximum absolute value and standardization proved the most efficient methods to scale explanatory variables. Borehole flow rate data is used to calibrate the results beyond standard machine learning metrics, thus adding robustness to the predictions. The southern part of the study area was identified as the better groundwater prospect, which is consistent with the geological and climatic setting. From a methodological standpoint, the outcomes lead to three major conclusions: (1) because there is no aprioristic way to know which algorithm will work better on a given dataset, we advocate the use of a large number of machine learning classifiers, out of which the best are subsequently picked for ensembling; (2) standard machine learning metrics may be of limited value when appraising map outcomes, and should be complemented with hydrogeological indicators whenever possible; and (3) the scaling of the variables helps to minimize bias arising from expert judgement and maintains robust predictive capabilities.

Rapid groundwater potential mapping in humanitarian contexts: improving borehole implementation in basement environments

Emergency responses in humanitarian contexts require rapid set-up of water supply. Boreholes are often drilled where the needs are highest and not where hydrogeological conditions are most favourable. The Rapid Groundwater Potential Mapping (RGWPM) methodology was therefore developed as a practical tool to support borehole siting when time is critical, allowing strategic planning of geophysical campaigns. RGWPM is based on the combined analysis of satellite images, digital elevation models and geological maps, obtained through spatial overlay of the two main hydrogeological variables controlling groundwater potential: water availability (WA) and reservoir capacity (RC). The WA associates hydrogeomorphological features to groundwater dynamic processes, while the RC reflects estimates of the hydraulic conductivity. RGWPM maps are produced through an overlay of WA and RC with the overall groundwater potential (GWP) characterized as ‘very low’, ‘low’, ‘medium’, and ‘high’, with each zone associated to a specific water supply option. The first RGWPM map was elaborated during a drilling campaign in Northern Uganda. The average yield for the eight boreholes sited ‘with’ RGWPM was 35 m3/h versus 3 m3/h for the 92 preexisting boreholes that were sited ‘without’ RGWPM. Statistical comparison of the classified yields of all hundred boreholes with the RGWPM predicted-yield ranges revealed a good correlation for the ‘low’ GWP unit, highlighting areas where well siting for motorised systems should be avoided. A rather poor correlation of 33% was found for the ‘medium’ GWP unit, believed to be artificially induced by the numerous hand pumps (low yields) located in potentially higher yielding areas.

Regional Mapping of Groundwater Potential in Ar Rub Al Khali, Arabian Peninsula Using the Classification and Regression Trees Model

Mapping of groundwater potential in remote arid and semi-arid regions underneath sand sheets over a very regional scale is a challenge and requires an accurate classifier. The Classification and Regression Trees (CART) model is a robust machine learning classifier used in groundwater potential mapping over a very regional scale. Ten essential groundwater conditioning factors (GWCFs) were constructed using remote sensing data. The spatial relationship between these conditioning factors and the observed groundwater wells locations was optimized and identified by using the chi-square method. A total of 185 groundwater well locations were randomly divided into 129 (70%) for training the model and 56 (30%) for validation. The model was applied for groundwater potential mapping by using optimal parameters values for additive trees were 186, the value for the learning rate was 0.1, and the maximum size of the tree was five. The validation result demonstrated that the area under the curve (AUC) of the CART was 0.920, which represents a predictive accuracy of 92%. The resulting map demonstrated that the depressions of Mondafan, Khujaymah and Wajid Mutaridah depression and the southern gulf salt basin (SGSB) near Saudi Arabia, Oman and the United Arab Emirates (UAE) borders reserve fresh fossil groundwater as indicated from the observed lakes and recovered paleolakes. The proposed model and the new maps are effective at enhancing the mapping of groundwater potential over a very regional scale obtained using machine learning algorithms, which are used rarely in the literature and can be applied to the Sahara and the Kalahari Desert.