
Geogenic groundwater contamination challenges universal access to safe drinking water. Understanding distributions of groundwater quality is important to identify hazards but a lack of data and appropriate modelling techniques hampers efforts. Here we use a modified random forest (fuzzy random forest) supported by recent water quality data to predict the spatial distribution of likely groundwater manganese exceedance and measures of prediction (un)certainty across Uganda. Relatively high certainty (< 1 % variation) of threshold exceedance (> 95 % probability) is identified across Southwest and North Uganda, highlighting locations to consider for appropriately-selected remediation and/or further monitoring for improved quality of groundwater derived drinking sources.