Where the Decision Tree Grows, Exploring the Use of Tree-Based Machine Learning Methods in Monitoring Well Site Selection

Abstract

Effective monitoring of groundwater is necessary for understanding the potential consequences of extraction on surface water features. Monitoring wells are often the most expensive part of a groundwater investigation. Thus, monitoring locations should be chosen to provide the most information regarding predictions of interest for scientists and/or decision makers. This investigation examines whether tree-based Machine Learning models can be used to identify informative monitoring locations in the context of a single pumping well and a stream. Existing information is contained in a basic numerical hydrogeologic model. Uncertainty is introduced through varied hydrologic property values and boundary conditions. Since the model generates both water level and stream flow data, the regression trees are trained on the ensemble of model outputs representing a range of plausible hydrogeologic systems. In addition, the groundwater model can be used to generate values of capture in the form of streamflow depletion, which cannot be measured directly. This value of capture can then be used in training these ML models as the new target, highlighting the impacts of well pumping. Results show that the built-in feature importance capability of regression trees allows for the efficient identification of optimal monitoring locations, only requiring a single run of each model in an ensemble. Future work could examine the use of tree-based ML models within the context of existing data and the inclusion of model and measurement errors.