Exploring Spatiotemporal Trends in Piezometer Network Data Using Self‑Organizing Maps

January 7, 2026

Large piezometer networks are implemented at mine sites for geotechnical and environmental monitoring. Groundwater head time series from these networks contain valuable information about aquifer stresses and hydraulic connections. Machine learning methods can efficiently extract information from large volumes of head data but are not yet common practice. To illustrate the utility of machine learning in a practical context, we use the self-organizing map (SOM) method to extract patterns from groundwater head time series for 85 piezometers at the Red Chris Mine in northern Canada. Aquifer response and hydraulic connectivity are uncertain at the site due to changing stresses and geological heterogeneity. We demonstrate the influence and choice of the map size hyperparameter. ‘Small maps’ provide insight on aquifer-scale hydraulic behavior and ‘large maps’ can highlight local anomalies like borehole leakage and repair. We evaluated SOM output using site-specific knowledge of geology and groundwater flow. The SOM results reveal spatiotemporal trends that were not readily apparent from previous site characterization studies.

Mee, E. 2025 “Exploring Spatiotemporal Trends in Piezometer Network Data Using Self-Organizing Maps” Mine Water and the Environment