Clustering of Temporal Profiles Using a Bayesian Logistic Mixture Model: Analyzing Groundwater Level Data to Understand the Characteristics of Urban Groundwater Recharge

The hydrogeological conditions of groundwater can be examined by carefully studying the patterns of fluctuations in groundwater levels. These fluctuations are spatially and temporally influenced by many complicated factors, including rainfall, topography, land use, and hydraulic properties of soils and bedrock (i.e., aquifers). In this paper, we develop methodology based on a Bayesian logistic mixture model to simultaneously 1) cluster profiles of groundwater level changes over time and 2) estimate the relationship between the characteristics of each cluster and environmental variables. We apply the methodology to analyze groundwater level profiles from 37 monitoring wells in Seoul, South Korea, and we find four clusters of wells. Using the estimated relationship between the clusters and the environmental variables, we discern the hydrogeologic conditions of each cluster. Thus, we gain a better understanding of the recharge and subsurface flow of bedrock groundwater in an urban setting and the vulnerability of groundwater to the inflow of potential pollutants from ground surface.