Identification of Hyporheic Extent and Functional Zonation During Seasonal Streamflow Recession by Unsupervised Clustering of Time‐lapse Electrical Resistivity Models

Hyporheic exchange influences hydrologic transport and water quality through transient storage, which extends solute transit time, and leads to mixing of surface water and groundwater. Despite its importance, estimating the extent and spatiotemporal variability of the hyporheic zone remains challenging due to limitations in assessing the subsurface with discrete point-scale sampling. Analysis of time-lapse electrical resistivity (ER) data from tracer studies has shown potential to ameliorate such limitations. However, its utility in objectively delimiting hyporheic extent and quantifying changes in surface-groundwater exchange has been impeded by reliance on qualitative analysis of hyporheic extent or the use of a priori assumptions about data quality and signal strength. This study applies a novel unsupervised clustering method to time-lapse ER models derived from a benchmark dataset collected throughout baseflow recession in a mountain stream. We demonstrate that unsupervised clustering of inverted ER model time series can delimit hyporheic extent by distinguishing solute transport signals from noisy background inversions and identify functional zones defined by unique transport characteristics. We found that the structure of these zones was stable even as discharge changed by an order of magnitude, likely due to morphological constraints in this steep, narrow valley. Compared to traditional methods utilizing a priori thresholds to delimit hyporheic extent, clustering is robust to unintentional variations in tracer breakthrough curves that are typical of field-based studies. Therefore, clustering of inverted ER models represents a more robust and data-driven functional zonation representation of hyporheic exchange than has been possible with point-scale sampling or transport modelling, which usually assumes a single well-mixed hyporheic zone.