Groundwater similarity across a watershed derived from time‐warped and flow‐corrected time series

Abstract

Information about catchment‐scale groundwater dynamics is necessary to understand how catchments store and release water and why water quantity and quality varies in streams. However, groundwater level monitoring is often restricted to a limited number of sites. Knowledge of the factors that determine similarity between monitoring sites can be used to predict catchment‐scale groundwater storage and connectivity of different runoff source areas. We used distance‐based and correlation‐based similarity measures to quantify the spatial and temporal differences in shallow groundwater similarity for 51 monitoring sites in a Swiss prealpine catchment. The 41 months long time series were preprocessed using Dynamic Time‐Warping and a Flow‐corrected Time Transformation to account for small timing differences and bias toward low‐flow periods. The mean distance‐based groundwater similarity was correlated to topographic indices, such as upslope contributing area, topographic wetness index, and local slope. Correlation‐based similarity was less related to landscape position but instead revealed differences between seasons. Analysis of variance and partial Mantel tests showed that landscape position, represented by the topographic wetness index, explained 52% of the variability in mean distance‐based groundwater similarity, while spatial distance, represented by the Euclidean distance, explained only 5%. The variability in distance‐based similarity and correlation‐based similarity between groundwater and streamflow time series was significantly larger for midslope locations than for other landscape positions. This suggests that groundwater dynamics at these midslope sites, which are important to understand runoff source areas and hydrological connectivity at the catchment scale, are most difficult to predict.

Spatiotemporal variability in hydrochemistry of shallow groundwater in a small pre‐alpine catchment: The importance of landscape elements

“…The topographic attributes were calculated based on a 6 × 6 m digital terrain model derived from LiDAR data. The 6‐m resolution is consistent with previous analyses on the relation between topography on groundwater level dynamics (Rinderer et al, , 2016, ) and is considered the optimum resolution for this catchment. We selected the site‐specific topographic attributes that are significantly correlated with the groundwater level dynamics (Rinderer et al, ): slope, TWI (Beven & Kirkby, ), accumulated area (c.f.…”

From Points to Patterns: Using Groundwater Time Series Clustering to Investigate Subsurface Hydrological Connectivity and Runoff Source Area Dynamics

“…A question remains whether this approach is applicable, if an extended groundwater monitoring network, like ours, is not in place? This study and previous work by Rinderer et al () showed that if groundwater is monitored in a few representative landscape positions, it can be transferred to other sites with similar environmental conditions — at least in catchments with steep slopes and shallow groundwater tables. As a consequence, the effort and costs for groundwater monitoring can be considerably reduced by an informed choice of representative monitoring locations based on landscape position.…”

“…Data-driven approaches, like the one presented here, depend on the amount of data available and the assumption that sites with similar landscape characteristics also have similar groundwater dynamics. At least for our study site with steep terrain and shallow groundwater tables, this could be proven (Rinderer et al, 2017). We, therefore, expect this method to be also useful for other catchments where the differences in groundwater level response dynamics are related to topographic indices or other spatial information.…”

“…Topographic characteristics (i.e., TWI, slope, upslope contributing area, and curvature) of the monitoring sites were derived from a 6 × 6 m DEM using a multiple flow direction algorithm (Seibert & McGlynn, 2007). Similar to Rinderer et al (2014Rinderer et al ( , 2016Rinderer et al ( , 2017, sites with a TWI >6, a local slope <30%, and an upslope contributing area >600 m 2 are defined as footslope sites. Upslope sites are defined as sites with a TWI <4, a local slope >50%, and an upslope contributing area <200 m 2 , whereas midslope sites have a TWI between 4 and 6, a local slope between 30% and 50%, and an upslope contributing area between 200 and 600 m 2 .…”

Runoff generation in a pre-alpine catchment: A discussion between a tracer and a shallow groundwater hydrologist

“…Distance-based similarity measures based on the integral between two time series capture similarity in the shape and amplitude of the time series and were correlated with topographic indices. Correlation-based similarity measures (based on the cross-correlation of two time series) revealed seasonal differences in the similarity in groundwater and streamflow time series (Rinderer et al, 2017). The cross-correlation between groundwater and streamflow time was higher during the growing season than during the dormant season, suggesting a stronger coupling of groundwater and streamflow and most likely a higher degree of connectivity between hillslopes and streams during the growing season than during the dormant season (Rinderer et al, 2017).…”