Groundwater dynamics along a hillslope: A test of the steady state hypothesis

Abstract

Appropriate conceptual simplifications and assumptions are a central issue for hydrological modeling, especially when those models serve as the foundation for more complex hydrochemical or ecological models. A common and often unexamined assumption in conceptual modeling is that the relation between groundwater levels and runoff can be described as a succession of steady state conditions. This results in a single‐valued, monotonic function between the groundwater levels and runoff. Consequently, the simulated rise and fall in groundwater levels always follow the dynamics of runoff. We tested this assumption with an analysis of detailed groundwater level data along two opposing hillslopes along a stream reach in a Swedish till catchment at Svartberget. Groundwater levels in areas close to the stream followed the dynamics of the runoff. The correlation between groundwater level and runoff decreased markedly for wells farther than approximately 40 m from the stream. The levels were often independent of streamflow: Upslope area groundwater could be rising when riparian groundwater and runoff were falling, and vice versa. There was a high degree of correlation between groundwater levels at similar distances from the stream. The median Spearman rank correlation between wells within 35 m from the stream was 0.86 and for wells located more than 60 m from the stream was 0.96. This indicated that there is a common hydrological pattern even in the upslope area that can be identified and modeled. Despite the widespread acceptance of the steady state assumption previously in this and other study catchments, our study shows that it is not valid for the investigated hillslope site. If the divergence from steady state, with potential ramifications for other processes such as runoff chemistry, is common, then it will be worthwhile to reconsider the appropriate range of applicability for the steady state hypothesis, and the alternatives to that hypothesis.