Predicting near-stream soil moisture using low-dimensional relationships in a small headwater catchment of the Coweeta Hydrologic Laboratory

Printer-friendly version

Spatial soil moisture patterns across the landscape have been shown to correlate best with topography during wet periods, but during dry periods are randomly distributed over space (Grayson et al., 1997).  Known as the preferred state hypothesis (Western et al., 1999), soil moisture correlation with topography or “spatial organization” is a function of mean hillslope wetness.  During wet periods, spatial organization is caused by greater precipitation relative to evapotranspiration (ET). This results in soil water redistribution occurring primarily in the horizontal direction towards the stream. During drier periods, ET is larger than precipitation leading to predominately vertical redistribution of soil water. Highly organized states can indicate significant soil water contribution to streamflow via subsurface flowpaths.  As hillslopes dry, patterns of soil moisture become more random indicating a lack of subsurface redistribution and possibly greater, more localized vertical redistribution. We hypothesize that these spatial patterns of soil moisture are linked to catchment runoff, so under highly organized states, streamflow will be greater than unorganized states. We also hypothesize that during drier periods with greater vertical drainage, shallow groundwater will be closely linked to streamflow.  Measuring soil moisture across an entire catchment requires intensive sampling, so to test relationships between catchment runoff and spatial soil moisture patterns we develop predictive models to estimate soil moisture at particular sites in our catchment. Our study is located along a northwest facing hillslope in Watershed 14 (WS14) of the Coweeta Hydrologic Laboratory in southwest North Carolina. Eighteen soil moisture plots are arranged along areas of topographic convergence perpendicular to the stream.  These plots are arranged into three transects of 6 plots to capture moments of topographic organization and disorganization.  Within each plot, synoptic soil moisture is measured bi-weekly to monthly using a handheld Dynamax ThetaProbe.  The portable soil moisture probe measures volumetric water content in the first 6cm of the soil.  In addition to this, three near-stream plots have datalogged soil moisture at two depths (0-30cm & 30-60cm) and shallow groundwater measured hourly. We limit our analysis of storage-runoff relationships to baseflow conditions to simplify drainage characteristics and predict soil moisture using three approaches: (1) an integral mass balance model that relates groundwater, soil moisture, and baseflow, (2) direct relationships between spatial soil moisture patterns and discharge, and (3) combining both mass balance and spatial models. Preliminary results show that the spatial soil moisture model tends to overpredict hillslope soil moisture, but by combining it with the mass balance model we can significantly improve our predictive ability.