How Direct Flux Measurements Can Improve Infiltration Estimates in Ephemeral Streams

Abstract:

In arid and semi-arid areas replenishment of groundwater resources is strongly dependent on surface water infiltration from ephemeral streams and creeks. An accurate estimation of water infiltration and aquifer recharge from these surface water features is paramount, but this task is subject to physical (streambed heterogeneity) and transient (flow variability) challenges of these watercourses. Although several methods are available to estimate stream infiltration, only a handful of them are suitable for ephemeral streams, all of which are indirect methods.

In this study we performed an infiltration experiment at the transect scale in an ephemeral stream at the time it was dry, thereby obtaining direct estimates of stream infiltration. Groundwater heads measured in a transect of piezometers orthogonal to the stream identified the development of a groundwater mound laterally displaced several meters into the streambank. The experiment was modelled using the two-dimensional integrated surface-subsurface model HydroGeosphere^® and calibrated with PEST using different combinations of groundwater heads and infiltration volume data. Although the model can be calibrated to produce a good match to measured groundwater heads, accurate predictions of stream infiltration can be made only if the heads used for model calibration capture the groundwater mound. The use of infiltration flux or volume during model calibration, instead, provides the best calibration results, and does not require knowledge of the position of the groundwater mound. Given that heterogeneity of streambed and streambank sediments leads to the possibility that the groundwater mound developed during flow events will be poorly characterised or perhaps missed altogether, we demonstrate that incorporating infiltration flux or volume data into the model calibration increases the ability of a model to simulate accurate estimates of stream infiltration during natural flow events.