Importance of stream infiltration data for modelling surface water-groundwater interactions

An infiltration experiment at the transect scale was performed in an ephemeral stream at the time it was dry, resulting in direct estimates of stream infiltration. Groundwater heads measured in a transect of piezometers orthogonal to the stream highlighted the development of a groundwater mound displaced several meters into the streambank. The experiment was modelled using the two-dimensional integrated surface-subsurface model HydroGeosphere (R) and calibrated with pilot points using different combinations of groundwater heads and infiltration volume data. Modelling results showed that despite the fact that groundwater heads were all below the streambed and that an unsaturated zone was present beneath the streambed, the stream and the aquifer become laterally connected within several hours of flow onset through a high permeability preferential path in the streambank. 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 thus estimates of stream infiltration during natural flow events. 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, incorporating infiltration flux or volume data into the model calibration increases the likelihood of obtaining accurate estimates of stream infiltration during natural flow events. (C) 2015 Elsevier B.V. All rights reserved.