Groundwater–surface water interaction: Influence of accurate river descriptions for groundwater simulations

Abstract

The focus of this paper is the system bias present in groundwater models using rectangular cross-sections to represent rivers, and especially wide or braided rivers. The accurate description of surface water-groundwater interaction is of key importance today as more and more regions manage surface and ground waters conjunctively. This is especially true in basins where the most significant groundwater recharge source is loss from surface water bodies. The bias is due to the inaccuracies of the relationship between depth and flow as well as in the model for bed conductance which should be a function of river stage. This is demonstrated via analytical examples as well as site specific models in New Zealand. The site specific models have been developed using the traditional Stream Package (STR) in MODLFOW, as well as the new Stream Flow Routing Package (SFR1). These are compared to a MIKE SHE model using MIKE 11 for the stream flow calculations. The STR, SFR1 and MIKE 11 all use Mannings equation for flow routing along the river. However, SFR1 and MIKE 11 determine bed conductance as a function of stage, whereas STR does not. The case study allows the direct comparison of the performance of the two different representations for river flow. System dynamics are more accurately achieved as a result of the more accurate relationship between flow and stage; which has direct influence on the gradient between the river and the groundwater system. Furthermore, MIKE 11 also allows the river time step to be disconnected from the groundwater time step, allowing for more accurate representation of flow events that are shorter than a groundwater time step. The case study shows how this approach can be fully integrated into large (1000’s km2) regional scale groundwater models. There are three key results from this work: (1) groundwater surface water interaction is more accurately simulated using realistic cross-sections compared to using rectangular cross-sections, (2) short term flow events have impact the groundwater dynamics and (3) the dynamics of simulated time series at wells and springs along rivers are much closer to what is observed.