Combined groundwater - surface water modeling with a lumped hydrological model

Abstract

The Lowland Groundwater-Surface water Interaction model (LGSI-model) is a lumped model concept which describes changes in storage via changes in the average groundwater depth and spatial variation in groundwater depth. These characteristics of the groundwater depth are used to calculate discharge. The spatial distribution of groundwater depths are described by means of a normal distribution that is related to the average groundwater depth within the catchment. The LGSI-model was first tested by Van der Velde et al. (2009) in the Hupsel brook catchment (6km2) and showed very promising results. The goal of the study described in this report is to apply the theory and model concepts developed by Van der Velde et al. (2009) to a larger catchment with a more complex geomorphology. For this purpose the Drentsche Aa catchment was selected. To build the lumped LGSI-model, the relations between groundwater depth distributions and fluxes can be calculated based on a spatially distributed groundwater. In this case, the spatially distributed model MIPWA, that includes the Drentsche Aa catchment, was used. Based on the MIPWA model results on a daily basis for the period 1989-2001, relations between storage above the surface (negative groundwater depths) and water fluxes between groundwater and surface water (e.g. stream discharge, tube drain discharge, saturated overland flow, evapotranspiration) could be formulated. These relations formed the basis of the LGSI-model of the Drentsche Aa Catchment. To improve the representation of the discharge generating processes within the catchment, the LGSI-model of the Drentsche Aa catchment was calibrated to a measured discharge series of the period 1989-2001 using the GLUE analysis. Due to the variations in the geomorphology of the catchment, the Drentsche Aa catchment had to be divided into two different areas. One low area with shallow groundwater tables reaching up to the surface and high, very dry areas where the groundwater tables never reach the surface. These two area types, or reservoirs, could be described by different statistical distributions: in the low areas, a normal distribution was valid, while in the high areas groundwater depth distributions were described by a gamma distribution. Both areas were coupled using a Darcy based equation which was a modification to the original model concept of Van der Velde et al. (2009). Results of the LGSI-model of the Drentsche Aa catchment were very promising and the Nash-Sutcliffe model efficiency for discharge and groundwater depth simulation for both low and high areas were respectively 0.76 (discharge), 0.78 (groundwater depth low area) and 0.87 (groundwater depth high area) over the calibration period 1989-2001 (all at a daily time steps). Peak discharges were slightly underestimated by the model, which is the result of inaccurate simulation of the groundwater depth, whereas recessions and periods with small discharge were modelled with high accuracy. A detailed analysis of the model stability was performed, by extending the model simulation of discharge to the period 1980-2010 (Nash- Sutcliffe discharge 0.64). The LGSI-model was also tested on a hourly time step to determine if the model could have a good model performance on a smaller timescale (Nash-Sutcliffe discharge 0.75). The results of the fast calibration using the GLUE analysis can be useful for evaluation of model parameters of the spatially distributed groundwater model. Since this type of calibration cannot be done for spatially calibrated models without very long calculation times, the LGSI-model could be used as a tool for calibration of these models. Another interesting feature of the LGSI-model concept is that it divides the total discharge of the catchment into its separate components (groundwater exfiltration in rivers, saturated overland flow, drainage). This offers the possibility to analyse the composition of peak discharge or determine the relative contributions of the discharge components throughout the year. It was concluded that the LGSI-model generated very good simulations of discharge and groundwater depths in the Drentsche Aa catchment. Due to these simultaneous and rapid model simulations of groundwater depth, storage and discharge, the LGSI-model concept provides additional insight in the discharge behaviour for a variety of groundwater depth distributions and parameter values. Moreover, the LGSI-model shows that knowledge of the groundwater system and the groundwater-surface water interaction processes that occur in a catchment is very useful, if not crucial, for good simulation and prediction of stream discharge of a catchment.