Sensitivity analysis and development of calibration methodology for near-surface hydrogeology model of Laxemar

This report describes modelling where the hydrological modelling system MIKE SHE has been used to describe surface hydrology, near-surface hydrogeology, advective transport mechanisms, and the contact between groundwater and surface water within the SKB site investigation area at Laxemar. In the MIKE SHE system, surface water flow is described with the one-dimensional modelling tool MIKE 11, which is fully and dynamically integrated with the groundwater flow module in MIKE SHE. In early 2008, a supplementary data set will be available and a process of updating, rebuilding and calibrating the MIKE SHE model based on this data set will start. Before the calibration on the new data begins, it is important to gather as much knowledge as possible on calibration methods, and to identify critical calibration parameters and areas within the model that require special attention. In this project, the MIKE SHE model has been further developed. The model area has been extended, and the present model also includes an updated bedrock model and a more detailed description of the surface stream network. The numerical model has been updated and optimized, especially regarding the modelling of evapotranspiration and the unsaturated zone, and the coupling between the surface stream network in MIKE 11 and the overland flow in MIKE SHE. An initial calibration has been made and a base case has been defined and evaluated. In connection with the calibration, the most important changes made in the model were the following: The evapotranspiration was reduced. The infiltration capacity was reduced. The hydraulic conductivities of the Quaternary deposits in the water-saturated part of the subsurface were reduced. Data from one surface water level monitoring station, four surface water discharge monitoring stations and 43 groundwater level monitoring stations (SSM series boreholes) have been used to evaluate and calibrate the model. The base case simulations showed a reasonable agreement between measured and calculated surface water discharges, but the model generally underestimates the total runoff from the area. The model also overestimates the groundwater levels, and the modelled groundwater level amplitudes are too small in many boreholes. A number of likely or potential reasons for these deviations can be identified: The surface stream network description in the model is incomplete. This implies that too little overland water is drained from the area by the streams, which creates ponded areas in the model that do not exist in reality. These areas are characterized by large evaporation and infiltration, contributing to groundwater recharge and reducing transpiration from the groundwater table, in turn creating high and relatively stable groundwater levels compared to those measured at the site. In order to improve the agreement between measured and modelled surface water discharges, the evapotranspiration was reduced in the model; in effect, this implied a reduction of the potential evapotranspiration. This probably caused a larger groundwater recharge and less transpiration during summer, thereby reducing the variations in the modelled groundwater levels. If the MIKE 11 stream network is updated, the potential evapotranspiration could be increased again, such that the modelling of groundwater dynamics is improved. The bottom boundary condition and the hydraulic conductivity of the bedrock may have a large effect on model-calculated near-surface/surface water flows in Laxemar. A sensitivity analysis shows that lowering the hydraulic head at the bottom boundary (located at 150 metres below sea level) lowers the groundwater levels in the Quaternary deposits, but also implies smaller surface water discharges. Lowering the hydraulic conductivity of the bedrock would increase groundwater flows to Quaternary deposits in groundwater discharge areas, which raises groundwater levels and reduces fluctuation amplitudes. An alternative model approach, using a deeper MIKE SHE model down to less fractured bedrock, may also be interesting to evaluate. It is recommended that the observations above are further evaluated in connection with the next modelling phase for Laxemar during 2008. A sensitivity analysis has been made on calibration parameters. The most important results from the sensitivity analysis are the following: The hydraulic conductivity in the saturated zone proved to be more important than all of the tested vegetation and unsaturated zone parameters. The second most important parameters were the hydraulic conductivity of the unsaturated zone (Ks) and the specific yield (Sy). A lower hydraulic conductivity in the saturated zone increases the peak surface water flows, decreases the base flows, and increases the groundwater head amplitudes and the groundwater head elevations. A lower hydraulic conductivity in the unsaturated zone (Ks) increases the surface water flows, and, to some extent, decreases the groundwater head elevations. A lower specific yield in the unsaturated zone (Sy) increases the surface water flows (although with a smaller effect than Ks), increases the groundwater head amplitudes, and to some extent, increases the groundwater head elevations. A method for performing the calibrations of future models is also presented based on the results from the base case simulations and the sensitivity analysis