Parameter uncertainties in groundwater modelling - A study on the effect of calibration method on parameter uncertainties in an inverse stochastic groundwater model

Abstract

Numerical models can be used to forecast the effects of underground constructions on groundwater conditions. Such forecasts always includes uncertainties, with values of used parameters being one common source of uncertainty. This study aims to assess how these parameter uncertainties and the resulting model outcome uncertainty are affected by either calibrating an inverse stochastic groundwater model against historic mean head observations or against a disturbance of the hydrogeological conditions. Two sections of the planned train tunnel Västlänken through Gothenburg are used as case studies to evaluate the effect that calibration method has on uncertainties. The first modelled section, Linné service tunnel uses historical mean head observations for the first case of calibration and leakage to the partially constructed tunnel with the effect on groundwater levels for the second calibration case. The other modelled section, Korsvägen, uses the same type of observations for the first calibration case, while the second case uses the effects of a pumping test. It was seen that the model over Linné service tunnel resulted in larger uncertainties for calibration case two compared to case one, while the model over Korsvägen resulted in less uncertainties for calibration case two compared to case one. This study shows that calibration method effects uncertainties deriving from used parameters and that more observations does not always mean less uncertainty.