## Samenvatting

In general, valley ecosystems are to a greater or lesser extent influenced by groundwater. Recharge by excess rainfall is the source for continuous groundwater flow towards seepage areas in river valleys. This groundwater flow appears as upwelling groundwater in wetland areas and results as base flow in rivers. In many cases natural groundwater systems have undergone important changes due to anthropogenic influences.

A numerical groundwater model is indispensable to describe these changes and the global flow dynamics of the groundwater system. A physically based model with spatially distributed input data offers the possibility to gain insight into the valley system dynamics.

The groundwater system of the Upper Alzette catchment, located in the south of Luxembourg, was modelled using the finite-difference groundwater model MODFLOW. After the set-up of the steady-state groundwater model an extensive sensitivity analysis was performed. The results formed the basis for a trial-and-error calibration process. Usually, calibration of a groundwater model involves an inverse modelling strategy, i.e. optimizing the model parameters - especially aquifer hydraulic conductivities - as a function of measured groundwater heads in a number of observation wells. Deviations of the simulated heads are expressed in one or more statistical goodness-to-fit measures, such as Root Mean Square Error.

However, in this study an additional calibration strategy was employed considering the water balances of a number of subcatchments. For each of these subcatchments the baseflow was estimated from river discharge time series using the Hydrograph Separation tool HYSEP. These estimated baseflows were used as additional observations for the calibration of a number of conductance parameters. While the simulated groundwater heads and total water balance remained nearly unchanged, this analysis clearly shows that the sub-balances differ considerably. The simulated baseflows for the sub-catchments are about 30% more accurate than the original results. This result not only indicates that considering sub-water balances in the calibration process greatly improves the model results, but also considerably increases the understanding of the groundwater dynamics.

After calibration the uncertainty of the model results was assessed. Using the results of the sensitivity analysis a spatially distributed 'range' for the simulated groundwater levels was defined, indicating the deviation at each location.

A numerical groundwater model is indispensable to describe these changes and the global flow dynamics of the groundwater system. A physically based model with spatially distributed input data offers the possibility to gain insight into the valley system dynamics.

The groundwater system of the Upper Alzette catchment, located in the south of Luxembourg, was modelled using the finite-difference groundwater model MODFLOW. After the set-up of the steady-state groundwater model an extensive sensitivity analysis was performed. The results formed the basis for a trial-and-error calibration process. Usually, calibration of a groundwater model involves an inverse modelling strategy, i.e. optimizing the model parameters - especially aquifer hydraulic conductivities - as a function of measured groundwater heads in a number of observation wells. Deviations of the simulated heads are expressed in one or more statistical goodness-to-fit measures, such as Root Mean Square Error.

However, in this study an additional calibration strategy was employed considering the water balances of a number of subcatchments. For each of these subcatchments the baseflow was estimated from river discharge time series using the Hydrograph Separation tool HYSEP. These estimated baseflows were used as additional observations for the calibration of a number of conductance parameters. While the simulated groundwater heads and total water balance remained nearly unchanged, this analysis clearly shows that the sub-balances differ considerably. The simulated baseflows for the sub-catchments are about 30% more accurate than the original results. This result not only indicates that considering sub-water balances in the calibration process greatly improves the model results, but also considerably increases the understanding of the groundwater dynamics.

After calibration the uncertainty of the model results was assessed. Using the results of the sensitivity analysis a spatially distributed 'range' for the simulated groundwater levels was defined, indicating the deviation at each location.

Originele taal-2 | English |
---|---|

Uitgeverij | Vrije Universiteit Brussel |

Aantal pagina's | 125 |

Uitgave | Research funded by ‘le Gouvernement Du Grand-Duché de Luxembour |

Status | Published - 31 dec 2007 |

## Keywords

- Hydrologische modellering
- Grondwater
- Alzette Rivier, Luxemburg