The Belgian Meuse is a river with a relatively low flood risk. The floodplains are narrow, and the estimated impact is not large until the flood reaches a magnitude of a once in a 200 years event. Flooding along the Belgian Meuse could however have an effect on the flood safety in the Netherlands, since there the flood defences are dimensioned on such extreme events. In addition, the climate is changing, causing more extreme rainfall in the Meuse basin. Therefore, knowledge of what will happen in case of extreme floods is becoming more important. Accurate numerical models and better estimates of extreme rainfall and flood events give the possibility to research flooding on the river Meuse. The goal of this thesis is to analyse extreme floods on the Belgian Meuse: What happens along this river under extreme conditions, and what are the consequences in the Netherlands? To analyse what areas along the Belgian Meuse are prone to flooding and how severe the flooding in these areas can be, a terrain analysis is carried out. This analysis shows that the river reach between Namur and the Dutch border is prone to flooding. The spatial limit that the flood could potentially have is not more than a few kilometres from the river axis, due to the valley shape of the floodplains. These flood prone areas have a potential storage capacity of 70 million m^3, and can thus have a significant influence on the flood wave. The valley shape of the Meuse ensures that there is no upper limit to the conveyance capacity, so the discharge at Eijsden has no cut-off level. A numerical model of the Belgian Meuse is built to simulate extreme floods. The new SOBEK 3.4 software, which combines SOBEK 1D and Flexible Mesh 2D, is used to build the model. Although the software is still under development, it has sufficient possibilities to model the area of research. The created flood model gives good results; it can reproduce the results from a similar but more accurate 2D model of the University of Liege. The resulting impact of the flooding on the discharge is a few percent damping, increasing gradually from nothing for once in 50 years wave to 7% for a once in 50,000 years wave. A sensitivity and uncertainty analysis are carried out to examine the uncertainties in the flood model. The first describes the selection of the parameters with the largest influence on the result. Two of these parameters, the main channel roughness and emergency measures, are further analysed in the uncertainty analysis. The main channel roughness gives the largest spread in discharge at Eijsden: a variation of 100 m^3 up or down. Placing sandbags has a small effect on the discharge, since these emergency measures will likely not be sufficient to protect the floodplains. A third parameter, the timing of the inflow from the side branches has a large influence too. It is recommended to use a smaller GRADE time step for the extreme events on the Meuse, since the timing is important for the fast genesis of floods in the Meuse basin. When combined with the uncertainties in the rainfall and run-off models, the contribution of the hydraulic uncertainty is minimal. When all GRADE uncertainties are incorporated in the design discharge curve, the impact of flooding along the Belgian Meuse leads to a maximum reduction of 100 m^3 compared to the case of no flooding. Considering flooding when determining the exceedance frequency curve is recommended, it gives a more realistic estimate for the discharge extremes. The hydrograph shape plays an important role here, since the wider wave compromises the effect of retention basins. It is thus also recommended to use a different, wider, hydrograph when considering flooding. The water level reduction due to flooding is estimated to be up to 30 cm along the Dutch Meuse. The current method of determining discharge extremes, HR2006, gives lower discharges than GRADE. When switching to the GRADE frequency curve with (or without) flooding and with uncertainty, the same safety standards will give higher design conditions. By considering flooding, these higher GRADE values are mitigated.