In deep polders in the Netherlands the dominant mechanism of salinization is via intense seepage through boils. These are local connections between the surface water and a deep saline aquifer. The study area of this research is the Lissertocht. This catchment, which is located inside the Haarlemmermeerpolder, contains a lot of boils that can cause problems for the main land use in this catchment agriculture. During summer the salinity is lowered by flushing the catchment with fresh water from outside the polder.

The objective of this thesis is to measure and model the salt concentrations and their sources in order to optimise the placement of sensors in a salinity monitoring network for optimal flushing in the Lissertocht catchment.

This thesis has provided insights into the detection of boils using different methods (Distribute Temperature Sensing (DTS), EC routing and visual inspection). The DTS method was the only method that is able to detect boils located at the bottom of the bigger main channels. Moreover, DTS is able to characterize the boil flux on a qualitative level and also on a quantitative level for a boil that is located near a pump; both have been done in this research. The boil flux is calculated with an energy balance and an 1D advection-diffusion model, the results correspond to values found in literature.

Data from CTD divers and sensitivity analysis have shown the influence of various processes on the spatial and temporal distribution of the salinity in the catchment. The sensitivity analysis in SOBEK also shows the areas which are sensitive to flushing, the flushing speed and the main pathways which the water from the inlets take. The complexity of the spatial and temporal distribution of the salinity is reduced to a small system of linear components using the Principal Component Analysis (PCA). For this case study, it was shown that over 90\% of the total variance in the system can be captured with three orthogonal linear components. A predictive model is built based on the PCA, this model is able to test the performance of a layout of sensors by reconstructing the principal components with virtual measurements. This shows that with a small number of sensors, a salinity monitoring network can be built that is able to predict the spatial and temporal distribution of the salinity in the catchment. Based on the knowledge from the measurements and modelling, three categories of key locations to monitor can be identified.

The optimisation of the sensor locations shows that a greedy algorithm is a very efficient way to solve the sensor placement problem. When more sensors are added with the greedy algorithm the performance of the monitoring network improves, but after six sensors the additional gain of an extra sensor is small in this catchment. The robustness analysis against disturbances shows that a monitoring network with more than three sensors is more robust. Moreover, the robustness analysis shows that a monitoring network is most sensitive to the locations of the boils.