This project will examine how infiltration beds level of performance decrease over time. In order to predict the substance load, which the infiltration beds are exposed to, a number of experiments have been made. According to these experiments it will be possible to predict the build-up rate of sediments on the urban surfaces in dry weather periods. The collected sediment is divided into particle fractions and the smaller fractions are tested for concentrations of six well-known heavy metals, which often are seen in stormwater runoff. The test results show a clear tendency where the smallest particle fractions have the highest concentrations and therefore, they are the most important particles to retain in the filter. For retaining these small particles by mechanical filtration, it is necessary to use a more heterogenic filter material, which often results in lower hydraulic conductivity. This lower conductivity could potentially lead to a longer time of contact, which could increase the level of retaining of pollutants by sorption. Furthermore, two experiments have been conducted in order to find a connection between the reduction of the hydraulic conductivity and the amount of sediment added. The first experiment
examines the pressure gradient over time and is measured by having a constant flow of water and a constant dosing of sediment. The other experiment tests the effect of the drainage of the water and if the smaller particles in the sediment are pulled further into the filter or whether they still form a clogging layer on top of the filter column. This is tested by scanning intact samples after test runs and the scanning indicates a deeper layer of clogging. The results from these experiments, together with observations and measurements from an infiltration bed in Bredballe are summarised in a numeric infiltration model, which describes water levels and infiltration quantities. Based on the experiments with reducing the conductivity, there is also incorporated a module which reduces the hydraulic conductivity based on load of substance per area of filter material. A calibrated model would be able to predict the intervals of maintenance needed to preserve a certain hydraulic conductivity. As there are no long-term data series of the hydraulic conductivity available for the infiltration bed in Bredballe, a Monte Carlo analysis has been conducted to illustrate the uncertainty in the numerical infiltration model.