Modelling the removal of microorganisms by slow sand filtration

Abstract

Slow sand filtration (SSF) is used for the removal of pathogenic microorganisms during drinking water production. The Dutch government has a drinking water policy, which includes an infection risk limit of 1 infection per 10,000 persons per year. Therefore, the concentrations of microorganism in the drinking water have to be determined. The microorganism concentrations in the influent as well as in the effluent water of SSF are in most cases below the detection limit. A model was developed by the RIVM, which calculates the removal efficiency of SSF under different conditions. This model was developed based on the data from pilot-scale experiments, which use indicator organisms in concentrations far above the detection limit. We obtained a better understanding of the influence of the Schmutzdecke on the head loss in the full-scale slow sand filters. A correction of the head loss was conducted for the parameters discharge and water temperature. With this correction we obtained a corrected head loss which is solely influenced by the Schmutzdecke. Most of the full-scale filters showed a seasonal dependent development of the corrected head loss. The seasonal fluctuation could be caused by temperature dependent predation rates. Moreover, the value of the initial corrected head loss is related to the rate of head loss increase. A small part of the full-scale filters did not have a significant corrected head loss increase. This could be due to tunnel forming worms, Eisenia (annelids). To improve the current model for the removal of microorganisms, some adjustments to this model are conducted. No relation was found between the removal of microorganisms and the head loss. There was a relation found of the sticking efficiency with the grain size and the water velocity. This was ascribed to the ratio of available surface sites which are favourable for attachment. The sticking efficiency was correlated to the grain size to the power of 2.5 and inversely correlated to the water velocity to the power of 2.9; α=F2*dc2.5/v2.9. Where F2 is the sticking factor, which is 0.0037 for E.coli and 0.00022 for MS2. With these new insights, we changed the location specific model to an universal model for all locations. This new model has a very strong dependency on the water velocity compared to the reference model. Besides, there is a different dependency on the grain size compared with the reference model. A smaller grain size is not always better according to the removal for the new model.