Probability of detecting and quantifying faecal contaminations of drinking water by periodically sampling for E. coli: a simulation model study.

Drinking water supply companies monitor the presence of Escherichia coli in drinking water to verify the effectiveness of measures that prevent faecal contamination of drinking water. Data are lacking, however, on the sensitivity of the monitoring programmes, as designed under the EU Drinking Water Directive. In this study, the sensitivity of such a monitoring programme was evaluated by hydraulic model simulations of contamination events and calculations of the detection probability of the actual sampling programme of 2002. In the hydraulic model simulations of 16-h periods of 1l h(-1) ingress of untreated domestic sewage, the spread of the contamination through the network and the E. coli concentration dynamics were calculated. The results show that when large parts of the sewage reach reservoirs, e.g. when they originate from the treatment plant or a trunk main, mean detection probabilities are 55-65%. When the contamination does not reach any of the reservoirs, however, the detection probability varies from 0% (when no sampling site is reached) to 13% (when multiple sites are reached). Mean detection probabilities of nine simulated ingress incidents in mains are 5.5% with an SD of 6.5%. In reality, these detection probabilities are probably lower as the study assumed no inactivation or clustering of E. coli, 100% recovery efficiency of the E. coli detection methods and immediate mixing of contaminations in mains and reservoirs. The described method provides a starting point for automated evaluations and optimisations of sampling programmes.