Conceptual modelling of E. coli in urban stormwater drains, creeks and rivers

Accurate estimation of faecal microorganism levels in water systems, such as stormwater drains, creeks and rivers, is needed for appropriate assessment of impacts on receiving water bodies and the risks to human health. The underlying hypothesis for this work is that a single conceptual model (the MicroOrganism Prediction in Urban Stormwater model – i.e. MOPUS) can adequately simulate microbial dynamics over a variety of water systems and wide range of scales; something which has not been previously tested. Additionally, the application of radar precipitation data for improvement of the model performance at these scales via more accurate areal averaged rainfall intensities was tested. Six comprehensive Escherichia coli (E. coli) datasets collected from five catchments in south-eastern Australia and one catchment in Raleigh, USA, were used to calibrate the model. The MOPUS rainfall-runoff model performed well at all scales (Nash-Sutcliffe E for instantaneous flow rates between 0.70 and 0.93). Sensitivity analysis showed that wet weather urban stormwater flows can be modelled with only three of the five rainfall runoff model parameters: routing coefficient (K), effective imperviousness (IMP) and time of concentration (TOC). The model's performance for representing instantaneous E. coli fluctuations ranged from 0.17 to 0.45 in catchments drained via pipe or open creek, and was the highest for a large riverine catchment (0.64); performing similarly, if not better, than other microbial models in literature. The model could also capture the variability in event mean concentrations (E = 0.17–0.57) and event loads (E = 0.32–0.97) at all scales. Application of weather radar-derived rainfall inputs caused lower overall performance compared to using gauged rainfall inputs in representing both flow and E. coli levels in urban drain catchments, with the performance improving with increasing catchment size and being comparable to the models that use gauged rainfall inputs at t