Stormwater pollutant loads modelling: epistemological aspects and case studies on the influence of field data sets on calibration and verification.

In urban drainage, stormwater quality models have been used by researchers and practitioners for more than 15 years. Most of them were initially developed for research purposes, and have been later on implemented in commercial software packages devoted to operational needs. This paper presents some epistemological problems and difficulties with practical consequences in the application of stormwater quality models, such as simplified representation of reality, scaling-up, over-parameterisation, transition from calibration to verification and prediction, etc. Two case studies (one to estimate pollutant loads at the outlet of a catchment, one to design a detention tank to reach a given pollutant interception efficiency), with simple and detailed stormwater quality models, illustrate some of the above problems. It is hard to find, if not impossible, an "optimum" or "best" unique set of parameters values. Model calibration and verification appear to dramatically depend on the data sets used for their calibration and verification. Compared to current practice, collecting more and reliable data is absolutely necessary.