Soutenance de thèse - Surveillance et modélisation des produits pharmaceutiques dans les eaux usées: charges quotidiennes et horaires dans les eaux usées hospitalières et urbaines

Daily and hourly loads of 15 pharmaceutical molecules at the inlet of a wastewater treatment plant have been measured over 3 years and modelled for both an urban catchment of 16 000 inhabitants and a hospital of 450 beds. Some molecules are never or rarely quantified. Daily loads range from 0.6 to 564 g/day depending of the molecule and the 24 h measurement campaign. Seasonal or weekly patterns are not identified. Pharmaceuticals hourly loads dynamics are distinctive from one another and from wastewater flow. The measured hourly loads are severely impacted by the random behaviour of the patients when the daily mass consumed is low. Thus, the average dynamics is difficult to identify. The main hypothesis to model pharmaceuticals loads in wastewater is that they result from the following steps: pharmaceuticals sales or distributions, human consumption, metabolism and excretion. Pharmaceuticals sales for the urban catchment and distribution for the hospital have been collected at different space and timescales (respectively 1, 6 and 223 pharmacies and daily, weekly and monthly). Larger scales are more reliable for magnitude but the variability of the smaller ones is closer to the variability observed in the measurements. The quantities of pharmaceuticals sold or distributed range from 0.4 to 1 600 theoretical patients per day. Associating measured daily loads with sales or distributions, no linear correlation is found. A minute time step stochastic model is proposed and applied to both sites. It produces reliable and accurate results for both daily and hourly loads. However, results are difficult to interpret when only a few patients are consuming a pharmaceutical. Also, the model does not reproduce the inherent specificity of the hospital. In addition, the model is also able to predict the domestic wastewater flow of an urban catchment with great accuracy for both daily volumes and dynamics.