Representative input load of antibiotics to WWTPs: Predictive accuracy and determination of a required sampling...

Predicting the input loads of antibiotics to wastewater treatment plants (WWTP) using certain input data (e.g. prescriptions) is a reasonable method if no analytical data is available. Besides the spatiotemporal uncertainties of the projection itself, only a few studies exist to confirm the suitability of required excretion data from literature. Prescription data with a comparatively high resolution and a sampling campaign covering 15 months were used to answer the question of applicability of the prediction approach. As a result, macrolides, sulfamethoxazole and trimethoprim were almost fully recovered close to 100% of the expected input loads. Nearly all substances of the beta-lactam family exhibit high elimination rates during the wastewater transport in the sewer system with a low recovery rate at the WWTP. The measured input loads of cefuroxime, ciprofloxacin and levofloxacin fluctuated greatly through the year which was not obvious from relatively constant prescribed amounts. The latter substances are an example that available data are not per se sufficient to monitor the actual release into the environment. Furthermore, the extensive data pool of this study was used to calculate the necessary number of samples to determine a representative annual mean load to the WWTP. For antibiotics with low seasonality and low input scattering a minimum of about 10 samples is required. In the case of antibiotics exhibiting fluctuating input loads 30 to 40 evenly distributed samples are necessary for a representative input determination. As a high level estimate, a minimum number of 20-40 samples per year is proposed to reasonably estimate a representative annual input load of antibiotics and other micropollutants.