Occurrence and fate of micropollutants in surface water are of growing concern. To take adequate actions, it is essential to rely on representative data. Therefore, it is necessary to assess uncertainties related to the different steps of the analytical procedure. In this work, Pierre Gy's theory of sampling is applied to micropollutants concentrations in surface water for three case studies. These cases are: a large river (Rhône River, discharge of 182.5 m3/s), a large wastewater treatment plant (WWTP, discharge of 0.5 m3/s) and a small hospital sewage network (CHUV, discharge of 0.006 m3/s). Micropollutants studied are atrazine, benzotriazol, carbamazepine, diclofenac, mecoprop, mepivacaïne, paracetamol, pymetrozin and sulfamethoxazol. For dissolved micropollutants the global sampling uncertainty is about 45% for the Rhône River and about 35% for the WWTP and the CHUV. The difference is explained by lower concentration in the river leading to greater analytical uncertainties. Sampling frequency and sampling volume have little effects on these uncertainties. They are rather related to discharge measurement (weighting error), instruments and their use (materialization error) and analytical error. For sorbed micropollutants, uncertainties of 2'000%, 140% and 160 % were calculated for the Rhône River, the WWTP and the CHUV respectively. These uncertainties are related to sampling volumes not large enough. They are easily reduced by increasing these volumes. Discharge is expected to affect uncertainties mainly if it affects micropollutant concentrations. In some cases, measured concentrations are not representative of the true concentration and can lead to misleading conclusions. To improve data reliability, sampling system could be enhanced. Moreover, sampling dissolved and sorbed pollutants separately could be beneficial.