The suitability of caffeine as a chemical marker for surface water pollution by domestic wastewaters was assessed in this study. Caffeine concentrations in influents and effluents of Swiss wastewater treatment plants (WWTPs, 7-73 and 0.03-9.5 microg/L, respectively) indicated an efficient elimination of 81-99.9%. Corresponding loads in untreated wastewater showed small variations when normalized forthe population discharging to the WWTPs (15.8 +/- 3.8 mg person(-1) d(-1)), reflecting a rather constant consumption. WWTP effluent loads were considerably lower (0.06 +/- 0.03 mg person(-1) d(-1)), apart from installations with low sludge age (< or = 5 d, loads up to 4.4 mg person(-1) d(-1)). Despite the efficient removal in most WWTPs, caffeine was ubiquitously found in Swiss lakes and rivers (6-250 ng/ L), except for remote mountain lakes (<2 ng/L; analytical procedure for wastewater and natural waters: SPE, GC-MS-SIM or GC-MS-MS-MRM, internal standard 13C3-labeled caffeine). Caffeine concentrations in lakes correlated with the anthropogenic burden by domestic wastewaters, demonstrating the suitability of caffeine as a marker. A mass balance for Greifensee revealed that approximately 1-4% of the wastewaters had been discharged without treatment, presumably on rainy days when the capacity of WWTPs had been exceeded. For Zürichsee, it could be shown that the monthly inputs of caffeine correlated with precipitation data. The depth- and seasonal-dependent concentrations in this lake were adequately rationalized by a numerical model considering flushing, biodegradation, and indirect photodegradation via HO. radicals as elimination processes and caffeine inputs as fitting variables.