This paper reports on a statistical review of a water quality monitoring programme aimed at estimating long-term pollutant loads discharged from waterways in and around Melbourne, Australia. Performance of the current programme was evaluated with respect to the required level of uncertainty to meet management objectives. Use of continuously measured turbidity was found to be an effective surrogate measure for estimating TSS, with errors in long-term load estimates of less than 5%. Where routine grab sampling was used instead, errors increased with sampling interval; a 3-day interval was required to maintain errors within 10% of the continuously measured load. For storm event sampling, auto-samplers were found not to be required, if only long-term load estimates are required. The importance of eliminating systematic errors, by ensuring frequent calibration and data verification, was demonstrated.