Marine fish farming is an important commercial practice in Hong Kong. Marine fish farms located in eutrophic coastal waters often face the threat of severe dissolved oxygen depletion associated with algal blooms and red tides. On the other hand, mariculture activities also contribute to pollution. The sustainable management of mariculture requires proper siting of the fish farms and stocking density control. Both of these are related to the carrying capacity of the water body concerned, which is mainly governed by its flushing characteristics. A simple method to determine the carrying capacity of a fish farm has been developed by using three-dimensional (3D) hydrodynamic modelling and its effective coupling with a diagenetic water quality model. A systematic methodology using numerical tracer experiments has been developed to compute the tidal flushing in a fish farm. The flushing time is determined from the results of a numerical tracer experiment using robust 3D hydrodynamic and mass transport models. A unit tracer concentration is initially prescribed inside the region of interest and zero elsewhere; the subsequent mass transport and the mass removal process are then tracked. The fish farms are usually situated in well-sheltered shallow embayments and may not connect directly to the open water. It is found that it is necessary to define both "local" and "system-wide" flushing times to represent the effectiveness of the mass exchange with the surrounding water body and the open sea respectively. A diagenetic water quality model simulating the sediment-water-pollutant interaction is employed to address the response of the water column and the benthic layer to pollution discharges. With the flushing rate reliably computed, the carrying capacity of the fish farm can be determined in terms of key water quality parameters: chlorophyll-a, dissolved oxygen, organic nitrogen and potential lowest dissolved oxygen level on a day of negligible photosynthetic production. The predictions are well-supported by field data.