The United States Environmental Protection Agency models CORMIX2 and UM were used to predict the ambient temperature rises from thermal discharges near Charlestown, Rhode Island, and Jamesport, Long Island, New York. Observations from two scaled model laboratory studies conducted at Alden Research Laboratories Inc. (ARL) were compared with temperature rise predictions from two numerical models. For the ARL diffuser at Jamesport, both numerical models underestimated the observed temperature rises. In the Charlestown study, UM produced results that were in good agreement with the temperature rises observed, whereas CORMIX2 underpredicted observations but remained within the model's stipulated ±50% error. However, UM was unable to model the plume shape properly, since it is primarily intended for alternating diffusers. The predicted wastefield shapes of CORMIX2 were similar to those observed in the laboratory but of greater surface area. Key words: multiport diffusers, thermal discharges, initial dilution.