A two-dimensional, depth averaged, computer code was used to model salinities for the Sabine-Neches Estuary in Southeast Texas, USA. The model bathymetry was set to the mean sea level of 1996, and channel friction was adjusted to give an optimised Manning's n value for the annual, 6-minute water level data at eight stations. The simulated velocities were checked using flow discharges measured at six estuary sections for a three-day survey period. Three dispersion methods were used to calibrate the model for salinity. The data used were a 6 month record for 1996 at seven estuary stations, and a 7-year partial record from 1990 to 1996 for two stations at the Upper and Lower ends of Sabine Lake. The first method used a constant salinity dispersion coefficient and the second used a velocity related coefficient; they were accurate to within an annual mean value of about 5 ppt. For the third method, a two dimensional dispersion function was derived, which modelled to within an annual mean value of about 1 ppt. The ultimate purpose for the study was to estimate salinity increase for Sabine Lake with designed, fresh water flow extractions from the catchment. A maximum salinity increase of about 2 ppt for a 50 year return period was estimated.