Daily maximum temperature frequency distributions are usually modelled using a single normal distribution. For coastal localities especially, this often provides a poor approximation. Another approach is to represent the distribution as a combination of maritime and continental airstreams with each distinct airstream having a characteristic normal distribution. This approach yielded realistic results for southern Australian coastal sites. Model parameters were calculated using a downhill simplex method of approximation and the goodness-of-fit was measured by the level of significance, α, using the x2 test statistic. Levels of significance in most cases exceeded 0.010, and in some cases surpassed 0.50. By comparison, in most cases a single normal model was rejected at a critical level of 0.010. Modelling coastal daily maximum temperature frequencies using a composite of two normal distributions has a physical basis, a characteristic other methods of depiction often lack. Potentially, this model could assist in climatological and weather forecasting applications. It may be used to synthesize maximum temperature frequency analyses at poorly represented sites, to assess impact of changes in the local atmospheric circulation patterns that may be associated with climate change, and as input to a statistically based scheme used in prediction of daily maximum temperature.