Emergency response systems in urban areas should be located to ensure adequate coverage and rapid response time. We develop a model for locating emergency vehicles on urban networks considering both spatial and temporal demand characteristics such as the probability that a server is not available when required. We also consider that service rates are not identical but may vary among servers and are dependent upon incident characteristics: corresponding districting and dispatching problems are also integrated in the location model. The model is applied using real data for locating freeway service patrol vehicles and results are compared with existing coverage and median models. Results show improvements in the mean response time particularly in cases of high demand for intervention when compared to 'traditional' models.