This paper describes the development and application of a modified Grouping Genetic Algorithm (GGA) used to identify sets of optimal ambulance locations. The GGA was modified to consider a special case with only two groups, and the reproduction and mutation schemes were modified to operate more efficiently. It was applied to a case study locating ambulances from a fixed set of alternative locations. The sites were evaluated using data of emergency medical services (EMS) calls summarised over census areas and weighted by network distance. Census areas serviced by the same selected location defined ambulance catchments. The results indicated alternative sites for ambulances to be located, with average EMS response times improved by 1 minute 14 seconds and showed the impacts of having different numbers of ambulances in current locations and in new locations. The algorithmic developments associated with the modified GGA and the advantages of using census areas as spatial units to summarise data are discussed.