Collection and traortation of municipal solid wastes (MSW) is one of the key functional elements in integrated solid waste management, which accounts for a considerable portion of the associated costs. Nowadays, increasing population (specially in developing countries), development of urbanization, and advancement of technology and industries along with environmental awareness have made the collection and traortation of MSW as one of the most challenging problems faced by municipal authorities and decision makers. Review of literature shows that one of the important categories of vehicle routing problems in which demand points are located on the network arcs is the capacitated arc routing problem (CARP). In order to develop a model to solve this problem a combined method based on geographic information system (GIS) was employed. This combined method consisted of linear programming model with integer variables for estimation of number and type of fleet vehicles and costs, a mathematical programming model to divide the study area into sub areas (partitions), and a heuristic model for vehicle routing in each partition. Computer codes were developed using Matlab for vehicle routing and obtaining some of the input parameters for partitioning. Other computer coded were developed using Lingo for estimation of vehicle fleet and partitioning. Capabilities of Arc GIS were employed to develop the network, to analyze the network and to establish the input parameters for heuristic method. The application of this combined method was investigated by analysis of a case study. The study area was Shazand city, located in the south west of Markazi province. This city has a population of about 20000 people and average MSW production rate of 10000 kg per day. In order to assess the performance of the model 3 scenarios were considered. Results of optimization showed a 16 percent reduction in total costs and less geo spreading compared to the existing system. In the third scenario it was assumed that all vehicles are permitted to use all arcs. Results showed an 18 percent reduction in total costs compared to existing system and 26 percent reduction in geo spreading compared to the second scenario. Overall, the results showed the capability and benefits of utilizing this combined model compared with traditional and empirical methods in different real and applied cases.