Abstract: The simulation of land use change in mountainous urban fringes is always a difficult problem. Existing methods, based on the traditional cellular automata (CA)-Markov coupling model, to capture and simulate such changes in this special areas suffer from notable deficiencies. In this paper, we explore, based on a case study of a mountainous urban fringe in Southwestern China, the performance of an improved method that combines the traditional CA-Markov model with multi-criteria evaluation (MCE). We develop an MCE based AHP model fitted by constraining conditions and factors in the study area. The MCE output is then used as the input of the CA-Markov model to simulate land use changes in the study area based on three periods (2006, 2011, and 2016). Meanwhile, the simulated results by the traditional CA-Markov model are also obtained. Both land use simulation results (the simulation images) are then compared against the observed land use in 2016 (the existing image in 2016) to examine the performance of the improved CA-Markov method. The results show that: 1) Establishing the missing link between the MCE process and the conventional CA-Markov model using land use suitability maps can significantly improve the performance of conventional CA-Markov model in mountainous urban fringes. The results from the case study show that the overall 93.15% simulation accuracy is much higher than that of the average 80% accuracy from most traditional CA-Markov models used in other similar areas. 2) The case study further indicates that from 2006 to 2016, the areas of cultivated land, garden, woodland, grassland, water body, and unused land in study area are decreasing annually. By contrast, the areas of urban land and rural settlement are increasing. The types with the most changes are cultivated land and urban land. This situation indicates that the rapid urbanization in mountainous urban fringe is still a mode of denotation expansion, in which a large amount of agricultural land, especially cultivated land, is occupied by urban construction. 3) The simulation shows that if the current rapid urbanization rate is maintained, then by 2030, the areas of cultivated land , garden , woodland , water body , unused land and rural settlement will be decreased by 6442.46, 923.53, 3239.17, 72.55, 41.26, 352.07 hm2, respectively. The increased land types mainly include urban land and grassland. We also argue the implications of the findings and suggest areas of further work.