Land use maps are among the most important information required layers in the managements of the catchments. Furthermore, utilizing forecasting methods of land use changes is of great importance in management and decision making of watersheds. In this regard, the aim of this research is the integrated use of Markov and Cellular automata models in the Multiple Criteria Decision Making (MCDM) method to model the land use changes. Those investigations were evaluated in the Talar Watershed in the north of Iran. The reason for selecting this site was the rapid changes occurred in this watershed in the recent years. Moreover, availability and completeness of the required data of this site is the other reason. Then, two Landsat 5 Thematic Mapper (TM) images of 2000 and 2010 with spatial resolution of 30 m and 1 SENTINEL-2 Multispectral Instrument (MSI) image of 2016 with spatial resolution of 10 m were prepared. The last image was used to evaluate the predicted land use based on the defined scenario. Furthermore, other required information layers including roads, villages, settlements, and protected areas maps were gathered from department of environment of Iran. Moreover, Digital elevation model (DEM) of Shuttle Radar Topography Mission (SRTM) was the other utilized information layer. Initially, geometric and radiometric corrections were accomplished on Landsat 5 TM and SENTINEL-2 MSI images. The land use maps of 2000 and 2010 were produced using maximum likelihood classification of those corrected images. Moreover, required layers for producing transition potential layer were prepared based on the available information. At the next satep, a scenario for forecasting land use changes was utilized to produce land use maps of the selected dates. This scenario was based on the increasing bare lands or degraded forests, increasing farmlands, and increasing settlement areas. Then, transition potential maps were produced using implementation of fuzzy transforms, weighted linear combination, and multiple criteria decision making methods. In this step, Markov chains models were utilized to generate transition probability matrix which is an indicative of the number of changing pixels and their area in the considered time range of the study. Finally, land use maps of 2016 and 2030 were produced by cellular automata using a neighborhood filter of 5 pixels. Validation of forecasted land use map of 2016 was performed by the classified SENTINEL-2 MSI image of that year. The validations results represented kappa coefficient of 0.84 which is a good result. Hence, the results suggested a good performance of the implemented model for forecasting land use changes of the study area. Furthermore, land use map of 2030 revealed that the settlement and bare land classes will increase significantly in that year. On the other hand, dense forest and degraded forest classes will be decreased which the last one can be due to transition of this class to the bare land class. Therefore, integrated use of Markov and Cellular automata models along with MCDM method can be lead to forecast the land use changes with an acceptable performance. Moreover, the results proved the proper performance of the selected methods.