بررسی مقایسه‌ای تغییرات کاربری اراضی با استفاده از مدل زنجیره‌ی مارکوف و سلول‌های خودکار(CA- Markov) در حوزه‌ی آبخیز خیاوچای مشگین‌شهر

One of the most important steps towards sustainable development is to protect the integrity of the land. Every year, a portion of the land changes, and removing such land from the production cycle causes irreparable damage. Since land use changes in the Khiav-Chai watershed are of great importance due to the specific conditions of this area, studying the spatial and temporal changes in land use provides valuable information to planners and managers for precise planning. Unplanned and unprincipled land use changes are considered significant and fundamental challenges for any country, and in turn, have highly destructive impacts on natural resources. Therefore, it is crucial to study and predict changes in land use. To model the land use changes in the study area, satellite images were used from Landsat 5 with the TM sensor and Landsat 8 with the OLI-TIRS sensor. By applying atmospheric corrections and using the supervised classification method with the maximum likelihood algorithm, the existing land uses in the region were classified into 6 categories. To generate the predicted land use change map for the year 2023, the land use maps of 1989 and 2007 were used as base and forward maps, respectively, and were input into the Markov chain model to predict the land use changes in 2023. Conditional probability plots, area transition matrix, and land use transition probability matrix were generated after modelling. Finally, the predicted map for the year 2023 was extracted using the STCHOICE tool. A fundamental limitation of the Markov chain in producing land use prediction maps is its inability to incorporate spatial information into the modelling process. To address this, cellular automata (CA_Markov) were integrated into the model to add the spatial element. The predicted map for 2023, including the spatial component, was generated by combining the cellular automata model with the Markov model and introducing the 1989 base land use map, the transition area matrix, and the conditional probability images. Then, land use maps for the next three decades (2033, 2043, and 2053) were predicted using the 1989 and 2023 land use maps. The accuracy of the maps was evaluated using the kappa coefficient and overall accuracy, while the validity of the maps was evaluated using the agreement and disagreement parameters. The satellite images of the years 1989, 2007, and 2023 for the study area were classified using the maximum likelihood method, and the land use map was extracted. It was found that the largest areas were occupied by rangeland, bare soil, and dry farming. The accuracy of satellite image classification was evaluated using an error matrix, and the kappa coefficient and overall accuracy were calculated as 72% and 82.87% for 1989, 83% and 88.40% for 2007, and 88% and 92.32% for 2023, respectively. Based on these results, it can be concluded that the classification accuracy of the images was acceptable. The analysis of land use changes in the study area showed that 119.7 ha of urban land, 354.42 ha of irrigated agriculture, 1039.05 ha of dryland agriculture, 2024.73 ha of bare soil, 3829.95 ha of rangeland, and 458.01 ha of snow cover remained unaffected between 1989 and 2007. Similarly, 123.12 hectares of urban land, 383.04 hectares of irrigated agriculture, 1282.32 hectares of dry agriculture, 2294.64 hectares of bare soil, 3704.04 hectares of rangeland, and 806.22 hectares of snow cover remained unchanged between 2007 and 2023. The area of land use in the predicted map for 2023 showed that urban land occupied 4.3%, irrigated agriculture 5.7%, dry farming 15.7%, bare soil 23.2%, rangeland 42.9%, and snow cover 8.2% of the study area. The accuracy evaluation results of the model showed an agreement of 0.84 and a disagreement of 0.16 between the predicted and actual maps, with a kappa coefficient of 0.88, demonstrating the model's relatively high ability to predict changes. Comparing the land use maps for 2033, 2043, and 2053 with the land use map for 2023, it was determined that over the next three decades, urban land use would increase by 45.9%, 46.9%, and 47.5%, and rangeland would increase by 10.9%, 7.6%, and 4.5%, respectively. In contrast, irrigated agriculture would decrease by 27.9%, 21.1%, and 14.9%, and dryland agriculture would decrease by 13.4%, 11.23%, and 9.3%, respectively.