تحلیل و پیش‌بینی پویایی‌های فضایی–زمانی تغییرات کاربری و پوشش اراضی در حوضه آبریز رودخانه کشف‌ رود

A B S T R A C T
Land use and Land cover (LULC) changes play a crucial role in the environmental dynamics and natural resource sustainability of arid and semi-arid regions. Driven by both natural and anthropogenic factors, these changes lead to alterations in ecosystem structure and function. In this study, past changes and future dynamics of LULC in the Kashafrud River Basin were analyzed and projected using a hybrid SVM–CA–Markov modeling approach. LULC maps extracted from GlobeLand30 data for the years 2000, 2010, and 2020 were classified into six categories: agricultural lands, forests, grasslands and shrublands, water bodies and wetlands, built-up areas, and barren lands. Transition probabilities among land cover classes were estimated using the Markov Chain model, while change potential maps were generated utilizing the Support Vector Machine (SVM) algorithm based on driving factors. Subsequently, the Cellular Automata–Markov (CA–Markov) model was employed to simulate spatial patterns and predict LULC distribution for the year 2037. Model validation demonstrated high accuracy, yielding a Kappa coefficient of 0.88 and an Area Under the Curve (AUC) of 0.946. The results indicate an expansion of agricultural lands and built-up areas coupled with a decline in natural land covers, particularly forests, throughout the study period; future projections confirm the continuation of this trend. These findings underscore the necessity for integrated land-use planning and sustainable natural resource management in the Kashafrud River Basin to mitigate environmental degradation and ensure ecological sustainability.
Extended Abstract
Introduction
Land Use and Land Cover (LULC) change, driven by natural processes and anthropogenic activities, serves as a primary indicator of environmental transformation, threatening sustainability by fundamentally altering ecosystem structure. These impacts are intensified in arid and semi-arid regions with limited resource availability, where ecological vulnerability is inherently high. Analyzing the spatio-temporal dynamics of LULC provides essential insights into past landscape transformations and supports the prediction of future land-use scenarios. Such analyses contribute to improved land-use planning and sustainable management of natural resources. Advances in remote sensing technologies and Geographic Information Systems (GIS) have considerably enhanced monitoring capabilities. These tools now enable more accurate modeling of land-use dynamics. Among available modeling approaches, Cellular Automata–Markov (CA–Markov) models have been widely applied to simulate spatial and temporal patterns of LULC change. However, incorporating machine learning techniques can further improve simulation accuracy and spatial realism. The Support Vector Machine (SVM) algorithm has demonstrated strong capability in modeling nonlinear relationships between land-use changes and their driving factors and in generating reliable transition potential maps. The Kashafrud River Basin in northeastern Iran is an environmentally sensitive region characterized by an arid and semi-arid climate, intensive land-use pressures, and its strategic importance for regional water supply. Despite numerous studies conducted within the Mashhad metropolitan area, comprehensive analyses at the basin scale remain scarce. Therefore, the objective of this study is to analyze spatio-temporal LULC dynamics in the Kashafrud River Basin and to simulate future land-use patterns using an integrated SVM–CA–Markov modeling framework.
 
 
 
Methodology
Land Use/Land Cover changes within the Kashafrud River Basin were analyzed utilizing GlobeLand30 datasets for 2000, 2010, and 2020 with a spatial resolution of 30m. The datasets were reclassified into six major land‑use categories, including agricultural lands, forests, grasslands and shrublands, water bodies and wetlands, built‑up areas, and barren lands. Transition probabilities between land‑use classes were estimated using the Markov chain module in the TerrSet software. Transition potential maps were generated using the SVM algorithm based on the selected driving variables, including distance to power transmission lines, distance to main roads, distance to rivers, slope, elevation, distance to industrial areas, and distance to faults. A total of 3,000 stratified random samples were used for model calibration and validation, and model parameters were optimized using the Radial Basis Function (RBF) kernel to enhance classification performance. Future land‑use patterns were simulated using the CA–Markov model with a 5×5 neighborhood configuration. Model accuracy was assessed by comparing the simulated 2020 LULC map with the observed map using the Kappa coefficient and the Area Under the Curve (AUC). After validation, land‑use patterns were projected for the year 2037.
 
Results and Discussion
The results of land use and land cover change analysis indicated that the Kashafrud River Basin experienced considerable transformations during the period from 2000 to 2020. All through this period, agricultural lands and built-up areas showed increasing trends, whereas natural land covers, particularly forests, as well as grasslands and shrublands, exhibited decreasing trends. This pattern reflects the increasing pressure of human activities on the natural resources of the basin. Model validation results indicated that the hybrid SVM–CA–Markov framework achieved high simulation accuracy. The Kappa coefficient reached 0.88 and the AUC value was 0.946, demonstrating strong agreement between simulated and observed land-use patterns. Simulation results suggest that built-up areas will expand significantly by 2037, increasing from approximately 388 km² in 2020 to more than 766 km². This expansion reflects rapid urban growth and infrastructure development within the basin, particularly in areas influenced by the Mashhad metropolitan region. Agricultural lands are projected to increase from approximately 4,448 km² to about 5,850 km², primarily through the conversion of barren lands into cultivated areas. Conversely, barren lands are expected to decrease from approximately 11,166 km² to about 9,494 km². Natural land covers are projected to decline further. Forest areas are expected to decrease from approximately 330 km² in 2020 to about 253 km² in 2037, while grasslands and shrublands are also predicted to follow a decreasing trend. These projected changes indicate that continuation of current land-use trajectories may increase pressure on soil and water resources, reduce infiltration capacity, increase surface runoff, and accelerate soil erosion processes. Moreover, degradation of natural vegetation may result in biodiversity loss and reduced ecological resilience across the basin.
 
Conclusion
The results indicate that the Kashafrud River Basin has undergone substantial land-use transformations over recent decades, largely driven by the expansion of agricultural and built-up areas and the decline of natural land covers. Future projections suggest that these trends are likely to continue, potentially threatening the ecological sustainability of the basin. The hybrid SVM–CA–Markov model demonstrated reliable performance in simulating LULC dynamics and successfully reproduced spatial patterns with high accuracy. Integration of machine learning techniques with spatio-temporal modeling approaches improved prediction capability and enhanced the robustness of simulation results. The findings provide a scientific basis for land-use planning and natural resource management in the Kashafrud River Basin. Application of the modeling results in urban development policies, water resource management strategies, and conservation programs can contribute to sustainable development in the region.
 
Funding
This research was financially supported by Ferdowsi University of Mashhad (Dr. Ali Shariati Faculty of Letters and Humanities) under Grant No. 61287. This funding included a thesis grant for the first author and financial support for the co-authors.
 
Authors’ Contribution
The authors’ contributions to this research are as follows:
The first author was responsible for conceptualization, methodological design, data collection and analysis, drafting the manuscript, and final editing.
The second, third, and sixth authors provided supervision of the research, validation of findings, critical review of the content, and approval of the final version.
The fourth author contributed to the development of the theoretical framework and structural editing; and the fifth author participated in preparing the analysis, discussion, and conclusion sections.
 
Conflict of Interest
Authors declared no conflict of interest.
 
Acknowledgments
We are grateful to all the scientific consultants of this paper.