This study aimed to analyze the spatial–ecological dynamics of land use/land cover in Shush County over a 33-year period (1990–2023) using an integrated remote sensing and metaheuristic modeling framework. Landsat satellite imagery, along with spectral indices NDVI and NDBI, was employed for data extraction. A hybrid ACO–SVM algorithm was applied for land use classification, and spatial metrics were used to quantify landscape changes. The classification accuracy was validated with a Kappa coefficient of 0.94.

The results indicate substantial transformations in land use patterns over the study period. Rangelands experienced a continuous decline at an annual rate of 8.74 km², while irrigated agricultural lands decreased by approximately 30% overall. In contrast, dryland agriculture expanded by 51%, and built-up areas increased by 53%. Spatial metrics analysis revealed intensified landscape fragmentation in natural ecosystems, with a pronounced reduction in mean patch size (MPS), particularly in rangelands (17.88 hectares per year).

Furthermore, temporal analysis of vegetation and built-up indices showed a declining trend in NDVI and an increasing trend in NDBI, indicating vegetation degradation alongside urban expansion. These changes are primarily attributed to shifts toward rainfed cultivation systems, expansion of rural and urban settlements, and increasing pressure on water and soil resources.

The ecological consequences of these transformations include biodiversity loss, accelerated soil erosion, reduced ecosystem productivity, and declining sustainability of agriculture-based livelihoods. The study highlights the necessity of integrated land management strategies, including rangeland restoration, sustainable agricultural planning, and development policies aligned with ecological carrying capacity. The proposed framework demonstrates the effectiveness of combining remote sensing, machine learning, and spatial metrics for long-term land use monitoring and environmental assessment.

Keywords

Spatial–ecological dynamics, Land use/land cover, ACO–SVM metaheuristic algorithm, Shush County