Investigation of the Support Vector Machine (SVM) algorithm for land use changes, a case study of Kerman Province, Iran

Document Type : Research Paper

Authors

1 Bachelor student, Shahid Bahonar University of Kerman, Civil Engineering Department

2 Graduate University of Advanced Technology

Abstract
Land use and land cover (LULC) play a crucial role in studying the ongoing worldwide environmental changes This study employs TM, ETM+, and OLI images acquired from the Landsat satellite over a thirty-year period (1990-2020) to analyze changes in land use. The Support Vector Machine (SVM) algorithm is utilized for land use classification, and a Median filter in the ENVI software was applied to the SVM algorithm results to produce smoother images. The study area is located in Kerman Province, Iran, encompassing the cities of Kerman, Zarand, Bardsir, and parts of Rafsanjan, Ravar, and Sirjan counties. An area with hot and dry climate. Due to the vastness of the consider region and the difficulty of detailed investigation owing to the decrease in accuracy four classes were selected for SVM classification: built-up areas, vegetation, water resources, and bare land (comprising soil and rock formations). The results indicated that the classification of OLI data performs better in terms of overall accuracy (more than 80%) and the kappa coefficient (more than 0.7). Specifically, the analysis revealed a 1.88% decrease in vegetation cover, a 0.19% increase in urban areas, and a 0.005% change in surface water bodies over the 30-year period. These changes are attributed to the influence of human activities, drought, poor management practices, and climate change on the dynamics of land use in the region.

Keywords

Subjects