Landsat Time-Series for Land Use Change Detection Using Support Vector Machine: Case Study of Javanrud District, Iran

Abstract

Changes in land use/land cover (LULC) affect the natural ecosystem and environment that can cause changes in soil resources, biodiversity, ecological function, and environmental components. Therefore, recognizing the trends of LULC changes plays an essential role in natural resources planning and management. This research aims to investigate the spatial and temporal variations of LULC using remote sensing and geographic information system in Javanrud district, Iran. For this aim, Landsat satellite imageries including Thematic Mapper (TM) of 2000, and 2010, and Operational Land Imager (OLI) of 2018, were acquired in the land use/cover changes in Javanrud district were analyzed. Geometric correction, topography correction, and radiometric correction were implemented to enhance the accuracy of the images. Finally, the support vector machine (SVM) method was employed to extract the LULC classes, and the images were categorized into five different classes, namely build-up, farmland, rangeland, woodland, and bare land. The results indicate that there was an increase in urban, farmland, and bare land types, while the extent of rangeland and woodland decreased. During the last 18 years, build-up land has been increased from 4.3km2 in 1990 to 6.4 km2 in 2018, farmlands and bare lands have seen an increase of about 34.22 and 16.86 km2, respectively, while rangelands and woodlands have been decreased by 15.93 km2, and 68.51 km2, respectively. Demand for the expansion of agricultural land, fuelwood and construction materials along with natural factors such as drought, trees disease, and forest fires are the major driving forces for the forest cover changes in this region.