Wetland cover change detection using multi-temporal remotely sensed data

Lake Urmia, located in northwest Iran, contains a number of wetlands significantly affecting the environmental, social, and economic conditions of the region. The ecological condition of Lake Urmia has degraded during the past decade, due to climate change, human activities, and unsustainable management. The poor condition of the lake has also affected the surrounding wetlands. This study analyzes the land cover change of one of the wetlands in the southern part of Lake Urmia, known as Ghara-Gheshlagh wetland, in the period 1989–2015 using post-classification change detection and machine learning image classification. For this analysis, three Landsat images, acquired in 1989 (TM), 2001 (TM), and 2015 (Landsat-8), were used for the classification and change detection. Support vector machine learning algorithm, a supervised learning method, is employed, and images are classified into four main land cover classes namely “water,” ”barren,” “salty land,” and “agriculture and grassland.” Change detection was carried out for pairs of years 1989 to 2001 and 2001 until 2015. The results of this classification show that there is a sharp increase in the area of salt-saturated land as well as a decrease in the area of water resources. Overall classification accuracy obtained were high for the individual years: 1989 (91.48%), 2001 (90.63%), and 2015 (88.6%). Also, the Kappa coefficients for individual maps were high: 1989 (0.89), 2001 (0.8742), and 2015 (0.84). After that, the land cover change map of the study area is obtained between 1989 to 2001 and then 2001 to 2015. The results of this analysis suggest that more efforts should be taken to effectively manage water resources in the region and point to potential locations for focused management actions within the wetland area.