Classification Lake Parishan water basin by random forest classification using Landsat satellite images

The global water scarcity crisis is affecting water management in various sectors, including water and agriculture. Aquatic basins and their surrounding areas have been encountered with serious challenges such as drying up of lakes and rivers, negative aquifer balance, changes in surrounding land use, increased cultivation of irrigated and horticultural lands, and changes in the pattern of cultivation from low-crop to high-water crops in recent years. Satellite imagery due to its wide spatial coverage, high resolution, low cost, rich time archive of satellite imagery and ease of use methods is a useful and efficient tool to help manage water and soil resources. In this study, four classes of soil, water and wet, urban and agricultural areas were selected. Then, two random forest classification methods and support vector machines are used to classify images. Classification methods were evaluated by calculating two indices of accuracy and Kappa coefficient using test data. The random forest classification in the four years, 2012, 2014, 2016 and 2018 and classification of support vector machines in two years, 2008 and 2010 have the most accuracy. Therefore, the random forest algorithm has worked well in separating the classes, especially in water basin, and can be used as a reliable method in this area.