Efficiency of different land use/land cover mapping methods in Kasilian representative watershed

In line with the study of land use and land cover, remote sensing technology has been welcomed by many researchers as a source of spatial information production and suitable tools, which shows the accuracy and validity of these maps. The purpose of this research is to evaluate and compare the accuracy of preparing land use maps using two methods of remote sensing and one method of visual interpretation of Google Earth images in the Kasilian representative watershed. In this research, after taking educational samples using Google Earth software and implementing them on the Landsat 9 image of 2021, classification of images was done in ENVI software, and the land use map was prepared based on training samples, Neural Network and SVM methods. In the method of visual interpretation, all land uses in Google Earth images were manually digitized and a land use map was obtained. Then, the accuracy of the map was checked for all three methods and the results showed that the map obtained from visual interpretation of Google Earth images with overall accuracy and Kappa coefficient of 100% was in agreement with the ground reality compared to Neural Network and SVM methods with overall accuracies of 87.6% and 88.2% and Kappa coefficients of 76% and 77.8%, respectively. However, due to the time-consuming visual interpretation method, especially for large watersheds, and the acceptable accuracy of Neural Network and SVM methods, it is suggested to use advanced methods to prepare land use maps, especially in large watersheds.