Đánh giá biến động thảm phủ và sử dụng đất khu vực miền Trung dựa vào thuật toán ước tính mật độ hạt nhân

Abstract

This paper introduces a new approach in constructing land use and land cover (LULC) in Central Vietnam from various remote sensing images. The approach is based on the algorithm of kernel density estimation (KDE) to classify 9 types of land cover including water surface, urban area, rice, crop, grass, fruit tree, bare land, forest and mangrove. The results show that the KDE method can create LULC maps with more than 90% accuracy. About 33 ± 18% of the area of ​​the study area appears to be changed in the form of the land cover, especially the forest and surface water, which increased sharply by 2680 km² and 570 km² respectively. Bare land and rice area decreased by 495 km² and 485 km², respectively, due to the conversion of cultivation and urbanization. In addition, the results also show that the use of global maps for local studies seems to have large errors and inconsistencies. These findings are essential for developing resource management strategies.

Key words: Land use/cover change, image classification, classification accuracy assessment, Landsat, ALOS/AVNIR-2, and ALOS-2/PALSAR-2.