MAPPING 10 YEARS OF LAND COVER CHANGE IN THE MEKONG DELTA

  • Ta Hoang Trung, Nguyen Van Tuan
Từ khóa: Land cover change, multi-temporal, machine learning.

Tóm tắt

In the last decade, the Mekong delta has seen a significant decrease in the total area of the wetland due to the expansion of the paddy field, and aquaculture farms. The expansion negatively affected the biodiversity of the region, threaten the environment. To conserve the biodiversity of the region and reach sustainable development, authorities have to monitor the transformation process effectively, thus an updated land cover map is necessary. This research used free satellite data, open-source library, and machine learning algorithm to create 10-meter spatial resolution land cover map of the region 2007 and 2017... In this study, Landsat 5 Thematic Mapper (TM), Landsat 8 Operational Land Imager (OLI), Sentinel-2, ALOS PALSAR, and PALSAR 2 mosaic, ALOS Global Digital Surface Model were employed to produce land cover map of the region by using Kernel Density Estimation classifier. Other ancillary data sources such as Open Street Map, regional geographical database was used as information supplement. To creating and validating the maps, 60,000 reference data points were created based on the field GPS photos as well as visual interpretation on Google Earth images. The overall accuracy of the maps is 82% and 84% in 2007 and 2017, respectively. The maps reveal the rapid loss of mangrove, mainly due to the expansion of aquaculture farms and the urban area between the two periods. The result also demonstrates the potentiality of automatically producing high-accuracy land cover maps in the large area.

điểm /   đánh giá
Phát hành ngày
2021-05-26