Analysis of Land Cover Changes in Northern Vietnam Using High Resolution Remote Sensing Data

Abstract

This study attempts to produce 15-meter resolution land cover maps over Northern Vietnam in 2007 and 2015 using multi-temporal and multi-sensor data including ASTER, Landsat, and PALSAR mosaic based on a kernel-based probabilistic classification method. Other ancillary such as SuomiNPP nightlight image, OpenStreetMap road network and SRTM30 were applied for additional information supplement. A number of about 60,000 reference data was built by field GPS photos as well as visual interpretation using Google Earth for training and validation. Results showed that the overall accuracy of the land cover maps is 81% and 89% in 2007, 2015 respectively. The results indicated many changes in areas of land cover types between 2007 and 2015 in Son La hydropower dam area and in selected sites for forest gain detection. The analysis showed that water area demonstrated an increasing trend while cropland area presented a decreasing trend in Son La hydropower dam area; and forest area experienced a rising trend whereas grassland area indicated a declining trend in the other selected sites. The results introduced a new high-resolution regional land cover data in Northern Vietnam for environmental modeling or other regional studies.