Analysis of Recent Vegetation Cover Changes Observed In Modis/Terra Data In Three Gorge Reservoir Area From 2001 To 2003

Abstract

Analysis of Recent Vegetation Cover Changes Observed In Modis/Terra Data In Three Gorge Reservoir Area From 2001 To 2003

Wang Jian
College of Environment and Spatial Informatics
China
Email: wjiancumt@sina.com

Zhang Jixian, Liu Zhengjun

Abstract
In this study, we mainly investigate the potential application of MODIS/Terra time series acquired between 2001 and 2003 for analyzing recent vegetation cover changes and seasonal variations in three gorges reservoir area based on decision tree classification and change vector analysis. First, the vegetation of the test area in this study is classified into six distinct biome types based on EVI, RED, NIR and MIR channels using decision tree classifier. Second, according to different typical vegetation types, we focus on exploring the mechanism responsible for the vegetation cover change in the RED and NIR channel reflectance space of the area. To reduce the cloud influence on the analytical results, a specifically designed cloud detection and smoothing technique is applied on the NDVI time series data. The singular spectrum analysis (SSA) method is then employed on the smoothed NDVI time series of specific surface biome types to reveal the seasonal variation characteristics of the years. In addition, change vector analysis (CVA) is also applied on pre-processed NDVI time trajectories for analyzing interannual vegetation cover changes. Since the length of the change vector indicates intensity of the change while its direction indicates the nature of the change, we classified it as four classes showing the vegetation cover changes in this area from 2001 to 2003, by calculating the magnitude and direction to evaluate the type of the vegetation cover changes. Finally, by using a geographical information system (GIS), the administrative boundary map is overlaid with the resultant vegetation cover change map so that the changed area at county level is estimated and its relationship with economic statistical data is also analyzed.