Vegetation monitoring using diferent scale of remote sensing data

１ＩｎｔｒｏｄｕｃｔｉｏｎＬａｎｄｓｃａｐｅｓｔｒｕｃｔｕｒｅｗｉｔｈｉｎｔｈｅＭｕＵｓＤｅｓｅｒｔ，ａｓｅｍｉａｒｉｄｒｅｇｉｏｎｏｆｎｏｒｔｈｅａｓｔｅｒｎＣｈｉｎａ，ｉｓｔｈｅｂａｓｉｃｆｏｃｕｓｏｆｔｈｉｓｓｔｕｄｙｃｏｎｃｅｒｎｉｎｇｌａ...

Vegetation monitoring using different scale of remote sensing data

This work sets out to simulate landscape model of Mu Us Desert in Inner Mongolia Autonomous Region of China at different spatial resolution using remote sensing images and distinguished landscape heterogeneity among different spatial resolutions. Landscape models were created from classification image of SPOT satellite data with 20m resolution and NOAA data with 1 km resolution. This study created landscape models of different scales by resampling the SPOT classified image using majority rule. The pixel resolution was increased from the finest scale of 20m by 20m up to 1000m by 1000m that was the coarsest spatial resolution. The Shannon diversity index was used to compare landscape models between different scales. At the finer scale the verify small patches such as deciduous forest, shrub and reedswamp with high vegetation coverage set on matrices with low vegetation cover (moving sand dune and sparse grassland) were verified. Broadening of scale resulted to the loss of small patches and at 1000m resolution, matrix classes were dominant. At 1km resolution of NOAA data, the matrix classes which greatly related to the topography of Mu Us Desert were detected. Diversity index decreased during scale broadening and the difference between SPOT 1km scale model and AVHRR data was not significant. The results showed that SPOT 20m model is good for the use of ecotone oriented revegetation planning, and NOAA 1km model is good for the seasonal and annual monitoring of each landscape unit, and revegetation planning at the regional level.