| With the development of remote sensing technology, multi and hyper spectral data has made considerable achievements in the fields of land use, ecology and environment in recent years. MODIS data are widely used in the research of regional land cover and use with the characteristics of multi-spectral information, the resolution and fast and long time series. Changes in vegetation cover and soil erosion is the current focus of the study. And how to estimate vegetation cover and vegetation cover management factor by remote sensing on large scale were particularly important in the study of vegetation cover and soil erosion. The problem of pixel unmixing must be solved for extracting vegetation coverage, because MODIS data have low resolution and a considerable number of mixed pixels.In this paper, monitoring vegetation coverage in the Three Gorges Reservoir Area from vegetation, bare soil, the abundance of the shadow by use of linear pixel unmixing model based on MODIS data and mixed pixel decomposition theory. Then the spatial distribution of maximum vegetation coverage and the interannual dynamic changes and its relationship with water and heat factors were analyzed in the Three Gores Reservoir Area from 2000 to 2009. Finally, value of vegetation cover management factor was estimated by using of abundance map of each component, and simple model was established between C factor and NDVI. The results showed that:(1) Spectral information of MODIS data can well measure the spectrum characteristics. Because the resolution is too high, the phenomenon of mixing exists in many pixels. Unmixing can be a good solution to this problem. In addition, it can be applied on long-term remote sensing monitoring on large scale and land use classification because of good time resolution and spatial resolution.(2) Unmixing is an effective method in the remote monitoring of land cover, as can effectively extract the information of surface feature. In the paper, vegetation, bare soil and shadow were extracted from MODIS images by using pure pixel index. The abundance maps of vegetation, bare soil and shadow in the Three Gorges Reservoir Area based on the linear spectral mixture model. Vegetation coverage was qualitative and quantitative verified by using a good linear relationship between NDVI and vegetation coverage and the same period high-resolution ETM+ images, respectively. The results showed that the vegetation coverage is line with the actual.(3) The vegetation cover in Three Gorges Reservoir Area is generally higher, areas of the annual maximum vegetation fraction, which was found to be more than 60%, accounted for 92.35% of the total reservoir area; The last 10 years, areas with increase in vegetation coverage is larger than the areas with decrease, but the number of pixels with significant increasing or decreasing trends only accounted for 7.16% of total number of pixels; Precipitation is the dominant factor of vegetation coverage inter annual fluctuations in Three Gorges Reservoir area,which accumulated precipitation from May to August is significantly related to the annual maximum vegetation fraction, showing a correlation coefficient of 0.805(p<0.01), but there are different correlation coefficients in different regions, having was significant positive correlation pixels only accounted for 13.81% of total number of pixels, mainly distributed in the western hilly farmland.(4) The value of vegetation cover management factor of the Three Gorges Reservoir Area was calculated based on abundance map of each component. The result was reasonable after test. This illustrates pixel unmixing was applied to estimating vegetation management factor is feasible. On this basis, a simple algorithm was established between C factor and NDVI. The changes of C factor were analyzed and the value of C factor was calculated by the weighted average rainfall erosivity in 2007 based on this method. |
PDF Full Text Request