Extraction Of Forest Disturbance During Recent Decade In Yunnan Province Based On MODIS Dataset And Landsat Imagery

Forest disturbance reflects the interaction and forest succession process of forest ecosystem and people,and it is also an important factor affecting the ecological environmentFor regional or larger scale,remote sensing is an effective method to study the carbon cycle and monitor the forest disturbance.By selecting Yunnan province as the study area, the paper adopt the disturbance index (DI) to extract forest disturbance from Landsat data imaged on 2002 and 2013 as well as MODIS products imaged from 2002 to 2013.On this basis, the paper selects three sampling methods of random sampling,stratified sampling, systematic sampling to establish the perturbation statistical relationship between the deforestation results of the MODIS data and Landsat data, and compares the monitoring precision of different sampling methods, then determine the most suitable sampling method to modify the disturbance area from MODIS data in order to improve the monitoring accuracy of low spatial resolution data.The conclusions of this paper are as follows:(1) Comparing with the disturbance of Landsat color image during two periods via visual interpretation, the disturbance index method based on tasseled cap transformation of Landsat data can be used to extract forest disturbance information accurately.(2) The paper constructs the MODIS disturbance index to extract forest change information, in which the instantaneous disturbance monitoring results is consistant with MODIS fire product.Google Earth analysis showed that the monitoring results of non-instantaneous disturbance have a higher precision.(3) Compared with the disturbance result detected from MODIS data and Landsat data, the MODIS data is feasible in dynamic monitoring of large forest disturbance areas.Due to the low spatial resolution of MODIS data,forest disturbance occurred in sub-pixel cannot be accurately monitored.(4) Random sampling,stratified sampling and systematic sampling method are used to extrapolate the disturbance area of the whole region by compositing forest disturbance monitoring results of high and low spatial resolution remote sensing image. The results show that the Newman optimal allocation method of the stratified sampling has the highest accuracy, and systematic sampling method can obtain a better result than random sampling, while these two sampling method depend on the sampling position. By using FORMA(Forest Monitoring for Action) data as the reference, compositing high and low spatial resolution image based on sampling method has a higher accuracy for extraction of forest disturbance, which improve the precision of about 18.29 percent compared to the result detected directly from the low spatial resolution images.(5) The forest disturbance area of the study areas in recent ten years are extracted from MODIS data,which are corrected based on the linear relationship between the MODIS-derived and the Landsat-derived forest disturbance. The results showe that forest disturbance area in Yunnan Province has a obvious temproal change characteristics.The annual average disturbance area is 23495.83 hectares, in which the disturbance area are larger in 2010,2012 and 2013 with the average annual disturbance area of 50366.67 hectares.The most severe disturbance occures in 2010, when the disturbance area is about 83000 hectares and accounting for 0.5 percent of Yunnan forest coverage area. The main disturbance region distributes in Dali, Lijiang, Yuxi.