Research Of Forestland Area Sampling Monitoring In Three Gorges Reservoir Area Based On Remote Sensing

With the rapid development of remote sensing, it’s broader and more convenient to achieve the remote sensing-image. The spectral resolution, temporal resolution and spatial resolution of also have a great improvement. So remote sensing image have widely used in forestry. Taking Three Gorges Reservoir Area as the research area, starting from management need of different scales, based on the high resolution, medium resolution remote sensing image, had a study on sampling method of high resolution image plot and the way to combine high resolution plot with medium resolution image.And this article designed forest land estimation system of two scales included integral scale and the county scale of Three Gorges Reservoir Area by high resolution image sample plot, also exported the forestland distribution as the supplement result of sample method through the integration of high resolution plot and full coverd medium resolution image, thereby Formed a complete sampling monitoring systemThis article designed to improve the efficiency of the national forestland ground investigation and provide reference to form the normalized forest resources sampling monitoring system by remote sensing. The research contents and conclusions of this article are as follows:(l)Designed a sample method of remote sensing plot for the integral forestland estimation in the Three Gorges Reservoir Area based on the forestland type division map from high resolution image classification in Three Gorges Reservoir Area, including:1) Decided 3km*3km 4km*4km and 5km*5km as the size of remote sensing sample plot through tendency analysis of variable coefficient, combined with random sampling, systematic sampling and stratified sampling, formed 9 sampling methods to repeat sampling. Calculated the estimating result of eight forestland types from the remote sensing plot and concluded that the results of systematic sampling are all better. Analysis the overall accuracy and stability of the sampling results, high resolution image acquisition and processing cost, chose the systematic sampling of 4km*4km plot for a optimal method.2)Divided the Three Gorges Reservoir Area into 6 regions, systematic sampling in each region base on the variable coefficient respectively, compared the overall accuracy of estimating result and sample cost with integral systematic sampling of 4km*4km plot. It is concluded that they showed the same effect in estimating but the sampling area of former is 15%more than latter. So finally determined the most suitable sampling framework for the three gorges reservoir area of forest land types monitoring: systematic sample 4km*4km plot for a separation distance of 16km.(2)Designed a sample method of remote sensing plot for the counties forestland estimation in the Three Gorges Reservoir Area based on the forestland type division map in Three Gorges Reservoir Area. In order to facilitate the sample frame control, used grid sampling in the form of encryption. Covered the Three Gorges Reservoir Area with 20 km x 20 km first-order grid and conducted the second-order grid encryption in counties according to the specific forest resources in each county, then laying remote sensing plot of counties in each encryption grid. Determined the remote sensing plot size of counties by 2km*2km through the center sample plot standard deviation. Calculated the estimating result of eight forestland types from the remote sensing plot in counties, the best estimating effect is closed forestland and shrubland. Except that some the forestland type in counties those area were too small to estimating well, the result is good as a whole.(3)Based on the sampling framework, combine the medium resolution image covering the research area and the of remote sensing plot to output forestland distribution map. monitoring. Took the integral systematic sampling of 4km*4km plot as a example and selecting the Three Gorges reservoir at Hubei region as research area, conducted the classification result of high resolution image to the supplementary information for the classification of TM image, then extracted training samples to complete support vector machine (SVM) classification to export forestland distribution map in Three Gorges reservoir at Hubei region. Test results by kappa coefficient test, Khat= 0.681, indicating that the classification results and the actual results have good consistency. According to statistical analysis of classification figure of the research area, the closed forestland area was 6.24×105 hm2, the classification accuracy was 95.09%. The other forestland area was 2.78×105 hm2, the classification accuracy was 86.90%, the result of classification was good.