Application Research Of Spatial Stratified Sampling In Forest-covered Monitoring

By means of remote sensing and geographic information system software,based on thehigh-resolution aerial remote sensing images,we did a sample survey on the forest-covered areain Siyang. And based on the sample survey, we estimated stand volume of Siyang with aerialphotograph volume table. We set100m*100m sampling frame in high-resolution aerial remotesensing images. By the priori knowledge of pre-sampling results, we used simple randomsampling, systematic sampling, spatial random sampling and spatial stratified sampling of spatialsampling on the base of95%reliability and85%sampling precision and compared the results ofsampling. The results show that the spatial stratified sampling is the most optimal samplingmethod, concluded as follows:(1) Significant autocorrelation of forest-covered area in Siyang is available to spatialsampling survey. Spatial autocorrelation of forest-covered area in Siyang qualified by Moran’s Iindex. Spatial autocorrelation coefficient of forest-covered area in Siyang is0.18, significancetest Z value is8.48,which is far exceeding the critical value2.58of the significance level of99%.Therefore, spatial autocorrelation of forest-covered area in Siyang is significant.(2) When ordinary Kriging interpolation results used as spatial stratified cofactors, theeffect is significant. Theoretically the higher the correlation of between stratification cofactor andsurvey objectives, the better the effect of stratification. Optimal stratification is based on thesurvey objectives itself. However, due to forest distribution of Siyang County is uneven, forinstance interspersed farmland, buildings, roads, appeared as a zonal distribution, which wouldcause great difficulties,and lead to large sampling error to the stratification. According,spatialstratification is a relatively better method. The space is divided into three stratification layers asordinary Kriging interpolation method of interpolation result used as a spatial stratified samplingstratification cofactor.(3) According to the result, spatial stratified sampling is a more realistic method ofinvestigation on the survey of forest-covered area. Forest resources tend to have a certaincorrelation in spatial, which considered and quantified by spatial sampling in order to reducesampling costs. Compared to the spatial random sampling, stratified sampling spatial not onlyinvolves the spatial autocorrelation, but also spatial heterogeneity, making the spatial stratifiedsampling result more precise.Spatial stratified sampling is proved as the optimal sampling method. Then, we estimatedthe stand volume of Siyang with the help of aerial photograph, concluded as follows:(1)Based on the spatial stratified sampling, forest stand volume of Siyang is2,836,400m~3,forest stand volume per unit area is65.83m~3/ha. According to the data identified by JiangsuProvince in2010, the result is creditable.Survey factor with significant spatial autocorrelationand spatial heterogeneity which makes the traditional forest resources survey analysis is notprecise enough proved by the estimation of forest stand volume of Siyang. The condition of high cost, time waste, manpower and material resources consumed of traditional investigation offorest resources will be solved with the development of3S technology.