Error Analysis On Spatial Estimation Of Regional Forest Carbon Distribution

Forest carbon storage and distribution is the key information on the forest ecosystem researchand relevant decision-making. The great significance of error analysis on spatial estimation ofregional forest carbon storage is to evaluate the estimated results, improve the estimation methodand reduce the estimation error.The major research object of this study is the spatial estimation of regional forest carbondistribution based on National Forest Inventory (NFI) data and the Landsat Thematic Mapperimage collected in Lin’an County, Zhejiang in2004. The input errors of spatial estimation werequantified by relative error, and then we analyze their error synthesis and error distribution.The error propagation model of the input error and the output error were constructed basis on this.The main results include as follows:1) The influence of Measurement environment and measurement personnel to the diametermeasuring error is significant; Follow the increase of classified tree height, tree height measuringerror tend to decrease.2) The population sampling error of the2004lin’an National Forest Inventory (NFI) data is9.18%, diameter measuring error is0.13%, tree height measuring error is0.08%, biomass modelerror is3.65%, geometric correction error of Landsat TM images is2.39%. Total input error is10.22%, output error is8.74%.3) The descending order of the contribution of error terms in the total input error is:population sampling error (81.06%)> biomass model error (12.81%)> geometric correction errorof Landsat TM images (5.49%)> tree height measuring error (0.65%)> diameter measuring error(0.02%).4) The regression equation between plot total input error and output error is: output error=7.984+1.18/plot total input error, R2=0.306, std. error of the estimate is0.462.5) The regression equation between input error of various error term and output error is:output error=0.578+0.06×biomass model error–0.117×tree height measuring error+0.440×tree diameter measuring error+0.014×geometric correction error of Landsat TM images, itsstandardized regression equation is: output error=0.406×biomass model error–0.215×treeheight measuring error+0.149×tree diameter measuring error+0.70×locating error of Landsat TM images coordinate correction, R2=0.166, std. error of the estimate is0.507.This paper offers certain reference function to the understanding of spatial estimation error ofregional forest carbon distribution and the improvement of estimation accuracy.