Abstract: It is very important to estimate forest volume in forest system.Taking Jiande as the research area,and using TM image(2007) and the fifth(2007) forest resource survey data,we established and evaluated the precise of the volume remote sensing estimation model,which was on pine trees with or without the discrete quality grades.Site quality grade according to the average height of the small class and the average age of the establishment of the status table is divided into three types good,medium and poor.Total volume of the sub-compartment is the dependent variable,and each individual remote sensing content is the independent variable.The results are:1.the first principal component analysis of R² Landsat TM image is the best,the correlation of determination is more than 0.54,the highest is 0.802;2.The reserved independent sample on the accuracy of the model is validated,without the discrete site quality grades,the overall level of quality estimation accuracy was 87.64%,with the site quality grades the overall level of quality estimation accuracy was 94.14%,95.32%,92.38% respectively,the classification modeling precision is much better than the unified modeling accuracy.The research results provide an improved method for the estimation of forest volume,and provide a reference for improving the accuracy of forest biomass and carbon storage estimation.

Key words: TM image, forest stock volume, site class, linear regression