Quantitative Estimation of Forest Stock Volume Based on Multi-Source Data

Abstract

Using SPOT-5, QuickBird, ALOS optical data with linear regression model, forest stock volume is quantitative estimated in Zhazuo Forest of Xiuwen County of Guizhou Province. The optimal ratio of band is chosen from the above three types of optical remote sensing data respectively by the criteria of mean residual sum of square called RMSq. Combined with GIS factors (tree species, land type, altitude, slope and slope aspect, forest age group, canopy density), the linear estimation model is confirmed. Four experiments are carried out using the three data respectively and multiple-source data. In this study, compared with the others, regression prediction model using ALOS data fits best. In this experiment, its root mean square error equals 0.623929, the total forecast relative error of the regression model equals 0.082878, and prediction error equals 0.75233. The ALOS ratio wave bands which finally enter the regression equation are AL1-AL2 /AL1+AL2, AL3/AL2, AL3-AL2/AL3+AL2. The use of multi-source data is also worthy of consideration.