Estimation Of Forest Stock Volume In Yanqing District Based On Sentinel-2 Images

Forest stock volume is one of the crucial stand parameters,which influences the ability of these forests to provide ecosystem goods and services.There is an urgent need for a new technology to improve traditional survey methods.In this study,the study area is Yanqing District of Beijing.Sentinel-2 images and digital elevation map were used as original data with the ninth national forest inventory data.Band reflectance,vegetation index,texture factor and topographic factor were selected as the alternative independent variables,and the forest stock volume of sample plots were taken as dependent variable.Through Pearson correlation analysis and Boruta feature selection,the optimal subset of variable factors was selected.Then the estimation model of forest stock volume was established by parametric and non-parametric regression method,and the applicability of the analysis model was compared.Finally,the optimal model was selected to invert the forest stock volume in study area,and the spatial distribution of forest stock volume was analyzed.The details and results are as follows:(1)Two parametric forest stock volume estimation models were established for the model samples,i.e.principal component regression analysis and partial least squares regression.Two non-parametric estimation models of support vector machine and random forest regression were also established.(2)Set the comprehensive evaluation index,use the modeling data to compare the advantages and disadvantages of the four methods.Finally the results show that the random forest regression method is the best.Its root mean square error is 12.15m3/hm2,and average relative error is 19.67%.(3)Inversion of forest stock volume in Yanqing area was carried out by Random Forest method.Among them,the forest stock volume of Yanqing was 2.563 million m3 with RF method,and the estimation accuracy was 83.51%.This indicates that the forest stock volume estimation technology of Yanqing area based on Sentinel-2 image has some reference value for forest resource monitoring and investigation.(4)Red-edge band has reference value in forest stock volume monitoring.The vegetation index of the red-edge band is highly correlated with the forest stock volume,and plays an important role in the construction of the average model.In object-oriented segmentation,the addition of red-edge band also improves the accuracy of segmentation.(5)The spatial distribution of forest stock volume in Yanqing is mainly distributed in the area below 600m from altitude.From the aspect of slope,it mainly distributes in the area without slope,followed by the southwest,north and northeast slopes.And from the aspect of slope,it mainly distributes in the flat,steep and slopes.