Pinus Tabulaeformis Plantations Above Ground Biomass Estimation In Huanglong Mountain Based On GF-2 Data

Forest resources are the material basis for the functional output of forest ecosystems.The status of forest resources reflects the resources of trees and woodlands in forest ecosystems.At present,the most commonly used forest resource survey method in China is still the national forest resources continuous inventory,forest resource planning and design survey and operation design survey system.The disadvantages of this system are long cycle times,high costs and data differences.In this context,how to quickly and accurately estimate forest above-ground biomass(AGB)is one of the hotspots and difficulties in forest resource survey research.In recent years,the remote sensing estimation method has become the main method for obtaining biomass information of different scales because of its fast and efficient characteristics.Domestic and foreign scholars use the data types mainly including low-resolution,medium-resolution and high-resolution optical remote sensing data when performing forest AGB inversion.This study takes the Shibao Forest Farm of Huanglong Mountain in Shaanxi Province as the research area.Firstly the research caculated Co-occurrence-Based eight texture features(Mean,Variance,Homogeneity,Contrast,Dissimilarity,Entropy,Second Moment,Correlation),and five vegetation indexes(Normalized Difference Vegetation Index,Green Normalized Difference Vegetation Index,Enhanced Vegetation Index,Difference Vegetation Index and Soil-Adjusted Vegetation Index),and combined with the measured data of the samples.Based on five methods including normal regression,stepwise regression,ridge regression,cable regression and principal component regression,and four texture windows(3*3,5*5,7*7,and 9*9),this study used the leave-one-out cross validation(LOOCV)to test the estimation accuracy of each model;Secondly,combined with the measured data and downloaded NPP data of China from Resource and Environment Data Cloud Platform of CAS,this study estimated the net production and timber assortments of Pinus Tabulaeformis plantations samples and analyzed the historical changes in net primary productivity.The research conclusions are as follows:1)Among the remote sensing factors derived from GF-2 images,9 factors were significantly associated with AGB(p<0.01).Among the 5 planting index factors,NDVI was significantly positively correlated with AGB,and the correlation was the highest;Among the 8 types of texture features,Dissimilarity is significantly positively correlated with AGB and has the highest correlation.At the same time,it is found that the correlation between the vegetation index factor and AGB is stronger than the correlation between texture features and AGB.In addition,it was found that there are serious multi-collinearity problems between these remote sensing factors.2)GF-2 data can achieve higher accuracy in the inversion of forest AGB.The model established by the five regression methods based on this study,LOOCV shows that the R~2of models are all greater than0.50 under the four texture windows.Among them,the best estimation results are the principal component regression model using 9*9 window(R~2=0.81,RMSE=0.43).The least accurate estimate was the normal regression model using 3*3 window(R~2=0.48,RMSE=0.95).On the whole,the principal component regression model has the highest accuracy in inversion;the window with the texture setting of 9*9 has the best prediction accuracy.3)According to the research,the empirical formula can estimate the net growth and timber assortments of Pinus Tabulaeformis plantations accurately.Based on the research of NPP data of Shibao Forest Farm from 2000 to 2010,this study found that the net primary productivity of Shibao Forest Farm from 2000 to 2010 not only showed a gradual growth trend,but also the scope of growth continued to expand.