基于哨兵2遥感影像的根河林区森林地上生物量估算

Estimation of Forest Above-Ground Biomass in Genhe Forest Area Based on Sentinel 2 Remote Sensing Images

我国大兴安岭林区在全球碳汇中扮演着重要角色,测算大兴安岭森林生态系统的碳汇功能具有重要意义。文章以大兴安岭根河林区为研究对象,对根河林区森林地上生物量(AGB)进行了系统研究。本研究结合2018年哨兵2号数据的植被指数、纹理特征和2018年森林资源连续清查固定样地数据,利用非参数模型(SVM和RF)来估算并比较高郁闭度的AGB。研究结果表明:(1)单波段中,红边波段的相关性最高,而植被指数与AGB存在负相关,NDVI的相关性最大,R=0.752。在所有的纹理特征中,以B5波段提取的方差纹理特征的相关性最大,R=0.557。(2)采用R2和RMSE来评价模型精度,发现随机森林的拟合精度相对较高。通过随机森林算法,基于红边波段得到的AGB模型,高于不加红边的AGB模型。加入纹理特征的模型高于没有纹理特征的AGB模型。说明加入红边波段和纹理特征会提高AGB模型估算精度。

Greater Khingan Mountains forest region in China plays an important role in global carbon sink.It is of great significance to measure the carbon sink function of forest ecosystem in Greater Khingan Mountains.Taking the Genhe forest region of Daxing’anling as the research object,this paper systematically studied the forest above-ground biomass(AGB)in the Genhe forest region.In this study,non-parametric models(SVM and RF)were used to estimate and compare AGB with high canopy density based on the vegetation index and texture features of Sentinel 2 data in 2018 and the fixed plot data of continuous forest inventory in 2018.The results show that:1 In the single band,the red edge band has the highest correlation,while the vegetation index has a negative correlation with AGB,and NDVI has the highest correlation,R=0.752.Among all the texture features,the variance texture feature extracted by the B5 band has the largest correlation,R=0.557.R2 and RMSE were used to evaluate the accuracy of the model,and it was found that the fitting accuracy of random forest was relatively high.Through the random forest algorithm,the AGB model based on the red edge band is higher than the AGB model without red edge.The model with texture features is higher than the AGB model without texture features.It shows that adding red edge band and texture features will improve the estimation accuracy of AGB model.