Study On Estimation Of Stand DBH Based On Multi-parameters

Diameter at breast height(DBH)is the diameter 1.3 meters from the root neck.It is an important tree measurement factor and an important factor for estimating forest stock,biomass,carbon storage and other forest indicators.In this study,Longquan City of Zhejiang Province is taken as the research area,and used the two kinds of survey data of 2017 province's forest resources,topographic data and Landsat-8 remote sensing images as data sources to estimate the stand DBH.The main research contents are as follows:Firstly,the topographic data and Landsat-8 remote sensing images were preprocessed.The processed topographic data,remote sensing data and forest resources survey data were extracted in subcompartments.A total of 19 independent variables were extracted,including elevation,slope,slope direction,soil layer thickness,humus layer thickness,age,canopy density,plant number per mu,band 2-7,normalized vegetation index,ratio vegetation index,difference vegetation index,enhanced vegetation index and red vegetation index.Then the correlation analysis of the extracted independent variables was carried out.Elevation,age,canopy density,plant number per mu,band 2-7 and difference vegetation index were retained as 11 independent variables.The independent variables are divided into two independent variable factor sets: ecological,biological factor sets and ecological,biological,remote sensing factor sets.Finally,13 empirical models were used to model and estimate DBH with age as the only independent variable.Then two independent variable factor sets were used to model and estimate the DBH of stand by multiple linear regression,generalized regression neural network and gradient lifting decision tree respectively,and the accuracy of the experimental results was compared and analyzed.The research results show that:(1)DBH was positively correlated with height,age and canopy density,and negatively correlated with plant number per mu,band 2,band 3,band 4,band 5,band 6,band 7 and difference vegetation index.(2)Among the 13 empirical models,the Gompertz model is the best one in DBH estimation.(3)Adding ecological and biological factors and remote sensing factors is conducive to DBH estimation.At the same time,multiple linear regression,generalized regression neural network and gradient lifting decision tree as estimation models have achieved good results in DBH estimation.The gradient lifting decision tree model has better generalization ability in DBH estimation.Taking ecological and biological factors,remote sensing factors as independent variables,the gradient lifting decision tree model has the best estimation effect.Its average absolute error is 1.3516 cm,average relative error is 12.47 %,root mean square error is 1.7731 cm,decision coefficient is 0.7207 and Theil inequality coefficient is 0.0730.(4)The best estimation effect of DBH is in the range of 15-20 cm.