Biomass Estimation Model Of Pinus Densata Forests In Shangri-La City Based On Landsat8-OLI By Remote Sensing

Take Shangri-La city of Yunnan province as the study area,a typical forest ecosystem of Pinus densata natural forests as the research object,the aboveground biomass of 116 sample trees had been investigated and constructed estimation model of the sample tree,then we combined with the sample plots survey data,calculated the biomass of 56 sample plots of Pinus densata forests;And with the Landsat8-OLI as data source of remote sensing,combined with the measured data of stand biomass,I constructed five regression remote sensing estimation model of the biomass of Pinus densata.This five model are as follows stepwise linear regression,nonlinear regression,regression linear simultaneous equations,random forest and cubist regression.We choosed the best regression model through analysis of fitting performance and prediction precision of estimation biomass of the Pinus densata forests based on different regression methods,and based on the best regression model inversion the biomass of Pinus densata forests in the Shangri La City.Based on the inversion results,I analyzed the distribution pattern of Pinus densata Forests Biomass.The research shows that:(1)The best estimation model of the biomass of sample tree for Pinus densata.The determination coefficient(R2)is 0.992,the root mean square error(RMSE)is 30.778,prediction precision(P)for 87.941%.(2)In the process of constructed the stepwise regression model,by changing the conditions(maximum probability of F),Model 5 was selected as the final model of stepwise regression,the determination coefficient(R2)is 0.608,The root mean square error(RMSE)is 33.388,prediction precision(P)for 49.521%.(3)Nonlinear regression model is the cubic model of biomass.The determination coefficient of the model(R2)is 0.425,the root mean square error(RMSE)is 36.625.From the independence test of model point of view: the total average relative error and absolute relative error,average relative error between-1%~49%,the predicting precision(P)is 68.005%.(4)Canopy density and joint equations for biomass estimation model by remote sensing.The results showed that: The determination coefficient of the model(R2)is 0.290,The root mean square error(RMSE)is 30.378,prediction precision(P)for 56.408%;The stepwise regression model with The determination coefficient(R2)is 0.608,the root mean square error(RMSE)is 33.388,prediction precision(P)for 49.521%.We can see that the Canopy density and joint equations model improved prediction precision of the model.(5)The randomForest regression model based on the decision tree algorithms.The determination coefficient of randomForest regression model(R2)=0.907,the root mean square error(RMSE)=17.738,so the models of biomass factors have better fitting performance.Model predictive precision(P)is 87.600%,the model higher prediction precision and has a high predictive ability.(6)The Cubist regression model based on decision tree algorithm,the determination coefficient of Cubist(R2)= 0.324,the root mean square error(RMSE)is 45.667.The model fitting performance is not ideal,the predictive precision(P)for 44.267%,model prediction ability is poor.(7)It can be concluded that randomForest regression model is with best model Through analyses and comparison of the determination coefficient(R2),the root mean square error(RMSE)and prediction precision(P)of the models.RandomForest regression model has the best fitting performance and higher prediction precision prediction precision and it is the best model of biomass inversion.(8)Biomass of Pinus densata forests Distributed at different altitude: Biomass of Pinus densata forests mainly distributed at altitude of 2800~4000m,the biomass accounts for 97.28%.Very few Pinus densata forests distributed 2000~2800m and 4000 m above,accounting for 2.72% of the total biomass;In terms of mean,when altitude of 4000 m the biomass of Pinus densata forests are maximum for 92.6902 t/hm2.at 22000~2400m are minimum 79.2764 t/hm2.Biomass of Pinus densata forests Distributed at different slope: Biomass of Pinus densata forests mainly distributed at slope of 6~45°,accounted for 94.44% of total biomass,least distribution on flat slope and risk slope the biomass accounts for 5.56% From the average point of view,the steep slope of the Pinus densata forests biomass was 91.2080 t/hm2.,the average value of the biomass flat slope was the smallest for 75.4577 t/hm2.Biomass of Pinus densata forests Distributed at different aspect: Pinus densata forests biomass accounts for 12%~17% of total biomass on sunny aspect and half sunny aspect;Pinus densata forests biomass accounts for 3%~27% of total biomass on shady aspect and half shady aspect;The average value from the point of view,northwest slope of biomass of Pinus densata forests average value are maximum for 113.6955 t/hm2.Southern slope of biomass of Pinus densata forests average value are minimum for 76.2036 t/hm2.The total biomass of Pinus densata forests was 11719611.90 t,the mean value was 86.8591 t/hm2.The actual average value was 87.7153 t/hm2.Between the estimated value and the measured value is only 0.86 t/hm2,between the measured value and the predicted value is very small,and the estimated biomass is more accurate.