Analysis Of Uncertainty And Error Sources Of Subtropical Forest Biomass Estimation Based On Airborne Lidar

Forest is the largest organic carbon pool in terrestrial ecosystem,which plays an important role in global carbon cycle and Slow down the rise of CO₂ concentration.Forest biomass is a key factor for calculating forest carbon storage,and its quantity and spatial distribution are important parameters for evaluating the carbon sink potential of forest ecosystems.At present,forest biomass estimation has become a research hotspot in the forestry field.However,two major challenges remain in the accurate estimation of forest biomass:（1）The allometric equation is the first key link in forest biomass estimation,and how does its choice affect the final biomass estimation result.（2）Due to low data quality and insufficient model accuracy,there are many uncertainties in the prediction of forest biomass on the tree scale,plot scale,and remote sensing scale.Therefore,quantifying the uncertainty in the biomass estimation process at different scales is of great significance for improving the accuracy of forest biomass estimation.Based on this,this study takes Longyan City,Fujian Province as the research area,using National Forest Inventory data and airborne lidar data,and on the basis of estimating forest biomass,firstly,the influence of different allometric equations on the error of the remote sensing forest biomass estimation model is analyzed.Then,the root mean square error is used as the uncertainty measurement index,and the error transfer method and Monte Carlo simulation method are used to accurately analyze and measure the uncertainty of forest biomass estimation on multiple spatial scales.In the analysis of the influence of different allometric equations on forest biomass estimation,the allometric equation of the main forest groups in China has a large error,but it is the smallest in the remote sensing biomass model error.On the tree scale,the uncertainty of biomass estimation caused by model residual error,model parameter error,and measurement error are 25.9%,4.3%,and 13.6%,respectively,and the total uncertainty is 29.6%.When transferred from the tree level to the plot scale,the corresponding biomass prediction error is reduced to 4.6%.On the remote sensing scale,in the uncertainty analysis of the biomass estimation at the pixel scale（plot size）,the uncertainty caused by the model residual error is 28.5%,while the model parameter error and the lidar data error cause the uncertainty.The certainty is small,5.6%and 9.7%respectively,and the total uncertainty is30.6%.In addition,the uncertainty of forest biomass estimation will be affected by the number of modeling samples.When the number of samples is 40,60,80,and 96 respectively,the uncertainty caused by the model residual error is 39.2%,31.5%,30.9%,28.5%.When the pixel-scale biomass estimation uncertainty is extended to multiple spatial resolutions,as the pixels become larger,the uncertainty caused by the model residual error and the lidar data error first rapidly decreases,and then stabilizes,while the model parameter error does not change significantly.In the entire forest biomass estimation of Longyan City,the total biomass is about252 million tons,the uncertainty of the total biomass is 77 million tons,and the relative prediction error is 30.6%.Through the above analysis of the results,the following conclusions are drawn:（1）When comparing the effects of different allometric equations on the error of the remotely sensed forest biomass estimation model,a simple analysis of the error of the remote sensing biomass model does not indicate the accuracy of the biomass estimation.In the estimation of forest biomass,more consideration should be given to the influence of the error of the allometric equation.（2）Among the uncertainty of biomass estimation at the tree scale,the model residual error has the greatest impact on the estimation result,followed by the measurement error,and the model parameter error has the least impact.Therefore,improving the accuracy of the allometric equation is of great significance for improving the estimation accuracy of forest biomass.When transferring from the tree scale to the plot scale,the uncertainty of the biomass estimation of the plot is much smaller than that of individual trees.（3）Among the uncertainties of biomass estimation at the remote sensing scale,model residual errors have the greatest impact on the estimation results,while model parameter errors and lidar data errors have little effect.In addition,the number of modeling samples of the remote sensing biomass model will affect the estimation of forest biomass.When the modeling data is small,the uncertainty caused by the model residual error is greater.Therefore,in the remote sensing of forest biomass monitoring,it is necessary to focus on improving the accuracy of the biomass model.（4）In the estimation of forest biomass in the study area,error transfer method,Monte Carlo simulation method,etc.can be used to realize multiple spatial scales and multiple spatial resolutions forest biomass estimation and uncertainty measurement based on airborne lidar data.