基于激光雷达的北京市主要林分因子联立模型构建与应用OA

[Objective]The purpose of this study is to explore the feasibility of establishing models for major stand characteristics based on LiDAR data to estimate the factors of forest patches,providing a demonstration for promoting the application of LiDAR technology in integrated monitoring of the national forest and grassland.[Method]Based on the LiDAR point cloud metrics and ground measured data of 1 966 forest plots in Beijing,the error-in-variable simultaneous equations were used to construct 8 forest factor estimation models for 13 forest types,including mean diameter at breast height(DBH),mean height,dominant height,stem number,basal area,stock volume,biomass and carbon storage.Additionally,based on the LiDAR point cloud metrics extracted from the 25 m×25 m grid cells within the forest patches in Beijing,the eight prediction models were used to estimate major stand characteristics of all forest patches.[Result]1)The LiDAR point cloud metrics that contributed the most to the estimation of major stand characteristics were 80%quantile of cumulative height and median point cloud height,followed by leaf area index.2)The mean prediction errors(MPEs)of eight major stand characteristics models for 13 forest types were less than 15%in either self-validation or cross-validation.3)Taking the forest as a whole,the determination coefficient(R2)of all eight prediction models were all above 0.7(excluding the stem number per hectare),the MPEs were less than 3%,and the mean percentage standard errors(MPSEs)were less than 40%,among which the MPSEs of the mean DBH,mean height and dominant height models were about 15%.4)According to the model inversion,the cumulative value of stock volume in all forest patches estimated by the volume model differed only by-1.79%from that obtained by the integrated monitoring of the municipality.The differences between the stock volume of forest patches and the integrated monitoring results in the three sub-populations were only 1.04%,-3.91%and-5.44%,respectively,which were all within the allowable error range of sampling survey.[Conclusion]1)The LiDAR point cloud metrics that contribute the most to estimating the major stand characteristics are percentile 80 of heights distribution and median height,followed by leaf area index.However,the point cloud intensity and density metrics have no significant effect.2)The method of error-in-variable simultaneous equations can be applied to construct the simultaneous models of major stand characteristics,which is able to solve both compatibility of different model parameters and error propagation of different stand characteristic estimates.3)The eight prediction models for 13 forest types can be used to estimate the major stand characteristics of forest patches in Beijing.4)The prediction accuracy of the major stand characteristics models based on LiDAR point cloud metrics can meet the technical requirements of forest resource inventory and monitoring,and the models can be applied in practice.