| The information of forest volume,biomass and carbon storage is the premise and foundation for scientific forest ecosystem management.How to efficiently and effectively monitor the annual dynamic change of forest resources is a research hotspot and technical difficulty in the field of natural resources survey.Optimizing sampling design and estimation methods to reduce survey samples is a key problem to be solved in annual monitoring of forest resources.In this study,the differences in biomass and carbon storage estimation between the single-tree model summation method and the expansion factor method were systematically compared,with the data from the sixth to ninth Continuous Forest Inventory of Sichuan Province in 2002,2007,2012 and 2017.The forest growth rate model of main tree species and the stand growth rate model of main dominant stands were established.The simultaneous model of stand volume,biomass and carbon storage for main dominant stands and the mixed effect model were established.Spatial autocorrelation analysis tool was used to analyze global spatial autocorrelation,incremental spatial autocorrelation and local spatial autocorrelation.Based on the sampling framework of CFI(Continous Forest Inventory),the sampling units were divided into five annual sampling units with different methods.In the selected 1/5 samples,combined with the spatial autocorrelation analysis,some samples from the selected 1/5 samples by stratified sampling were selected to form annual sampling plots.Generalized sampling estimation and probability sampling estimation were applied for sampling efficiency analysis.Data and annual monitoring results were updated according to the established growth model system.The existing forest resources investigation and sampling was optimized to improve the efficiency of survey in order to meet the different spatial and temporal distribution characteristics of annual monitoring demand.The main results are as follows:(1)There is no statistical difference between the estimation results of biomass and carbon storage by single tree model method and that by expansion factor method.Single tree model method is more accurate to estimate on the level of plot.The expansion factor method can be used as a supplementary method under the conditions of limited data,such as visual investigation and remote sensing plots.18 species(group)tree-level DBH and volume growth rate model,13 main arbor type stand-level volume growth models,and 10 major arbor volume,biomass and carbon storage combined models were established.Fitting determination coefficient R~2 are greater than 0.9,which can meet the simulation and updating date.The growth model system solves the compatibility problem among the volume,biomass and carbon storage,and provides an effective reference for remote sensing interpretation of volume update.(2)Before conversion,the stand,forest,scatter and side resources density all have a positive skew distribution.The global Moran’s I index,z-score and P-score of the standing and forest resources densities were all greater than 0.10,40.00 and 0.00respectively.There is a very significant positive spatial correlation between the stand and forest resources density distribution.The spatial autocorrelations of the scatter and side resources density with distance are unstable.At the significance level ofα=0.01,the high-value areas are mainly concentrated in Aba Prefecture,Garze Prefecture and Liangshan Prefecture,while the low-value accumulation areas are mainly concentrated in Sichuan Basin region,and the heterogeneous accumulation is staggered between the high-value accumulation areas and the low-value accumulation areas,while the transition zone between the three Prefectures region and the basin region is randomly distributed.(3)The average estimation accuracy of stochastic sampling,second-stage systematic sampling and second-stage cluster sampling is 94.70%with 95%reliability,lower than continuous forest inventory estimation accuracy.It can meet the requirement of accuracy of 95%,when forest volume exceeds 500 million m~3.The mean correlation coefficient R between the estimated value of the standing volume,biomass and carbon storage and observations of systematic sampling,stochastic sampling and cluster sampling is 0.97,0.99 and 0.96,respectively.The relative difference RD%is-0.28,-0.14 and-0.14,respectively.The goodness of sampling scheme in the order of the largest to the smallest is stochastic sampling,systematic sampling and cluster sampling with comprehensive analysis.Spatial stratified sampling,which uses spatial distribution pattern information,significantly reduced the number of plots to 1.68 per 10000,comparing with the average rates of CFI sampling and the 1/5 of plots sampling,which are 13.73 per 10000 and 2.75 per 10000,respectively.With 95%reliability,the mean of the estimation accuracy of spatial stratified sampling is 93.48%,the mean of correlation coefficient R between spatial stratified sampling and observations published is 0.96,and the mean of relative difference RD%is 0.35%.The clustering pattern analysis can effectively reduce the variance within each stratification,which could be used as the prior information for spatial stratified sampling.Spatial stratified sampling,significantly reduced sampling ratio to 1.68 per10000 from 13.73 per 10000,comparing with the sampling ration of CFI,which dropped 95.46 percent in the number of plots surveyed.The sampling is mainly concentrated in the area with large human disturbance,which is of large coefficient of variation.The analysis of temporal and spatial distribution characteristics of forest resources can be used for annual monitoring sampling design optimization and data which can provide practical reference for low-cost,rapid and accurate annual output of forest resources.After meeting the annual monitoring requirement,it is necessary to maintain the forest resources continuous inventory system at a certain time interval,investigate all samples,and evaluate the sampling reliability of annual monitoring. |
PDF Full Text Request