| Forest is a spatially dynamic and continuously changing ecosystem.Forests provide a number of necessary resources and services,including timber and non-timber products(e.g.,wildlife habitats,carbon and water cycle,ecological diversity,etc.)which are required for human society.Subtropical forests have high species richness,high forest productivity and complex structure of forest.Quantitative measurements of forest structure parameters of subtropical forests are required to understand forest ecological mechanisms,promote regional ecological developments,maintain biodiversity and enhance regional carbon balance.Traditionally,forest structure parameters are measured by conventional field inventories,which is often time-consuming,costly and often limited in spatial extent.Remote sensing technology can quickly,accurately and efficiently provide the multi-spatial,multi-temporal and multi-resolution information on forest structure parameters and volume distribution,greatly reducing the labor intensity and inventory cost.As a developed active remote sensing technique,airborne LiDAR can provide spatially explicit three-dimensional(3D)forest structure with high accuracy.Previous studies have demonstrated that the standard metrics(SM)(e.g.,height percentiles)were be effectively used to estimate forest structure parameters.However,these metrics usually neglect a mechanism explanation of summarize complex canopy characteristics and the information about continuous distribution of forest structure were not fully exploited by these metrics.Moreover,the standard metrics tend to be strongly inter-correlated.The canopy metrics(CM)derived from the canopy vertical profiles effectively exploit spatial forest structure information which describes the heterogeneity of forest structure to estimate forest structure parameters combining the mechanism explanation.This study was conducted in a north subtropical forest to investigate the synergistic use of the canopy metrics(CM)derived from the canopy vertical profiles and standard metrics(SM)to estimate forest structure parameters(i.e.,DBH,Lorey’s mean height,stem density,basal area,volume and aboveground biomass);Quick,timely and accurate estimation of forest volume distribution using airborne LiDAR data has contributed to the appropriate regulation of structure composition,decision-making support of forest management and accurate prediction of forest yield production.This study firstly employed the LiDAR metrics to estimate total volume.The three prediction methods of Weibull parameters,i.e.,parameter prediction method(PPM),percentile-based parameter recover method(PPRM)and moment-based parameter recover method(MPRM)were then used to directly or indirectly estimate Weibull parameters,and the mean tree volume was further calculated by predicted Weibull parameters.Then the stem density were predicted by means of total volume divided by mean tree volume.The tree volume distributions were finally estimated by predicted Weibull parameters and predicted stem density based on the three methods of PPM,PPRM and MPRM.The results of this study are as follow:(1)The forest structure parameters were estimated using airborne LiDAR data and results showed that the estimation accuracies of Lorey’s mean height(R~2=0.65-0.92,rRMSE=5.13–12.79%)and aboveground biomass(R~2=0.58-0.84,rRMSE=12.19–28.42%)models were the highest,followed by volume(R~2=0.45-0.80,rRMSE=12.02–19.22%),DBH(R~2=0.50-0.78,rRMSE=8.59–13.31%),basal area(R~2=0.44-0.74,rRMSE=11.34–20.48%),whereas stem density(R~2=0.44-0.71,rRMSE=18.68–29.86%)models were relatively lower.All models were established according to forest types(all plots,coniferous forests,broadleaved forests and mixed forests).The results showed that in general,the accuracies of forest type-specific models(R~2=0.50-0.92,rRMSE=5.13–28.42%)were relatively higher than general models(R~2=0.44-0.79,rRMSE=8.54–29.86%).Furthermore,the fitted models of the forest structural parameters were relatively more accurate for coniferous forests(R~2=0.58-0.84,rRMSE=8.59–26.55%)than broad-leaved forests(R~2=0.53-0.92,rRMSE=6.39–28.42%)and mixed forests(R~2=0.50-0.87,rRMSE=5.13–28.52%).(2)Comparing the capability of standard metrics(SM)and canopy metrics(CM)based models and combination models for estimating forest structural parameters,the results showed that the inclusions of canopy metrics improved the estimation accuracies of forest structure parameters and the combo models obtained the R~2 of 0.44-0.88 and rRMSE of 6.94–29.26%.Actually,the voxel size had effects on extracted metrics and final estimation models.Thus,this study also investigated the sensitivity of voxel sizes,and the results revealed that the resolutions of 5 m×5 m×0.5 m was the most optimal voxel size for estimating forest structure parameters in this study.(3)The results of study on estimating forest volume distribution using airborne LiDAR data forest volume,Weibull parameters(based on PPM),percentiles(based on PPRM)and moments(based on MPRM),the accuracies of forest type-specific models(R~2=0.49-0.91,rRMSE=10.65-39.81%)were relatively higher than general models(R~2=0.32-0.80,rRMSE=23.60-49.70%).The tree volume distributions were also generated by airborne LiDAR data in this study.The results showed that the relationship between predicted and reference volume distributions showed a relatively high agreement when the frequencies were scaled to ground-truth stem density(mean Reynolds error index (?)_R=28.07-41.92,mean Packalén error index (?)_P=0.14-0.21)comparing with the LiDAR-predicted stem density((?)_R=31.47-54.07,(?)_P=0.15-0.21).In addition,the tree volume distributions based on the PPRM(average (?)_R=37.75,average (?)_P=0.17)performed better than the MPRM(average (?)_R=40.43,average (?)_P=0.18)and the PPM(average (?)_R=41.22,average (?)_P=0.18). |
PDF Full Text Request