Prediction Of Subtropical Forest Parameters Using Airborne Laser Scanner

Light Detection and Ranging (LiDAR) is one of the most promising technology in forestry, which shows potential for timely and accurate measurement of forest biophysical properties over time. The aim of this study is to explore regression models relating variables derived from airborne laser scanner for estimation of various forest metrics, and describe why laser height metrics derived from airborne laser scanner are highly correlated with tree height, volume, basal area and above ground biomass. The study area was sub-tropical forest in central Yunnan Province, in which through processing the airborne LiDAR data and analysising the measurements of tree height outside, the inversion of forest parameters using airborne LiDAR technology was done. This study explored several regression models relating variables derived from airborne laser scanner for estimation of various forest metrics, and discussed the results of prediction concluding accuracy of them. These prediction models used 78 plots with a 7.5 m or 15 m radius in Kunming of Yunnan province. Two series of variables were provided from the airborne laser scanner data, one related to canopy height and one related to canopy density. They were used as independent variables in the regressions. The stepwise regression analysis was used to select various independent variables. Finally several prediction models were obtained, and the sampling test was used to test accuracy.The results showed that:(1) The variables and regression coefficients selected by stepwise regression analysis were able to achieve significant level, and get good prediction results. It indicated that forest parameters of different forest types can be predicted accurately using the regression models.(2) To different forest types, the difference of tree height estimation errors was insignificant, the average relative errors were controlled in 0.01 ~ 2.6%. Sampling test accuracy of all plots and coniferous forest plots were 0.86 and 0.84, respectively. It indicated that laser data can estimate the forest height with high accuracy, and different species composition of forest effected the results weakly.(3) To different forest types, the average relative errors of basal area estimation were 11.2%~17.5%. There was no obvious relationship between basal area of mixed forest and laser variables. The coefficient of determination of basal area prediction model for different forest types in decreasing ordered was: broad-leaved forest> coniferous forest> all plots.(4) To different forest types, estimation errors of volume were small, the average relative errors were 0.1%~8.9%. Different types of forest volume estimation error is small, the average relative errors were 0.1%~8.9%. In which the coefficient of determination of volume prediction models for different forest types in decreasing order was: mixed forest> broad-leaf forest> coniferous forest> all plots.(5) To different forest types, the biomass estimate errors were significantly different, the average relative error of coniferous forests was the smallest, followed by errors of broad-leaved forest and all plots.{6}The accuracy results of sampling test showed that: the precision of different forest parameters in decreasing order were: Lorey's tree height> Volume> aboveground biomass. And the precision of sampling test on plots of a single forest type was higher than sampling test on all the plots.