Remote Sensing Diagnosis Of Forest Canopy Height And Forest Aboveground Biomass Based On ICESat-2 And GEDI

Forests are a key component of the terrestrial biosphere providing a variety of ecosystem services,such as environmental purification and soil and water conservation.Forest canopy height is one parameter of the forest structure,and can partly reflect the level of forest ecological health and productivity of forest ecosystem to a certain extent.Forest biomass is a key index to measure the carbon sequestration capacity of forest ecosystem.The accurate acquisition of forest canopy height is an effective way to reduce the estimation error of forest biomass and reduce the uncertainty of scientific research in regards to the carbon cycle.Based on ICESat-2/ATLAS and GEDI data of two different models distributed in the last two years,we focused on the estimation of forest canopy height and forest biomass at the regional scale.This study focuses on the further study of the ecological health module under the interdisciplinary direction of‘remote sensing diagnosis of environmental health'.The research was carried out based on photon counting spaceborne Li DAR ICESat-2/ATLAS data,large spot waveform spaceborne Li DAR data GEDI data,ZY-3 multispectral and stereo images,Sentinel-1 data,multi angle MISR data,landsat8 OLI data and field measured data,using a variety of statistical learning and machine learning regression models,remote sensing diagnosis of forest canopy height and forest biomass in the study area.The main contents of this study comprise the following three aspects:(1)Retrieval of forest canopy height using ICESat-2/ATLAS and ZY-3 stereo imagesFirstly,the photon classification information provided by ICESat-2/ATLAS ATL08 data and the photons'longitude,latitude and absolute elevation information provided by ATL03 data were collected.Based on the two kinds of data,the discontinuous canopy height model CHM dataset was extracted.Finally,the CHM data set was used as the training sample to construct the inversion model of forest canopy height in Nanning city.The independent accuracy verification results showed that R~2 was 0.51 and RMSE was 3.38 m.(2)Remote sensing diagnosis of forest biomass in Nanning City based on optical and SAR dataBased on the forest inventory data and multi-source remote sensing data,the remote sensing diagnosis index system of forest biomass including optical remote sensing factor,texture factor,terrain factor and Sentinel-1 SAR was constructed.Based on multi-source stepwise regression,random forest and support vector machine,the remote sensing diagnosis results of forest biomass witn 30 m resolution in Nanning city were obtained.The accuracy of the three models was compared and analyzed.The results showed that the random forest model had the highest accuracy for remote sensing diagnosis of broad-leaved forest biomass,R~2 was 0.578,RMSE was 18.3250 Mg/ha.Finally,the random forest model suitable for the study area was selected for the next step of forest biomass estimation with 275 m resolution.(3)Remote sensing diagnosis of forest biomass in Nanning City based on GEDI and MISR dataThe highest accuracy canopy height RH95 corresponding to 95%of cumulative echo energy was selected.Combined with multi angle MISR data,the forest canopy height of Nanning city was inversed.Finally,the statistical model of forest biomass estimation based on single forest canopy height is compared with the remote sensing diagnosis model based on forest canopy height and multiple factors(optical remote sensing factor,texture factor,terrain factor and SAR factor).The results show that the former can significantly"underestimate"the forest biomass,while the latter can improve the accuracy.The innovation of this study includes the following three points:(1)A new method of retrieving forest canopy height based on ICESat-2/ATLAS and ZY-3 stereo images was proposed,which is of great significance for forest canopy height retrieval at regional and global scales.(2)Using two newly released spaceborne Li DAR ICESat-2 and GEDI data,remote sensing diagnosis of regional forest canopy height and forest biomass is realized,which fills the gap in related research of spaceborne lidar since ICESat/GLAS ceased working in 2009.(3)We constructed a diagnostic index system of forest biomass in Nanning City,which includes four kinds of remote sensing factors.The multi-source remote sensing diagnosis of forest biomass in Nanning City based on optical remote sensing data,SAR data,spaceborne Li DAR data was realized using machine learning models.