Study On Canopy Height Estimation Based On ICESat-2/ATLAS Photon Counting LiDAR Data

Forest ecosystem is an important part of terrestrial ecosystem,which plays an important role in the process of restraining global warming.Forest canopy height is not only the basis for estimating the carbon sink capacity of the whole forest ecosystem,but also the basis for studying forest biological productivity,net primary productivity,carbon cycle and global change.The traditional method of artificial forest structure parameter survey has the characteristics of heavy workload and low efficiency,which is difficult to achieve multi spatial scale and multi time scale promotion.With the rapid development of remote sensing technology,including optical remote sensing,Synthetic Aperture Radar(SAR)and Light Detection And Ranging(LiDAR)remote sensing,it has been widely used in forest structure parameter estimation.Optical remote sensing data reflect more horizontal structure information of forest canopy because of its low penetration,which affects the estimation accuracy of forest vertical structure parameters,SAR has a certain penetration on forest canopy,but it is greatly affected by terrain factors and tree species,which seriously limits its application in forest canopy parameter inversion.As a rapid developing remote sensing method in recent years,LiDAR technology can quickly and accurately obtain the three-dimensional structure and ground information of forest canopy,so it has unparalleled advantages in the extraction of forest canopy parameters.The purpose of this study is to carry out the research on LiDAR forest canopy parameter inversion method,mainly from the following three aspects:1)Research on the filtering noise algorithm of ATLAS photon data,and according to the characteristics of noise photons under daytime and nighttime observation conditions,filtering the noise photon of ATLAS photon data in forest research area;2)Classification algorithm for ATLAS photon data is proposed and the classification method is based on the elevation information of ground photons and canopy top photons.3)Research discussed the forest canopy parameters estimating by ATLAS photon data.The research could provide suggestions for ATLAS photon data processing and its application in forestry.The main conclusions are as follows1)This paper proposes a multi-level adaptive photon cloud filtering noise algorithm,which is mainly improved in two aspects.For one side is to take the noise distribution characteristics of daytime observation conditions and nighttime observation conditions as the main research object,so as to make it more targeted to the noise characteristics of the sun.For the other hand is to improve the search domain of the filtering noise algorithm for forest research,so as to make its search domain more consistent with forest.The distribution of signal photons in forest region can improve the filtering noise accuracy of photon cloud data.The results show that the proposed multi-level adaptive photon cloud filtering noise algorithm and the DRAGANN photon cloud filtering noise algorithm can effectively retain the signal photons and divide the noise photons.The study proposed noise filtering algorithm and DRAGANN algorithm both achieved the mean F value of 0.79.2)For different observation time,the two algorithms perform better filtering noise level in daytime.For strong and weak beam data,strong beam data and weak beam data have no obvious influence on the noise filtering accuracy of photon cloud.For different forest coverage,when the forest canopy density is 61%-80%,both algorithms show the best filtering noise results.3)A multi-criteria photon cloud data classification algorithm is proposed to estimate the forest terrain and forest canopy height.Based on the elevation information of signal photons,signal photons are classified as ground photons and canopy top photons.The results show that the multi-criteria photon cloud classification algorithm and NASA official photon cloud classification algorithm can effectively classify ground photons and canopy top photons.The accuracy of the multi-criteria photon cloud classification algorithm is R2=0.82,RMSE=2.39 m at the ground photon section level,R2=0.69,RMSE=5.41m at the canopy top photon section level.4)For different observation times,the inversion accuracy of ground photon and canopy top photon under night observation conditions is better than that under daytime observation conditions.For strong and weak beam data,the multi-criteria photon cloud classification algorithm has better ground inversion accuracy under strong beam conditions,and better canopy top photon inversion accuracy under weak beam conditions.For different forest coverage,as the forest coverage increases,the accuracy of the two algorithms for inverting the understory at the photon level and the section level has decreased.The difference in observation time and the temperature zone of the study area are the main factors that affect the accuracy of the canopy top inversion by the two algorithms.5)For the canopy height percentage,the canopy height inversion algorithm proposed in this study and NASA's canopy height inversion method,in the case of H70,the canopy height inversion algorithm proposed in this study and NASA's official algorithm both show better forest canopy height inversion accuracy.6)For different sampling windows,the optimal sampling windows of the canopy height estimation algorithm proposed in this study and that proposed by NASA are both 20m.For different forest coverage,when the forest coverage is 81%-100%,the two algorithms perform better estimation accuracy of forest canopy height.For different temperature zones,the two algorithms show better estimation accuracy of forest canopy height in temperate zone.For different observation time,whether it is daytime or nighttime observation conditions,as long as the noise filtering algorithm can effectively filter noise photon data.Daytime observation data can also provide scientific and effective data for estimating forest canopy height.The results show that ATLAS photon data can be an important data source for estimating forest canopy height parameters,and provide a reference for the new generation of spaceborne LiDAR in forest canopy parameters estimation.