Biomass Estimation Of Woody Plants In Elm (Ulmus Pumila) Sparse Forest Grassland Ecosystem Of Otindag Sandy Land Based On UAV Platform

Grassland with sparsed elms in Otindag Sandy Land is a kind of special zonal vegetation type in the middle of Inner Mongolia,It has long been a powerful ecological barrier to maintain the ecological security of the Beijing-Tianjin-Hebei region.As a key component of sandy land ecosystem,woody layer plants play an important role in water and soil conservation,so as soil nutrient improvement.This study was carried out in the long-term ecological positioning observation of elm sparse forest grassland in Otindag Sandy Land located in Xilingol league,Inner Mongolia autonomous region.The UAV remote sensing platform can be used to obtain high-resolution vegetation images quickly,3D reconstruction of vegetation in the sample grounds is finished by Agisoft Photoscan software,and the data of Digital Orthophoto Mape,Digital Surface Model,Digital Ground Model and vegetation’s Canopy Height Model can be generated efficiently.The parameters of single wood structure were extracted by image segmentation and geographical processing.Besides,based on the existing empirical equations of species,a model was constructed to explore an efficient and accurate method for estimating the biomass of vegetation in sandy land ecosystem.The main research results are as follows:(1)The Unet neural network model combined with watershed algorithm can effectively segment most of the crown range in the orthophoto image,with an overall segmentation accuracy of 77.57%,but it is still not possible to completely distinguish the single tree crown.(2)The height,crown breadth and canopy projected area were selected as the biomass estimation parameters.The height of a single tree in the vegetation canopy height model was calculated by the ray-projection algorithm and the local maximum algorithm.The crown breadth was calculated by constructing the minimum external rectangle method and the canopy projected area was calculated by the polygon segmentation method.Three structural parameters were successfully extracted from 75.52% elms and 82.05% shrubs,with the highest average extraction accuracy of height up to 78.51% and the highest average extraction accuracy of crown breadth up to 86.92%,and the projected area of crown of elms was 0.1578 ～118.774 ㎡.(3)Combined with the existing empirical equation of the biomass of the main woody plants in Otindag Sandy Land,22 elms’ DBH inversion models and 5 shrubs biomass prediction models were fitted,respectively.The biomass of sample land was estimated by effectively extracting the parameter value of single wood from UAV image,which is 70.63% of the estimated value of ground survey data.Biomass estimation was generally underestimated.