From a research paper by Qin Ma, etal in Scientific Data.
Here we produced an individual tree dataset including tree locations, height, crown area, crown volume, and biomass over the entire New York City, USA for 6,005,690 trees. Individual trees were detected and mapped from remotely sensed datasets along with their height and crown size information. Tree biomass in 296 field plots was measured and modelled using i-Tree Eco. Wall-to-wall tree biomass was mapped using relationships between field measurements and remotely sensed datasets and downscaled to individual trees.
Validation using field-plot measurements indicated that our mapping products overestimated tree number, mean tree height and maximum tree height by 11.1%, 8.6%, and 5.3%, respectively. These overestimations were mainly due to the spatial and temporal mis-match between field measurements and remote sensing observations and uncertainties in tree segmentation algorithms.
This dataset enables the evaluation of urban forest ecosystem services including regulating urban heat and promoting urban health, which can provide valuable insights for urban forest management and policy making.
LiDAR is an active remote sensing technology that can provide precise 3D quantification of tree structures18. Unlike optical imagery-based urban tree mapping that is often influenced by cloud cover19, shadows20, and saturation effect21, the LiDAR dataset can penetrate through vegetation surface and precisely delineate individual tree structures22.
Many studies have successfully quantified tree height, crown size, and crown volume for individual trees from LiDAR data23,24,25,26. Individual tree segmentation from LiDAR data is usually the first step for tree structure estimation. A number of tree segmentation methods have been developed, including point-cloud based classification using top-down27 and bottom-up28 strategies; and Canopy Height Model (CHM) based tree segmentations, such as marker-controlled watershed segmentation29.
For the complete paper CLICK HERE.
