Utilizing Single Photon LiDAR Data for Estimating Individual Tree Attributes

Single Photon LiDAR (Light Detection and Ranging) is a novel and promising technology that can make laser scanning faster and cheaper. Compared to typical linear mode LiDARs (LML), SPL (Single Photon LiDAR) can be operated from higher altitude which means wider bandwidth on ground and a larger scanning area at once. Due to capability of SPL systems to create denser point clouds than current typical LML systems, the flight altitude can be higher in SPL which means quick remote sensing data collations abilities over large areas. Additionally, SPL can penetrate thin clouds and fog which gives airborne ALS better time frame as flight can be operated earlier in the morning than with LML. To the best of authors knowledge, this is pioneering research in Finland to analyse the applicability of SPL in Finnish forests and compare it with LML dataset. This thesis focuses on applying and comparing two LiDAR systems (SPL and LML) for extracting individual tree level (ITD) forest inventorying attributes and generating canopy height models in mature forests. Results were validated over 49 field measured plots, located in southern boreal forest. Additionally, the suitability of two crown segmentation methods (local maxima and watershed) were tested in both datasets. Watershed segmentation method yielded more accurate results for tree density and height estimation in both LML and SPL datasets. Tree density was underestimated by 4.7% (rRMSE: 32.3%) for all species. Comparing tree density estimation in different species, it was most accurate in deciduous plots (rRMSE: 17.0%, rBias: -9.5). Tree height estimation with SPL was highly correlated (R 2=0.93) with field-measured height and reliability accurate with underestimation of 3.4% (rRMSE of 7.0%). Comparing the tree height estimation in different species, it was most accurate between pine plots (rRMSE: 1.1%, rBias: 4.9%). In this research, SPL represented reliable and usable point cloud data for forest remote sensing and quality similar to LML. As expected, SPL had more deviation and higher bias compared to LML in tree density but yielded more accurate results for height estimations. Further studies with more accurate geolocated plots and individual tree maps are required. The hypothesis, the applicability of SPL data for forest inventorying and extracting tree density and tree height in mature forests, is valid.