ESSDD Earth System Science Data Discussions ESSDD Earth Syst. Sci. Data Discuss. 1866-3591 Göttingen, Germany 10.5194/essd-2025-340 BorFIT: A Novel LiDAR-Based Training Dataset for Individual Tree Segmentation and Species Detection in northern boreal Forests Schladebach Jacob https://orcid.org/0009-0001-8381-2228 1 Heim Birgit https://orcid.org/0000-0003-2614-9391 1 Enguehard Léa https://orcid.org/0000-0002-2144-8264 1 Wieczorek Mareike https://orcid.org/0000-0002-3180-1607 1 Broers Jakob 1 Jackisch Robert https://orcid.org/0000-0001-5696-8721 2 Gloy Josias 1 Hao Kunyan 1 Tretton James 1 Gorshunova Anna 1 Kruse Stefan https://orcid.org/0000-0003-1107-1958 1 Alfred Wegener Institute Helmholtz Centre for Polar and Marine Research, Potsdam, Germany Technical University of Berlin, Berlin, Germany 20 08 2025 2025 1 31 Copyright: © 2025 Jacob Schladebach et al. 2025 This work is licensed under the Creative Commons Attribution 4.0 International License. To view a copy of this licence, visit https://creativecommons.org/licenses/by/4.0/ This article is available from https://essd.copernicus.org/preprints/essd-2025-340/ The full text article is available as a PDF file from https://essd.copernicus.org/preprints/essd-2025-340/essd-2025-340.pdf

BorFIT is a novel training data set designed to assist in the segmentation of individual trees and the detection of species from LiDAR point clouds, thus contributing to deep learning-based forestry applications. Recent advancements in AI-supported individual tree detection have shown significant progress; however, satisfactory results remain elusive in dense and structurally-complex boreal forests. We compiled a training data set designed to remedy this issue. It comprises 384 LiDAR point clouds, each with an area of 20 m &times; 20 m, in the form of reference plots, with up to 200 manually segmented and species classified trees per point cloud. We carried out LiDAR surveys at 146 sites between 2021 and 2024 in East Siberia (Yakutia), northwest Canada, and Alaska (USA), selected along a bioclimatic gradient to represent the circumboreal region. From each LiDAR transect derived point cloud, we extracted a minimum of four reference plots (each 20 m &times; 20 m) based on maximum tree heights within the plots to systematically sample the apparent tree density gradient. We manually segmented identifiable trees within each reference plot point cloud leading to 16,530 individual trees in total. Following segmentation, we trained four randomForest classifiers to predict the species of every segmented tree. The predicted tree species include: <em>Picea mariana</em> (Britton, Sterns Poggenb.), <em>Picea sitchensis</em> ((Bong.) Carri&egrave;re), <em>Picea glauca</em> ((Moench) Voss), <em>Pinus contorta</em> (Douglas ex Loudon), <em>Abies lasiocarpa</em> ((Hook.) Nutt.), <em>Larix laricina</em> ((Du Roi) K.Koch), <em>Betula papyrifera</em> (Marshall), <em>Betula neoalaskana</em> ((Regel) Ashburner McAll.), <em>Populus balsamifera</em> (L.), <em>Populus tremuloides</em> (Michx.), <em>Pinus sylvestris</em> (Thunb.) and <em>Alnus glutinosa</em> ((L.). The data offer the means for 3D space analysis of species distribution and stand structure around the circumboreal region. Furthermore, it can be used as a training data set for artificial intelligence (AI) applications and thereby improve our understanding of the boreal forest&rsquo;s vegetation reorganization in response to significant global warming.