FORA: An Open-Source, Browser-Based Platform for Individual Tree Detection and Forest Metric Extraction from UAV LiDAR Data

Most of the operational Forest LiDAR processing is being done on desktop platforms that require unique applications, specific knowledge of programming and computers that are likely expensive and high-speed access to the Internet and therefore not accessible to many developing countries. This paper documents the use of an open-source web-based application called FORA (FORest Analysis) to perform iTreeDetection (ITD), ABA Metric Extraction, allometric estimation for 8 Turkish Forest Species and estimation of Aboveground Biomass for Pinus brutia in an entirely client-side application. FORA was validated using seven UAS LIDAR data files that are available through the PANGAEA repository (Kruse et al., 2025) that were collected in the boreal zone and tundra - taiga transition zone of Yakutia, Russia (56,252–149,300 m², 344–877 points/m², Canopy % cover of 0.60–0.95 & Mean Ht. 8.8–14.0m) with varying structures.

FORA was subjected to two tests: a default scenario (Cell = 0.5 m, minH = 2 m and radius = 3 m) and a matching scenario (Cell = 0.25 m, minH = 0.5 m and other PANGAEA parameters). The results were compared with the lidR datasets (Dalponte2016, Silva2016, Watershed) and the published PANGAEA segmentation. Sixteen area-based ABA metrics (i.e., height percentiles H5–H99, density ratios D1–D9, CC1.3, variability statistics) showed zero variation between the two scenarios. This demonstrated absolute parameter invariance, i.e., all ABA metrics taken from the distribution of point cloud heights do not depend on the segmentation parameter values. In the ITD, the R² value for mean tree height was 0.948 (RMSE = 0.88 m) for FORA in the matched scenario compared with the PANGAEA reference and comparable with Silva2016 (R² = 0.917) and Dalponte2016 (R² = 0.919). Correlation of crown diameter was higher than that of all the lidR algorithms. Three forest structural classes (closed, moderately closed, sparse) were distinguishable using CC₁.₃ .

These results demonstrate that browser-based forest LiDAR processing achieves research-grade accuracy for both individual-tree and area-based characterization without installation or programming.