Howland Research Forest twelve-method aboveground biomass projection comparison: analysis code and derived data products

Analysis code and derived data products for a comparative assessment of twelve methods for projecting forest aboveground biomass (AGB) at the Howland Research Forest, Maine, USA. This deposit accompanies the manuscript Weiskittel et al. (2026), Forestry, in preparation at the time of deposit; a related identifier linking to the journal DOI will be added as a new version on acceptance. Methods compared: area-based LiDAR (Chapman-Richards and Schumacher), LiDAR with Acadian yield curves, individual tree crown (ITC) segmentation with strata averaging and with FVS-NE and FVS-ACD, terrestrial laser scanning with Acadian projection, sample-based field inventory, multi-attribute yield-curve matching (Tompalski et al. 2016, 2018), wall-to-wall directional 3D point-cloud projection (Kohek et al. 2022) on ALS decimated to 25 pts m squared, and two hybrid methods anchoring the 2017 ITC tree list with default and TLS-calibrated expansion coefficients. Key result: across 10, 20, and 50 year horizons the ITC + FVS-ACD baseline remains the most cost-effective method (0.88 US dollars per acre, 10.0 percent RMSE at 10 years). Of the four directional methods added in the v4.5 revision, the ITC-2017-anchored Hybrid A is best (24.7 percent RMSE); the operational value of the directional 3D family is individual-tree resolution rather than improved area-based accuracy. Contents: the full R analysis code repository; six ORNL G-LiHT AGB rasters (2012-2025); the 2025 MELiTE validation AGB raster; the 2017 wall-to-wall ITC tree list (208,632 trees) plus 2012 and 2021 vintages; the 2025 terrestrial laser scan stem table (4,417 stems); field inventory plot and tree tables; method 9-12 result objects with stratified-calibration, holdout, and bootstrap refinements; the v4.5 manuscript and supplemental figures; and the 25 pts m squared decimated ALS tiles. See README.md and data_dictionary.csv for the full inventory and field descriptions.