lsog-ne

FIA-based late-successional / old-growth (LSOG) classification for the U.S. Northeast. Started in Maine; designed to extend to New Hampshire, Vermont, and New York as data become available.

Built on the proxy scoring framework in R/fia_lsog_analysis_v3.r. The v3 script already supports a multi-state STATE_CODES vector and produces regional comparison figures when more than one state is processed.

Project arc

• Phase 1 (done): Cardinal project layout, v3 baseline import, GitHub remote.
• Phase 2: 2020-2024 evaluation panel + unorganized-territory subset; STDAGE imputation; multi-state run for ME, NH, VT, NY.
• Phase 3 (blocked on raster availability): plot-level recalibration against Hagan et al. (2024) LiDAR LSOG classes for the Maine unorganized townships.
• Phase 4 (in progress this session): cross-validate the v3 proxy against the ORNL DAAC dataset 2498 ("Mature and Old-growth Forest Probability Maps for the Conterminous US") at the FIA plot level, statewide for each state in STATE_CODES.

Layout

• R/ Analysis scripts.
  • fia_lsog_analysis_v3.r v3 baseline classifier (multi-state aware)
  • fia_lsog_analysis.R v1 reference
  • phase4_helpers.r shared scoring + raster utilities
  • phase4_ornl2498_extract.r Phase 4 pipeline
• scripts/ Shell helpers.
  • download_ornl2498.sh wget + .netrc downloader for the 3 COGs
  • refresh_earthdata_netrc.sh interactive password updater
  • submit_phase4.sh sbatch template (uses default account)
• data/fia/ FIA tables (gitignored). Cardinal already has ~/fia_data/ with ME tables; symlinked here.
• data/rasters/ Hagan and ORNL rasters (gitignored).
• output_v3/ Phase 1 baseline figures and CSVs.
• output_legacy/ Earlier outputs (will populate in a follow-up commit).
• output_phase4/ Phase 4 results (created by phase4 scripts).
• docs/ Methodology notes, comparison tables, source PDFs.
• logs/ SLURM logs (gitignored).

v3 baseline result (Maine, 2014-2018)

Statewide:

• Transitioning LS: ~1,328,000 ac (7.9 percent)
• Late-Successional: ~171,000 ac (1.0 percent)
• Old-Growth: ~22,000 ac (0.13 percent)
• All LSOG: ~1,521,000 ac (9.0 percent)

Compare with Hagan/Thompson (Maine unorganized territories only, 9.5M ac):

• Transitioning LS: 17.2 percent
• LS + OG: 4.2 percent
• All LSOG: 21.4 percent

References

• Bruening, J.M. et al. 2026. Mature and Old-growth Forest Probability Maps for the Conterminous United States. ORNL DAAC. https://doi.org/10.3334/ORNLDAAC/2498
• Hagan, J. et al. 2024. Using LiDAR to Map, Quantify, and Conserve LSOG in Maine. Our Climate Common. https://ourclimatecommon.org/lsog-project/
• Thompson, J. et al. 2026. Pathways for Protecting Maine's Remaining Old-Growth Forests. Harvard Forest / University of Maine.

Running on Cardinal

source /etc/profile.d/lmod.sh && module purge && module load gcc/12.3.0 gdal/3.7.3 R/4.4.0
cd ~/LSOG
ln -sf ~/fia_data data/fia    # use the already-staged FIA tables
Rscript R/fia_lsog_analysis_v3.r

For the Phase 4 ORNL comparison see docs/PHASE4_PLAN.md.