PERSEUS Forest Intelligence

Precision-forestry decision support for the conterminous US. Pick or draw an area, see its localized forest health, risk, and value, then run multiple growth-and-yield models across contrasting economic, policy, and climate scenarios and get a multi-criteria recommendation and a downloadable report. Built as the assessment-and-values engine of the Guo et al. (2026) integrative forest-health framework, with a near-term disturbance-surveillance layer.

Live app: https://holoros.github.io/perseus-forest-intelligence/ Deployed version: v1.38 Stack: React + Vite + MapLibre, static SPA, auto-deployed to GitHub Pages from main.

Two views, one tool

A header toggle sets the interface for the audience:

• Landowner (default): a simplified path. Pick your state or find your area (Forest near me, AOI upload, coordinate jump), then use three tabs: Compare areas, Build a run, and Forest health. Research-only map controls are hidden.
• Researcher: the full surface. All map data layers (forest structure, ownership, stumpage/products, future risk), the carbon-trajectory map mode, point inspection, and the full tab set (Engine compare, RD trend, Engine spread, Stumpage, LANDIS, Landowner yields, Faustmann rotation).

What it does

• Compare areas: how a state or drawn area stacks up against similar places on priority forest area, stress, resilience, and climate exposure.
• Build a run: the on-demand flow. Choose an area and data source (FIA / TreeMap / upload inventory), select models (FVS, CBM, CEM, yield curves, LANDIS), build management x climate scenarios, set market prices, ecosystem-service payments, and a policy driver, then get a real multi-model ensemble, per-acre economics (timber + carbon + ecosystem services), a six-criterion multi-criteria scorecard, a plain-language recommendation, and a one-click HTML/PDF report. The free tier resolves the precomputed PERSEUS series client-side; the subscriber path dispatches the same run-spec to the OSC Cardinal HPC cluster.
• Forest health: priority forest area (stress x resilience) for 48 states and ~219k FIA plots, with current vs RCP4.5/RCP8.5 projections, ownership and ecoregion breakdowns, and a per-state species/vulnerability drill-down. Map units: surface, hexes, counties, ecoregion.

Develop / build

npm install
npm run dev        # local dev server
npm run build      # static bundle -> dist/
npm run preview    # serve the built bundle

Deploy

Continuous deploy: .github/workflows/deploy-pages.yml builds with Vite and publishes dist/ via the Pages artifact on every push to main. Pages Source must be set to "GitHub Actions". A clean build reproduces the deployed site, so pushes to main are non-regressive. To deploy: commit to main and push.

A backend launch-architecture scaffold (Supabase + Cloudflare Pages + Paddle, with Cardinal as the compute layer) lives in run-service/. It is configuration and code templates only; standing it up requires creating the accounts and supplying credentials (see run-service/README.md). The static app runs fully without it.

Data

public/api/ is the canonical data source of truth. Key files:

• series/{ST}.json — per-state multi-model trajectories by management and metric.
• meta.json, states.json — catalog and per-state coverage.
• yield_curves_by_l3.json — FIA Chapman-Richards yield curves by forest type x EPA L3 ecoregion x ownership.
• hrr_*.json — forest health/risk/resilience by state, grid, county, ecoregion, hex, owner.
• landowner_by_county.json (and _ecoregion, _hex as they are produced) — ownership composition from the USDA FS forest-ownership raster (RDS-2025-0045).
• stumpage.json, faustmann_rotation.json, landowner_yields.json, landis_stratified.json.
• geo/us_counties.geojson — CONUS county boundaries for the county map unit.

Raster overlays are in public/raster/, produced by scripts/50_raster_image_overlays.sh.

Methods and provenance

• Yield-curve methodology: docs/yc_engine_provenance.md and the in-app methods page.
• Inventory stratification analysis: public/methods/inventory-stratification/.
• Per-state HTML reports: public/reports/{GA,ID,IN,ME,MN,OR,WA}_report.html.

Known limitations

• Economics: timber value now uses real per-state blended stumpage ($/m3, from api/econ_params.json) with a user-selectable discount rate. The carbon price, ecosystem-service payments, and policy multipliers remain illustrative placeholders. Timber value is the optimal single-rotation (Faustmann) NPV at rotation age R*, with the perpetual land expectation value (LEV) also reported, gross of establishment and management costs.
• Climate scaling: RCP pathways currently share the baseline yield curves for most engines; calibrated CEM climate scaling is in progress, so historic and RCP may read similarly until it lands.
• Western climate band: CSI rasters cover the eastern domain; 18 western/plains states use a modeled transfer (flagged domain=modeled).
• perseus_db export: public/api/ is canonical; re-syncing the upstream DB export to reproduce the deployed data is tracked separately.

Citation

If you use PERSEUS Forest Intelligence in research, cite the underlying perseus_db data product and the CRSF yield-curve pipeline. Contact: Aaron Weiskittel (aaron.weiskittel@maine.edu).

Contributing

See CONTRIBUTING.md. The Pages workflow deploys main automatically; verify a clean npm run build before pushing.