Skip to main content
Glama
Dendritic Forest Management (HDFM)

Dendritic Forest Management (HDFM)

Dendritic Forest Management

DFM checks

Open-source landscape planning, beginning with existing roads and waterways.

Status: software pilot and working research, not peer reviewed and not field-validated.

Aim: keep the woods working, and reconnect old-growth genetics in a dendritic pattern. Old growth of each place's own species, regrown along its rivers, valleys and ridges and linked through working woodlots, farms and ranches so that its populations can exchange genes again. The dfm-core engine checks the structure that aim needs: corridors that hold through every harvest plan, and an old-growth spine that never counts a plantation, or habitat recorded as mostly introduced species, as old growth. Whether genes then flow along the spine is an open field question (research status). Public site: dendriticforest.com.

Core output: the old-growth spine. The engine's central result is the old-growth spine: the corridor of retained forest along a landscape's streams, valleys and ridges, mapped whole, tested as a network and projected as it ages toward old growth. Structural connectivity only; not a claim about species movement, genetics, regulatory compliance or old-growth condition. Age is reported, never condition. Viridis Conservation (viridisconservation.com) offers professional old-growth corridor mapping for land trusts, towns and landowner groups, built on this same open engine.

Viridis DFM helps land managers compare forest-retention options with traceable inputs, explicit assumptions, maps and portable reports. It is currently a geometric screening pilot. It does not establish ecological viability, regulatory compliance or an optimal forest plan.

Agents can start at llms.txt (this repository), dendriticforest.com/llms.txt and AGENTS.md, then try it from your agent.

Start with the working application

To reproduce the worked examples step by step, with the output you should see, follow QUICKSTART.md. To cite the software, use CITATION.cff.

The web application includes:

  • Draw or import roads, waterways, planning boundaries, forest, open-water polygons and management units.

  • Record data sources, dates, reuse terms and field notes.

  • Save and reopen projects, with protection against conflicting saves.

  • Compare waterway buffer and proposed retention scenarios using the same saved input revision and calculation method.

  • Calculate forest within retention proposals after modeled road-surface and supplied open-water exclusions.

  • Export maps, comparison spreadsheets, printable reports and reproducible assessment records.

  • Run the same bounded analysis offline from a command line.

git clone https://github.com/jdhart81/hdfm-framework.git
cd hdfm-framework/web
npm ci
npm test
npm run dev

Open http://127.0.0.1:4173/map. Local projects are stored in .data/dfm.sqlite; no hosted account is required. A fictional watershed is included for exploration. See self-hosting and deployment instructions for Docker, PostgreSQL and authentication requirements.

The Viridis-hosted pilot currently requires owner access; it is not a public signup service. Self-hosting remains available independently.

Related MCP server: Sardinia Forest Dieback EO MCP Server

Try it from your agent

A coding agent, or you, can install the engine and check harvest plans against the corridors of a bundled fictional woodlot. This needs Node 22 or later and no account.

git clone https://github.com/jdhart81/hdfm-framework.git
cd hdfm-framework/packages/dfm-core
npm ci
node --input-type=module -e "
import {checkConnectivitySync} from './src/index.mjs';
import {woodlot} from './fixtures/woodlot.mjs';
for (const plan of [[], ['harvest-3'], ['thin-3']]) {
  const r = checkConnectivitySync(woodlot({treatments: plan}));
  console.log([plan[0] ?? 'no plan', r.status, ...r.reasons, ...r.warnings].join(' '));
}"

Expected output:

no plan pass
harvest-3 fail Link core-A–core-B is lost (caused by harvest-3). Treatment harvest-3 overlaps retained habitat or a core by 12000 m² without a recorded light-treatment permission.
thin-3 pass Treatment thin-3 overlaps retained habitat or a core as a permitted light treatment (12000 m²).

The woodlot has two cores joined by a 120 m corridor that crosses a road at a verified crossing, checked at a 100 m minimum width. A clearcut across the corridor (harvest-3) breaks the link, and the check names it. The same footprint recorded as a permitted single-tree selection (thin-3) keeps the link and is reported as a warning. packages/dfm-core/test/agent-try-it.test.mjs fails if this output changes.

To check a Landscape Package of your own, run this from the same folder with the path to your file:

node --input-type=module -e "
import {readFileSync} from 'node:fs';
import {fromLandscapePackage, checkConnectivitySync} from './src/index.mjs';
const r = checkConnectivitySync(fromLandscapePackage(JSON.parse(readFileSync(process.argv[1], 'utf8'))));
console.log(r.status, JSON.stringify({reasons: r.reasons, lostLinks: r.lostLinks, pinchedLinks: r.pinchedLinks, warnings: r.warnings}, null, 1));" path/to/landscape.json

A pass means mapped retained habitat at least the minimum width still links the same cores after the plan. It does not establish species movement, gene flow, regulatory compliance or old-growth condition (research status). dfm-core is not on npm yet, so run it from this checkout. npm test in packages/dfm-core runs the full engine suite in about 3 minutes.

An MCP client (Claude Desktop, Cursor and others) can run the same check and the spine analyses through the local MCP server instead.

Prompts to paste into a coding agent (for example Claude Code) opened in this repository:

  1. Read AGENTS.md and docs/RESEARCH_STATUS.md. Then run the dfm-core corridor check on the woodlot fixture with no plan, with harvest-3 and with thin-3, using the commands in the README section "Try it from your agent". Tell me which link each plan keeps or loses, which unit is responsible, and what a pass does not establish.

  2. I have a DFM Landscape Package at ./my-woodlot.json. Validate it against packages/dfm-schema/landscape-package.schema.json, run checkConnectivitySync from packages/dfm-core on it, and list the status, lost and pinched links, and every warning about missing sources or dates. Keep the file on this machine and do not commit it.

  3. Using packages/dfm-core/fixtures/watershed.mjs, draft the old-growth spine with deriveSpine and test it with spineNetwork. Explain the core links, the single points of failure and the warnings in plain language, and say which results rest on assumed rather than verified crossings.

The foundation

Waterways organize riparian study zones. Existing roads identify access and crossing questions. Forest, terrain, boundaries, management objectives and field evidence determine how these networks can inform a plan. A road surface is not retained forest, and a waterway centerline is not automatically a wildlife corridor.

See the system mapping, release notes and pilot assessment scope.

Research toolkit

The existing Python package remains in hdfm-framework/. It explores patch networks, entropy scores, widths, genetics and synthetic climate scenarios. It is separate from the working web application's geometric analysis.

The research toolkit is experimental. A minimum spanning tree minimizes summed edge length on its supplied graph, not necessarily the implemented composite entropy score. Synthetic comparisons and passing software tests do not prove ecological superiority. Known issues remain in area budgets, solver-result handling and population interpretation. Consult research status before using research outputs.

cd hdfm-framework
python -m pip install -e .
python -m pytest -q

The research package and web application have independent versions. Web release 0.6.0 does not certify or change the research package's scientific claims.

Open source and the Viridis service

The software is MIT licensed. Users can inspect methods, self-host and export their work. Viridis's proposed paid service adds data preparation, managed project operation, assessment delivery and support. Pricing and scope should be agreed for each initial pilot. Billing, subscriptions, teams, public signup and large-scale PostGIS processing are not enabled in this release.

Code adoption and GitHub activity are not evidence of customer demand or ecological benefit. Our next milestone is a real landscape assessment reviewed by its intended user.

Part of the Viridis open commons

DFM is one of five open projects that build on each other. Each one is useful alone; together they carry a piece of land from a map to a verified result.

Project

What it does

Where

Dendritic Forest Management (DFM)

Starts from existing roads and waterways and checks whether retained forest stays connected as harvests are planned. MIT.

hdfm-framework · dendriticforest.com

VergeCommon

Lets neighbours run a local conservation co-op: private mapping, parcel-by-parcel consent, evidence and shared decisions. Runs DFM inside its woodland projects. AGPL-3.0.

verge-common · vergecommon.com

Global Soil Intelligence Project (GSIP)

Builds an open soil spectral reference library, with phone and web capture. MIT code, ODbL data.

gsip

ViridisOS

Turns verified results into recomputable conservation modules and certificates anyone can check. Apache-2.0.

viridisos

Viridis Research

The public, machine-checked research record behind the methods. CC BY 4.0.

viridis-research

How they connect: DFM finds where a corridor runs. The corridor almost always crosses property lines, so VergeCommon gives the neighbours along it a private place to organize and consent. GSIP adds soil evidence from the same ground. ViridisOS lets anyone recompute what was claimed, and Viridis Research holds the proofs those checks rest on. Viridis Conservation (viridisconservation.com) offers professional old-growth corridor mapping for land trusts, towns and landowner groups, built on this same open engine.

Built something with one of these? Add the badge and tell us in an issue, so we can link back:

[![Part of the Viridis open commons](https://img.shields.io/badge/Viridis-open%20commons-2e7d32)](https://github.com/jdhart81/hdfm-framework#part-of-the-viridis-open-commons)

Contribute

Start with CONTRIBUTING.md. Useful contributions include reproducible bug reports, geometry fixtures, source documentation, GIS interoperability and independent scientific review: foresters testing the check on a woodlot they know, ecologists and geneticists helping design the field test of gene flow, and botanists building per-region native-status sources for planting records. Keep tests of implementation separate from claims about conservation outcomes.

Part of the Viridis open-source commons

Taking a retention option to the people who share the land? VergeCommon handles consent-based mapping and group decisions, and GSIP adds soil observations. The handoff between them is manual today.

This project is one of a connected set of open tools from Viridis. Each stands on its own; together they run from field evidence to checked decisions.

Conservation commons

  • VergeCommon: community conservation co-ops with consent-based land mapping, field evidence and group decisions

  • Dendritic Forest Management (DFM) (this repository): compare forest-retention options around roads and waterways, with traceable inputs and portable reports

  • Global Soil Intelligence Project (GSIP): contribute soil observations and explore open soil maps with explicit uncertainty

Verifiable science

  • Viridis Canon: machine-checked Lean 4 mathematics for ecological decisions, with a DOI for every result

  • ViridisOS: open reference runtime for DOI-backed decision modules and recomputable results

  • Viridis Research: the nightly public research record, each claim labeled by what a verifier actually checked

  • ProofLens: see what a Lean statement means before reading the proof

Agent tools

Browse every project on the Viridis profile. If this one is useful to you, a star helps other people find it.

License and attribution

Application code uses the MIT license; third-party notices remain with vendored libraries. Dataset licenses and project branding are separate. Research materials describe the HDFM agenda of Justin Hart / Viridis and remain available in the repository; this release does not independently validate manuscript claims.

Archived releases are on Zenodo: concept DOI 10.5281/zenodo.23112435, which resolves to the latest release (v0.6.0, git tag v0.6.0). Cite with CITATION.cff.

AI use. Much of the code and documentation was written with AI coding assistants, mainly Claude (Anthropic) and some OpenAI Codex; those commits carry a Co-Authored-By trailer or a claude/ or codex/ branch name in the history. Justin D. Hart (Viridis LLC, ORCID 0009-0008-3082-2482) reviews the changes and is the accountable author. No AI tool is listed as an author.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      A
      quality
      C
      maintenance
      Provides read-only tools for planning, writing, and auditing lesion network mapping and connectome-based localization papers, including design classification, methods checklists, phrasebooks, limitations, and manuscript auditing.
      9
      1
      MIT
    • A
      license
      A
      quality
      B
      maintenance
      Enables an AI assistant to run audit-style GIS checks on real spatial files for carbon projects: stratum areas in an equal-area CRS, declared vs. calculated area, overlaps, plot placement, forest loss by year and attribute completeness. Every result reports the CRS used and what data was missing, so answers come from actual geoprocessing rather than guesses.
      7
      MIT