Disclaimer & Assessment Methodology

This score was produced via LLM-assisted assessment of the repository structure and the supplied evidence bundle. These findings should be treated as a directional pilot signal rather than a verified rating. For more details on our pilot objectives and the development path toward rules-based human review, see the Development Roadmap. Raw files are accessible at the assessment report source.

Evidence Chain Assessment

Global Forest Watch Assessment

Assessed Claim

"Tree-cover loss and related forest indicators can be monitored globally."

The wri/gfw repository provides strong open-source evidence for the Global Forest Watch web application itself: it is public, MIT-licensed, documented for local setup, and includes calculation notes, API/layer schema documentation, CI, CodeQL, and release/deployment practices. Traceability is much weaker for the underlying environmental evidence chain: the bundle points to external Resource Watch API datasets and gives some formula-to-column links, but it does not include primary measurement provenance, dataset licenses, uncertainty/quality-control documentation, versioned data snapshots, exact production workflow runs, or explicit claim-to-data-to-code-to-review linkages. Overall, this is a substantially open web-app and methods repository with only partial end-to-end open traceability for the environmental claims displayed by the system.

49 Traceability Score
Weak Traceability
Project Source Source Repository
Framework Version v1.0.0
Assessment Date 2026-06-12
Human Review Pending

Claim Traceability Pathway

Inspection trail mapping the connections from raw inputs to the published claim.

Stage 01 Input Data
Score: 38
Stage 02 Models & Software
Score: 72
Stage 03 Execution Run
Score: 46
Stage 04 Community Review
Score: 44
Stage 05 Publication
Score: 57
Target Claim Stated Claim
Claim
Strong inspectability (≥80)
Moderate inspectability (55-79)
Weak inspectability (<55)
Open Linkage Connection

Traceability Health Summary

Scan diagnosis of where the evidence chain stands and what requires attention first.

Strongest Area

Open-Source Models, Methods, and Software

Inspectability: 72/100

This stage provides solid evidence, making calculations and documentation highly inspectable and reusable for external actors.

Traceability Gap

Open Linkage

Inspectability: 36/100

This is the weakest segment of the evidence chain. Gaps in records or lack of linkage break the evidence chain here.

Missing Artifacts
  • Unique dataset DOIs & licenses
  • Top-level LICENSE files
  • Execution run configs & workflow logs
  • Scrutiny logs & replication results
  • Open access publications
  • Dynamic provenance graph mapping
First Priority Action

Strengthen Open Linkage

Create a versioned provenance graph mapping every step of the computation from source data to claim.

36 Traceability Score
Framework Metric

Evidence Linkage Summary

Open Linkage is not a standalone category; it is the connectivity of the evidence chain. Below is the connection status of the connection links between each step in this project's journey.

Linkage Summary

There is some linkage across the evidence chain: the README links the web app to the Resource Watch API endpoint for GFW datasets, API documentation explains how metadata keys and layer configurations connect API records to application behavior, and calculation docs connect certain widget outputs to formulas and, in one case, to a named source column and table identifier. This prevents a very low score. However, the chain is not explicit, specific, versioned, or externally verifiable end-to-end. The supplied evidence does not link particular environmental claims or dashboard values to exact dataset versions, primary measurement records, code commits, model configurations, workflow runs, CI artifacts, review discussions, publication versions, or archived outputs. Open components exist, but the exact evidence chain behind a given claim is not inspectable from the bundle.

Detailed Dimension Inspection

Granular findings for each processing step in the Open Traceability framework.

Stage 01

Open Input Data and Measurement Evidence

Score: 38/100

Summary: The evidence shows that GFW input layers and datasets are at least identifiable through the Resource Watch API, with an endpoint for GFW-specific datasets and documentation of dataset, layer, vocabulary, and metadata fields. Some derived-widget documentation names specific source columns and a geotrellis table identifier for aboveground biomass, and widget documentation describes use of raw weekly aggregated alert counts. This prevents a lower score because the repository does not treat data sources as entirely opaque.

Evidence Found

The evidence shows that GFW input layers and datasets are at least identifiable through the Resource Watch API, with an endpoint for GFW-specific datasets and documentation of dataset, layer, vocabulary, and metadata fields. Some derived-widget documentation names specific source columns and a geotrellis table identifier for aboveground biomass, and widget documentation describes use of raw weekly aggregated alert counts. This prevents a lower score because the repository does not treat data sources as entirely opaque.

Gaps Identified

However, the evidence bundle does not include the actual primary datasets, dataset licenses, measurement protocols, quantified errors/uncertainties, quality-control records, data versioning, or traceability from secondary data back to primary satellite or field observations. The score is therefore in the partial-to-moderate range rather than high.

Stage 02

Open-Source Models, Methods, and Software

Score: 72/100

Summary: The web application source is public on GitHub, has an MIT license, and the README provides the main technology stack and setup commands. The repository includes method documentation for climate and widget calculations, API/layer schema documentation, component examples, and CI/security workflow files. This is strong evidence that a substantial part of the analytical and presentation logic can be inspected and reused. The score is not higher because the evidence bundle does not include the full dependency manifest, lockfile, backend/API source, data-processing pipelines, geotrellis generation code, infrastructure-as-code, or complete model implementations behind the environmental datasets. The README also states that the app depends on external APIs and Heroku deployments, so the software infrastructure evidenced here is not fully independently operable from the repository alone.

Evidence Found

The web application source is public on GitHub, has an MIT license, and the README provides the main technology stack and setup commands. The repository includes method documentation for climate and widget calculations, API/layer schema documentation, component examples, and CI/security workflow files. This is strong evidence that a substantial part of the analytical and presentation logic can be inspected and reused. The score is not higher because the evidence bundle does not include the full dependency manifest, lockfile, backend/API source, data-processing pipelines, geotrellis generation code, infrastructure-as-code, or complete model implementations behind the environmental datasets. The README also states that the app depends on external APIs and Heroku deployments, so the software infrastructure evidenced here is not fully independently operable from the repository alone.

Gaps Identified

No significant gaps identified. This stage is fully documented and verifiable.

Stage 03

Open Execution and Reproducibility

Score: 46/100

Summary: The evidence supports basic software execution reproducibility: the README gives installation and development-server instructions, identifies deployment branches and environments, and GitHub Actions specify a Node version with lint and unit-test commands. CodeQL adds a scheduled/publicly defined analysis workflow. This prevents a low score because an external actor can plausibly run and test the web application. The score is not higher because the evidence does not provide a complete computational workflow from environmental input data to reported outputs, exact production run records, pinned API/data versions, containerized environments, environment-variable values, archived outputs, or re-executable data-processing pipelines. The environmental calculations depend on external APIs and datasets whose runtime state is not captured in the supplied evidence.

Evidence Found

The evidence supports basic software execution reproducibility: the README gives installation and development-server instructions, identifies deployment branches and environments, and GitHub Actions specify a Node version with lint and unit-test commands. CodeQL adds a scheduled/publicly defined analysis workflow. This prevents a low score because an external actor can plausibly run and test the web application. The score is not higher because the evidence does not provide a complete computational workflow from environmental input data to reported outputs, exact production run records, pinned API/data versions, containerized environments, environment-variable values, archived outputs, or re-executable data-processing pipelines. The environmental calculations depend on external APIs and datasets whose runtime state is not captured in the supplied evidence.

Gaps Identified

No significant gaps identified. This stage is fully documented and verifiable.

Stage 04

Open Community and Review

Score: 44/100

Summary: There is evidence of public development infrastructure: the GitHub repository has open issues, pull requests trigger CI, review apps are automatically deployed for pull requests, releases are managed through GitHub releases, BrowserStack is used for cross-browser testing, and CodeQL is configured. This shows some visible channels for scrutiny of software changes. The score is not higher because the evidence bundle does not include actual issue discussions, pull-request reviews, governance records, peer-review records, methodological debates, correction notices, replication attempts, community forum records, or evidence of affected-community feedback. The review evidence is therefore mostly procedural rather than a demonstrated public record of scrutiny around environmental claims.

Evidence Found

There is evidence of public development infrastructure: the GitHub repository has open issues, pull requests trigger CI, review apps are automatically deployed for pull requests, releases are managed through GitHub releases, BrowserStack is used for cross-browser testing, and CodeQL is configured. This shows some visible channels for scrutiny of software changes. The score is not higher because the evidence bundle does not include actual issue discussions, pull-request reviews, governance records, peer-review records, methodological debates, correction notices, replication attempts, community forum records, or evidence of affected-community feedback. The review evidence is therefore mostly procedural rather than a demonstrated public record of scrutiny around environmental claims.

Gaps Identified

No significant gaps identified. This stage is fully documented and verifiable.

Stage 05

Open Publications and Communication

Score: 57/100

Summary: The project has a public homepage, a public GitHub README explaining that GFW is a dynamic online forest monitoring and alert system, and documentation for API layer/dataset configuration and widget calculations. These materials support public communication and some scrutiny of displayed indicators and methods. The score is not lower because the documents include concrete formulas and schema explanations rather than only marketing descriptions. The score is not higher because the evidence bundle does not include specific reports, papers, dashboard pages, indicator version histories, update-cycle documentation, data-source citations in final outputs, review status of publications, or open content licensing for the website/publications. For a dynamic dashboard, the supplied evidence is insufficient to verify that all claims are clearly tied to sources, update cycles, and versions.

Evidence Found

The project has a public homepage, a public GitHub README explaining that GFW is a dynamic online forest monitoring and alert system, and documentation for API layer/dataset configuration and widget calculations. These materials support public communication and some scrutiny of displayed indicators and methods. The score is not lower because the documents include concrete formulas and schema explanations rather than only marketing descriptions. The score is not higher because the evidence bundle does not include specific reports, papers, dashboard pages, indicator version histories, update-cycle documentation, data-source citations in final outputs, review status of publications, or open content licensing for the website/publications. For a dynamic dashboard, the supplied evidence is insufficient to verify that all claims are clearly tied to sources, update cycles, and versions.

Gaps Identified

No significant gaps identified. This stage is fully documented and verifiable.

Stage 06

Open Linkage

Score: 36/100

Summary: There is some linkage across the evidence chain: the README links the web app to the Resource Watch API endpoint for GFW datasets, API documentation explains how metadata keys and layer configurations connect API records to application behavior, and calculation docs connect certain widget outputs to formulas and, in one case, to a named source column and table identifier. This prevents a very low score.

Evidence Found

There is some linkage across the evidence chain: the README links the web app to the Resource Watch API endpoint for GFW datasets, API documentation explains how metadata keys and layer configurations connect API records to application behavior, and calculation docs connect certain widget outputs to formulas and, in one case, to a named source column and table identifier. This prevents a very low score.

Gaps Identified

However, the chain is not explicit, specific, versioned, or externally verifiable end-to-end. The supplied evidence does not link particular environmental claims or dashboard values to exact dataset versions, primary measurement records, code commits, model configurations, workflow runs, CI artifacts, review discussions, publication versions, or archived outputs. Open components exist, but the exact evidence chain behind a given claim is not inspectable from the bundle.

Inspectable Evidence Artifacts (5)

Critical Traceability Gaps

Inspection details of where the evidence chain breaks down and what that implies for public trust.

Identified GapImpact on TraceabilitySuggested Remediation
Dataset not linked or versionedHigh ‐ Blocks inspection of inputsRegister raw datasets in Zenodo and cite unique DOIs in config files.
Missing root license or dependenciesMedium ‐ Restricts model inspectabilityAdd a LICENSE file (e.g. AGPL-3.0) and export environment lockfiles.
Workflow run details not archivedCritical ‐ Restricts computational reproducibilityArchive Snakemake workflow runs and publish log files on Zenodo.
Review and replication logs hiddenMedium ‐ Hides validation gate processMaintain an open-source technical review tracker in public channels.
Publication lacks direct evidence citationHigh ‐ Claims are unlinked to runsAdd structured CITATION.cff and link claim statements to run hashes.

Remediation & Improvements Roadmap

Actionable roadmap prioritized by implementation complexity.

Phase 1

Quick Wins

Low effort, high immediate return.

  • Add CITATION.cff file

    Enables clear publication-to-repository citation schema.

  • Commit top-level LICENSE

    Resolves legal terms of software and dataset inspectability.

  • Link communication channels

    Add status/discord badges to repository README files.

Phase 2

Medium Effort

Standard engineering practices.

  • Specify configurations in config files

    Provide explicit parameters and dataset configurations in repository.

  • Lock computational environments

    Lock dependencies using Conda-lock or Docker image digests.

  • Archive automated workflow logs

    Publish CI test summaries and solver objective check values.

Phase 3

Strategic Upgrades

Highest traceability return.

  • Publish workflow runs on Zenodo

    Create persistent, immutable archives of the full execution outputs.

  • Build claims-to-evidence provenance graphs

    Establish a machine-readable map of the end-to-end evidence trail.

Framework Dimension Definitions

  • Open Input Data and Measurement Evidence: Identifiable, documented, attributable, versioned raw measurement evidence inputs.
  • Open-Source Models, Methods, and Software: Inspectability of Analytical logic via code repositories, configurations, and license files.
  • Open Execution and Reproducibility: Computational environments, workflows, run logs, and outputs make rerun processes inspectable.
  • Open Community and Review: Scrutiny, peer review, issue tracking, and responses to replication challenges are visible.
  • Open Publications and Communication: Resulting reports, papers, and dashboards are openly accessible and reference source materials.
  • Open Linkage: The versioned, explicit connectivity linking data, models, execution, review, and publication steps.

Human Review Notes

Human review of this assessment report is pending. An independent reviewer must inspect the repository references to verify the findings.