Methodology, data sources & credibility

MineCheck produces a desktop EIA/ESG compliance draft for a Ghana mine. It is built to be transparent: every figure is screening-level, every requirement that cannot be resolved from a desk is flagged, and the science behind it is published and peer-reviewed. This page explains exactly how it works and what it does — and does not — do.

1 · The screening pipeline

• You select a demo mine or enter a concession — a point plus commodity (the box is sized to the commodity: ~6 km for a lease, a 25-acre block for a small-scale licence), the pillars of the concession plan pasted in decimal degrees, degrees-minutes-seconds or UTM 30/31 N metres (WGS84), or pillars clicked directly on the map. The polygon is drawn, its area computed and — for a small-scale licence — checked against the 25.2-acre cap. Ghana National Grid plans must be converted to UTM/WGS84 first.
• MineCheck samples a grid of points across the concession, clipped to the lease polygon where pillars are supplied.
• Each point is screened on Ghana-resolution spatial data, then results are aggregated worst-case with coverage counts.
• Findings are mapped to the Ghanaian regulatory and international responsible-mining frameworks, producing a conformance matrix, a desktop-readiness score and an honest gap register.

2 · Data sources (screening-level)

• Land & water (via Resilicity SafeGround): OSM water/wetland/road layers, MERIT Hydro stream network, WDPA protected areas, JRC Global Surface Water, ISRIC SoilGrids/WoSIS, SRTM terrain, GHSL built-up, curated Ghana contamination & coastal-erosion datasets.
• Galamsey / illegal-mining context (via SafeGround's 4-layer check): documented hotspots (Pure Earth Pollution Atlas, EPA shutdown notices, IMCIM, Minerals Commission registers), the galamsey-affected river corridors (Pra, Ankobra, Birim, Offin, Tano — HydroRIVERS), a district-prevalence table, and — where Earth Engine is active — Sentinel-2 NDVI-delta detection of clearance in the last six months (Forkuor et al. 2020). A red status needs satellite confirmation; a documented context floors a site at amber. Weights (0.15 / 0.20 / 0.10 / 0.55) are expert-elicited priors, not yet calibrated against EPA decisions — so every live report prints the four layer scores at the worst point and re-runs the status with each weight perturbed ±25 %, stating whether the status is robust or weakly held. Calibration against the Authority's historical galamsey decisions is a pilot deliverable.
• Flood & runoff (via Resilicity FloodGuard): national flood-susceptibility grid and region-specific ambient air quality (NO₂/PM2.5).
• These are indicative screening data, not surveyed or monitored values. A live engagement ingests the operator's GIS lease polygon, WRC permits and monitoring records.

3 · Frameworks mapped

Ghana Environmental Protection Act 2025 (Act 1124, which renamed the regulator the Environmental Protection Authority and repealed the 1994 Act 490) with the Environmental Protection (Environmental Assessment) Regulations 2025 (LI 2504) — in force since 23 June 2025; Reg 55 revoked the 1999 LI 1652 and its amendments — Minerals & Mining Act 703, Water Resources Commission (Act 522) — mapped alongside GISTM, IRMA (2018), ICMM, IFC Performance Standards, ISSB IFRS S2, GRI 14 (2024) and TNFD, plus commodity-specific instruments (ICMI Cyanide Code and Minamata for gold; Forestry Commission off-reserve consent and IFC PS6 for bauxite; WHO manganese guideline for manganese; etc.). LI 2504 rows follow the gazette text on epa.gov.gh: mining is a First Schedule undertaking (items 118–133), screening under Reg 6–8 sets the assessment level (PER, Reg 9–11, or full EIS), scoping runs Reg 13–17, the EIS follows Third Schedule Part B (Reg 18–21, Authority review Reg 26), decisions run on the Reg 28 clock (60 days for an EIA), permits are issued per phase for 24 months (Reg 30), the reclamation bond must be posted within six months before commencement (Reg 40), the EMP is due 18 months after commencement and every 36 months thereafter (Reg 33), the Annual Environmental Report at year-end (Reg 38), and assessment reports must be prepared by an Authority-certified practitioner (Reg 49). Licensed small-scale gold is a fifth profile on the same engine: Act 703 Part V (ss.81–99, as amended by Act 900 and Act 995), LI 2176 Reg 204 (≤ 25.2 acres), Act 1140 (2025 — GoldBod as sole buyer / exporter of ASM gold), Act 703 s.96 mercury purchase against the Minamata National Action Plan, the WRC buffer policy and the Reg 40 bond; engineered-TSF (GISTM), ICMM and listed-company disclosure rows are marked Not applicable while mercury, cyanide, water and reclamation rows stay in force. AKOBEN status: AKOBEN is an administrative programme of the Authority — neither Act 1124 nor LI 2504 gives it a statutory basis — publicly documented disclosure cycles date from the early 2010s and we found no published 2025–2026 cycle, so AKOBEN rows are framed "where the EPA runs an assessment cycle" and never inferred from a submission. Requirements that do not apply to a commodity (e.g. tailings/cyanide for dry-mined bauxite) are marked Not applicable rather than flagged as gaps. Siting inside a gazetted forest reserve is treated as a bar, not a gap: mining in forest reserves has been banned since the Environmental Protection (Mining in Forest Reserves) Revocation Instrument, 2025 annulled LI 2462 / LI 2501 (10 December 2025). Watercourse proximity is read against the WRC Riparian Buffer Zone Policy (2013).

4 · How readiness is scored

The desktop-readiness score is the share of applicable framework requirements with desktop/GIS evidence already in place (addressed 1.0, partial 0.5, needs-primary 0.15, gap 0; not-applicable excluded). It is a drafting-progress measure, not a compliance score — a deliberately modest number, because most of a real EIS depends on field work.

5 · What MineCheck deliberately does NOT do

• It does not fabricate primary baseline data (field ecology, hydrology, ambient air, social baseline).
• It does not produce stakeholder-consultation or free, prior & informed consent (FPIC) records.
• It does not replace the EIS prepared by an EPA-certified environmental practitioner (Act 1124, s.31 register), nor the independent audits GISTM/IRMA require.
• It is not an Environmental Impact Statement and does not constitute regulatory clearance.

The gap register in every report lists exactly what remains and who must close it. MineCheck cuts the desktop drafting; the practitioner adds the field data and signs off.

6 · The reclamation & phytoremediation estimate

The air-quality co-benefit of a reclamation planting is estimated with the leaf-scale deposition-velocity (Vd) method developed and published by Dr. Aduse-Poku:

• Aduse-Poku, M. (2025). Quantifying pollutant absorption potential of façade climbing plants. Urban Ecosystems (Springer). link
• Aduse-Poku, M. (2024). Methodology for quantification of pollutant absorption by climbing plants. ScienceDirect. link

Figures are order-of-magnitude screening estimates with transparent, tunable assumptions, calibrated with site leaf-area and ambient data for a live engagement.

7 · Two lenses, one engine

Every screening can be read two ways, and the switch never re-screens the site:

• Proponent draft — for the EIA consultant or mine environmental team: the full conformance matrix (Ghanaian statutory and voluntary international standards), the desktop-readiness scorecard, the reclamation outline and the gap register.
• EPA reviewer lens (view=reviewer) — for the case officer who receives the submission: a desktop triage signal, the site flags ordered by severity with "ask the proponent for", and a tick-able completeness checklist of the statutory requirements only (LI 2504, AKOBEN with its zero-tolerance criteria, Act 703, WRC). Voluntary lender standards are omitted because they are not part of the EPA's decision. Ticks and notes stay in the reviewer's browser; the sheet prints and exports to CSV.

The reviewer lens is site-informed triage, not a document audit: MineCheck does not read the EIS, and nothing it produces is an EPA determination.

8 · Data handling

• What you send: a point or the corner pillars of a concession plan, a commodity and an optional site name. No applicant names, no documents.
• What the server does with it: screens it in memory against the spatial engines on the same server, caches the result for at most one hour (so a lens switch does not re-screen), then discards it. Nothing is written to a database.
• Logs: the report endpoint is excluded from the web-server access log and the application runs without an access log, so coordinates and pillars are not retained in server logs. Rate-limit counters hold only the requesting IP address and a timestamp for the limit window.
• In your browser: ticks, notes, the review-stage and clock fields on the reviewer sheet are stored only in your browser's local storage. Print the sheet or export the CSV for the file.
• Sister engines: screening calls go to SafeGround and FloodGuard on the same server over the loopback interface; they do not leave the machine.
• Demos: anonymised; live screening of a demo hides the real district and hotspot names.

Relevant to the Data Protection Act 2012 (Act 843): the tool is designed so that an officer's screening of an application creates no personal-data record on Resilicity's side.

9 · Statutory pack revision

Every report prints the tool version and the statutory-pack revision it was produced under (currently LI 2504 pack rev. 2026-09-07). Changes to any Ghanaian-law row are logged, dated, in docs/STATUTORY-CHANGELOG.md in the repository — most recently: LI 2504 replacing LI 1652, the Reg 4(1) Second Schedule siting bar, the forest-reserve ban after the LI 2462 revocation, the AKOBEN framing, and the small-scale gold profile. Corrections from the Authority are applied there first, then to the rows.

10 · Who builds it