← atlas · detector results

Worked example: images + cores → block model → draft report

Everything below was produced by the pipeline from public GEOMAR SO268 data with no manual geology. It shows the SHAPE of the pilot deliverable; it is not a resource statement.

block model

Per-image output for the whole corpus (34k rows, geolocated, with the GEOMAR reference values): so268_per_image_coverage_v3.csv

DRAFT — Polymetallic Nodule Mineral Resource Estimate, SO268 German (BGR) area corridors

*Prepared by Ocean Map software 2026-09-21 for Competent Person review. NOT a signed resource statement. Structured to ISBA/21/LTC/15/Rev.1 Annex V (CRIRSCO-derived) Table-1 items; S-K 1300 §229.1302 mapping noted in brackets.*

1. Summary [S-K 1300 (b)(1)]

Imagery from 12 OFOS profiles (40,476 images; GEOMAR SO268/1–2, 2019; CC-BY) and 36 box cores (0.25 m², Schoening & Gazis 2019) were used to estimate nodule abundance over the German-area survey corridors. Nodule coverage was read from every image by a convolutional model validated on held-out profiles; coverage and nodule size were calibrated to core abundance; abundance was kriged onto 500 m blocks; blocks were classified by data spacing.

ClassArea (km²)Mean abundance (wet kg/m²)Block sd (kg/m²)Tonnes (wet)Tonnes (dry, 28 % moisture assumed)
Measured26.015.7±3.4408,654294,231
Indicated44.014.9±3.7655,685472,093
Inferred66.515.5±4.21,027,751739,981

2. Sampling techniques and data [Annex V T1 §1; S-K 1300 (b)(3)]

  • **Imagery:** towed OFOS, ~1.6 m² footprint per frame after colour/light-cone correction (Mbani & Greinert 2023), 10-s cadence,
  • USBL-positioned. Frames flagged by altitude/blur are excluded (quality flag in per-image table).

  • **Cores:** 36 box cores; total nodule mass, surface coverage and median nodule size recorded per core; buried fraction median 8 %.
  • **Coverage reader:** ResNet-34 trained on 19,683 labelled frames; held-out on 9,096 frames from 4 unsampled
  • profiles: MAE 2.95 % coverage, R² 0.65. Known limitation: packed facies (>30 % coverage) MAE ≈ 8 %.

    3. Estimation methodology [Annex V T1 §3; S-K 1300 (b)(9)]

  • **Coverage → abundance:** linear model on the 36 cores, 5.82 + 0.255·coverage% + 0.098·size_cm² (residual sd 2.72 kg/m², LOO R² 0.40).
  • Abundance reported on a WET basis as measured on deck; dry tonnes use an assumed 28 % moisture (NORI convention) pending site moisture determination.

  • **Spatial estimation:** per-image abundance aggregated to 100 m nodes, ordinary kriging with a spherical variogram
  • (sill/range/nugget [17.71, 12625.158, 3.012]) on 500 × 500 m blocks (UTM 11N). Block uncertainty = √(kriging variance + calibration variance).

  • **Cut-off:** none applied (all blocks > 4 wet kg/m²). **Top-cuts:** none. **Factoring of abundance:** none.
  • 4. Classification criteria [Annex V T1 §4]

    Distance from block centre to nearest data node: ≤0.5 km Measured, ≤1.5 km Indicated, ≤3.5 km Inferred. These are illustrative spacings for a corridor survey; a grid survey would follow the NORI precedent (cores ≤7 km + photo infill ≤3.5 km → Measured).

    5. Data gaps and recommended work [S-K 1300 (b)(15)]

    Highest block uncertainty lies at corridor edges and in the 1574–1577 km northing gap between the two survey blocks (see sd map). Recommended: (i) 6–10 box cores at high-sd blocks with laser-scaled deck photographs to tighten the coverage→kg/m² calibration; (ii) AUV photo lines at 3.5 km spacing across the inter-corridor gap; (iii) backscatter acquisition to constrain buried nodules.

    6. Competent Person statement

    *To be completed and signed by the CP after review of the data, methods and this draft.*

    Attachments

    block_model_v0.csv (per-block abundance, sd, class), block_model_v0.png, coverage_v3_all_images.csv (per-image), FINDINGS-calibration-v1.md.

    Draft 2026-09-21 — Ocean Map. Data © GEOMAR/PANGAEA CC-BY.