1. Why stockpile volume accuracy is a financial issue, not a technical one
Replace with your own drone capture of a stockpile site
The conversation about drone stockpile surveys usually starts in the wrong place — with technology specifications and accuracy statistics. Before any of that matters, the operator needs to understand why the client is asking for this service. And the answer is almost always money.
Stockpile volume measurement errors compound into write-offs. A quarry that consistently underestimates its ore pile by 5% on monthly audits is not just producing inaccurate reports — it is misvaluing its inventory for financial reporting, miscalculating its ore feed to the processing plant, and accumulating discrepancies that surface at year-end as reconciliation losses. Documented cases show that a single quarry underestimating by 10,000 tonnes can face year-end write-offs in the range of $60,000–$70,000. In at least one documented case, switching to precise drone-based measurement revealed an additional $100,000 worth of material that manual methods had missed entirely.
When you deliver a drone stockpile survey, you are not selling a map. You are selling financial accuracy. The client’s mine manager, CFO, and auditors are the audience for your deliverable — not just the site surveyor.
2. Accuracy — what the numbers actually mean
In a formal validation study by Archer Western comparing drone data with robotic total station measurements, the average volume difference across all drone processing methods tested was 1.1%. An independent field test produced a volume within ±2.6% of a meticulous tape-and-total-station reference survey. The accuracy range in professional practice is approximately 1–3% when the workflow is executed correctly.
2.1 The variables that control final accuracy
| Variable | Low-error outcome | High-error risk |
|---|---|---|
| GSD / flight altitude | ≤3 cm/px at 80–100 m AGL | GSD >5 cm/px at high altitude |
| Overlap | 80% frontal, 75% sidelap minimum | Below 70/60 — reconstruction gaps |
| GNSS correction | RTK fixed or PPK processed | Standalone GPS or RTK float |
| Base plane selection | Correct method for terrain type | Wrong method for site geometry |
| Polygon edge placement | Precisely at material toe line | Inside material or over surrounding debris |
| Pile texture | Dry, lit, granular surface | Wet sheen, uniform colour (coal, sand) |
| Checkpoints | 1–2 independent validation points | No independent check — unverifiable |
| Repeat survey consistency | Same base plane, same polygon | Different annotation on each visit |
3. Pre-flight planning — the foundation of a good volumetric survey
3.1 GSD selection and flight altitude
| Flight altitude (AGL) | Approx. GSD (M4E) | Practical use |
|---|---|---|
| 60 m | ~1.6 cm/px | Small piles <1 ha, high detail, short battery range |
| 80 m | ~2.2 cm/px | Standard stockpile survey — recommended for most sites |
| 100 m | ~2.7 cm/px | Large stockyards >5 ha, acceptable accuracy |
| 120 m | ~3.3 cm/px | Large area coverage — borderline for volumetric accuracy |
| 150 m+ | >4 cm/px | Not recommended for stockpile volumetrics |
For tall stockpiles (8–15 m), reduce planned altitude by the estimated pile height to ensure top-of-pile GSD remains within tolerance.
3.2 Overlap settings
Frontal overlap: 80% minimum, 85% recommended for complex or steep pile geometries. Sidelap: 75% minimum — do not go below 70%. For the M4E at 80 m with these settings, a 10-hectare site typically requires one battery and approximately 300–450 images.
3.3 Mission boundary — include the ground
3.4 Time of day and sun angle
Fly when the sun is at 30–50° elevation — in tropical Guyana that means 07:00–09:00 or 15:00–17:00. Slightly raking light brings out surface texture and produces cleaner reconstructions, especially on dark-coloured materials.
4. Equipment setup — base station, checkpoints, and pre-flight
Replace with your field photo showing DJI RTK 3 deployed at a stockpile site
4.1 Base station positioning
The base station should be positioned centrally for maximum radio link coverage. For PPK, the base position matters for absolute accuracy, not radio range. If on a known benchmark: enter coordinates immediately in the correct geodetic reference frame. If on an unknown point: allow minimum 30 minutes static averaging — 60 minutes is better.
4.2 Checkpoint placement
Minimum standard for mining audit-quality volumetrics: 2 checkpoints per survey area at different elevations. Measure with RTK rover, record all coordinates in the same CRS as the survey.
Horizontal RMSE ≤ 3 cm for survey-grade mining work
Vertical RMSE ≤ 5 cm for survey-grade mining volumetrics
If vertical checkpoint residuals exceed 8 cm, reprocess with PPK backstop data before issuing the volumetric report
Document checkpoint residuals in the deliverable report — this is what auditors and mine managers inspect.
5. Mission execution
- Power on DJI RTK 3. Allow 5 minutes for satellite acquisition. Confirm base is logging raw observations (PPK backstop).
- In DJI Pilot 2: enable RTK, select ‘Fixed RTK’. Wait for Fixed solution stable for 60 seconds before launching.
- Note RTK fix time and satellite count. Photograph the RC Plus 2 RTK status screen as a timestamped record.
- Confirm PPKRAW.bin logging is enabled in DJI Pilot 2 settings.
- Load pre-planned mapping mission. Verify: altitude = 80 m AGL, frontal overlap = 80%, sidelap = 75%, terrain follow enabled if required.
- Launch. Monitor RTK status throughout. Note any Float transitions with timestamp.
- Sustained Float (>2 min) triggers PPK processing as primary accuracy method.
- On return: do not power off the base station until PPKRAW.bin is confirmed in the image folder.
- Download images from M4E. Verify PPKRAW.bin is present — hard check before leaving the site.
- Download base station raw file (DAT from DJI RTK 3, RINEX .obs from Emlid RS2+).
- Measure independent checkpoints with GNSS rover. Record all coordinates with timestamp and fix status.
- Review flight log for RTK solution history. If Float detected, confirm PPK processing for full dataset.
- Pack base station. You cannot recover the base log once overwritten. Collect it now.
6. Processing in Agisoft Metashape — the full workflow
Replace with your own Metashape processing screenshot
- File → New. Save to a folder named with date, site, and client. Consistent naming is critical for repeat-survey archives.
- Workflow → Add Photos. Import PPK-corrected images (if using PPK) or original RTK-geotagged images.
- In the Reference pane: set coordinate system to WGS84 / UTM Zone 21N for Guyana. Verify camera positions appear correctly on the map view.
- Workflow → Align Photos. Settings: Accuracy = High, Reference Preselection = Source, Adaptive Camera Model Fitting = enabled.
- Target: >95% of images aligned. Below 90%: check for motion blur, insufficient overlap, or PPK geotag mismatch.
- If using GCPs: add markers, assign coordinates, mark each in 4–5 images, then run Optimize Cameras.
- Workflow → Build Dense Cloud. Quality = High. Depth Filtering = Mild (preserves pile edge detail).
- Processing time: 20–60 minutes for 300–500 images on a modern laptop. Enable GPU in Tools → Preferences.
- Inspect for holes over smooth, reflective, or dark material surfaces. Areas with holes produce unreliable volume estimates.
- Workflow → Build DEM. Source = Dense Cloud, Interpolation = Enabled, CRS = WGS84 / UTM Zone 21N.
- Also build the Orthomosaic (Workflow → Build Orthomosaic) — used for polygon digitisation and client deliverable.
- Export both as GeoTIFF for archival and client delivery.
7. Base plane selection — the most important decision in the workflow
Replace with a diagram or annotated screenshot showing Linear Fit vs Lowest Point vs Triangulated
7.1 Linear Fit (best fit plane)
Calculates a flat plane in 3D that best fits the edge points around the pile base. Correct for standalone piles on sloped or undulating ground. The safest default when unsure.
7.2 Lowest Point
Creates a horizontal plane at the elevation of the lowest edge vertex. Only correct for piles on perfectly flat, level ground — concrete bays, level compacted pads. On sloped ground, this systematically overestimates volume.
7.3 Triangulated Surface
Joins all edge vertices with triangles to create an irregular 3D base surface. Most accurate for adjacent piles with shared boundaries or heavily disturbed ground at pile edges.
Isolated pile on sloping ground → Linear Fit
Pile in flat concrete bay or level pad → Lowest Point
Adjacent piles with shared boundaries → Triangulated
Complex terrain with truck traffic disturbance → Triangulated
Repeat surveys comparing to previous flight → Previous survey DEM as base
When unsure → Linear Fit is the safest default
Always annotate your choice in the deliverable so the client’s team uses the same method next time.
8. Volume calculation in Metashape
- Switch to Ortho view. Select the Draw Polygon tool. Trace the pile boundary precisely at the material toe line, using 8–15 vertices.
- One polygon per individual pile. Do not draw a single polygon around the entire stockyard.
- Close the polygon. Right-click inside and select Measure Volume.
- Select the appropriate base plane method (see Section 7).
- Metashape displays Cut volume (material above base) and Fill volume (depressions below). Material Volume = Cut volume.
- Record volume in m³ with base plane method, vertex count, and timestamp.
- Repeat for each pile. Export via File → Export Report.
- Cross-reference DEM elevation at each checkpoint location against the surveyed elevation.
- Target: ≤5 cm absolute vertical difference at each checkpoint.
- If any checkpoint residual exceeds 8 cm: do not issue the volumetric report. Reprocess with PPK or re-fly.
- Record all checkpoint residuals in the deliverable documentation.
9. DJI Terra — the fast-turnaround field option
DJI Terra is included with the M4E purchase and supports the complete stockpile workflow including Local PPK processing, DEM generation, and stockpile volume tool. For simple sites, it delivers fully acceptable results faster than Metashape.
- PPK processing: Settings → GNSS → Local PPK. Add base station file, enter base coordinates, calculate.
- Reconstruction: High quality. A 300-image mission reconstructs in approximately 20–30 minutes.
- Volume tool: Draw polygon at pile toe, select base plane method, read output.
- Export: DEM as GeoTIFF, orthomosaic as GeoTIFF, volume report as PDF.
10. Converting volume to tonnage
Tonnage = Volume (m³) × Bulk density (t/m³). The bulk density figure must match the stockpiled (loose) state of the material.
| Material | Typical bulk density (t/m³) | Notes |
|---|---|---|
| Alluvial sand (dry) | 1.5–1.7 | Higher when moist |
| Alluvial gravel (dry) | 1.4–1.6 | Depends on particle size distribution |
| Crushed limestone | 1.3–1.5 | Loose stockpiled state |
| Laterite / iron ore | 1.8–2.2 | Highly variable — confirm with client |
| Gold-bearing alluvial ore | 1.6–2.0 | Bulk density for run-of-mine material |
| Coal (bituminous, loose) | 0.7–0.9 | Low density — easily underestimated |
| Construction fill | 1.5–2.0 | Depends on clay/rock fraction |
| Crushed concrete | 1.2–1.5 | Variable with debris composition |
For audit-quality reporting, document the bulk density source (lab measurement, client-provided, or handbook reference) alongside the volume figure in every deliverable.
11. Deliverable format — what the client report should contain
Replace with a screenshot or mock-up of your actual report template
A professional stockpile volumetric report is a documented, auditable technical record. It is not a screenshot and a table. The standard deliverable package includes:
- Report PDF: Survey metadata, CRS, GNSS method, flight parameters, checkpoint residuals, base plane method per pile, volume in m³, bulk density source, tonnage per pile, site total.
- Spatial data: Orthomosaic and DEM as GeoTIFF, point cloud as LAS, volume polygons as shapefile or DXF.
- Tabular data: CSV with Pile ID, volume, bulk density, tonnage, base plane method, survey date.
1. Base plane method — and whether it matches the previous survey
2. Independent checkpoint residuals — the primary accuracy evidence
3. GNSS correction method and whether fixed solution was achieved throughout
4. Bulk density source and internal consistency with accounting records
5. Whether polygon boundaries are consistent with the previous survey
12. Complete workflow checklist — field to report
- Flight altitude set for target GSD ≤3 cm/px (typically 80 m AGL for M4E)
- Frontal overlap ≥80%, sidelap ≥75%
- Mission boundary extends 20–30 m beyond outermost pile on all sides
- Terrain following enabled if pile height or site elevation variation >5 m
- Sun at 30–50° elevation — early morning or late afternoon
- Site coordination: clear airspace over stockyard during mission window
- DJI RTK 3 on known point or averaged position (≥30 min if no benchmark)
- Base station logging active — verify DAT file indicator
- PPKRAW.bin logging enabled in DJI Pilot 2 settings
- RTK Fixed solution confirmed ≥60 sec before mission launch
- 2 checkpoints measured with GNSS rover, coordinates recorded
- RTK solution status monitored throughout (Float periods timestamped)
- Mission completes all flight lines — no partial data over pile areas
- Base station remains powered until aircraft has landed
- PPKRAW.bin confirmed in image folder before packing
- Base station DAT / RINEX log downloaded before departing site
- PPK processing completed if any RTK Float periods in log
- Photos aligned in Metashape — ≥95% alignment rate
- Dense cloud built at High quality, Mild depth filtering
- DEM and orthomosaic generated, exported as GeoTIFF
- Checkpoint residuals checked: H ≤3 cm, V ≤5 cm RMSE
- One polygon per individual pile, vertices at material toe
- Base plane method documented and consistent with previous surveys
- Volume in m³ recorded per pile with method noted
- Tonnage calculated using client-confirmed bulk density
- Report PDF, GeoTIFF, CSV, and shapefile prepared
- Checkpoint residuals included in report documentation
- All files archived with date, site, and client reference
13. Common errors and how to avoid them
| Error | Consequence | Prevention |
|---|---|---|
| Mission boundary too tight | Cannot establish base plane; all volumes unreliable | Extend 20–30 m beyond outermost pile |
| Wrong base plane method | Systematic volume error of 5–15% | Follow Section 7 guide; document method |
| Polygon redrawn on repeat surveys | Volume ‘change’ that is not real | Save polygon as shapefile; reuse every visit |
| Overlooking RTK float periods | Random geotag errors in reconstruction | Review flight log; use PPK backstop |
| Photogrammetry over wet shiny material | Reconstruction holes over pile surfaces | Survey in dry conditions; increase overlap |
| Noon overhead sun | Loss of surface texture; noisy tie points | Fly early morning or late afternoon |
| Bulk density from handbook, not client | Tonnage error independent of volume accuracy | Confirm density in writing before reporting |
| No independent checkpoints | Cannot defend accuracy if challenged | Minimum 2 checkpoints per site every survey |