Hobart, Tasmania · bare earth with buildings removed, flow paths traced · Showcase
Accuracy
Accuracy
Accuracy is measured on every job and stated per job. Measured figures by scan type, each with its basis and its limits.
By scan type
Measured,
then stated.
Every GroundMap delivery carries an accuracy note: the check points used and held out of processing, the statistics, and what was not tested. The figures below are measured on showcase data, by scan type. Each has its basis and its limits, and none is quoted as the accuracy of another dataset or job.
| Drone and ground scanning | El Teniente underground study: within 1 cm on 98.7 % where two or more scans overlap. Basis ↓ |
|---|---|
| Government aircraft LiDAR | 43 State survey marks held out of GroundMap's processing: median −3 mm, 33 of 43 within 10 cm, RMSE 0.12 m. Statistics and limits ↓ |
| GroundMap Fusion, Hobart | On stable open ground, checked independently of the alignment, the 2011 survey agrees with the 2024 survey to a median −6 mm (NMAD 64 mm, 95th percentile 157 mm; 94.4 % within 15 cm): relative agreement between surveys, not absolute accuracy. Basis ↓ |
| Client jobs | Measured on its own check points, held out of processing, and stated in the job's accuracy note. How ↓ |
Drone and ground scanning
Scan to scan,
within 1 cm.
Terrestrial, mobile and underground laser scanning, and drone capture: registration measured wherever scans overlap.
- El Teniente underground study: within 1 cm, where two or more scans overlap
- 98.7%
| El Teniente underground study | Underground laser scanning. Within 1 cm on 98.7 % where two or more scans overlap. 44 underground laser scans registered without GPS: each scan within 1 cm of the shared surface, where two or more scans overlap (scans used in the registration; not held out). The El Teniente study Showcase: GroundMap processing of public survey data. Survey data: Chilean underground mine dataset, Advanced Mining Technology Center (AMTC), Universidad de Chile (Leung et al., IJRR 2017; CC BY-NC-ND 3.0 CL). Processing: GroundMap. No affiliation with the mine operator or the dataset authors. |
|---|
These figures are registration agreement between scans, not absolute accuracy against control. On a job, the same registration figures are reported for the job's own scans, beside the residuals at its control and check points.
Government aircraft LiDAR
Held-out
survey marks.
Finished ground checked against levelled State survey marks that never entered GroundMap's processing.
- Median difference, finished ground minus mark
- −3mm
- Marks within 10 cm
- 33of 43
- RMSE (RMSEz 0.116 m)
- 0.12m
Full statistics
| Check points | 43 levelled State survey marks: 12 flush with the ground, 31 set in rock. Third-order levelling (42) and fourth-order (1), heights on AHD. |
|---|---|
| Median difference | −0.003 m |
| Mean difference | −0.019 m |
| Standard deviation | 0.115 m |
| NMAD | 0.078 m |
| Standard error of the median (≈ 1.25 × NMAD / √n) | 0.015 m |
| RMSEz | 0.116 m |
| 95 % (ICSM v2.0, 1.96 × RMSEz) | 0.23 m |
| 95th percentile of absolute difference | 0.25 m |
| Within 0.10 m | 33 of 43 (76.7 %) |
| Within 0.30 m | 42 of 43 (97.7 %) |
| Largest difference | −0.371 m |
| Checked | 20 October 2026 |
Difference = finished ground minus mark. Readings of 20 October 2026. At 0.23 m (95 %), the method check is within the 0.30 m vertical limit ICSM v2.0 sets at 95 % confidence.
How the marks were used
A check of the method against State survey marks held out of GroundMap's processing.
- Held out of GroundMap's processing. The State's marks were split in two by their State site number, by a rule fixed before any mark was read. The held-out half was never used to align, correct or tune anything, and no height was read from it until the check.
- Reading rules. One reading per mark: the median height of the finished ground returns within 1.5 m of it, minus the mark's AHD height. A mark counts only with at least five ground returns within 1.5 m; no wider window is used. Open, relatively flat ground only (slope × 0.30 m within 0.05 m, about 9.5°; under 30 % of returns within 3 m more than 2 m above the ground). A mark that fails a rule is listed with its reason and left out of every figure.
- Levelled marks decide. Only levelled marks flush with the ground or set in rock decide the figures. Kerb, pit and box marks sit beside a step or below the surface, and never decide.
- Discs excluded by a rule fixed before any figure was printed. Discs set in concrete, and marks with no description, were to be left out if their median differed from the flush and rock marks' by more than 0.05 m. It did: +0.077 m against −0.003 m. The reading picks up the grass and concrete around a disc rather than the ground, so discs are left out of every figure on this page.
- The finished ground as output. The readings were taken on the finished ground exactly as GroundMap's processing writes it out, and a sample of readings was cross-checked against a separate export of the same ground to within 1 mm.
What it does not prove
- No quality level claimed. The check tests the method; no ICSM quality level or accuracy class is claimed for it.
- Not an ICSM-compliant assessment. The figure is computed in the ICSM v2.0 form, but the number, distribution and photographic record of the check points fall short of section 3.6 for the area the marks span.
- Not a certificate for any survey. No source survey holds enough marks to be judged on its own, and the check certifies no survey, area or job.
- Not a comparison with the sources. No claim is made against the source surveys' own stated accuracy.
- Control use not known. Whether a source survey's provider used any of these marks as control is not known.
- Not tested: horizontal accuracy, vegetated ground, slopes over about 9.5°, ground beside structures, and change between the levelling and the flight.
- Check point error not included. The State marks publish no height uncertainty.
Survey marks: LIST Survey Control, the LIST © State of Tasmania, CC BY 3.0 AU. Aircraft LiDAR: public surveys published on theLIST (State of Tasmania) and obtained via ELVIS (Geoscience Australia), under the licences recorded with them (theLIST: © State of Tasmania, CC BY 3.0 AU). Processing: GroundMap.
GroundMap Fusion, Hobart
Relative
agreement.
The 2011 survey (Mount Wellington 2011) aligned to the 2024 survey, then checked against it on stable open ground, independently of the alignment: how closely two surveys agree, not how close either is to the ground's true height.
- Median, 2011 minus 2024, stable open ground after alignment
- −6mm
- NMAD
- 64mm
- 95th percentile of absolute difference
- 157mm
On stable open ground, checked independently of the alignment, the 2011 survey agrees with the 2024 survey to a median −6 mm (NMAD 64 mm, 95th percentile 157 mm; 94.4 % within 15 cm): relative agreement between surveys, not absolute accuracy.
The figure above is checked on open ground independently of the alignment; it measures consistency between surveys. The Hobart model has no figure against survey marks: no levelled mark flush with the ground or set in rock, of those held out, lies within its 2024 survey.
Showcase: GroundMap processing of public survey data, not client work. Survey data: Hobart City 2023/24 (captured by Fugro for the State of Tasmania; theLIST; CC BY 3.0 AU), Climate Futures Derwent 2008 (captured by DiMAP for ACE CRC; ACE CRC licence) and Mount Wellington 2011 (no licence on record), via ELVIS. Processing: GroundMap. The Hobart showcase
Every job
The accuracy
note.
Each delivery carries a note with its own check points, statistics and limits.
Check points
Independent of control and calibration, on open, level, hard ground, spread across the area: at least the count ICSM v2.0 section 3.6.4 sets for the area, and 30 or more wherever an ASPRS figure is wanted.
Rules first
Counting rules, the point types that decide and the figure the contract asks for are fixed and recorded before any point is read. A rule changed after a number is seen is void.
Read and report
Each point read from a TIN of single returns (ICSM v2.0 section 3.4.1) and on the delivered surface, reported in both forms: ICSM non-vegetated vertical accuracy at 95 %, and RMSEV (ASPRS Edition 2). Every point left out is listed with its reason.
Sign-off
The note states what was not tested and who reviewed it. A registered surveyor is engaged wherever a job needs survey sign-off.
Accuracy statement
Standards
and credit.
| Download | Accuracy statementPDF · 2.2 MB |
|---|---|
| National form | ICSM Australian Airborne LiDAR Acquisition Specification, Version 2.0 (2025), sections 3.4 to 3.9: non-vegetated vertical accuracy at the 95 % confidence level (1.96 × RMSEz). |
| International form | ASPRS Positional Accuracy Standards for Digital Geospatial Data, Edition 2 (2023): RMSEV, reported as found. |
| Datums | GDA2020 / MGA; heights on AHD. |
| Credit | Survey marks: LIST Survey Control, the LIST © State of Tasmania, CC BY 3.0 AU. Aircraft LiDAR: public surveys published on theLIST (State of Tasmania) and obtained via ELVIS (Geoscience Australia), under the licences recorded with them (theLIST: © State of Tasmania, CC BY 3.0 AU). Processing: GroundMap. |