Oil & Gas Equipment & Services Data: Wells, Steel, Sensors and Barrels · Head-to-head
NSTA Open Datasets on data.gov.uk (Wellbores, Fields, Licences) vs NORA 3D Multimodal Oil & Gas Dataset
Which oil & gas equipment & services data: wells, steel, sensors and barrels data fits your job: NSTA Open Datasets on data.gov.uk, or NORA 3D Multimodal Oil & Gas Dataset. API, files, or your warehouse. Daily, weekly, or hourly.
NSTA Open Datasets on data.gov.uk (Wellbores, Fields, Licences)
NORA 3D Multimodal Oil & Gas Dataset
Coverage, side by side
| NSTA Open Datasets on data.gov.uk | NORA 3D Multimodal Oil & Gas Dataset | |
|---|---|---|
| Geographic | United Kingdom Continental Shelf offshore plus England/UK onshore petroleum areas, publishable in ED50, WGS84, ETRS89 or British National Grid | A single 15 m x 7.5 m x 5 m mock-up at the SENAI ISI-SIM facility in Brazil, in scanner-local coordinates |
| Temporal | Live register; each wellbore row keeps its own spud, total-depth and completion dates while statuses read present tense | Fixed capture of March 2026; version 2 published 30 March 2026 and already superseded by a newer cut |
What each contains
Pick by fit, not by loyalty.
| NSTA Open Datasets on data.gov.uk | NORA 3D Multimodal Oil & Gas Dataset | |
|---|---|---|
| Publisher | UK North Sea Transition Authority (NSTA), regulator of the United Kingdom Continental Shelf | SENAI Innovation Institute for Sensing Systems (ISI-SIM) with Shell Brasil, released on Zenodo |
| Subject lens | — | One full-scale piping-and-instrumentation mock-up captured as annotated LiDAR, paired with an as-designed CAD model, P&ID diagrams and 18 HDR panoramas |
| Geographic coverage | United Kingdom Continental Shelf offshore plus England/UK onshore petroleum areas, publishable in ED50, WGS84, ETRS89 or British National Grid | A single 15 m x 7.5 m x 5 m mock-up at the SENAI ISI-SIM facility in Brazil, in scanner-local coordinates |
| Temporal coverage | Live register; each wellbore row keeps its own spud, total-depth and completion dates while statuses read present tense | Fixed capture of March 2026; version 2 published 30 March 2026 and already superseded by a newer cut |
| Detail level | — | Per-point semantic class and instance ID across a consolidated cloud, plus per-component CAD entities and diagram symbols |
| Modalities | Tabular attributes joined to point and polygon geometry | Point clouds (raw multi-scan e57, annotated ply), parametric CAD, PDF diagrams, HDR imagery |
| Scale | 922 datasets; the top-holes layer alone holds 9,500 wellbore origins in roughly 17 MB | Roughly 7.7 GB across four bundles devoted to a single scene |
| Field documentation | Verified during research, column by column, following PPDM 'What is a well' | Verified against the record; the complete class taxonomy ships only inside the metadata bundle |
| Best for | Mapping, screening and citing UK activity by well, block, subarea or field | Training and benchmarking perception models on realistic industrial geometry |
What each does better
the NSTA registry
Four coordinate systems, one estate. Features publish in ED50, WGS84, ETRS89 and British National Grid, which matters when joining legacy survey files against modern GIS stacks. Because it is a live register, statuses read present-tense: a 2025-spudded appraisal well appears as Abandoned Phase 3 with its complete date spine intact.
the NORA 3D Multimodal Oil & Gas Dataset
Supervision raw scans never include. Every point in nora.ply carries both a semantic class and an instance ID, so repeated identical components stay separable - the difference between 'flanges are here' and 'that specific flange'. That is ready-made ground truth for semantic segmentation, instance segmentation and object detection.
As-built meets as-designed. The registered cloud pairs with a clean parametric CAD model of the same structure and P&ID diagrams of the same process, which makes digital-twin alignment - fitting reality onto design - a benchmark exercise rather than a scavenger hunt across unrelated releases.
Full scale, real messiness. The subject is a genuine 15 m x 7.5 m x 5 m industrial mock-up scanned from 18 positions under a Design-of-Experiments acquisition plan, with HDR panoramas aligned to each position for camera work layered on top of geometry. See LiDAR point clouds for why per-point labeling is the scarce commodity in this industry.
Where they're equivalent
More than their shapes suggest. Both field dictionaries are research-verified rather than inferred - a grade many cataloged datasets never earn. Both score above the catalog's 7.81 average (9 and 8 respectively). Both are anchored to formal standards - PPDM well definitions on one side, ISO 10628 and ISA 5.1 diagram conventions on the other. And neither is a time series: the registers hold event dates per wellbore rather than monthly aggregates, and NORA holds a single instant. These two answer where and what; when-and-how-much needs different tools.
The verdict
Verdict: sample both, pick by fit - they are different instruments pointed at the same industry.
Accept the ceiling: reference geometry only - no subsurface well files, no production series, hard stop at UK jurisdiction.
Take NORA if your question is perceptual: segmentation baselines on industrial geometry, instance-level tagging prototypes, as-built versus as-designed alignment experiments, sensor and camera placement studies. Accept the ceiling: one scene, one capture, no time dimension and no production plant.
Journalists and academics and market researchers lean toward the registers; data scientists and ML engineers reach for NORA first and the registers when a model needs somewhere real to deploy.
Sample both, pick by fit. See NSTA Open Datasets on data.gov.uk · See NORA 3D Multimodal Oil & Gas Dataset
Or take both in one feed
Yes - as a train-here, deploy-there pipeline rather than a row-level merge. The registers say where the industry's assets are and who operates them; NORA teaches a model what such assets look like at close range. A defensible workflow: train and benchmark perception on NORA's labeled scene, then apply the trained model to inspection captures of North Sea assets identified in the registers.
Three alignments decide whether the combination holds. First, frames: national coordinate systems versus a scanner-local millimetre frame - any overlay needs an explicit georeferencing pass. Second, vocabularies: the administrative wellbore fields and the component-class taxonomy share no codelist, so the join stays conceptual, never row-level. Third, vintage: a live register against a static capture that a newer version has already superseded - pin the exact cut before citing either. Browse the rest of the shelf at the oil gas equipment services data hub.
Datadory ships either record alone or both merged onto one calendar, delivered daily, weekly, or hourly - your call. Or take both in one feed.
API, files, or your warehouse. Daily, weekly, or hourly.
Fair questions
Do the two datasets cover the same ground?
Only in the sense that both describe physical oil and gas assets. The registers cover the whole United Kingdom Continental Shelf in four coordinate systems at row-per-well granularity; NORA covers fifteen by seven and a half by five meters in Brazil at per-point granularity. There is no shared identifier between a UK well origin reference and a point index inside the mock-up, so nothing joins at the row level.
Which one should feed a machine-learning model?
For perception - segmenting, detecting, tagging components - NORA, without competition: it is the only member of the pair shipping per-point labels and instance IDs. For anything geographic - geocoding assets, screening acreage, enriching a map layer - the registers win, and they hand a deployed model somewhere real to be applied. Many teams train on NORA and place with the registers.
Does either dataset carry time series worth trending?
Neither, honestly. The registers carry event dates per wellbore - spud, total depth, completion - plus present-tense statuses, so histories can be reconstructed one well at a time but no monthly aggregate series exists. NORA is a single capture instant, and a newer version of the record has already superseded its second cut. Trend work belongs to the production and pricing feeds elsewhere in the catalog.
Can Datadory deliver both datasets together?
Yes - alone or merged onto one calendar, delivered daily, weekly, or hourly, your call. Each arrives normalized to its documented field dictionary with sample rows for inspection. The combination work is conceptual alignment - coordinate frames reconciled, label vocabularies kept separate, vintages pinned - handled in the merge. Or take both in one feed.