Why it matters
Without AI aids, unexpected discoveries and unplanned risk zones stall works, trench encounters defy scheduling, and unreported features complicate post-excavation requirements.
From design to safety
Design
AI-guided site potential assessment
The AI reads HER data, geology, topography, known sites, and predictive models to generate a potential map and a risk-scoping report; it exports a formal risk zone list with coordinates, depths, and confidence scores for integration into the written scheme of investigation and site plans.
- HER data
- geology
- topography
- predictive models
Build
Inform trench planning and access routes
The system outputs risk-aware trench layouts and access corridors, guiding written scheme of investigation updates and alignment with site-prospecting findings; it records traceable decisions in context sheets, plan revisions, and hazard signage placements.
- topography
- known sites
- risk zones
Operation
Dynamic field targeting and updating
During surveying, the AI re-weights targets based on live geophysics, drone survey outputs, and archive checks, updating context sheets and plans; it creates provisional trench logs, sampling priorities, and on-site alert notes for supervisors.
- geology
- topography
- HER data
Maintenance
Data governance and archive handover
Post-excavation, predictive model recalibration uses archive deposition data, ensuring HER and planning records remain traceable and auditable; it assigns data owners, records provenance, and generates metadata for HER submissions and post-excavation reports.
- predictive models
- HER data
Safety
Trench safety through predictive zoning
The system flags buried-services risk zones and unstable trench sectors, triggering mitigations and sign-off by the qualified person; it logs safety drill outcomes, service plan updates, and contingency actions within the safety case and site diary.
- risk zones
- known sites
Part by part
- 1
magnetometer array
- Watch for
- Magnetic anomalies indicating discretely buried features or service routes.
- AI action
- Detect anomalies; trigger targeted GPR transects and cautious trench entry.
- 2
ground-penetrating radar antenna
- Watch for
- Subsurface layer discontinuities suggesting walls, ditches, or voids.
- AI action
- Update predictive map; propose trench sections and sample priorities.
- 3
drone-mounted LiDAR system
- Watch for
- Topographic rise/fall or micro-relief changes signaling features.
- AI action
- Refine topographic models; adjust trench grid and access points.
- 4
borehole log and sampling kit
- Watch for
- Stratigraphic interrupts or unexpected fills
- AI action
- Correlate with GIS layers; adjust context recording strategy.
- 5
archival HER data binder
- Watch for
- Inconsistent historic references or missing site records
- AI action
- Flag gaps; prompt staff to locate primary sources and update context sheets.
- 6
predictive-model database
- Watch for
- Model drift or new variable inputs
- AI action
- Retrain with new data; refresh risk-zones and potential maps.
Worked example
Development-site potential mapping for a 12 ha coastal agglomeration, phases 1–3 (Iron Age to Roman)
Illustrative scenario with sample data.
A development site near a historic port requires potential modelling to guide IWIs, trench layout, and archivist handover.
Signal in
HER records: port episode 9, site x coords 42.13N; geology: calcareous sands; topography: elevated terrace; known-sites: enclosure HUM-07; predictive-model: coastal settlement likelihood high; risk zones: shallow gas line.
System out
AI-generated risk map with priority trench sectors and updated IWI plan; briefing note to planning archaeologist and curator for HER submission.
Human sign-off: Project manager reviews model outputs with the archaeologists and HER curator before finalising trench designs and reporting.
How to roll it out
Phase 1
Weeks 1 to 4
Baseline data integration from HER, geology, and topography for initial potential map.
Phase 2
Weeks 5 to 8
Model calibration with early geophysics; adjust risk zones and trench layouts.
Phase 3
Weeks 9 to 12
Full integration into written scheme of investigation and site-specific plans.
Where people stay in charge
- Decisions about risk zones and trench priority must be signed off by the planning archaeologist or HER lead curator.
- All data used for predictive modelling must be auditable with provenance and version control.
- Trench design changes derived from AI outputs require project-manager sign-off after field-crew consultation and safety review.
Questions owners and crews ask
What data feeds the predictive model in the design phase and how is it validated?
The model consumes HER data, geology, topography, and known sites; validation occurs through cross-checks with historic maps and specialist input from the planning archaeologist.
How does AI influence trench design and recording in the build stage?
AI informs trench layouts, access routes, and context sheet priorities, then staff verify changes during trench plan reviews with the project manager and site supervisor.
What happens if new data contradicts the model during operation?
New data triggers reweighting of targets and possible plan amendments after a structured review by the field supervisor and GIS lead.
Who signs off post-excavation assessments informed by AI predictions?
The post-excavation assessment is signed off by the client’s heritage consultant and the local authority planning archaeologist.