Prospection, Prediction and Management of Archaeological Sites in Alluvial Environments

1. Prospection, Prediction and Management
of Archaeological Sites in Alluvial
Environments
Keith Challis, Mark Kincey and Andy J Howard,
IBM Vista, University of Birmingham
U B

2. Outline
• Study Areas
• Prospection
• Assessing Preservation
• Predictive Management
• Critique

3. Study Areas
Three, c.300km2
study areas
• MTV Derbyshire
• MTV Newark
• LTV Gainsborough
30k ha terrace
35k ha floodplain
144 SAM
11,222 HER records
2254 Aimee (NMR) records

4. Prospection

5. Prospection
• Traditionally focused
on known sites
• Conventional
photography
• Landscape scale
• Filling gaps
• Newer digital
techniques

6. Prospection
Conventional Aerial
Photography
• Extensive opportunistic
flying
• Good years reveal much
• Not Systematic
• Limited by crop response
• Visible spectrum(390-
750nm)

7. Prospection

8. Prospection
Multispectral and
Hyperspectral
• No systematic surveys
• Research suggests huge
potential
• Extended visibility beyond
visible spectrum
• Rapid area surveys
• Satellite imagery

9. Prospection

10. Prospection
Airborne Lidar
• 3D record of topography at
very high resolution
• Systematic survey by
Environment Agency
• Upstanding and buried sites
• Assessment of preservation
• Change detection (multi
temporal survey)

11. Prospection

12. Assessing Preservation
Hyperspectral Techniques
• Adaptation of vegetation
monitoring
• Ratios of reflectance in
different spectral bands
reflect vegetation vigour
• NDVI (Normalised
difference vegetation)
• Tasselled Cap

13. Assessing Preservation

14. Assessing Preservation
Airborne Lidar Intensity
• NIR reflectance enhanced
detection of vegetation
and soil properties
• Not a robust indicator
• No systematic collection
• Much work to be done

15. Assessing Preservation

16. Predictive Management

17. Problem
• Information on heritage assets resides in
expert hands
• Issues of availability / confidentiality
• Discrete, not continuous record
• Articulated need for “red flag” mapping
• How to achieve this without alienation of
some stakeholders

18. Goals
• To provide interpreted
information to non-expert
users
• Models rooted in
knowledge base
• Not to usurp the HER as
a source of data or to
undermine curatorial
prerogative

19. Approach: User-focused
• Understand what users
need, how they think and
work
• Model the knowledge-
based approach of expert
users “topsight” (Gelernter
1992)
• Presentation of results
structured to fit the real-
world and in a user friendly
medium

20. Approach: Simplify
• Inductive (data driven)
rather than deductive
(theory driven)
• Simplify and summarise
(the detail is in the HER)
• Validate through user
feedback (rather than test
and quantify)

21. Approach: Model Objectives
The completed models will provide per parcel scores for:
• The predicted archaeological potential of all land parcels.
• The aggregate bearing potential and value of all land parcels.
• The susceptibility of individual land parcels to field evaluation
techniques.
• The likely physical condition of buried cultural remains.
• The risk of encountering buried waterlogged organic remains.
• The level of impact that different forms of extraction may have on
the archaeological record
• The importance of archaeology in the light of regional priorities.
• The likely mitigation needs in the light of PPG 16 guidance

22. Method: Predictive Models
• Classic predictive
modelling
• Big, empty,
heterogeneous areas
(2500km2 /
21 sites)
• Assess and weigh
environmental factors
• Weights inform model

23. • Such models are
inappropriate for the
TV
• c. 40% of land parcels
contain a record
• Eg. Newark, 1254
parcels out of 5012
Method: Predictive Models

24. Method: Our Data Model
• OS MasterMap® as a
spatial framework
• Raster based GIS
models
• 50m grid (200k cells)

25. Method: Model Building
• Source data is rasterised
• Simplified scores are
applied or extracted
• Models are based on
weighted means of
scores
• Blank areas filled using
landscape classification
and spatial modelling
Terrace: Score = 3

26. Method: Per Parcel Results
• Calculations reclassified
to 5 level scale from low
risk to high
• Aggregated model scores
devolved to level of an
OS MasterMap® TOID
• Built up parcels, water
and parcels less than 1ha
in extent excluded

27. Delivery
• Data for stakeholder GIS (tables of values for each TOID)

28. Delivery
• Interactive using embedded Google Earth application

29. Critique
• “Topsight” is not necessarily the same thing as predictive
modelling or risk mapping
• Modelling period based activity and intensity of activity is
problematic
• It would be possible model individual classes of
monument with clear geographic preferences (eg burnt
mounds)
• Perhaps general models are the most helpful
• The meaning of results is imprecise and open to
misinterpretation

30. Concluding Thoughts
Prospection
• Non-photographic techniques offer huge potential, but uptake issues
(availability, cost, education)
• In fact, in times of limited finances reliance on traditional techniques may
not be cost-effective
Predictive Management
• Need for strategic management of heritage assets is axiomatic
• System adoption requires clear joined-up thinking at high level
• Possible conflict with aggregate resource assessments in England
• Do we need another level of information?