The Gold Standard of Archaeology | Geoarchaeological Core Correlation, Facies Boundaries and Sampling Gaps

1. Start with the archaeological question and landscape model

Back to contents

Decide whether the cores need to locate a channel, reconstruct a buried surface, test deposit survival or connect excavation contexts. That question determines spacing, depth, orientation and the evidence needed for correlation. Place each borehole in a common coordinate and elevation system. A depth of three metres below ground level is not comparable across a sloping or made landscape until surface elevation and datum are known.

2. Treat every core as a sampled column

Back to contents

Record drilling method, diameter, run lengths, recovery, refusal, compaction, disturbance and open intervals. A one-metre drive that recovers seventy centimetres has thirty centimetres whose position or character is uncertain; stretching the recovered material to fill the run fabricates continuity. Mark the top and base conventions used by the rig and laboratory. Tool changes can alter apparent texture or smear material across a boundary.

3. Log facies with observable attributes first

Back to contents

Describe grain size, sorting, colour, structure, inclusions, organic content, carbonate reaction, contacts and post-depositional features before naming an environment. “Grey silty clay with diffuse organic streaks” is an observation. “Tidal-flat deposit” is an interpretation requiring wider evidence. Standard terms improve comparison, but a code should not conceal uncertainty or force gradual change into sharp units. Retain photographs and, where useful, split-core scans or subsamples.

4. Distinguish boundaries, transitions and omissions

Back to contents

A sharp contact can mark erosion, rapid environmental change, truncation during drilling or simply the boundary between recovered runs. A diffuse contact may reflect gradual deposition, bioturbation or sampling resolution. Refusal can occur on gravel, masonry, compact clay or an obstruction. Do not draw a facies boundary through an unsampled interval as if its exact position were observed. Bound it between the last secure material above and below.

5. Correlate several properties in elevation space

Back to contents

Compare facies successions, contact elevations, particle size, geochemistry, microfossils, magnetic susceptibility and dated horizons where appropriate. Similar sediment can form at different times, while the same depositional surface can vary laterally in texture. Draw more than one plausible correlation and state which observations favour each. A cross-section should distinguish observed core intervals, interpolated boundaries and extrapolation beyond the sampled area.

6. Worked example: a buried channel between four cores

Back to contents

Four invented boreholes cross a development site. BH1 contains peat from 2.1 to 2.4 metres above datum over sand. BH2 has peat from 1.8 to 2.2 metres; BH3 recovers organic silt at 1.6 to 1.9 metres with a twenty-centimetre recovery gap below; BH4 contains only gravel above 1.4 metres. Colour matching alone might join all organic units as one level. Pollen and a dated tephra instead link BH1 and BH2, while BH3 lacks the marker and contains younger pottery in the overlying fill. The section therefore shows a secure BH1–BH2 correlation, a provisional BH3 alternative and no peat boundary at BH4. The inferred channel margin lies between BH3 and BH4, not on a precisely observed line.

7. Use ages to constrain, not decorate, correlation

Back to contents

Radiocarbon, luminescence, tephra, artefacts and historical records date different events and materials. A bulk organic sample may contain mixed ages; a shell can be reworked; a luminescence sample may not have been fully reset. Plot calibrated ranges and stratigraphic order rather than single central dates. If an age conflicts with the logged sequence, investigate context, residuality, contamination and laboratory assumptions before moving a boundary to fit it.

8. Changed-context transfer check: an urban basement site

Back to contents

Move from an open floodplain to a city block with basements, utilities and made ground. The transferable method still records each sampled column, ties it to elevation and separates observation from correlation. The new context adds truncation and anthropogenic heterogeneity. Two nearby cores can legitimately differ because one crossed a cellar backfill and the other preserved soil beside a wall. A smooth natural facies model would erase the very archaeology the coring was meant to detect.

9. Know the limits and common misconceptions

Back to contents

Dense coring does not guarantee complete recovery, and close spacing does not remove small-scale heterogeneity. Identical colours are not time-equivalent markers. A line on a section is not a directly observed surface between boreholes. Absence from a core is weak evidence when the target is thin, patchy or smaller than the sampling footprint. Automated clustering and interpolation can organise data, but they cannot decide whether units are genetically or chronologically equivalent without archaeological reasoning.

10. Audit the correlation from rig to section

Back to contents

Archive borehole coordinates, ground levels, drilling logs, recovery per run, photographs, sample depths, laboratory methods, facies codes, dating records and the versioned correlation model. Keep original logs unchanged when interpretations are revised. Use symbols for refusal, disturbed recovery and unsampled intervals. Test how channel margins or buried surfaces move under alternative correlations and sampling-gap placements. The final section should let another analyst see where evidence ends and interpolation begins.

11. Frequently asked questions

Back to contents

How many cores are enough? There is no universal number; spacing follows feature scale, heterogeneity, access and decision risk. Can two similar peats be correlated by colour? Not reliably without stratigraphic and independent support. Does refusal mean bedrock? No; it records that the chosen method could not advance. Can a gap be assigned to the layer above or below? Only as an explicit interpretation with bounds. Should every core be sampled in the laboratory? Targeted analysis is often appropriate, but selection rules must be recorded.

12. The gold standard keeps the buried landscape conditional

Back to contents

Good core correlation turns narrow, imperfect columns into a testable landscape model. It preserves elevations, recovery and facies observations; uses independent evidence to constrain links; and shows uncertainty where cores do not decide the geometry. The goal is not the smoothest cross-section. It is the most defensible explanation of what may connect, what may change laterally and where another borehole or excavation would reduce the uncertainty that matters.