Engineering & Technology · Geotechnical Engineering & Ground Systems

Ground investigation, geological context and sampling limits

A borehole shows what was encountered along one narrow path. A foundation or tunnel affects a much larger volume. Geotechnical investigation connects those different scales without pretending that an unseen volume has been fully inspected.

Review force balance, normal and shear stress, strain, density versus unit weight, fluid pressure and hydraulic head. Read graph axes and convert kN/m² to kPa, metres to millimetres and seconds to days. Ground below the surface is spatially variable; the lessons introduce every soil-specific symbol and keep saturated, hydrostatic and loading assumptions explicit.

Back to the chapter

After reading, you can

Build a provisional ground model from geological context and bounded observations, and identify a useful next investigation.

Start with the questions the project needs answered

A ground model describes the arrangement and behaviour of soil, rock and groundwater relevant to a project. It combines observations with interpretations. Begin with the proposed loads, geometry, movement tolerances and construction stages: a shallow pavement and a deep tunnel need different information even at the same address.

Geological history offers testable explanations. River deposition may leave changing channels and lenses; weathering can weaken rock unevenly; old filling or excavation may conceal man-made boundaries. Maps, historical records, nearby performance and surface exposures help choose hypotheses. They do not replace direct evidence at the affected location.

Combine methods and preserve what each actually observed

Borehole logs record material and water observations with location, elevation, depth, date and method. Samples enable laboratory identification and behaviour tests. In-situ penetration or loading tests probe response without removing the whole material. Geophysics can cover more ground, but the measured signal needs interpretation and may admit several geological explanations.

Sampling changes confinement and can disturb fabric or water content. A specimen suitable for particle-size identification may be unsuitable for estimating undisturbed compressibility. Label recovery, handling and test conditions so an attractive number is not detached from the material state that produced it.

Groundwater is also a time-dependent observation. A water level observed during drilling is not necessarily an equilibrated seasonal maximum, and separated permeable layers may have different heads. Repeated, appropriately interpreted observations address a different uncertainty from adding another strength test.

Horizontal distance 0–40 m at 11 SVG units/m; depth 0–6 m at 40 units/m, increasing downward. At x=0 m the observed clay contact is 2 m deep; at x=40 m it is 4 m deep. Dashed line between them is one assumed linear contact, predicting 3 m at x=20. A hollow point at x=20, depth=5 m represents a proposed counterexample observation, not existing field data. The drawing contains no verified continuous stratum.

Choose new evidence for its decision value

The dashed contact in the figure is a hypothesis, while the orange contacts are supplied observations. A continuous-looking drawing must retain this distinction. Separate uncertainty about the material arrangement from uncertainty in properties and measurement. Averaging cannot make an unrecognized weak lens disappear.

Additional investigation should challenge the uncertainty that could change the decision. A new hole at a suspected buried channel, a targeted sample or time-series water observation can be more informative than repeating a familiar test elsewhere. Update the model and its implications while retaining the earlier evidence and reasoning.

Challenge an original interpolation

Suppose constructed logs at distances x=0 and 40 m show the top of a clay layer at depths 2 and 4 m below a common horizontal datum. A linear assumption gives z=2+0.05x, so at x=20 m the predicted contact is 3 m deep. This is arithmetic for a hypothesis, not an observed boundary.

A proposed additional log might instead find that contact at 5 m. It would contradict the simple straight-line model locally; it would not automatically prove the clay is absent elsewhere or that every property value is wrong. A changed layer thickness could alter settlement estimates or a tunnel’s expected face conditions.

Two successful endpoint logs provide no defensible numerical probability that the midpoint lies at 3 m. A confidence statement would require an explicit spatial model, relevant data and assumptions. The immediate useful result is a clearer, testable information gap.

Try the reasoning

Why can a perfect laboratory test still give an unsuitable project parameter?
Show the answer and reasoning
The tested material or its disturbance, stress, drainage or loading condition may not represent the affected ground.
Does another borehole eliminate geological uncertainty?
Show the answer and reasoning
It adds another bounded observation. Its value depends on location, depth, method and which hypothesis it tests.
What should remain visible when the contact interpretation changes?
Show the answer and reasoning
The original logs, coordinates, dates and methods, the former model, and the evidence and reasoning behind its revision.

What to carry forward

Treat the ground model as an explicit interpretation that new evidence can challenge.

References

Teaching resources and research supporting this lesson’s explanations and examples.