Soil depth and nearby boreholes

Nearby observations refine a model

Bilby estimates soil depth by combining SGU’s soil-depth model with usable nearby well/borehole records. Depth is the estimated thickness of soil above rock, in metres. A nearby drilled observation is evidence about that location; it does not become a measurement under your planned house.

Only usable observations take part

A record needs a finite non-negative distance and depth and measurement reliability above 0 and no greater than 1. The result shows the usable count and nearest usable distance separately from all records considered.

Distance, consistency and agreement

Each well starts with weight = reliability × exp(−distance/350). Before averaging, an individual raw weight is limited to 65% of the total raw weights. The resulting weighted mean and spread describe the well evidence. This is a cap on raw weights, not a guarantee that no observation exceeds 65% after renormalisation.

Well influence combines proximity strength, sample strength (0.68 for one, 0.82 for two, 1 for three or more), consistency and agreement with SGU. Variable/mixed soil or terrain variation above 8 m applies a 0.85 factor. Final influence is at least 0.15 and at most 0.68 for one well or 0.85 for multiple wells. Combined depth = well mean × influence + SGU estimate × (1 − influence).

The range half-width is the largest of 1 m, the remaining SGU range contribution, 1.5 × well spread, 0.75 + nearest distance/350, the single-record allowance, 45% of model/well disagreement and a variable-context allowance. Depth and limits round to 0.1 m; the lower limit cannot be below zero.

Fictional equal observations

Assume SGU predicts 4 m with a 3–5 m range, and two reliable wells each report 4 m at 100 m distance, reliability 1, with no variable context. Each raw weight is exp(−100/350) ≈ 0.7515. The well mean is 4 m and spread zero. Influence is about 0.7515 × 0.82 = 0.6162. The combined estimate remains 4.0 m because the sources agree.

The distance allowance 0.75 + 100/350 ≈ 1.0357 m is the largest half-range term. Rounded bounds are 3.0–5.0 m. The two close reliable observations, full consistency and agreement produce high confidence. If there are no usable wells, the same SGU 4 m and supplied 3–5 m range remain, but confidence is very low and status partial.

Confidence is not a statistical probability

High confidence needs at least two reliable observations within 350 m, consistency ≥0.75, agreement ≥0.65 and no variable context. Medium needs a close reliable observation or at least two observations, a nearest distance ≤750 m and consistency ≥0.55. Spread above 4 m, strong disagreement or nearest distance above 750 m forces low. Without SGU, well-only results are downgraded one level.

With only SGU, the fallback half-range is at least 1 m or 25% of depth unless usable supplied limits exist. With neither source there is no estimate. A narrower range is not proof of uniform ground across the plot. The record’s water/energy-well type is context; the blend uses its supplied reliability rather than assigning a price or an extra depth adjustment solely from well type.

Source and method dates

Method soil-depth-combined-v1 and its listed/coordinate displays were reviewed 18 October 2026. The source check date records when evidence was processed, not when the ground was drilled. SGU model and well records have different observation dates and coverage. SGU soil depth SGU wells