B4 ยท Publication Volume 9

Exploration under Cover

signal dispersion, masking, contamination and sampling media

Residual, mechanical, hydromorphic, biogenic and contamination signal pathways
Residual, mechanical, hydromorphic, biogenic and contamination signal pathways

Learning objectives

After this lesson, you should be able to classify plausible signal pathways through cover; distinguish masking, dilution, dispersion and contamination; select a sampling medium from a process hypothesis; define interpretation domains before analysing anomalies; use robust statistics without ignoring spatial support; and design a residual-versus-transported test with explicit decision boundaries and quality controls.

Start with a field problem

A regional soil survey over flat sandy cover contains a cluster of high copper values. The cluster follows a subtle drainage depression rather than mapped bedrock strike. It could represent a buried source expressed through groundwater, mechanically transported fine sediment, iron-oxide scavenging, windblown contamination, a change in sampling moisture or analytical batch, or a chance extreme within a mixed population.

Calling it an anomaly is only the start. Exploration under cover asks three linked questions: what material was sampled, which process connected it to a possible source, and over what domain is the comparison valid? A threshold calculated across incompatible regolith units can make an ordinary material boundary look exceptional.

Core process model

Five signal pathways are useful hypotheses:

  1. Residual: weathering leaves a signal close to its bedrock source, although vertical redistribution and volume loss may modify it.
  2. Mechanical: particles or clasts are moved by gravity, water, wind or ice; signal direction and distance depend on pathway and grain host.
  3. Hydromorphic: dissolved, colloidal or nanoparticulate components move with groundwater and precipitate, adsorb or exchange at chemical boundaries.
  4. Biogenic: roots, soil organisms or other biological processes redistribute components across depths and materials.
  5. Anthropogenic: roads, drilling, agriculture, infrastructure, waste or previous sampling introduce or redistribute material.

These pathways can coexist. A ferruginous grain may be mechanically transported and later acquire a hydromorphic coating. A plant may access groundwater that crossed a transported channel fill. The pathway model must specify source, carrier, direction, transformation, sink and the spatial support represented by a sample.

Masking places low-signal material between source and observation. Dilution mixes source material with a larger background mass. Dispersion spreads a signal beyond its source through a defined mechanism. Contamination introduces material unrelated to the geological process under investigation. Each predicts a different geometry and set of controls.

A robust univariate score is


z_r=\frac{x-\operatorname{median}(x)}{1.4826\,\operatorname{MAD}(x)},

where MAD is median absolute deviation. It reduces sensitivity to extremes but does not solve mixed populations, censoring, closure, spatial dependence or inconsistent media. Background must be defined within process-relevant domains.

Evidence and measurement

Begin with a regolith-landform map, not an assay map. Define domains by material origin, geomorphic position, hydrology, particle size and sampling method. Use pilot transects across known boundaries to test which medium and fraction respond coherently. Record depth, horizon, moisture, vegetation, surface lag, disturbance and collection footprint.

Quality controls address different errors. Field duplicates test local sampling plus preparation variability. Preparation duplicates test subsampling. Analytical duplicates test measurement repeatability. Blanks test contamination; reference materials test bias and drift. Randomise or balance batches so geography is not confounded with laboratory order. Preserve censored values and detection limits rather than replacing them silently with zero.

Interpret multielement patterns through mineral hosts and compositional constraints. A high ratio may result from a low denominator. Iron- and manganese-rich phases can scavenge several elements. Grain-size or organic-matter variation can dominate total concentration. Compare raw concentration, host-normalised relations and spatial process predictions.

Worked example

In a synthetic soil domain, copper is 80 mg/kg, the domain median is 20 mg/kg and MAD is 5 mg/kg. Then


z_r=\frac{80-20}{1.4826\times5}=8.1.

The value is statistically extreme within the declared domain. It does not prove a vertical bedrock source. Suppose the high samples lie along a buried palaeochannel and have rounded ferruginous grains with a provenance different from local saprolite. A mechanical or hydromorphic transported model becomes at least as plausible as a residual model.

Design two predictions. The residual model predicts increasing signal toward shallow saprolite, consistent immobile-element affinity with local bedrock and limited displacement across the residual boundary. The transported model predicts alignment with channel architecture, size- or coating-specific signal, downstream or groundwater-directed asymmetry and a provenance change. A shallow profile transect with separate matrix, clast and coating analyses can discriminate them better than denser surface sampling alone.

Misinterpretations and uncertainty

Do not choose a sampling medium solely because it worked elsewhere. Process, depth, climate, cover thickness, target mineralogy and analytical method may differ. Do not move a threshold across domain boundaries or use a global percentile as a geological fact. Spatial smoothing can create coherent patterns from sparse or clustered samples and should display data support.

Absence of a surface anomaly is not absence of a source: masking, non-reactive host, deep water table, insufficient analytical sensitivity or incompatible medium can all suppress expression. Conversely, a strong anomaly may be transported far from source. Historical disturbance and access bias must remain mapped, not deleted after results appear.

Practical investigation

For a covered-terrain teaching map, delineate at least three sampling domains before viewing chemistry. For each domain, nominate a medium, depth, size fraction, spacing logic, duplicate frequency and the pathway hypothesis it tests. Include a no-sample zone where the chosen medium is invalid.

After receiving synthetic results, map raw values, domain-relative robust scores and quality-control outcomes separately. Build residual, mechanical and hydromorphic source models. Rank observations by discriminating power and propose a follow-up transect that crosses rather than follows the predicted signal pathway.

Mastery check

  1. Why is an anomaly threshold conditional on regolith domain?
  2. How do masking and dilution differ?
  3. Which quality-control sample tests field heterogeneity?
  4. Why can an iron-rich fraction carry a displaced signal?
  5. What would falsify a residual-source interpretation?

Sources and further reading