11.3 Integrating Direct and Indirect Evidence for Exploration Decisions

Exploration decisions are made under uncertainty. Indirect prospecting—geochemistry, geophysics, remote sensing, and geomorphology—measures a response caused by geology but does not expose the target itself. Direct prospecting—outcrop mapping, trenching, pitting, underground exposure, and drilling—observes or samples rock. Direct evidence reduces ambiguity, yet a mineralized intercept still does not by itself establish continuity, recoverability, legal access, or economic value.

Build an Evidence Chain

A robust target links observations through a testable mineral-system model:

1. Source and fertility: Is there a plausible metal and fluid source?
2. Pathway: Do faults, contacts, intrusions, or permeable horizons provide transport?
3. Trap: Is there a chemical or structural setting where ore could precipitate?
4. Preservation: Has erosion or later alteration removed or obscured the target?
5. Response: Do mapped alteration, pathfinders, chargeability, magnetics, gravity, or spectral data behave as predicted?

Convergence matters more than the loudest anomaly. A chargeability high can be barren pyrite, graphite, clay polarization, cultural noise, or a sulfide target. Its priority rises when it coincides with favorable host rock, mapped alteration, coherent copper-gold geochemistry, and a structural intersection. Conversely, strong gold in one grab sample may be a nugget or selective sampling artifact.

Make Data Comparable

Before integration, check coordinate system, survey elevation, sample support, units, analytical method, detection limits, QA/QC status, and acquisition date. A soil result in ppm cannot be overlaid naively with a rock assay in percent. A 100-metre airborne magnetic cell has different spatial support from a two-metre channel sample. Separate below-detection values from true zeros and distinguish measured points from interpolated surfaces.

Use transformations appropriate to the distribution. Exploration geochemistry is often log-normal; robust percentiles or median absolute deviation may identify anomalies more reliably than an arithmetic mean plus two standard deviations. Define the background population by lithology or regolith domain so naturally metal-rich ultramafic soils are not mistaken for ore-related nickel anomalies.

Target Ranking Matrix

A transparent score can combine weighted criteria, provided the scores do not conceal fatal flaws:

Do not average away a fatal constraint. No mineral right, an excluded protected area, irreparable sample contamination, or geometry incapable of economic scale can stop a target even if its weighted score is high.

Stage-Gate Choices

At each gate, define the decision before collecting more data:

• Abandon: evidence contradicts the model, likely scale is immaterial, or a fatal legal, environmental, safety, or social constraint cannot be resolved.
• Further exploration: the concept remains plausible, but one or more high-value uncertainties can be tested economically.
• Advance toward development studies: continuity, geometry, metallurgical response, infrastructure, tenure, environmental baseline, and commercial assumptions are sufficiently mature for resource definition and study-level engineering.

A next program should target uncertainty. If geometry is unknown, tighter soil sampling may add little; oriented drilling or a geophysical inversion may have higher information value. If assay bias is suspected, more drilling without laboratory investigation merely multiplies bad data.

Expected-Value Example

A proposed scout hole costs PHP 6 million. The team estimates a 20% chance that success would add PHP 45 million of incremental project value before probability weighting, and an 80% chance of no incremental value. Ignoring secondary information value:

$EV = 0.20(PHP45\text{ M}) - PHP6\text{ M} = PHP3\text{ M}$

The positive expected value supports drilling if the probabilities and consequences are defensible and the program is permitted. If a new mapping result reduces success probability to 8%, $EV = 0.08(45)-6 = -PHP2.4$ million; the correct decision may be redesign or abandonment. Money already spent is a sunk cost and must not inflate the probability.

Decision Record

Document the data cut-off date, deposit model, alternatives, assumptions, QA/QC exceptions, risks, approvals, and the evidence that would reverse the decision. This makes the result reproducible and allows later teams to learn whether the model, data, or judgment failed.

Record the Decision Boundary

For each target, state what evidence would advance, recycle, hold, or abandon it before seeing the result. Record data quality, alternative explanations, expected information gain, cost, access, safety, rights, environmental constraints, owner, and expiry date. This prevents a visually impressive anomaly or sunk expenditure from silently changing the gate. A drill recommendation should specify the hypothesis, geometry, expected intersection, deviation tolerance, sampling and downhole-survey plan, and what each possible outcome means.