Every anomaly is an anomaly relative to something, and choosing that something is the whole job. Four thresholds are computed here and they will not agree — the spread between them is the honest measure of how well separated your populations are. When they agree closely, the anomaly is real. When they scatter, the data are telling you the background is not one population.
Mean plus two standard deviations is the threshold everybody was taught, and on geochemical data it is usually wrong in a particular direction: the anomalous samples are in the dataset, so they inflate the mean and the standard deviation, and the threshold rises to exclude the very thing you were looking for. The robust estimators do not have that problem — a median and a median absolute deviation barely move when you add outliers, which is exactly why they are here.
Before believing any anomaly, find out what the same sample gives twice. Field duplicates measure everything at once — the sampling, the splitting, the preparation, the instrument — which is why they are worth more than any certified standard for deciding whether a two-fold difference between two sites means anything.
How much of an element is a lot depends entirely on which element. Twenty ppm of copper is nothing; twenty ppm of gold is a mine. Comparing against average crustal abundance puts every element on the same footing and turns a number into a judgement.
| element | crust | ore grade | × |
|---|
The last column is the concentration factor an ore body represents — how many times over nature had to concentrate the element before it became worth mining. It is why aluminium ore is a four-fold enrichment and gold ore is a thousand-fold, and why the two are found by completely different means.
The element you want is often the hardest to see — too rare, too erratic, too expensive to assay everywhere. A pathfinder travels further, disperses more evenly, or costs less, and points at the thing itself.
A survey that cannot hit the target is expensive in the worst way — it produces a clean result and a false conclusion. The rule is geometric, not statistical: to be certain of crossing a target, the line spacing must be smaller than the target's width across the lines, and the sample interval smaller than its length along them.
Whether a rock's whole chemistry says what it is belongs to Rock chemistry, and what the assay laboratory can and cannot actually measure is in Assays and detection. Once there is an anomaly worth drilling, the metallurgy is Ore processing.