What the survey data say

Five questions we could answer once the layers were assembled. Plain language first, numbers second. Every figure is computed from data on disk.

—
agreement between two independent surveys on regional copper
—
how much copper a partial digest misses
—
slope lost when one sample stands for a region
—
waste-water association with copper, geology held fixed
The route from field samples to a defensible statement. Each box is a step that had to be finished before the next one meant anything.

1Do independent surveys agree about where metals are high?

Two surveys sampled the same countries in the same years without sharing samples: GEMAS took agricultural topsoil, LUCAS took a wider land-use sample. If their regional patterns match, the pattern belongs to the land rather than to one team's sampling.

We averaged GEMAS within small European regions, then compared its ranking with the LUCAS ranking in the same regions. One means perfect agreement, zero means no agreement at all.

reproduces does not reproduce
Rank agreement between the two surveys, by element, over the same regions. Hover any bar for the sample size.
Each dot is one region: horizontal is the GEMAS average, vertical is the LUCAS value, both on log scales. The dashed line is perfect agreement.

2How much does the laboratory method change the answer?

Sweden analysed the same landscape five ways: one total measurement that dissolves everything, and four partial digests that reach only part of the metal. Because the terrain is shared, any gap is chemistry of method, not geology.

middle half of samples 10th to 90th percentile median
Copper in Swedish till by analytical method. Bars cover the middle half of samples, the pale stem the 10th to 90th percentile.
At sites sampled twice within 100 metres, how much copper each partial digest recovers compared with the total measurement. One would mean no difference.

3How badly does a regional average blur a local relationship?

Replacing a soil measurement with the average of its region throws away local variation, and a relationship measured against it gets weaker. That is not a reason to avoid regional data, but it has to be corrected, and the correction depends on how many samples each regional average contains.

How much of a true slope survives when regions are represented by averages of n samples. The dashed line is the one-sample case.

4After geology, does industrial pressure show up?

Geology explains most of the regional copper pattern: two surveys agree at — across — regions. The useful question is what remains. We held the geology estimate fixed and asked whether each pressure indicator still tracks copper.

copper mg/kg no data
Copper by European region, the unit used in this analysis. Pale is low, warm is high.
before accounting for geology after accounting for geology
Rank association between each pressure indicator and regional copper. Positive means more pressure goes with more copper.

5Which elements travel together?

Elements that share a source move together. Geogenic pairs follow parent material; anthropogenic pairs follow use. The pattern differs between a soil survey and a sediment survey, which is itself informative.

Rank correlation between elements within one survey. Blue is positive, orange negative; the diagonal is one by definition.

6Big picture: what actually drives the numbers

Five separate questions, one composite answer. For seven elements at once: how much of the regional pattern is geology, how much is pressure, and which pressure touches which element.

Where the variation comes from

explained by geology added by pressure unexplained
Share of the regional ranking of each element accounted for by the geology estimate, then by pressure indicators on top of it. Hover a bar for the counts.

Which pressure touches which element

positive after geology negative after geology nothing
Each cell is the rank association between one pressure indicator and one element, with geology held fixed. Not one pressure story but several.

Was it just city size?

population density waste-water load
Each pair is the association with one element after geology is held fixed. The two predictors are only weakly related to each other, which makes the comparison informative rather than circular.

Does soil inherit the rock beneath it?

Agreement between bedrock copper and the nearest till sample, split by how far apart they are. Sample counts differ by band, shown on hover.

Which elements move as a group

positive negative
Elements as nodes, rank correlation as edges. Thicker means stronger; hover a node to isolate its connections, click to pin it.

7Can we see industry, mining and deposition in the soil?

At point level, in 1,680 European topsoil samples: does metal concentration fall as you move away from a source? Distance is straight-line, from each sample to the nearest facility, mine, large waste-water plant or reported water discharge.

Median concentration in each distance band, expressed relative to the closest band so elements of different abundance can share one axis. Hover a point for the raw median and the sample count.
higher near the source lower near the source
Rank association with distance, within region so the geological baseline is held fixed. Small numbers mean the soil does not know the source is there.

What this does not settle

Agreement between surveys does not prove either is accurate: both could share a bias. The method comparison is Swedish till and does not transfer elsewhere. The attenuation curve uses copper spread measured in the LUCAS design and assumes that spread is representative. The pressure results are associational and waste-water load also tracks city size, so it is an index of urban intensity rather than a measured cause. None of this yet uses the biology.

close ✕