Two national soil surveys, 2,334 sites with sequencing, one question: what actually shapes which microbes live where? This report walks through the answer piece by piece, with the data interactive at every step.
LUCAS samples farmland and countryside across the European Union; RMQS samples France on a regular grid. Both record soil chemistry at the same place they take a DNA sample, which is rare and is what makes this comparison possible.
Hover any point to read its acidity. The map is a plain projection with country outlines, deliberately simple so the sampling pattern is the visible thing.
Take the whole community of a sample, reduce it to a single axis of variation, and that axis is almost a pH meter. In LUCAS the first axis explains 24.6% of all differences between samples and correlates −0.92 with pH; in RMQS it explains 21.9% and correlates −0.86.
Points animate in by acidity, so the gradient draws itself. Lower pH sits at the left; the community axis responds almost linearly across four pH units.
Fit all the variables together and ask what each one adds that the others do not. pH keeps a large share. Copper keeps a small but real one. Land use, which looks powerful on its own, keeps nothing.
Bars animate when you switch. Land use falls from 0.36 to zero because European farmland sits on alkaline soils and forest on acid ones: the two variables are nearly the same variable.
Choose a metal and see its relationship with the community. Then switch on the pH control and watch how much of it was acidity wearing a metal's coat.
A widely used European soil dataset reports cadmium values that disagree with an independent geochemical survey by a factor of twelve in Finland and ten in Poland, while agreeing in France and Italy. That pattern, wrong in some countries and right in others, is what a scrambled column looks like.
A value of 1.0 means the two surveys agree. Bars are drawn against a log scale anchored at 1.
On sandy, weakly binding soils the same metal load sits in solution and the community responds. On carbonate and clay-rich soils it is held, and the community barely moves.
Values are partial correlations with acidity held constant, so a bigger bar means copper still mattered after pH was removed. Unconsolidated soils lead; carbonate and siliciclastic soils sit near zero.
Acidity first. It is the strongest single predictor in both surveys and survives every control we applied.
Metals second, and mostly copper. Once pH is accounted for, copper keeps one to one and a half percent of independent explanatory power; zinc and lead keep less and are not consistently detectable.
Land use is a proxy. Its apparent 36 percent is soil chemistry in disguise.
Availability chemistry did not help. Free metal ions, dissolved metal and reactive pools all performed no better than total metal, and usually worse.
Data quality is a result in itself. The cadmium finding came from an internal coherence check that costs nothing to run and should be routine.