Geochemical surveys
Measured element concentrations in soil or sediment. These are the observations the maps are built from.
GEMAS Ap
The Geochemical Mapping of Agricultural and Grazing Land Soil survey sampled European farmland at two depths and analysed the same sites for the same elements. The Ap layer is the ploughed horizon, so it carries both the parent material signal and the fingerprint of fertiliser, manure and atmospheric deposition. We use it as the main European comparator.
GEMAS Gr
The grazing land layer of the same survey, sampled where the soil has not been ploughed. Because Ap and Gr come from the same countries with the same analytical protocol, the difference between them shows what cultivation does to a metal inventory, separate from any change in geology.
USGS NGDB
The United States National Geochemical Database collects soil, stream sediment and rock analyses made by many laboratories over several decades. It is the largest single source here, and also the most heterogeneous: sample media, digestion and detection limits change between contributing studies. We use it for spatial pattern, not for absolute levels.
USGS DS801
A single USGS data release of soil geochemistry, processed to one consistent set of element columns. Smaller than the national database but internally comparable, so it is used where a uniform method matters more than spatial coverage.
NGSA Australia
The National Geochemical Survey of Australia sampled catchment outlet sediments, which integrate the erosion products of the whole catchment above each site. The result is a coarse but genuinely continental grid. It describes source terrain rather than local topsoil, so it should not be read as a soil contamination map.
FOREGS
The Forum of European Geological Surveys baseline sampled stream sediment, floodplain sediment and topsoil across Europe with one protocol. It predates GEMAS and covers ground GEMAS does not, which makes it useful as an independent check on regional patterns rather than as a second measurement of the same sites.
Tellus topsoil
The Geological Survey Ireland Tellus programme sampled topsoil on a dense national grid, with a consistent analytical suite. Its value here is resolution: a single country at a site density that continental surveys cannot approach, which shows how much structure is lost when data are averaged into regions.
LUCAS 2018
The Eurostat LUCAS Soil module sampled topsoil across the European Union and recorded land cover, land use and management at each point. We use the 865 sites that also carry the full covariate set: pH, organic carbon, texture, cation exchange capacity, carbonate, nitrogen, phosphorus, potassium and bulk density. Metals for LUCAS are regional medians from the same programme, drawn as a separate polygon layer, and are not point measurements.
RMQS
The French Soil Quality Monitoring Network, a 16 km grid of sites resampled on a fixed cycle. It provides extractable metals, exchangeable cations and oxides at 1,760 sites, all analysed in one laboratory. That uniformity is why it carries the structural modelling on this site: geology, pressure and population can be compared without a method correction.
Industrial pressure
Where industry and waste water sit, used to ask whether measured metal follows reported activity.
E-PRTR facilities
The European Pollutant Release and Transfer Register records reported releases from large industrial facilities, including heavy metals to air, water and land. It measures what operators report, not what reaches the soil, so we treat it as a pressure indicator and test whether soil metal responds to it.
BASOL sites
The French inventory of polluted or potentially polluted sites, with the activity that caused the designation. It marks where contamination is already suspected, which is useful for placing known hotspots but biased for any test of whether contamination is widespread: a site only appears after someone looked.
UWWTD agglomerations
Urban waste water treatment agglomerations, from the European Environment Agency reporting under the Urban Waste Water Treatment Directive. We use the load and size of each agglomeration as a proxy for urban intensity. It correlates with soil copper in our analysis, and it also correlates with city size, so the association is not proof of cause.
Mining
Locations of mines and mineral occurrences, used as a geological and disturbance covariate.
Mindat mines
Mindat localities record mines, prospects and mineral occurrences worldwide, with the commodity mined where known. We use them to mark mineralised ground, which is both a geological source of metals and a disturbance. Recording effort varies by country, so absence of a mine is not evidence of absence.
Sequencing
Sampling locations of published 16S studies, used to place sequence data next to measured chemistry.
ENA samples, EU + NA
Sample records from the European Nucleotide Archive for 16S studies in Europe and North America, plotted at their stated coordinates. They show where sequence data exist, which is the question that decides whether a place can be studied at all. The plotted subset is a fraction of the archive.
Soil archives
Sites with measured soil carbon and radiocarbon, used for context on soil age and carbon stocks.
ISRaD sites
The International Soil Radiocarbon Database holds measured soil carbon and radiocarbon profiles. Radiocarbon gives an age, so these sites say how long the soil has been accumulating carbon, which is context that metal concentrations alone cannot provide.
1000 Soils Pilot
A pilot collection of soil profiles with linked chemistry, mineralogy and biology. Its value is depth and completeness at a small number of places, the opposite trade to the continental surveys.
Microbial taxonomy
GTDB and NCBI Taxonomy
Genome Taxonomy Database and NCBI names supply the genus and phylum labels used throughout the genera explorer, and the mapping between them. Where the two disagree we keep both, because a name conflict usually marks a real taxonomic revision rather than an error.
BacDive
BacDive collects phenotype data for type strains: Gram stain, oxygen requirement, cell shape, motility, sporulation, growth temperature, pH optimum and genome size. Type strains are the reference definition of a species, so these values describe what the taxon is, not what every isolate does. Strain counts are shown with each value.
Trait databases
metaTraits
metaTraits mines trait statements from the literature and harmonises them into comparable categories, reporting the fraction of underlying observations that agree. A value at 60 percent agreement is a majority habit, not a rule, and the explorer shows that fraction rather than hiding it.
BactoTraits
BactoTraits is a curated database of bacterial traits assembled for ecological analysis, covering morphology, metabolism and environmental preferences. It is used as a third opinion on the same attributes, so a trait that two independent compilations agree on can be trusted more than one that only a single source reports.
Madin et al. 2020
A published compilation of bacterial and archaeal trait values, assembled from named primary sources. It adds coverage for genera that the other databases miss, at the cost of being a snapshot: it does not update as type strains are revised.
Habitat and function
MicrobeAtlas
MicrobeAtlas reanalyses public 16S surveys and reports the fraction of samples in which each genus is found, by habitat. It is observation, not culture, so it answers where a genus actually appears in sequence data. Sampling effort is uneven across habitats, and the explorer prints the sample count behind every percentage.
Omnicrobe
Omnicrobe applies text mining to the microbiology literature and extracts habitat, phenotype and use terms per taxon. It captures what authors say, including taxa nobody has sequenced in the habitat they discuss. It is literature derived, so a missing term can mean either no study or a study that never used the word.
FAPROTAX
FAPROTAX maps taxonomy to function using cultivated representatives, and reports three tiers: function assigned at genus level, at species level, and inherited from family or order. The inheritance tier is a guess by construction. The explorer keeps the tiers separate so a predicted nitrification does not look as firm as a measured one.
Literature and claim checking
Europe PMC
Europe PMC indexes abstracts and full text for the life sciences. The magnifier links on this site run a live search for the genus or the element in question, so a reader can go from a claim to the papers behind it without a separate search.
Paperclip
Paperclip retrieves full-text passages for a stated claim and judges each one as supporting, refuting or neutral. It is a check on our wording, not on our data: it answers whether the literature says what our sentence says. Where it finds almost nothing relevant we report that as silence, not as disagreement.
Our own analysis outputs
Regional reproducibility
Two independent surveys covering the same regions were compared element by element. Six of seven elements agree closely; cadmium does not. The output is the per-element correlation, the number of shared regions, and a permutation test, so the claim that the map is real can be checked rather than believed.
Digestion method comparison
Sites measured twice within a hundred metres, once by total digestion and once by a partial extraction, isolate the method from the place. Partial digests return 18 to 22 percent less copper. The ranking of sites survives, the absolute level does not.
Variance budget
For each element, the share of regional variation explained by geology is compared with the share added by anthropogenic pressure. Geology carries most of it, and the pressure terms add a few percent at most. The point of the output is the comparison, not a single number.
Genus metal effects
A cross-study synthesis: for each genus and metal, the median reported effect, the interquartile range, and the number of studies. One study is shown as one study. These are field associations and dose responses, not laboratory tolerance measurements.
Paperclip claim results
The stored verdicts behind every paperclip on this site: the exact claim sentence, the passages retrieved, and how each was judged. Storing them means the page shows what the check found at the time it ran, and a reader can audit the wording.