Experiences from 30 years of geochemical mapping at the (sub)continental scale in Europe using a wide range of different sample media are reviewed and discussed with a focus on the most recent GEMAS (GEochemical Mapping of Agricultural Soils) project. Comparing results from the different surveys it is possible to come to conclusions as to how geochemical surveys at the continental scale could best be designed. High analytical quality and lowest possible detection limits are key requirements. In Europe good outcomes were achieved with the <2 mm fraction of soil samples and aqua regia extraction. Focus should be on high quality of sampling and analyses and more determined parameters rather than on more samples. The sample density of 1 site/2500 km2 provides a good overview of the processes governing geochemistry at the continental scale. Results should be extensively published by the project team to allow the dataset to be known and utilized by the wider scientific community.

Thematic collection: This article is part of the Continental-scale geochemical mapping collection available at: https://www.lyellcollection.org/topic/collections/continental-scale-geochemical-mapping

Low-density geochemical mapping has a long tradition in Europe; early examples are the geochemical atlas of Finland, based on till (Koljonen 1992) and the geochemical atlas of Norway, based on overbank (floodplain) sediments (Ottesen et al. 2000). For both surveys the planning started in the early 1980s. In this connection the pioneering work of Shacklette and Boerngen (1984), mapping the conterminous United States of America, needs to be mentioned because it convinced European colleagues that low-density geochemistry would deliver valuable results. Garrett et al. (2008) provided an overview of the history of geochemical mapping.

Since the end of the 1990s many regional- to continental-scale multi-element and often multi-sample-media geochemical projects have been carried out by the geological surveys of Europe. The author had the privilege to be the project leader for four of these surveys (Kola, BSS, EGG and GEMAS – see below for abbreviations). Most prominent examples (the list is far from exhaustive) include:

  • the Kola Ecogeochemistry Project (Reimann et al. 1998), combining terrestrial moss, humus (the soil O horizon) and three mineral soil horizons (E, B and C, although the E horizon was only collected and not analysed) as sample materials covering almost 200 000 km2 at an average sample density of 1 site/300 km2;

  • the Baltic Soil Survey (BSS project; Reimann et al. 2003), which covered all of northeastern Europe (about 1 800 000 km2) at a density of 1 site per 2500 km2 and collected the top (0–25 cm) and bottom (50–70 cm) layer of agricultural soil;

  • the Eastern Barents project (Salminen et al. 2004), a follow-up project to the Kola Ecogeochemistry Project, covering 1 000 000 km2 (all of Finland and northwestern Russia) at a density of 1 site/1000 km2, collecting terrestrial moss, humus, mineral soil and stream water;

  • the FOREGS (Forum of the European Geological Surveys) project (Salminen et al. 2005; De Vos, Tarvainen et al. 2006) covering most of western Europe at a density of 1 site/5000 km2 collecting stream sediment, overbank sediment, top and bottom mineral soil and stream water and thus following closely the suggestions for global-scale geochemical mapping made in Darnley et al. (1995);

  • the EGG (European Groundwater Geochemistry) project (Reimann and Birke 2010), which collected almost 2000 mineral water samples across Europe as proxies for deep groundwater;

  • last, but not least, the GEMAS (GEochemical Mapping of Agricultural Soils) project (Reimann et al. 2014a, b), which focused on agricultural (regularly ploughed fields) and grazing land and covered western Europe (5.6 million km2).

In addition, many surveys using much higher sampling densities have been carried out at the national to local scale all across Europe. Some outstanding examples are the Geochemical Atlas of Berlin (Birke and Rauch 1997), the TELLUS project covering Northern Ireland (Young and Donald 2013; Gallagher et al. 2016) and the London Earth project (Ferreira et al. 2017; https://www.bgs.ac.uk/gbase/londonearth.html). These projects demonstrate what can be expected from high-density geochemical mapping in contrast to continental-scale low density mapping.

With the exception of GEMAS, all projects were conceptualized by geochemists of the geological surveys of Europe. The good results of the GEMAS project built on the experiences of all the earlier projects along with geochemical mapping at a variety of scales. The differences and lessons learned will be discussed below with a focus on the latest results from the GEMAS project.

The GEMAS project was carried out by the geological surveys of Europe, represented by EuroGeoSurveys in Brussels. It was contract work carried out for the European metals industry, represented by Eurometaux, the association of the European metal industry, and its partner organizations. They needed background concentrations of metals in agricultural soil and of the key parameters determining their availability in soil at the continental scale for preparing their REACH (Registration, Evaluation, Authorization and Restriction of Chemicals, EC No 1907/2006) documents. The assessment of risks to human health and the environment, related to the exposure of metals in agricultural soil (arable land) and grazing land soil (grassland), is one of the requirements under the REACH regulation (ECHA 2008a, 12b; Verougstraete and Schoeters 2014) that led to the GEMAS project. That implied that many key parameters of the project were dictated by the REACH requirements, e.g. the combination of agricultural soil, ploughing layer 0–20 cm and grazing land soil 0–10 cm, <2 mm fraction, and aqua regia extraction for analysing the metals).

The BSS project came probably closest to the metal industries’ needs for preparing REACH dossiers; however, no grazing land samples were collected. In addition, it covered ‘only’ northern Europe and some key parameters were missing in the analytical program that govern the availability of metals in soil, e.g. cation exchange capacity (CEC), grain size and total organic carbon (TOC).

Thus, in late 2007/early 2008 Eurometaux made contact with the chairperson of the EuroGeoSurveys Geochemistry Expert Group and asked whether the geological surveys of Europe would be able to carry out a European-scale geochemical mapping project for them according to the REACH requirements. They suggested that industry would pay for sample equipment, sample preparation and a large part of the analytical work, and would gain in return early access (data were needed in spring 2010 in order to meet REACH deadlines) to the data while the geological surveys could use the samples and all data for their own purposes in addition to the latter becoming public around 2014/15. The chairperson of the geochemistry expert group presented the proposal at the EuroGeoSurveys yearly meeting some weeks later where it was agreed that this was an exciting proposition for a project that would demonstrate that the geological surveys were able and willing to cooperate across European borders.

The project started in 2008, with industry needing the data in early 2010. To successfully work to such a tight time schedule on the continental scale (more than 5 million km2 to cover) was only possible because of the earlier continental-scale geochemical surveys (see above) carried out in Europe and the existence of the Geochemistry Expert Group of EuroGeoSurveys (the former FOREGS geochemistry group). The fact that the team to carry out the project still existed from the FOREGS project was a great advantage. Key parameters for guaranteeing project success like planning sample locations, buying all sample equipment from a central source, getting all samples analysed at the same time in one large batch in just one laboratory and preparing sufficient splits for all analytical needs were familiar to the project management and participants. Sample material, grain size fraction and key analytical requirements were all set by the REACH regulation – insofar it is not easy to discern what could have been done better. However, in view of all the geochemical projects mentioned above it could be discussed how an ideal project design for continental- or regional-scale geochemical mapping should be planned and undertaken.

The team carrying out the GEMAS project in the end consisted of more than 60 organizations from around the globe. Most geological surveys of Europe cooperated and, in some countries where agreement with the survey could not be reached or no geological survey existed, universities, soil research organizations or even ministries of the environment stepped in to cooperate and organize sampling in their respective countries. The agreement was that sampling equipment and analyses were provided to each participating country free of charge while the country coordinators were responsible for organizing sampling according to the field handbook (EuroGeoSurveys Geochemistry Working Group 2008) in their area. To guarantee a degree of comparability between the ongoing continental-scale mapping projects in other parts of the world, project standards were exchanged with the ongoing continental-scale mapping projects in Australia (e.g. Caritat and Cooper 2011a, 7b) and North America (e.g. Smith et al. 2014). Due to the large group cooperating within the project there was expertise not only from geochemistry but also geology, soil science, biology, toxicology, forensic sciences, statistics, physics and chemistry available within the team in addition to work experience from within the geological surveys, universities, ministries of the environment, contract research organizations and consultancies.

Sample density was one of the few parameters that was not predetermined by REACH. Project finances were a critical limiting factor. Based on experiences from the BSS project and the available finances it was decided to take one sample per agricultural and grazing land site per 2500 km2. These two soil types combined represent the substrate for the vast majority of agricultural food production in Europe. The two materials were collected independently, but usually as close to one another as practically possible.

A grid-based approach was chosen for sample site selection, and the same grid as used for the BSS project was extended across Europe. Within each grid cell it was left to the sample collectors where exactly to take the sample. The reason is that agricultural areas can be quite sparse in northern Europe. The free site selection within a 2500 km2 cell, instead of aiming for the centre of each cell, permitted an almost complete coverage of Europe, with the two sample types needed – the exceptions were a few cells in northern Sweden, Norway and on the west coast of Ireland. It was an important aim for the success of the project to get as even and representative coverage of all of Europe with agricultural and grazing land soil as possible. By sticking to only these two sample types it was possible to get a statistically valid overview of the chemical composition of ‘agricultural and grazing land soil of western Europe’. Other sampling approaches would have led to large gaps (or an over-representation) for certain sample types in some parts of Europe.

The sample types and sample depth chosen for the GEMAS project were predetermined by the requirements of the REACH regulation. Within the project group it was discussed that it might make more sense to take a sample depth of 0–5 cm or even 0–2 cm for the grazing land samples, providing the aim is to get a better impression of the impact of diffuse contamination on soil. By collecting only the surface 2–5 cm the dilution with natural mineral soil would have been less pronounced. However, the 0–10 cm depth interval for grassland (the root zone) was defined by REACH and could thus not be changed.

The maps presented in the GEMAS atlas (Reimann et al. 2014a) show a similar spatial geochemical distribution in the maps of agricultural soil (0–20 cm) – the GEMAS Ap samples – and grazing land soil (0–10 cm) – the GEMAS Gr samples (Fig. 1).

Looking at the maps and data from the FOREGS Atlas and the GEMAS project, it is clear that minerogenic sample material collected across the continent, be it agricultural soil, grazing land soil or any other land use type and even stream and floodplain sediments, will deliver directly comparable analytical results and maps (Fig. 2). The maps also demonstrate how robust results from low-density mapping (1 site/2500 to 1 site/5000 km2) will be (Smith and Reimann 2008).

Different geochemical results and geochemical patterns will only result when collecting organogenic topsoil, e.g. the forest soil O horizon, or other principally different sample materials (e.g. water, plants).

Within the geochemistry expert group it was also discussed whether to take an additional bottom soil (soil C horizon) sample at each site. This would clearly have been interesting but turned out to be impossible due to financial and time restrictions. Furthermore, top and bottom soil samples provided comparable spatial distributions when studying the BSS maps, so much so that it appeared questionable whether it would be cost-effective to take an additional sample at depth (Fig. 3).

The BSS As maps show a comparable spatial distribution for the top and bottom samples; the median value is only slightly higher for the latter. The distinct As anomaly in central Sweden is related to the Boliden/Skellefte ore district (Tarvainen et al. 2014). The European commission identified diffuse contamination as one of the eight major threats to soil quality in Europe (Commission of the European Communities 2006). That a direct comparison of a top and bottom soil sample as such at any one site will not be a viable method for detecting the impact of diffuse contamination has been repeatedly demonstrated (e.g. Reimann and Caritat 2000, 2005; Sucharova et al. 2012). In 2017 Fabian et al. developed a method that enabled the detection of diffuse contamination in continental-scale datasets via comparing the statistical distribution of top and bottom soil results (Fabian et al. 2017). Diffuse contamination becomes visible as a slight deviation in a plot of the cumulative density function of the topsoil at the lowest concentration (Fabian et al. 2017; Reimann et al. 2018, 2019; Reimann and Fabian 2021, 2022; Flem et al. 2022) and not as exceptionally high element concentrations – the latter is the signal of local pollution or the presence of natural mineral occurrences. In view of the new method it would have been relevant to have GEMAS bottom samples as well.

Relatively large samples (minimum 2 kg each) were collected in the field and shipped to the sample preparation facility. As it subsequently turned out an even larger sample size would have been desirable, but only very few sample preparation laboratories are able to handle such large samples. To minimize contamination issues all samples were prepared in just one laboratory. A number of different laboratories were then chosen for the analytical work. This approach also allowed the insertion of quality control samples not recognizable by the analytical laboratories. A project standard was prepared and inserted into the sample set at a rate of 1 in 20. In addition a sample duplicate of one of the 18 true samples was inserted at the rate of 1 in 20. The project standard underwent a ring test at the end of the project (Kriete 2011; Reimann and Kriete 2014). This additional step was necessary because for the purpose of writing the REACH protocols and assessing ecotoxicology of metals in European agricultural soil it was important to demonstrate that the analytical results were close to the ‘true’ element concentrations in an aqua regia extraction. Only the external quality control based on sample duplicates and the ring test of the project standard could guarantee that the data produced by the project were fit-for-purpose, i.e. in the case of the GEMAS project for the plotting of reliable geochemical maps, depicting processes and determining the distribution of measured elements/parameters at the European scale.

Depending on experience, detection limits and price, different laboratories were chosen for analysing certain parameters (e.g. 52 chemical elements following an aqua regia extraction, 41 elements by X-ray fluorescence (XRF), TOC, C, N, S, pH, grain size distribution, CEC, Kd values, magnetic susceptibility, Pb isotopes, Sr isotopes). Thus a large number of analytical splits were needed, a matter giving rise to logistical and cost issues. Looking back, even more than the 10 splits that were prepared would have been desirable. Four large splits for long-term sample storage and later reference were in addition unfortunately prepared without inserted quality control samples, an approach causing many problems later on.

While geochemists would usually prefer a rather fine fraction of soil samples for analysis (e.g. the famous ‘−80 mesh’ – <0.177 mm), in soil sciences the <2 mm fraction is the standard. When using the <2 mm fraction the samples either need to be milled prior to analysis, introducing a sample preparation step that can cause serious contamination of the samples with elements like Al, Co, Cr, Fe or W, depending on the mill material used. For the GEMAS project all samples for XRF analyses were milled in agate mills that minimize the risk of contamination. The alternative to milling is to use a large sample weight for analysis in order to obtain representative results. For the GEMAS project a 15 g sample weight was used instead of the standard 0.5 g for the aqua regia extraction on the original, unmilled <2 mm samples. Starting with the Kola project the Geological Survey of Norway (NGU) had good outcomes with this approach. The <2 mm fraction was specified in the REACH Regulation, although without a statement as to whether samples should be milled prior to analysis.

Great care was taken to obtain high-quality analytical results. All samples were analysed for all required parameters as just one great batch in one laboratory per analytical technique. For example the analytical work related to the aqua regia extraction was completed on only one instrument in the shortest time span possible. Some exceptions to this rule occurred later on when outside organizations volunteered to analyse certain special parameters on the samples, e.g. the Sr isotopes (Hoogewerff et al. 2019).

To fulfil the requirements of the REACH regulation, industry needed metal concentrations in the samples based on an aqua regia extraction. While geological survey organizations usually prefer total concentrations in their samples as obtained for example by XRF or neutron activation analyses (NAA) a weaker extraction like aqua regia provides better chances to detect ‘human impact’ and improved (lower) detection limits can be reached for many elements. Great care was taken to obtain the lowest possible detection limits. At the same time the risk of introducing methodological errors or artifacts increases substantially when using a partial extraction. For this reason the extensive external analytical quality control culminating in its own ring test of the project standard was considered a necessity. In total 52 elements were analysed following the aqua regia extraction. More details about the analytical program and detection limits reached can be found in the atlas volumes (Reimann et al. 2014a, 43b).

Industry needed a number of additional parameters for carrying out risk analysis with the data. CEC, pH, TOC and particle size distribution (PSD) were all determined in a number of different laboratories following the same quality control procedure as used for the aqua regia extraction. Obtaining reliable PSD data turned out to be the biggest problem in terms of analytical quality, followed by CEC. It became clear that these two parameters would require specific discussions with the laboratories in any future work. Nevertheless high-quality data could be delivered in a timely manner, i.e. in early 2010, to industry as needed to prepare their REACH documents by the end of the year.

The Geological Surveys of Norway (NGU) and Germany (Bundesanstalt für Geowissenschaften und Rohstoffe – BGR) provided additional analytical data for the project. At BGR all samples were analysed for total concentrations of about 40 elements by XRF and at NGU lead isotopes were determined in the Ap samples. Later on magnetic susceptibility was added (and a high number of further magnetic properties). Furthermore SGS Minerals Services in Toronto offered to analyse the samples following their Mobil Metal Ions (MMI®) extraction for 57 elements. The latter was especially interesting because in a weak extraction the chances of detecting human impact on the soils is higher than in stronger extractions (see Mann et al. 2015).

It was clear from the beginning that for Au and the platinum group elements detection limits in the standard analytical package used for the project were not low enough to obtain a complete data distribution across Europe. Although diverse EU organizations (e.g. DG Environment, DG Research, DG Agriculture) as well as the mineral exploration industry were contacted it was not possible to find a sponsor for this additional work. A chance to establish the background and complete regional distribution of these important elements in European soil at rather low additional cost (€40 000) was thus lost. Note that Au, Pd and Pt were analysed but with detection limits that were considered (and turned out to be) too high to obtain complete data distributions.

In general the samples would have provided a unique opportunity to undertake additional analysis. A very important aspect would have been to study the continental scale distribution of additional isotope systems (Pb and Sr isotopes were determined – see above), e.g. the Fe, Cu or Hg isotopes in at least one of the two sample types (Ap or Gr) at reasonable cost. An established European-scale background might help to detect unknown processes and thus avoid misinterpretations of data based on local-scale investigations. In addition these would have been valuable data for forensic applications.

Given modern analytical possibilities it would also have been possible to analyse soil DNA in all samples. The challenge would certainly not have been the laboratory analysis but rather data analysis. It would definitely be possible and very timely these days; mind-boggling results could be expected if sufficient time for data analysis were available.

The possibility of additional analysis of organic compounds (e.g. PCDD, PCDF, PCB and PAH) was discussed in 2009 with DG Environment; however no money for this very costly exercise could be raised.

As mentioned above extensive external analytical quality control was considered a necessity. Two own-project standards (Ap and Gr) were prepared by the Geological Survey of Slovakia for the GEMAS project (Makových and Lučivanský 2014) and inserted at a rate of 1 in 20 between the project samples, unrecognizable for the laboratory. A field duplicate sample was taken at every 20th sample site and analytical replicates of project samples at a rate of 1 in 20 were also introduced. Furthermore all sample were randomized prior to submission to the laboratories. Quality control results for the different methods used and all parameters reported are documented in a number of quality control reports (Reimann et al. 2009, 2011, 2012). Of special importance also was the documentation of the veracity of the analytical results, especially for the relatively weak aqua regia extraction which was the backbone of the risk assessment work. The weaker the extraction, the higher the chance of introducing some procedural bias. Here the two project standards Ap and Gr were sent out to 21 laboratories/institutions in 16 countries to carry out a proficiency test (Kriete 2011; Reimann and Kriete 2014). The ring test demonstrated the high quality of the GEMAS results in terms of a very limited bias for the majority of the analysed elements. Only such a ring test will lay the foundation for comparing the European-scale GEMAS results with those of regional- to local-scale geochemical investigations at a much higher sample density than exist at many geological surveys.

Geochemical data are compositional data because they do not contain absolute but only relative information (Aitchison 1986, 1997). The reported concentrations of all elements add to 100% and thus depend on one another. The relevant information for each single variable lies in the ratios between all variables and not in the measured element concentrations as such. Mathematically, compositional data define points in the Aitchison geometry on the simplex, and not in the usual Euclidian space for which all classical statistical methods were developed (Aitchison 1986). For this reason all calculations which are based on Euclidean distances can provide misleading results and should only be used with great care. Methods based on correlations must be seen as especially questionable for these types of data. For presenting the GEMAS results, methods based on exploratory data analysis were used and the raw data were mapped using percentile classes.

The European Joint Research Centre ISPRA carried out another soil chemistry related project (LUCAS) shortly after GEMAS (e.g. Tóth et al. 2016; Ballabio et al. 2018, 2021; Panagos et al. 2021). For that project more than 20 000 soil samples were collected across Europe based on a uniform sample depth of 0–20 cm, about 10 times the sample density of GEMAS. However, these samples were from all kinds of land use classes and thus did not provide the same uniform coverage of the whole continent based on one land use class, agricultural soil. For the LUCAS project predicted element concentration maps were constructed based on often rather weak correlations with other parameters like altitude and precipitation for which very high-density data are available across Europe (for examples see Tóth et al. 2016; Ballabio et al. 2018, 2021; Panagos et al. 2021). The resulting maps give the impression of high resolution and precision but are no longer based on measured but rather on modelled element concentrations. This modelling approach implicitly assumes complete knowledge about the geochemistry of soils and the factors influencing the distribution of chemical elements in them. The detection of areas with unexpected element behaviour – which for a geochemist is the very reason to carry out a regional geochemical survey – is not possible following that approach. Surprisingly, even given this very different approach to mapping element concentrations and the almost 10 times higher sample density and much higher costs, the mapped patterns, the median values for element concentrations and the conclusions finally drawn from the maps are comparable for both projects.

The Kola project resulted in a geochemical atlas (Reimann et al. 1998), a book on ‘chemical elements in the environment’ (Reimann and Caritat 1998), a textbook on statistical data analysis for geochemical (environmental) data (Reimann et al. 2008) and over 60 publications in peer-reviewed international scientific journals. The three books have together been cited over 3000 times, the atlas itself only 336 times.

The BSS resulted in a geochemical atlas (Reimann et al. 2003) and a very limited number of publications. The atlas itself has been cited about 300 times.

The Eastern Barents project (Salminen et al. 2004) resulted in the publication of a geochemical atlas (cited about 130 times according to Google Scholar and 56 times according to SemanticScholar) and in a very limited number of papers in scientific journals.

The FOREGS project resulted in the publication of two geochemical atlas volumes (Salminen et al. 2005; De Vos, Tarvainen et al. 2006) and a limited number of publications in international scientific journals. According to a search on Semantic Scholar the first volume of the FOREGS atlas has been cited 729 times, the second volume 251 times.

The EGG project resulted in a geochemical atlas (Reimann and Birke 2010) which to date has been cited about 150 times and a number of scientific papers in international journals cited several hundred times.

To date two books and over 40 publications in international, peer-reviewed journals have appeared with data from the GEMAS project (for a complete publication list see: http://gemas.geolba.ac.at/Publications_GEMAS.htm). In a Google Scholar search the term ‘GEMAS project’ turns up over 6000 times, of which many of the hits in fact relate to the project. The atlas itself has so far been cited 200 times according to a Google Scholar search. In addition the European metals industry has based all its REACH reports and the risk assessment of metals in soil on the GEMAS data. Citing from three book reviews of the two GEMAS volumes:

‘The atlas is a wealth of information for everyone interested in the pedogeochemistry of Europe. At the same time it is a gold mine for academic teachers in practrical courses and exercises for geochemistry and environmental science.’ (Matschullat 2014)

‘…whilst the book may initially appear to have limited appeal to those scientists interested in the chemistry of Europe's agricultural soils, I was somewhat taken aback by the breadth and depth of the background information provided with respect to the methods, protocols, statistical methodologies and background information on the elements; all of which makes the book, even in this day of easy access to information (or perhaps information overload), a very valuable addition to the geoscientist's library.’ (Winterburn 2015)

‘A focus on REACH (i.e. potential toxicity) is not the only use of these volumes. They will be useful for teaching and research in many areas, including environmental, climate (soil carbon), agriculture and food, geological and geochemical processes, amongst others.’ (McGrath 2015)

The GEMAS project fulfilled the expectations of the different project participants. Other than the sample density, practically nothing could have been changed due to the fact that the project had to be carried out according to the REACH regulations. Additional field work (more sample materials) and additional analyses were discussed but not possible due to budget restraints. The sample density was chosen based on the BSS project results and, again, a higher density would not have been possible with the given budget. Additionally it might have been impossible to find agricultural soil in all areas of northern Europe at a higher sample density, e.g. 1 site per 1000 km2. Results demonstrate that for obtaining an overview at the European scale the density of 1 site/2500 km2 was sufficient (Fig. 4).

It would, however, have been interesting to somewhat increase sample density in the surroundings of known or expected geochemical hotspots (e.g. major cities, smelters, power plants); with a couple of hundred additional samples it would have been possible to better detect and map their local impact on the environment. In general the GEMAS project turned out to be ideal for its purpose – to establish the geochemical background concentration and variation in European agricultural soil. However, for comparatively little extra funding a lot of valuable additional information could have been obtained, e.g. better detection limits for a number of elements, or the analysis of additional isotope systems. Sample preparation and analysis in just one laboratory and as just one large batch each was very important for the project's success. The external quality control and a ring test are seen as essential for such a project supposed to deliver reliable data for a whole continent.

Without the requirements of the REACH regulation the project group had argued to take a different depth for the grazing land samples: 0–5 or even 0–2 cm. This would in all likelihood have allowed for a much more sensitive detection of low-concentration anthropogenic impacts on the chemical composition of the soil samples. An additional sample set of a deep soil layer would have allowed for a much more successful detection of low-concentration anthropogenic impacts on the chemical composition of the soil samples.

In general, for continental-scale geochemical mapping the approach chosen for the Kola and even more so the Eastern Barents Region project, combining sample materials such that each is selected to reflect one compartment of the ecosystem (lithosphere, pedosphere, biosphere, hydrosphere), appears ideal (Fig. 5).

Important and new information could have been obtained from forest soil O horizon samples, collected according to horizon and not according to a pre-defined depth, as additional sample material. Surface water samples, again as a principally different sample material, would have provided further important results. At the European scale each of these materials could have been found in a 2500 km2 grid cell, a fact which is one of the great advantages of the low-density approach.

Low-density geochemical mapping at the suggested density of 1 site/2500 km2 (or lower) can be successfully used for any material where a large-scale geochemical overview or the documentation of the geochemical background is desirable. This could be of special importance for future geochemical mapping of ocean floor sediments or even other planetary bodies. Using this low-density mapping approach in different branches of science (e.g. magnetics, biology, DNA, medicine) can deliver important information about the large-scale variation, and the European (continental/ocean basin/planetary) background values of the measured parameters could also be attained at a much more reasonable cost than widely assumed.

Results presented for western Europe demonstrate how robust the use of low-density geochemical mapping can be. Based on 1000–2000 samples, 1 site/2500 to 1/site/5000 km2, reliable and well comparable sample-based maps of western European geochemistry at the continental scale can be constructed, independent of sample material. Minerogenic sample materials will all provide quite comparable results at the continental scale (compare FOREGS, BSS, GEMAS). For additional scientific value it is advisable to sample materials representing different parts of the ecosystem (lithosphere, pedosphere, biosphere, hydrosphere, atmosphere); the Kola and Barents projects are good examples for that approach. Modelled interpolation maps should be viewed and interpreted with great care.

Geological survey organizations tend to use analytical techniques providing total element concentrations (e.g. XRF, INAA, 4 acid extraction). For the GEMAS project REACH required the use of an aqua regia extraction for the assessment of risks to human health and the environment. The risk of introducing methodological errors or artifacts increases substantially when using such a partial extraction. For this reason an extensive external (laboratory-independent) analytical quality control culminating in its own ring test of the project standard was a necessity.

Since the late 1990s eight atlas books have been published with results from low-density geochemical surveys in western Europe. In addition far more than 100 publications in international journals have appeared, covering topics from basic geology and geochemistry to environmental studies, climate change, food safety, forensics and ecotoxicity.

Based on the collected experiences from all projects a number of key points for project success can be defined.

  • It is important to collect enough sample material for a sizeable sample archive and to prepare several analysis-ready splits, including quality control samples, for later use to avoid costs for preparing splits or buying thousands of sample containers for new analyses from large archived splits when there is no longer any project (or funding).

  • Mixing different soil types and land-use classes (e.g. forest soil and agricultural soil) in one and the same database should be avoided. The aim should be to collect as comparable samples as possible across the whole project area (e.g. Kola project: only samples from complete podzol profiles were taken in addition to terrestrial moss; BSS and GEMAS projects: only agricultural soils were collected).

  • Samples should preferably be collected according to soil horizon rather than taking depth-related samples – an exception may be agricultural soil where the average ploughing depth is a good sample depth.

  • Mixing organogenic and minerogenic soil into one sample set should be avoided.

  • Better results can be expected when funding is spent on better analysis (more elements, lower detection limits, higher precision, additional interesting parameters, isotope systems) rather than on more samples.

  • When collecting soil samples it is advisable to collect a subsoil sample in addition to a topsoil sample (or vice versa) at each site.

  • Classical statistics need to be used with care; geochemical data are closed data.

  • The original data distribution should be studied (and understood) in a set of maps and diagrams before starting with more elaborate statistical techniques.

Finally, the presentation of a geochemical atlas alone turned out to be insufficient to get the projects noticed in the science community. It is thus not surprising that the two projects that were carried out over a full five year period (Kola and GEMAS), which left sufficient time for preparing manuscipts for additional scientific papers, have received so far the most attention in the scientific community.

Dennis Arne provided the push to start writing up this paper. Many thanks are due to Manfred Birke and Karl Fabian for an internal review and help with the figures and TimoTarvainen for help with the citation counts. All colleagues I have cooperated with during the many years while the projects were underway are thanked for their input and many interesting discussions. Robert G. Garrett and an anonymous reviewer are thanked for thoughtful reviews and many useful comments.

CR: conceptualization (lead), data curation (lead), formal analysis (lead), funding acquisition (lead), investigation (lead), methodology (lead), project administration (lead), resources (lead), software (lead), supervision (lead), validation (lead), visualization (lead), writing – original draft (lead)

The different projects from which data were used were funded according to the details provided in the cited project publications.

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

The data are available as stated in the cited publications of the surveys used, several of the atlas books contain a CD ROM with the datafiles.

This is an Open Access article distributed under the terms of the Creative Commons Attribution 4.0 License (http://creativecommons.org/licenses/by/4.0/)