Determining the lithogeochemical background concentrations of 39 elements in Bavarian rocks

An approach is presented to establish a state-wide basis for geochemical background values from the available data of the Geological Survey of Bavaria. For this purpose, typical element background concentrations of rock units in Bavaria were statistically evaluated and presented in a lithogeochemical map derived from the geological map at a scale of 1:25,000. The elements comprise 39 geogenic main and trace elements (SiO2, Al2O3, Fe2O3, MnO, MgO, CaO, Na2O, K2O, TiO2, P2O5, Li, Be, Sc, V, Cr, Co, Ni, Cu, Zn, Ga, As, Rb, Sr, Y, Zr, Nb, Mo, Cd, Sn, Sb, Cs, Ba, La, Ce, Tl, Pb, Bi, Th, and U). The distributions of element concentrations in the lithogeochemical units follow closely lognormal patterns in a large majority of cases. Statistical parameters (10th, 25th, 50th, 75th, 90th percentiles) of investigated elements were determined using the cenfit function of the NADA package in the open-source R program. The investigation, based on 8,838 analysed samples, provided data for about 2/3 of the area of Bavaria. The lithogeochemical map with medians (50th percentiles) and background values (90th percentiles) of the investigated elements is available in a web map application. Thus, the regional geogenic background values of the investigated elements in Bavaria are publicly available for a large variety of applications, such as environmental issues and applied research questions.


Introduction
Geological services often have extensive geochemical data sets that have been collected over a longer period of time. An evaluation of these data with respect to typical element distributions in the rock units and their representation in lithogeochemical maps can provide an important basis for geochemical investigations. The prerequisite for such an approach is the statistical proof of a normal distribution of the underlying data sets with a sufficiently high number of samples.
Geochemical rock data are important for many applications. For example, element concentrations define a suitability of rocks as raw materials. Then, geochemical rock properties influence the chemical composition of groundwater. The soil geochemical properties, in turn, depend strongly on the chemical properties of underlying rock.
In Germany, soil protection laws prescribe legal measures if soils or the underlying rock exceed specified concentration limits for contaminants. Therefore, it is important to distinguish whether potential pollutants are of geological or anthropogenic origin, because natural exceedances of limit values are exempt from legal procedures.
The main objective of the study was to apply a valid approach to determine geological background values of inorganic trace elements for environmental purposes to a typical dataset of geochemical samples from the geological survey and link these data to the geological map. A special attention was given to potential pollutants and their availability to users via a web application.
Medians and 90th percentiles of element concentrations are considered as robust parameters for geochemical distributions (Licht 2020). However, the scope of this study is limited to a regional scale; this is due to a relatively small number of samples (i.e., 8838 samples) covering an area of about 72,000 square kilometers, i.e., about one sample per 10 square kilometers (Demetriades et al. 2018). For that reason, regionalization of lithogeochemical background 1 3 207 Page 2 of 10 variation within lithogeochemical units was not addressed in this study.
In Bavaria, background values have been determined for soils (Geuß et al. 2011) and groundwater (Wagner et al. 2011). First steps to determine background values for rocks in parts of Bavaria-at a scale 1:200,000-were made by Linhardt and Zarbok (2005). Similar studies were carried out in Poland (Zglobicki et al. 2011), for the European scale (Salminen et al. 2005;De Vos and Tarvainen 2006), and for Australia (Reimann and Caritat 2017).
This paper presents an approach to mapping lithogeochemical background concentrations of rocks, based on detailed geological maps (1:25,000) by the Geological Survey of Bavaria. The concluding data are accessible via a web map browser and a web map service. Thus, valuable basic information is available for numerous applied geochemical research questions.

Data basis
Sampling and laboratory analysis were performed by the Geological Survey of Bavaria. An overview is given below. Further details can be found in Linhard and Zarbok (2005) and Tammen et al. (2020).

Sampling
Samples were taken during mapping campaigns in the field as part of the geological survey and from drill cores and reserve samples stored in the drill core archive of the Bavarian Geological Survey. The samples were taken over a period from the 1980s until today. To achieve the best possible area coverage, about 2000 additional samples were taken for the present project. For the geochemical analysis, the rock samples are freed from possible weathering crusts by sawing in the laboratory. After subsequent crushing in the jaw crusher, an aliquot is ground in the ball mill with agate insert.
The analytical data used for the investigation are stored in the Bavarian Soil Information System. All samples are assigned to the geological units of the geological map. Analysis data of 8.838 samples from the Bavarian rocks were available for the investigation.

Laboratory analysis
X-ray fluorescence analysis is used to determine the elemental contents in solids. This method works independently of the solubility of the various elements. A few grams of the finely ground sample material are annealed at 1050 °C to quantitatively release volatile components such as water, organic carbon, and CO 2 from carbonates that would interfere with subsequent melt tablet production. The weight loss of the sample that occurs during this process is calculated as the loss on ignition (LOI). For the production of the fused tablets, the annealed sample is mixed with flux and melted in platinum crucibles in a fusion unit under constant swirling at 1200 °C and then poured into tablet form. This ensures homogeneous element distribution and a flat surface as a prerequisite for the subsequent measurement.
All other elements were analysed by inductively coupled plasma mass spectrometry (ICP-MS): Li, Be, Sc, V, Cr, Co, Ni, Cu, Zn, As, Rb, Sr, Y, Mo, Cd, Sn, Sb, Cs, Ba, La, Ce, Tl, Pb, Bi, Th, and U. For this purpose, a few hundred milligrams of the finely ground sample material were digested at about 180 °C with an acid mixture of mainly hydrofluoric acid and perchloric acid. This is done in pressure-stable, sealed vessels (autoclaves) with Teflon inserts to achieve the high temperature required for better dissolution of various poorly soluble minerals by the resulting pressure. After fuming the acids, the dissolved sample is measured as a dilute acid solution.

Lithogeochemical units
The lithogeochemical map of Bavaria was derived from the Geological Map (1:25,000) of Bavaria. For this purpose, the mapped units of the geological map were joined into larger units containing one or more geological units. Criteria for the grouping of mapped units are similar lithological compositions, e.g., sandstones, mudstones, or limestones are each grouped, based on the assumption that the larger units will have a similar pattern of lithogeochemical composition. The resulting lithogeochemical map of Bavaria contains 184 units, assembled from about 2500 units of the geological map. Figure 1 shows the same section of the geological map with 14 units (Fig. 1a) and the resulting lithogeochemical map with 9 units (Fig. 1b).

Statistical analysis
According to Helsel (2012), geochemical data are often normally or lognormally distributed. A simple approach to visualizing the distribution of large datasets is by probability plots (Lepeltier 1969;Sinclair 1976), where normally distributed data values lie on a straight line. Figure 2 shows a typical lognormal distribution of arsenic concentrations in a lithogeochemical unit in the probability plot, and in Fig. 3, the same data are presented in a frequency plot. No clear outliers were identified in this dataset.
All geochemical datasets were examined with respect to distribution patterns. Most distributions were rather close to normal or (predominantly) lognormal (left-skewed) distributions. Based on visual inspection of the probability plot, any outliers were removed from the datasets. The cause of the outliers may be analytical errors, incorrect assignments, or atypical lithologies such as condensates, mineralization, or weathering residues that do not represent the "typical" lithologies of the respective lithogeochemical units. The distribution of element concentrations is represented by Box-Whisker plots. Several methods are available for determining statistical measures (i.e., 10th, 25th, 50th, 75th, and 90th percentiles), such as Kaplan-Meier (KM), Maximum Likelihood Estimation (MLE), and Regression on Order  (Helsel 2012). In the study presented here, the KM method was used, which is implemented in the cenfit function of the R program. The analysis was performed only for datasets where at least 50% of the data were above the detection limit and more than ten measured values were available to ensure statistical reliability (Walter et al. 2012). A major reason for using R for the analysis was that it allows automated analysis of large amounts of data in a standardized procedure.

Background concentration of elements
Statistical parameters were determined for 93 out of 184 lithogeochemical units, for which sufficient analyses were available. The area coverage of these units is about 2/3 of Bavaria. Trace elements could be evaluated in 54 lithogeochemical units with an area coverage of about 55%. To make the data publicly available, the lithogeochemical map with the statistical measures of all investigated parameters is published as a web map service and in the web browser of the Bavarian Environment Agency: https:// www. umwel tatlas. bayern. de (see Geologie → digitale Lithogeochemische Karte 1:25.000; only available in German).
Due to the large number of elements investigated, only details for the trace elements arsenic, chromium, nickel, and copper are presented below. Figure 4 shows the box-whiskerplots for these elements.
The percentiles of elements determined for the lithogeochemical units can be spatially represented using the lithogeochemical map of Bavaria described above. Figures 5 and  6 show map views of the background concentrations of the trace elements As, Cr, Ni, and Cu (see Fig. 4). The 90th and 50th percentiles of lithogeochemical units are presented in classes of concentrations. White areas represent the lithogeochemical units that could not be evaluated due to a lack of analytical samples.

Discussion
The geochemical background values of the lithogeochemical units of Bavaria were determined from 8838 samples of total content analysis for major and trace elements. The evaluation of concentration distributions showed that the majority of the distributions are lognormal. Normal distributions allow the determination of valid concentration ranges for most elements in the assigned units.
Due to the available sample density and the requirement of at least ten normally distributed samples in the lithogeochemical units for a sufficiently valid statistical evaluation, it was not possible to generate area-wide evaluations for the investigated parameters for all of the 184 differentiated  Statistical parameters given for lithogeochemical units always refer to defined lithologies and not to the chemical composition of the respective units as a whole. For example, if a unit consists of sandstones and mudstones, a separate statistical evaluation was performed for these two lithologies. In the same way, element concentrations of unconsolidated rocks are determined with respect to different main grain size fractions (e.g., clay, silt, sand, and gravel).
The samples were collected over a long period of time (abt. mid-1980s to present) and as part of various projects with different objectives. Therefore, the dataset may have some limitations in terms of representativeness concerning analytical methods, detection limits, and sample distribution with respect to rock material selection. For example, mapping geologists may have been interested in determining the geochemical content of local features, such as, for example, concretions or mineral precipitates rather than the predominant typical lithology. This issue was addressed by pre-selecting samples based on lithology and excluding concentration anomalies from the dataset. Anomalies may be due to possible analytical errors, mismatched samples, or special samples from lithologies not representative of the units. Due to the large sample quantities and, in many cases, unavailability of the sampling protocols, the causes of the anomalies cannot always be clarified for the individual sampling sites. However, anomalies accounted for a rather small portion of the total data set (less than 1% of all analytical values), so that the determined rank-correlative values are considered representative for the defined units.
Lithogeochemical units were defined by aggregating the geological units in the Geological Map of Bavaria (1:25,000), assuming that the units have a similar geochemical composition. However, this basic assumption cannot be Fig. 6 Medians (50th percentile) of selected trace elements (arsenic, chromium, nickel, and copper) of surface rocks of Bavaria validated for geological units with too few samples. Therefore, the statistical analysis for some of the geological units may be biased to some degree. However, analysis of data of lithogeochemical units, in which the same lithologies of different geologic units were present, mostly showed very good agreement. Only in a few cases, different geochemical signatures were found. In these cases, the lithogeochemical units had to be separated subsequently.
Lithogeochemical units with highly variable lithologic compositions (such as sediments with a wide variety of catchment areas such as fluvial sediments in a highly varied geological landscape), may have less significant element distributions. This can be seen in large ranges of the Box-Whisker plots of these units.

Conclusions and outlook
Background concentrations for 39 elements were determined by geochemical analyses of rock samples from all over Bavaria. An important prerequisite for establishing valid background concentrations is the statistical proof that the samples have a characteristic distribution, as shown in Fig. 2 and 3. The evaluation of the data collective with predominantly normal or lognormal distributions has shown that the data set largely satisfies this requirement.
Based on these findings, maps with typical background values of elements were successfully produced using the Geological Map of Bavaria 1:25,000. These maps can serve for various applications, especially in the environmental field, such as: -Soil protection: delineate areas with elevated geogenic background concentrations of elements versus areas with anthropogenic influences -Hydrogeology: correlate the hydrogeochemical composition of groundwater to the lithogeochemical background of aquifers -Soil science: linking the soil geochemical composition to the lithogeochemical properties of the underlying rocks.
In addition, knowledge of the characteristic distribution of geochemical parameters in lithogeochemical units and their delineation in maps offers numerous applications in research.
In the future, additional samples will be added to the database to provide better area coverage. It is estimated that additional 2,000 samples are needed to achieve near 100% coverage. In addition, the maps shown in Figs. 5 and 6 still need to be included in the web browser. So far, the percentiles of the background concentrations of the elements are given only as numerical values.