4.1 Definition of lithological groups
Based on the information that could be retrieved about the specimens (rock type, mineralogical composition, modal composition, textural description, etc.) we classified the samples in broad rock types, following the classical criteria of classification for sedimentary, igneous and metamorphic rocks (e.g. Charmichael, 1989; Carmichael & Klein, 2020). We adopted the classical petrological (e.g. Jahns & Kudo, 2020) and sedimentological (e.g. Bissell et al., 2020) terminology, commonly in use in geological literature, but we preferred always terms describing the mineralogical composition and microstructure rather than the petrogenesis. For example we preferred the terms “feldspar gneiss”, “mica-bearing gneiss” and “schists” to the terms orthogneiss and paragneiss, because many physical properties are mainly related to mineral and modal composition and rock texture. We defined in this way 26 rock types (see Additional file 1: S2 “lithology list”) for which we had obtained a sufficient number of physical properties.
Subsequently, we grouped the 26 rock types in 22 lithological groups, establishing a correlation between each lithological group and the 69 lithology types that are listed in the legend of the geotechnical map of Switzerland (Federal Office of Topography, 2006) version 1/2000, map scale of 1:500,000, as shown in Additional file 1: S2 “lithology list”. This correlation was necessary in order to represent the distribution of physical properties with statistical significance (namely density and velocity) on a map. Water, ice, unconsolidated debris and unconsolidated fine-grained deposits are also listed in the table for completeness, but no data were collected for them and the values that were used to compile the maps (see Sect. 5) were taken from literature and refer to worldwide data.
While for the majority of rock types the classification in a lithological group has been straightforward, in some cases the attribution to a specific lithological group has been less certain and somehow subjective. This difficulty arises because in SAPHYR we are interested in lithologies that are representative of the whole continental crust (including its deepest parts) with regard to physical properties of rocks, while in the geotechnical map (Federal Office of Topography, 2006) the legend is designed to represent the technical use of the shallow subsurface (extraction of mineral resources, building projects) and the rocks are subdivided according to their formation (genesis).
In order to explain the compromising efforts we quote, f.e., the case of conglomerates and breccias. The two lithotypes are jointly mapped in the geotechnical map (Federal Office of Topography, 2006), but the two terms in a broad sense refer to rocks of different origin: conglomerate is a coarse-grained clastic sedimentary rock that is composed of a substantial fraction of rounded to sub-angular gravel-size clasts; breccia is a rock composed of broken fragments of minerals or rock cemented together by a fine-grained matrix and it may derive from igneous or sedimentary rocks (e.g. Shukla & Sharm, 2018). As a result of the different origins and consequently different mineral and textural composition, the physical properties of the breccias and conglomerates are statistically quite different (see the example of density in Fig. 6a) with breccias showing a density range twice as large as that of conglomerates. Thus for our purposes it makes sense to keep the two rock types separate in the dbase. The treatment of this and analog cases in the map preparation phase is explained later in the text.
Another example of a case where the lithology classification we adopted was more detailed that in the geotechnical map (Federal Office of Topography, 2006) is the case of tonalities and granodiorites. Both rock types are relatively abundant in the laboratory data from literature, but not frequently enough encountered in surface outcrops to be mapped separately at the scale of the geotechnical map (Federal Office of Topography, 2006) and, therefore, neglected in its legend. We kept tonalites and granodiorites as separate groups in the dbase even if it is impossible to distinguish their outcrops on the maps. The same case applies to rock types typical of lower crust in the Ivrea-Verbano Zone such as kingizites, stronalites (Bertolani & Garuti, 1970), granulites, that are quite important for the representation and modeling of the lowermost part of the continental crust (see Fig. 6b).
In most of the cases, however, it was not necessary to add details and subdivisions to the lithology list of the geotechnical map (Federal Office of Topography, 2006). On the contrary, more often it was necessary to group geotechnical-lithological units into broader rock types units. A good example is the case of sandstones, the composition of which may vary over a broad range, depending on the nature and amount of the cementing matrix, on the grain size and the sorting, on the degree of compaction and therefore on the porosity, or on the interlayering with other sediments. Such a fine distinction (interlayering, compaction) is beyond our knowledge from the single laboratory sample, or too arbitrary to be applied systematically to the whole data population. Therefore, we used the term “sandstone” in its broader sense, with the consequence that many physical properties measured at outcrop conditions are distributed over a wide range of values.
In terms of representativeness and for visualization of physical properties on GIS maps, an advantage to group the original lithology types of the geotechnical map (Federal Office of Topography, 2006) into general lithology groups (in total 20 groups referring to solid rocks and three more to unconsolidated deposits, ice and water) is the increase of data population for each lithology available. Once our data were grouped in rock types, it was straightforward to identify the rock types of which we did not have a sufficient number of samples (a minimum value of 7 for each rock type was arbitrarily chosen) in order to perform statistical analysis (as described in Sect. 2), and for which new sampling and measurements was needed, as described in Sect. 3.3.
4.2 The multi-layered tabular dbase
The available data have been collected and organized in Microsoft Excel spread sheets where each sample is identified through its name and geographic coordinates. We choose the form of spread-sheets because it provides the easiest way to import them directly into ArcGIS (ESRI 2011. ArcGIS Desktop: Release 10. Redlands, CA: Environmental Systems Research Institute), the selected geographic information system (GIS) for working with maps and geographic information in this study. ArcGIS was then used for creating maps of the areal distribution of the physical properties, as explained later in Sect. 5 dedicated to maps description. To localize the samples, we adopted the Swiss geographic coordinate system, therefore all latitudes and longitudes have been recalculated in Cartesian coordinates. Altitude, or depth in case of borehole or tunnel data is also recorded.
We kept two separate dbase parts, one for boreholes and one for outcrops. The dbase on boreholes refers to data reported by NAGRA reports and regards deep boreholes (more than 1 km depth) in the Swiss plateau. The dbase referring to outcrop data also contains data from tunnels and shallow boreholes that are identified by a special code. As borehole samples are collected at different depth, their physical properties (or at least part of them) might be influenced by the burial history, and equally the properties collected from outcrop samples might be affected by the exhumation history.
In the same spread-sheets we assembled information about rock type, geologic unit or stratigraphic unit, and tectonic unit, together with all the physical properties measured on the sample and their respective error estimates, when available. The source of the information (literature reference, public or private report, unpublished data with laboratory and authors’ names) was also recorded for each data. No data of unknown origin was inserted in the dbase.
In the dbase, data have been grouped on the basis of the physical property and on the basis of the measurements procedure. Error estimates for each properties were recorded as well, when present in the original documents, or calculated for new measurements. When necessary, data were recalculated in SI units.
4.3 Physical property processing
Each data collected in the spread-sheets represent a local point source of information. The aim of the dbase is not only to collect in a systematic way the physical properties available from laboratory experiments, but to also present them in the perspective of the assembled data of same kind (physical property) from same lithology to provide the user of the dbase with information about the representativeness of the individual data and finally to display the data on bulk density or ultrasound velocity as thematic maps.
The primary goals of the processing are to identify the range and distribution of the values and their uncertainties for every physical property and every lithology. To establish those distributions and to obtain uncertainties we take into account the reliability and uncertainty of the individual data and, in particular, search for and identify outliers that result from wrong information.
By analyzing the distributions with a statistically significant number of data for each property and rock type, it was possible to identify several cases with mistaken reports of lithology, location or obviously erroneous measurement reports. In some cases of wrong lithology association, f.e., due to local naming, outliers in more than one property could be cross-linked and thus the data could be corrected. In all other cases, the outlier data was marked in the dbase as problematic, neglected for further processing and not used for statistical treatment for the preparation of thematic maps. The density values were of particular usage to detect “outlier data” for reasons of wrong lithology association as these data were clearly falling outside the reasonable tails of Gaussian distributions.
4.4 Variation in density and Vp within a lithology
There is a large variability in the number of data among different physical properties and among different lithologies. When observing the number of data for each measured property as shown in Fig. 4, it is evident that the dbase contains a large amount of data for density while data for other properties such as, for example, ultrasound velocity and especially shear waves velocity data, are rare.
Density is the most often measured physical property and consequently it is the most abundant property in the dbase. Moreover, density is a simple scalar quantity and as such it does not contain the complications linked with tensor properties, namely the directionality of the measurement. For these reasons, density was used first to develop the statistical treatment of the data. The grain and bulk density data of each lithology group were used to calculate mean, median, and range distribution. When possible, the most frequently occurring, or repetitive, value in each sample of data, the mode, was calculated too.
The physical property distribution diagrams for each lithology are provided in Additional file 1: S5 “distribution of physical properties for each lithology”. Here we discuss only a few selected examples with the intent to illustrate some key observations and conclusions that can be taken from such diagrams.
Case 1
Granites—the ideal case.
Grain densities for granites are a good example of a high-quality data set. Figure 7 shows the data distribution of samples from outcrops (68 data) and boreholes (113 data). Both data sets show a narrow distribution (standard deviation of 48 kg/m3) almost symmetric around the mean value (2640 kg/m3 for boreholes, 2670 kg/m3 for outcrops). The definition of granite is well constrained in the geological terminology. It refers to a specific chemistry and mineralogy and the rock for its nature is almost non-porous and exhibits no or very minor anisotropy. Moreover, the data collection is abundant, which makes the statistic treatment more solid and representative. The distributions of outcrops and boreholes data are quite similar, both for the reasons mentioned above and for the equally large number of data. In the case of granites, the average value of 2652 kg/m3 with a standard deviation of 50 kg/m3 was adopted to represent all outcrops of the units n. 50 and 51 of the geotechnical map (Federal Office of Topography, 2006) (see Sect. 5). Note further, that this value corresponds very well with the average bulk density of 2670 kg/m3 used in standard Bouguer gravity correction for the near-surface crystalline basement rocks.
Case 2
Dolomites—boreholes versus outcrops data quality.
The case of dolomites (Fig. 8) is a good example of different data quality in outcrops and borehole data. While we have quite a large number of bulk density data from boreholes (117) only a limited number of density data derive from outcrop samples (17). As a result the distribution of outcrops data is a bit scattered, but the average values are quite similar. Nevertheless, standard deviations are in both cases relatively small, probably resulting from a relatively precise definition of the term dolomite that refers to a compact sedimentary carbonate rock that contains a high percentage of the mineral dolomite. Combining outcrop and borehole data, we obtain an average value of 2797 g/cm3 with a standard deviation of 76 g/cm3 (for 134 samples) and these values were adopted as representative for the geotechnical units n. 47, 48, 49 of the geotechnical map (Federal Office of Topography, 2006).
Case 3
Ultramafic rocks—alteration and weathering.
Grain density distributions for peridotites and serpentinites present two clearly separated peaks (averages are 3192 and 2733, respectively) with a quite extended overlap area (Fig. 9) between them. The data has been confirmed by quality screening to be reliable and precise. Serpentinites are the alteration product of peridotites. Peridotites gradually transform into serpentinites due to more or less alteration of olivine (single crystal end members mineral density of 3270 kg/m3 for fosterite up to 4390 kg/m3 for fayalite; Chen et al., 2002) into serpentine polymorphs (mineral density of 2520–2650 kg/m3, e.g. Deer et al., 2013, and references herein), while the excess Fe goes into magnetite. The overlapping area between the two peaks is the result of this gradual transition from pure peridotite to serpentinite. Almost all outcropping peridotites are affected by variable degree of serpentinizaiton, because the process of exhumation implies decompression and fracturing and quite often fluid circulation is involved. Even a tiny amount of serpentine reduces drastically the density of the rock and affect many other physical properties, such as magnetic and seismic properties (e.g. Christensen, 1966; Fujii et al., 2016; Horen et al., 1996; Kern & Tubia, 1993; Toft et al., 1990). Therefore we believe the samples with density below 3.1 kg/m are serpentinites with relicts of peridotites. We choose the value of 3.1 as threshold on the basis of the density of olivine. It has to be noted that serpentine is stable also at great depth (Ulmer & Tromsdorff, 1995) therefore serpentinized peridotites are not only a characteristic of outcrops, but they can also be found at depth, therefore we think it is important to show also the values of “mixed” peridotites and serpentinites.
Serpentines and peridotites show quite pronounced maxima in their distribution curves that allow the attribution of two quite distinct values of density: 3193 kg/m3 and 2726 kg/m3 for peridotites and serpentinites, respectively. Standard deviations in both cases are quite large reflecting the above mentioned effect of progressive alteration and change in modal composition.
Case 4
Breccias and conglomerates—subdivision not representable in the geotechnical map (Federal Office of Topography, 2006).
As already mentioned, in most of the cases it was necessary to group individual lithology types of the geotectonic map into broader lithological groups (see Additional file 1: S2 “lithology list”). Only in few cases and for specific reasons we created subdivisions within the lithological types. One such example is given by breccias and conglomerates that are joined in one lithological unit in the geotectonic map (23: breccia and conglomerate, 46: calcareous breccia or conglomerate; 64: sericite-rich conglomerate and breccia) but they represent rocks of different origin. Especially a breccia may have a variety of rocks of different origins, including sedimentary breccia, tectonic breccia, igneous breccia, and impact breccia. The various origin of the breccias is reflected in a density distribution that is significantly wider compared to that of conglomerates (Fig. 6a). Nevertheless, for sedimentary breccias the difference with conglomerates is only in the shape of the clasts (more rounded in conglomerates and angular in breccias) deriving from a less mature sediment in the case of breccias. In this case the two terms largely overlap, as may be seen with the two density distributions. We decided to keep breccias separate from conglomerates in the dbase, but to adopt a common average and a large standard deviation for representation in the GIS maps.
4.5 Data presentation
The data distribution has been a valid instrument to identify outlier data and to determine and validate the criteria we adopted in the grouping of lithologies (see Additional file 1: S2 “lithology list”). For each lithology group the standard deviation was calculated assuming a Gaussian distribution around the mean value. Arbitrarily we choose a minimum number of 7 data over which to perform statistics. In other words, rock types with less than 7 measurements from different samples, f.e., of bulk density were not considered for statistical treatment and no average and standard deviation is listed in the dbase. This statistical treatment of the data for each group of rock types and of each physical property was primarily used for the assessment of the individual data quality and completeness (Sect. 4.2) and for the assessment of physical property value range, distribution and uncertainty. Subsequently, it provided also the basis to transform the data point collection into a representation of maps of physical property distribution.
The use of a GIS allows also an easy visualization of complementary data as in the examples shown in Fig. 10 for all properties measured in a sample from an outcrop (inset a) or from a borehole sample (inset b). By recording coordinates and depth, in case of boreholes, it is easy to show from the same geographical locations many data collected on different lithologies at various depth.
Statistical information for each lithology group are presented for selected physical properties in the tables in Additional file 1: S4 “dbase of physical rock properties” and in diagram format in Additional file 1: S5 “distribution of physical properties for each lithology”. Statistical data include number of available data (population size), minimum, median and maximum values, standard deviation (σ) and standard deviation interval (σ-interval). Moreover, a somewhat subjective description of the binned data distribution may provide some information about the general quality and reliability of the data.
The great amount of density data collected both from literature and by new measurements allowed the attribution of average values to the rock formations and the drawing of a surface rock density distribution map of Switzerland (Fig. 11). In parallel and in analogy with the density map, we developed a surface P-wave velocity distribution map (Fig. 12). We did not produce an analogous map for S-wave velocity because the data in literature for S waves are scarce and the sampling does not allow a reliable extrapolation from single sample data to whole rock formation.
The tables in Additional file 1: S4 “dbase of physical rock properties” contain the individual values that were used to prepare the maps in Figs. 11 and 12. Note that data of some lithology groups (e.g. evaporites, stronalites and kinzigites) that mainly outcrop outside Switzerland do not appear in Figs. 11 and 12. Since rocks are defined as agglomerations of minerals, and minerals are naturally occurring crystals, ice as a naturally occurring crystalline substance has been considered as a rock (lithology group 21 in geotechnical map). As laboratory data for ice collected in the Swiss Alps was not available, the data reported in tables and figures have been exceptionally collected solely from worldwide literature. With reference to the dependence of ice density and velocity on temperature, only data collected in the range of − 40 °C to 0 °C have been considered, representing Swiss Alps temperature conditions. Unconsolidated sediments (e.g. sand, gravel, pebbles, stones and debris blocks), that are abundant at the surface, and therefore appear in many areas on the geotechnical map (Federal Office of Topography, 2006), have been considered also as a “lithology group” and again due to lack of lab measurements density data was collected from literature reporting in situ measurements collected mainly in Switzerland (Anselmetti et al., 2010) but also worldwide (Mavko et al., 2009 and references herein; Schön, 2015 and references herein).
Figures 11 and 12 display the surface distribution of average values per lithology group of bulk rock density and P-wave velocity at surface conditions, respectively, over the entire Swiss territory. In Fig. 13 we also display the standard deviation of density for different lithological groups. Note that the standard deviation in this case does not refer to uncertainty estimates but rather it indicates the range of variability of density for each lithology group that may result from various conditions and the evolution of the rock (see, f.e., Fig. 9 and explanations in text). The effect on density from compaction by overlying mass is greatest for unconsolidated sediments as, f.e., those in the overdeepend Alpine valleys (Kissling & Schwendener, 1990) and hence, these lithology groups show the largest standard deviation (Fig. 13).