[go: up one dir, main page]
Next Article in Journal
In Search of Phytoremediation Candidates: Eu(III) Bioassociation and Root Exudation in Hydroponically Grown Plants
Previous Article in Journal
LiDonit®—A Potential Secondary Raw Material for Ceramic Applications in Concentrated Solar Energy
Previous Article in Special Issue
Leveraging Domain Expertise in Machine Learning for Critical Metal Prospecting in the Oslo Rift: A Case Study for Fe-Ti-P-Rare Earth Element Mineralization
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Minerals
Volume 14
Issue 8
10.3390/min14080753
minerals-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Derivation of Predictive Layers Using Regional Till Geochemistry Data for Mineral Potential Mapping of the REE Line of Bergslagen, Central Sweden

by
Patrick Casey
*,
George Morris
and
Martiya Sadeghi
Geological Survey of Sweden, Box 670, SE-751 28 Uppsala, Sweden
*
Author to whom correspondence should be addressed.
Minerals 2024, 14(8), 753; https://doi.org/10.3390/min14080753
Submission received: 16 April 2024 / Revised: 12 July 2024 / Accepted: 22 July 2024 / Published: 26 July 2024
(This article belongs to the Special Issue Editorial Board Members’ Collection Series: "Critical and Strategic Minerals")
Browse Figures
Figure 1
<p>Bedrock map of the REE line taken from the SGU 1:1,000,000 map from SGUs database. Coordinates are based on Swedish SWEREFF-99TM system. The red points represent an up-ice coordinate transformation of original sample locations to roughly 16 km NNW based on inferred transport distance. Location the REE line is shown in the inset map of Sweden in a red rectangle.</p> "> Figure 2
<p>The study area with soil depth shown in meters. The thickest sediments are typically associated with fluvial or lacustrine areas. Arrows demonstrate ice-flow direction as measured from orientation of striations in bedrock.</p> "> Figure 3
<p>Flow chart defining mineral systems from critical processes through mappable proxies.</p> "> Figure 4
<p>(<b>A</b>–<b>C</b>) Biplots of PC scores for PCs 1 through 4 for the ILR transformed all-element data. Individual points represent individual till samples, and colors represent their K-means cluster membership.</p> "> Figure 5
<p>Interpolated factor scores from the all-element data. (<b>A</b>). PC1 shows spatial correlation to mafic (high factor scores) and felsic (low factor scores) bedrock. (<b>B</b>). High scores along PC2 demonstrate possible correlations to the more evolved 1.85–1.75 Ga granites and pegmatites. (<b>C</b>). High factor scores show special affinity to the Norberg REE mineralizations. Black arrows indicate ice direction.</p> "> Figure 6
<p>Clusters membership of individual till samples for all-element data plotted over the bedrock map of the REE line. REE mineralizations are shown in red crosses.</p> "> Figure 7
<p>Biplots of trace-element data after ILR transformation for PCs 1 through 5. (<b>A</b>). PC1–PC2. (<b>B</b>). PC1–PC3. (<b>C</b>). PC1–PC5.</p> "> Figure 8
<p>(<b>A</b>–<b>C</b>) Interpolated results for principal component analysis of the trace-element data. (<b>A</b>). PC1 demonstrates a rough mafic (low scoring) and felsic (high scoring) divide between till samples. (<b>B</b>). Positive PC2 scores demonstrate association with 1.85–1.75 Ga granites and pegmatites. (<b>C</b>). PC5 shows correlation with known REE deposits. Red crosses are known REE mineralizations. Arrows indicate ice direction.</p> "> Figure 9
<p>Clustering results from the trace-element data plotted over the bedrock map of the REE line.</p> "> Figure 10
<p>Clusters 2, 3, and 6 shown overlain on the interpolated factor scores of PC5 of the trace-element data.</p> "> Figure 11
<p>Principal components of till samples from each of the three clusters highlighted in <a href="#minerals-14-00753-f011" class="html-fig">Figure 11</a> showing association with mineralization and positive loadings on the fifth principal component of the trace-element data. (<b>A</b>). Cluster 2, (<b>B</b>). Cluster 3. (<b>C</b>). Cluster 6.</p> "> Figure 12
<p>Interpolated results of the alteration index of till samples within the REE line with known non-REE bearing mineralizations shown as red points (sulfide bearing) or green triangles (Fe-oxide).</p> ">
Versions Notes

Abstract

:
With the increasing need for rare-earth elements (REEs) to reach the goals of the ongoing green energy transition, new and innovative methods are needed to identify new primary resources of these critical metals. This study explores the potential to use a non-biased, uniform till dataset to generate evidentiary layers that describe these critical factors and geochemical anomalies to aid mineral potential mapping (MPM) for REEs using machine-assisted methods. The till samples used in this study were collected from the “REE Line”, a sub-region within the Bergslagen lithotectonic province, Sweden, where numerous REE mineralizations occur. Multiple approaches were used in this study to isolate geochemical anomalies using multivariate methods, namely principal component analysis (PCA) and K-means clustering. Additional factors for classifying till samples were also tested, including alteration indices. Using known REE occurrences in Bergslagen as validation points, the results demonstrated the usefulness of multivariate methods applied to till geochemistry for predictive bedrock mapping, and to identify potential areas of REE mineralization within the REE line. The results of the alteration indices showed that the till geochemistry demonstrated similar levels of alteration when compared to the underlying bedrock, allowing for a regional alteration map to be generated. These results show that regional-scale till sampling can provide low-cost data for mineral exploration at the regional scale and generate usable evidentiary layers for GIS-based MPM.

1. Introduction

With the ongoing green transition and increasing need for high-tech consumer and industrial devices, rare-earth elements (REEs) are in high demand, with consumption of these elements expected to increase by 65% by 2035 [1]. Currently, the European Union is wholly dependent on the import of these critical elements and the need for domestic sourcing has become more apparent over the last decade. The REE group consists of lanthanide elements, with yttrium and scandium generally grouped with REEs due to similar chemical behavior. Despite the designation of “rare”, the abundance of the most abundant REE, cerium, is roughly similar to copper in the crust and the least abundant, lutetium, is 200× more abundant than gold [2]. What is rare about these elements are the processes that concentrate them into exploitable deposits, and these processes are not fully understood. REEs are typically divided into two groups: light rare-earths (LREEs), La-Eu, which are more abundant in the crust than heavy rare-earth elements (HREEs), Gd-Lu, due to the larger ionic radius of LREEs and thus the more favorable incorporation of smaller HREEs into the mantle minerals.
Sweden has an intimate association with REEs. Many were first isolated from minerals sourced from Swedish mines, including the famous Ytterby pegmatite mine outside of Stockholm, where eight of the REEs were discovered [3]. Also well known are the Bastnäs mines in the Bergslagen mining region of central southern Sweden, where the elements La and Ce were discovered [4], which became the first economic mines for REEs in the 19th century. The Bergslagen region hosts the so-called “REE Line”, a southwest to northeast trending series of Fe-REE ± Cu ± Mo ± Bi ± Au mineralizations, which includes the famous Bastnäs deposit, as well as numerous other smaller REE mineralizations [5].
Identifying new mineable sources of REEs is key for keeping up with the increasing demand for critical materials for the manufacture of green technology, and innovative methods are needed to identify new prospective areas. The “Exploration Information System” (EIS) [6] is a method for transforming criteria used to identify deposits using the mineral systems approach method (MSA) [7] into predictive layers to be utilized as proxies within mineral prospectivity mapping (MPM). The MSA utilizes a holistic approach to the discovery of mineral systems, treating deposits as a small part of the larger systems that lead to their formation, such as energy or mass transfer, transportation, and depositional systems and incorporating these factors into a unified model for formation. As part of the EU-funded “EIS Project”, work is ongoing to create a user-friendly geographic information system (GIS) tool to create prospectivity maps using a variety of methods including artificial intelligence (AI) and machine learning (ML). These computer-assisted methodologies are powerful tools in aiding in MPM [6] and their application in exploration for REEs can be an important complement to this method, and the generation of useful evidentiary layers for training these models is an important aspect of EIS. Sweden in particular is an ideal test bed for these methodologies due to the large amounts of data and knowledge collected through mineral exploration and by the Geological Survey of Sweden (SGU), including geochemical data from glacial tills. A preliminary MPM on REE mineralization in this area has been attempted by [8] but this approach did not include geochemical data from till.
Glacial till is created during the advance of glaciers and ice sheets as they grind down the bedrock substrate over which they advance. This process pulverizes the bedrock and transports it downstream in the ice-flow direction, where it is ultimately deposited during the retreat of the glaciers. While till may superficially be considered a soil, unweathered till is best interpreted as simply a mélange of rock in a more finely powdered and distributed form while still retaining a composition of the bulk rock from which it originated [9]. The geochemical signature of the eroded substrate is carried along with the till, and these signatures may contain anomalies that can aid in the identification of areas of potential interest for mining, such as the Björkdal gold deposit in northern Sweden, which was identified using till geochemistry [10]. Sampling of glacial deposits, such as glacial tills, and boulder tracing, or the following of mineralized boulders to their bedrock origin, make up the processes known as drift prospecting. These prospecting methods have been used throughout previously glaciated terranes to identify potential mineralizations including copper, iron, diamond [11], and iron oxide–apatite [12] mineralizations. National surveys organized by geological surveys often conduct regional-scale sampling (~7–10 km2 per sample), while prospecting surveys typically reach camp to deposit scale [11]. Till surveys typically include geochemical analysis of various fractions, typically <63 µm or finer, and may include heavy-mineral analysis to identify indicator minerals typical of mineral deposits [11].
As nearly 75% of the land area in Sweden is covered by glacial tills [13], Sweden is an ideal region for till sampling. The Geological Survey of Sweden (SGU) conducts the Regional Till Sampling Program, which aims to create a uniform dataset of a large suite of geochemical data from till in Sweden. By collecting till samples with a roughly even spatial distribution over the entire country, any inherent biases in a dataset that is focused on samples taken in direct support of exploration can be removed and a relatively unbiased geochemical database is created. Spatial variations in geochemical signatures identified within till, in particular using multivariate associations, have been used to identify areas of potential Li mineralization in brown and greenfield areas [14], and further potential indicators of mineralization, or pathfinder signatures, may be able to be isolated from till data.
In this study, detailed analyses of the SGU regional till dataset will be used to determine regional geochemical trends from which mappable critical mineralization processes may be identified. By analyzing these data with multivariate analyses, including PCA and K-means clustering, along with alteration chemistry, a better understanding of the regional geochemistry of the geology of the unexposed bedrock in the REE line can be created, and regional-scale geochemical anomalies related to processes that can lead to REE mineralization can be identified. While higher sampling density studies are needed to target specific mineralizations [11], the results of the compositional data analysis of the SGU dataset can be compared to known mineralizations to identify any potential regional geochemical trends that may influence the formation of these deposits. As a part of the EIS project, this study serves to provide a background case study [6] on the generation of predictive layers and data preprocessing. The results of this study are to be incorporated as part of further tests of the EIS software (experimental version 0.5.1) through the creation of mineral potential maps for REEs in Bergslagen.

2. Geological Background

The Bergslagen region is located in southern central Sweden, within the Fennoscandian shield, and comprises over 6000 known mineralizations occurring predominantly within a Paleoproterozoic (1.92–1.87 Ga) sequence of felsic supracrustal volcanic rocks. The region was moderately metamorphosed during the Svecokarelian event, ranging from greenschist facies in western Bergslagen through upper amphibolite facies through the east. A wide range of mineralizations occur, including the large magmatic iron oxide–apatite mineralizations at Grängesberg and Blötberget, large base-metal deposits rich in Zn, Pb, Cu ± Ag, and Au such as Zinkgruvan and Garpenberg as well as thousands of smaller deposits. The history and geology of Bergslagen has been extensively described in the other literature [15,16] and as such this geologic background will focus on the REE line.
The study area for this project is the REE line shown in the map in Figure 1 as generally defined by Jonsson and Högdahl [17]. The dominant lithologies are 1.92–1.87 Ga granitoids to syenites, as well as supracrustal felsic rocks, dominantly rhyolitic to dacitic in composition. Younger granites intrude throughout the study area, such as the 1.8 Ga granite seen in the left side of the map in Figure 1, a rapakivi-like granite dating to the Sveconorwegian event, and the granites of the 1.85–1.75 GP-suite whose emplacement led to a later stage of mineralizations [15]. Gabbroids and dioritoids, likely pre- to syn-depositional with supracrustal sequences, occur throughout the REE line. Mineralization is found in the felsic supracrustal sequences, typically within intercalated biogenic carbonate beds that have been variably altered and metamorphosed.
REE mineralizations here are generally skarn type, epigenetic, massive to disseminated magnetite Fe deposits with lesser sulfides, typically Cu with associated Bi, Co, Mo, and other minor sulfides [17]. The REE deposits are typically found in altered carbonate bodies, which occur along a narrow stretch of the REE line (Figure 1). This particular mineralization style is referred to as Bastnäs-type mineralization [18], and it has been subdivided into two mineralization styles: mineralizations enriched in LREEs such as at the Bastnäs mines, just west of Skinnskatteberg, and HREE-enriched mineralizations such as the Norberg district [19]. REE mineralization of the first type is hosted within REE silicates: ferriallanite, cerite, and the fluorcarbonate bastnäsite, found within silicate skarns [20]. In the iron mines of the Norberg district, such as at the Malmkärra and Östanmossa mines, the mineralization is Y- and HREE-enriched, and excess F in the system led to the dominance of F-rich REE minerals such as bastnäsite, parisite and fluorobritholite-(Ce, Y, Nd) [21]. REE mineralization within the mines typically occurs as pods of REE minerals, up to a meter in thickness and striking over 10 s of meters [19]. While these bodies of REE ore are small, diffuse mineralization of REE minerals typically occurs within the entirety of the ore deposits and often into a diffuse alteration halo in the country rock surrounding the mineralizations [17,19], which provides a larger geochemical footprint of REE enrichment surrounding the deposits.
Mineralization within Bergslagen is associated with Mg, K, Ca, and Na alteration, where these elements are enriched in altered rocks. The iron and sulfide mineralizations within the REE line are typically strongly magnesium altered with the destruction of feldspars and building of calc–silicate skarns with Mg mineral assemblages. Early models for the formation of the Bastnäs mineralizations coupled this alteration with the intrusion of late-stage granitic bodies and their associated fluids [18]; however, this model has fallen out of favor. While there is still debate over the formation of the Bastnäs deposits, the currently favored model proposes magmatic-sourced high-temperature fluids reacting with biogenically produced carbonate beds forming the deposits [5] before the intrusion of the granitic bodies.

3. Quaternary Geology

Quaternary Geology of Bergslagen

The landscape of the Bergslagen region has been invariably altered by repeated glaciations, with the last glacial maximum (LGM) occurring during the Weichselian at approximately 20 kya when the Scandinavian ice sheet reached an extent as far as northern Poland and Germany [22]. The landscape was altered repeatedly by the ice sheets, with the final retreat of the ice occurring after the Younger Dryas advance at around 11.5 kya, leading to the deposition of the uppermost tills in Sweden [23]. Ice sheets in the Bergslagen region were warm-based during the younger dryas deglaciation as evidenced by elongated glacial landforms including crag-and-tail formations, rouche moutonnée, and drumlins. The general orientation of these features, as well as striations in the bedrock, give an ice-flow direction of north–northeast within the Bergslagen region [24]. In northern Sweden, where glacial ice was cold-based, an ice transport distance of 6 km is assumed, while in the southernmost regions of Sweden, where the ice was warm-based, transport of up to 40 km is assumed [13]. Some tills sampled in the region were subsequently inundated by marine waters and are occasionally covered by marine clays if they are situated beneath the former highest coastline. When sampling care is taken to avoid sampling tills that have been reworked by post-glacial processes. After the retreat of the ice sheets, typical podzol horizons formed within the till, with the C-horizon considered to be generally unweathered [11,24]
Till thickness varies throughout Bergslagen, with thinner tills deposited on ridges and promontories, and thicker tills deposited in topographical lows such as valleys and depressions. Within the present study area, tills are estimated to have a maximum thickness of around 10 m in topographical lows and 1–3 m on topographical highs. The map in Figure 2 demonstrates the average soil thickness throughout the study area.

4. The Mineral Systems Approach to Mineral Potential Mapping and Its Application to Bergslagen

The formation of a mineral deposit is the result of numerous complex Earth systems acting in concert and a deep understanding of these critical systems is key for creating a formational model for the targeted mineralization type [7]. Critical factors in the MSA include source, transport, trap, modification, and preservation. Detailed geological studies are thus required for modeling a mineral system to identify targetable criteria for exploration, and in particular which factors of these models are “mappable”, or translatable, into a GIS system for use as evidentiary layers in MPM [6,25]. The small size of deposits within Bergslagen, particularly of the REE-rich deposits of the Bastnäs type, requires an understanding of the systems that led to the formation of these deposits as a whole to identify prospective areas at a regional to district scale using the MSA model. Despite the small nature of these deposits, the REE line presents a unique test bed in Sweden for the mineral systems model as the area has numerous REE deposits in a relatively small area; these REE deposits have been well studied by SGU and other researchers, and the area has been well covered during the SGU till sampling campaigns.
A mineral systems model for skarn-type REE mineralization in Bergslagen presented in [26] is as follows:
  • Lithostratigraphic factors: REE skarn deposits in Bergslagen form within supracrustal inliers of felsic volcanic rocks of calc–alkaline affinity and in proximity to banded iron formation. Skarn or limestone contacts with the supracrustal rock, indicating favorable conditions for REE mineralization.
  • Geochemical factors: spatial associations of elements related to skarn-type mineralizations, i.e., Fe, Mg, REE+Y, and P.
  • Geophysical factors: REE mineralization demonstrates positive associations with areas of high magnetic intensity based on aeromagnetic surveys. Highly magnetic rocks stretch along the length of the REE line, representing the magnetite-bearing mineralizations.
  • Alteration: Rocks in Bergslagen are variably altered, with regional-scale alteration stronger in the vicinity of mineralizations. Alteration is typically ±K, ±Mg, or ±Na, and occasionally lesser Fe alteration. Thus, tools such as the alteration index [27] or the chlorite–pyrite–plagioclase index (CCPI) [28], which examine variations in the ratios of K, Mg, Na, and Fe originating from mineral alteration, can be used to identify the degree of alteration.
  • A key factor in the formation of mineralizations in Bergslagen is the circulation of hydrothermal fluids, which requires a heat source. Heat generated by intrusive or synvolcanic rocks, with ages from 1.97 to 1.87 Ga and 1.8 Ga, is proposed to have acted as a driver for hydrothermal circulation.
  • Structural factors: Spatial distribution of REE mineralization in Bergslagen shows a relationship to the D2 deformation phase, or at least processes coeval with D2 [29]. The pattern of magnetic anomalies in the REE line is s-folded, with known REE deposits occurring within these s-folds, more specifically where the fold pattern turns to a more northerly direction.
In summary, since most skarn-related REE mineralizations in Bergslagen are found in this area, and less so elsewhere, it seems likely that (1) the presence of banded iron formations, skarn, and limestone, (2) the phase of intense and wide-spread hydrothermal alteration, and (3) processes at the time of D2 together were somehow crucial to the formation of REE mineralizations. The BIFs, with local skarns and carbonate rocks, may have served as trap rocks during alteration since their formation [8]. The phase of hydrothermal alteration may locally have led to the initial LREE enrichment of the country rocks as described elsewhere [30]. Metamorphic to metasomatic processes during M2/D2 may have formed fluids that released the REEs from the country rocks and then precipitated them in trap rocks at certain structurally favorable sites.
These factors can be translated to a GIS mineral system approach, which is summarized in the flowchart shown in Figure 3. From these processes, this paper aims to extract the proxies that can be identified using till geochemistry to build evidentiary layers for MPM (Table 1).

5. Materials and Methods

5.1. Till Sampling Methodology

Till samples were collected from hand-dug pits from the C-horizon of organic free mineral soil at a depth of around 1 m depending on the depth of plant activity. Till sampled from this depth is typically representative of the youngest ice advance [11]. Approximately 1 sample is collected from each hand-dug pit per 6.5 km2 with an average spacing of 2.5 km between till samples. Care is taken in areas where areas were inundated by seawater after glaciation to sample undisturbed tills below marine clays or glacio-fluvial sediments. Properties of the till such as color, rough grain size estimates, and general composition of clasts larger than 3 cm are noted. Field interpretation of types of till are conducted, with a simple binary of traction till or meltout till assigned to each sample. Samples are vacuum-dried at 60 °C and sieved into three fractions: >2 mm, 2 mm–63 µm, and <63 µm. The fine fraction (<63 µm) is analyzed as this provides mono-mineralic grains, which reduces the noise induced by larger lithic (multimineral) grains, and thereby improves the inter-sample comparability within the dataset. The fine fraction is tested with 10% HCl solution to determine if large amounts of carbonate minerals are present.

5.2. Geochemical Analysis of Till

The analytical routine is modified from Swedish Standard method SS 02 83 11 [31], adapted to suit large numbers of samples and to use modern equipment. Lab routines and practices follow ISO 17025 [32] Significant emphasis is placed on quality control, including the development of in-house standards, to ensure a level dataset, lab-blind analysis (sample randomization) to remove systematic bias, and field repeats to monitor analytical precision and sample handling [31].
Elements are analyzed from a 7 M HNO3 partial leach extraction. The analytical package returns 54 elements from this extraction [32]. Analyses were carried out at SGU’s Uppsala lab using a Perkin-Elmer Elan 5000 up to 2007, then a Perkin-Elmer Elan 9000 thereafter [33].
Data used in this study were collected in 2002, 2003, 2005, 2006, 2008, and 2017–2020. Prior to 2008, no analyses were conducted on Nb or Zr and these elements were excluded from analyses conducted in this study. The only geochemical analyses for REEs prior to 2008 were conducted on La and Y. As excluding analyses lacking the full suite of REEs would limit the usefulness of this study, the relationship between Y and La and the other REE was studied.
Examination of samples with full suites of analyzed REEs from post-2008 samples show Y concentrations display strong correlation with HREEs, and La with the other LREEs (Y/∑HREE r2 = 0.8877, La/∑LREE r2 = 0.913). The correlation of Y and HREEs as a proxy for REE mineralizations in Swedish bedrock has been observed previously [34] and thus Y and La should act as suitable proxies for the full suite of REEs to enable the use of pre-2008 till geochemical data. Elements that were included in this study are Al, Ba, Be, Bi, Ca, Cd, Co, Cr, Cu, Fe, K, La, Li, Mg, Mn, Mo, Na, Ni, Pb, Rb, Se, Sn, Sr, Th, Ti, Tl, U, V, W, Y, and Zn.

5.3. Data Treatment Methodology

5.3.1. Transformation of Sampling Coordinates

A key point of this study is to interpret the geochemistry of the till as a representation of the bedrock from which the material was sourced. As till has been transported by glaciers, a sense of where the bedrock from which the till sample originated must be obtained for the geochemical data analyzed here to have any application for spatial analysis.
Data were taken from the SGU’s regional till geochemistry database from samples collected both within and outside the boundaries of the REE line shown in the map in Figure 1. Ice-flow direction was estimated using linear landforms: drumlins and crag-and-tails, identified in LiDAR data, as well as measurements of glacial striations in bedrock, which gave an ice-flow direction of 160° (Figure 2). Linearity of ice flow was established using measurements of striations and linear landforms such as crag-and-tail and drumlin features.
For this study, which is located in the central region of Sweden, a linear transport distance of 16 km was inferred. This inference was based on the distance of the occurrence of multiple till samples in a roughly linear pattern that demonstrated positive reactions to carbonate tests performed with HCl on the <63 µm fraction. This roughly NE-striking linear grouping of till samples were collected approximately 16 km down-ice direction from the NE-striking large carbonate beds seen in the southwest near Nora, as shown in the bedrock map in Figure 1.
The original geographic coordinates of each sample site were subsequently transformed to move them 16 km up-ice to better reflect the origin of the material, and the points in Figure 1 represent an inferred source of the material comprising the till according to this approximation. This transformation is simply an approximation of the source of the material and uncertainty is expected due to potential variations in local ice-stream dynamics, and potential reworking of tills due to post-glacial processes.
In the southwest upper corner of the map, the all-element data were masked due to poor analytical results for Na. The southwestern border of the study area is dominated by a 1.8 Ga porphyritic, alkaline syenite granite with textural and geochemical similarities to rapakivi-type granites. Initial PCA and clustering results for the till samples from the region surrounding the REE line demonstrated that till samples associated with this granite overwhelmed geochemical anomalies for all principal component results due to anomalous REEs and other felsic-associated elements, and as such, till data with the rapakivi-like signature were removed to allow for identification of geochemical anomalies related to mineralization. As these granites are associated with the post-mineralization Sveconorwegian event [35], they post-date mineralization and removing these samples allows analysis of tills assumed to originate from bedrock formed during Svecokarelian in Bergslagen. These rapakivi-influenced samples, and their associated unique clustering results, provided a check for the transport distance inference discussed earlier. After the transport distance correction was applied to these samples, their potential origin was located over the extended rapakivi-like granites.

5.3.2. Exploratory Data Analysis

Exploratory analysis was conducted to determine the distribution of the data including skewness (Table 2). Routine criteria for determining which elements were to be included were used, i.e., where >70% of analyses for a given element should not return values below the detection limit, and data should follow a normal distribution [28]. For analyses under the detection limit that met the >70% criteria, one-half of the detection limit was substituted.
Simple histograms were created from non-transformed data to determine the population distribution of elemental abundances in the samples. As, Bi, Ca, Cr, Cu, Mo, Ni, Rb, Ti, Tl, and W showed higher standard deviations than the mean. Generally, single populations were observed; however, positive skew was observed for multiple elements due to the presence of outliers and extreme values.

5.4. Principal Component Analysis

Regional geochemical till sampling of large regions where mineral deposits are relatively small, such as along the REE line, often leads to the need to identify cryptic geochemical signals to locate potential anomalies for further exploration. Many of the economical deposits of iron and sulfides within the REE line occur along a short strike, often 100 s of meters, while the REE mineralizations occurring within these host mineralizations are only 10 s of meters in length and only up to around a meter wide, with haloes of diffuse REE mineralizations occurring throughout the deposits [19]. Single-element signals in the till are therefore unlikely to be useful for the exploration of REEs. Multi-element signatures are key, and the reduction in the dimensionality in the data is necessary.
PCA is used to reduce the dimensionality of variables, in this case, the elements within the till data, into a smaller group of uncorrelated data into their principal components. This allows for the interpretation of the data based on the extent of variance that is captured by each principal component, with the first principal component capturing the highest amount of variance, with subsequent PCs explaining smaller proportions of variance within the dataset. In this study, the optimum number of PCs needed to explain the variance in the dataset was based on the “Kaiser rule” [36] where all retained PCs had a value of >1.
Till geochemical data are of a compositional nature, meaning that data are a part of a constant sum, i.e., summing to 100%. In the case of much geochemical data, data are often expressed in weight percent, or parts per million. Correlations in the constant sum simplex space lead to spurious correlations due to the constraint of a constant sum due to the closed nature of the data. [37]. To remove the issue of the problem of closure within compositional data, log transformation of the data is the standard means to “open” the data system [38]. Various log transformation methods have been suggested, including additive log ratio (ALR), centered log-ratio (CLR) transformation [38], and isometric log-ratio (ILR) transformation [39]. The REE line till data were opened using isometric log-ratio transformation followed by PCA analysis of each transformed dataset using ioGAS software v8.2. ILR transformation of data requires an n − 1 function to reduce the dimensionality of the data by “sacrificing” one element in the dataset. While the use of immobile elements has been suggested [35], in this case, a residual was created by subtracting the sum of all data in ppm from the sum of the total (1,000,000) to prevent the loss of any single element during analysis. Biplots of the principal components were created with ioGAStm.
The issue of what constitutes a significant factor score within PCA is often debated [40]. Within the REE line till data, most factor scores were <0.5. Therefore, in this study, a factor score of <0.2 for a specific element is considered to be significant, while scores between 0.1 and 0.2 were considered to have moderate significance. The results of the factor loadings were plotted in ArcGIS pro and interpolated using simple IDW methodology, power 2, with a variable search radius based on a maximum of 12 points.

5.5. K-Means Clustering

K-means clustering is a useful method for an unsupervised optimization technique to group geochemical samples for which there are multiple possible solutions [41]. Clustering of samples does not inherently indicate that geochemical samples can be divided into groups of “most alike” samples based on their geochemistry [42], but instead minimizes in-group variation and maximizes variation between the groups by defining a centroid for each potential cluster and uses an iterative process to place each sample into the best-fitting cluster. In this study, K-means clustering was conducted on the ILR transformed data of both all-element and trace-element till data for the rapakivi-like granite-excluded dataset, with the iterative method [43] used for K-means clustering conducted in ioGAStm software v 8.2 with a maximum of 12 clusters using a random seed of 8,675,309 and 25,000 tries.

5.6. Alteration Index

Alteration within Bergslagen is regionally extensive and as noted previously, varies from K, Mg, Na, and Ca alteration [44]. To identify the extent of alteration as represented within the till, the Ishikawa alteration index (AI) was used using molar abundances of the oxides:
A I = 100 × K 2 O + M g O K 2 O + M g O + N a 2 O + C a O
The Ishikawa alteration index is a method to find the ratio changes in major rock-forming elements during alteration to sericite and chlorite originally developed for the identification of zones of alteration around VMS deposits [27]. Higher values for the AI indicate higher levels of alteration, with the highest values indicating a total replacement of feldspars by sericite or chlorite [45]. Values for the AI were interpolated along the surface of the REE line. For comparison, the alteration index of the lithological samples collected in situ was used for the comparison of the predictive surface expressed by the interpolation of values from the till samples [44]. While as noted above, partial leaching does not provide total concentrations of the elements in the sample, a test of the AI of the samples in eastern Bergslagen with both XRF and nitric acid data available showed a very strong positive correlation (r2 = 0.84) and the partial leaching data were considered acceptable for use in this study.

6. Results

6.1. All-Element Data

6.1.1. PCA Analysis of All-Element Data

PCA analysis of the masked till data from the REE line gave nine PCs that met the Kaiser criteria of eigenvalues >1. These nine PCs account for 73% of the variance of variance within the dataset (Table 3).
Factor loadings for the first 7 principal components are seen in Table 4 and PCs 1–4 are shown in biplots in Figure 4. The first PC for the all-element data, accounting for 19.5% of the total variance, is strongly controlled by the mafic and felsic components of the bedrock from which the till was sourced. Positive loadings on the first PC are associated with Zn-Pb-Al-Ni-Li, probably indicating a mafic control while negative loadings show an association of W-U-Ag-Rb-Se, which are most likely related to high levels of fractionation of more felsic magmas that formed the granites [46]. When interpolated, positive PC scores show an affinity for areas with more mafic rock, while negative scores are better associated with felsic rock. Positive scores on the second PC show a relationship between Co-Sn-Sr-Mn, which may point to a younger, more evolved granite.
Positive scores on PC2 show spatial proximity to the younger 1.85–1.75 Ga granites (Figure 5), which are more highly evolved and related to later Mo, Sn, and Bi mineralizations in the region [46]. A grouping of Na, Mg, and K on the negative scale of PC2 may point to alteration-related processes. The third PC shows a Y-Cr-P-Ba association with positive scores. Y and P are commonly mineralogically associated in the REE line, and Ba in particular is associated with mineralization near Bastnäs [19]. When the scores are interpolated, some evidence of a dispersal pattern following the ice direction is seen in map B in Figure 5.
PC4, which explains 9% of the variance, shows an association of Fe-K-P-Th-Y-W, with positive loadings and negative loadings showing a Cu-Mn-Cr-Rb association. Positive loadings on PC4 are potentially related to alteration and mineralization, while the negative loadings may represent less-altered rocks. When interpolated, a potential relationship between PC4 and the Norberg deposits is observed (Figure 4C), and the distribution of strongly positive scores is similar to that of the alteration map seen in Figure 11.

6.1.2. Clustering Results of All-Element Data

K-means clustering of the ILR transformed till data was conducted using both all-element and trace-element data in ioGAS with 25,000 random tries and a maximum of 12 clusters with a random seed of 8,675,309. The optimal number of clusters were selected using the elbow method, which identified nine optimal clusters for the all-element data, which are shown in Figure 6. Visual inspection of clusters in relation to the underlying lithologies and mineralizations was carried out. Cluster 6, shown in violet in the biplots, can be seen to be closely associated with negative loadings on PC1 and these clusters plot on the map in relation to felsic rock, while the more mafic-associated clusters show a relationship with the mafic rock. Several clusters are represented by the samples collected within the vicinity of the known REE mines, and no clear association between clusters and mineralization was seen in these data. Cluster 2 appears to show some relationship to carbonate occurrences in the southwestern part of the REE line from Nora through the Bastnäs mines, as shown in Figure 6.

6.2. Trace-Element Data

6.2.1. PCA Results Trace-Element Data

The results of the PCA for the trace-element data yielded nine PCs with eigenvalues that satisfied Kaiser’s rule accounting for a 79% of the variance (Table 5).
The first PC shows an association of Se-Ag-La-Y-Cd-V-Pb with positive factor scores and negative factor scores showed an association of W-U-Rb-Sn-Th-Zn (Table 6). The positive associations on PC1 are likely a mafic association while the negative scores are associated with felsic, possibly more fractionated granites. When interpolated over the REE line, the positive scores show good association with areas of higher occurrence of mafic bedrock. The second PC (Table 6) demonstrated a positive association of Be-Mo-Co-Sr-Rb-Ag, probably a more felsic association associated with the 1.85–1.75 Ga granite suite, while the Bi-Th-Cu-Ni association along the negative axis reflects a mafic association, as seen in PC2 of the all-element data. Similar associations are seen through PC4, and no spatial relationship to known REE deposits was seen. PC3 showed an association of W-Mo-Bi, which may be related to highly fractionated granites, or may indicate the sorting of heavy minerals.
PC5 showed a departure from the elemental associations of typical mafic and felsic seen in the lower PCs. Here, a Ba-W-Be-Y-Cu-Zn association is seen with positive loadings. The presence of elements such as Ba, W, and Be, which in lower PCs demonstrated an association with elements associated with the fractionation of felsic granites instead of showing a relationship to sulfide-associated elements and Y in PC5, indicates a possible connection with mineralization. Scheelite (CaWO4) and barite (BaSO4) are common accessory minerals in mineralization throughout the REE line and minerals containing Be such as gadolinite (Y2Fe2+Be2Si2O10) are major sinks for REEs in Bergslagen [19]. When the factor scores for PC5 for individual samples are interpolated, a spatial association is seen with known REE mineralizations and the general distribution of carbonate rock, which can host mineralizations.

6.2.2. Trace-Element Clustering

K-means clustering of the trace-element data produced eight optimal clusters within the REE line. The distributions of the clusters are color coded in the biplots of the PCA scores seen in Figure 7 and plotted with respect to their sampling location in Figure 8. It can be observed in Figure 7 that the optimal clusters are consistent with the principal component loadings for each factor.
The K-means cluster membership for each sample was then plotted onto a map of the REE line. Three clusters showed an apparent spatial relationship to known REE mineralizations, particular those near the Bastnäs area, clusters 3 and 6, as shown in Figure 9. Additionally, these clusters are tied to areas of high factor scores along the fifth PC. In Figure 9, it can be noted that clusters 3 and 6, and to some extent cluster 2, plot over strongly positive factor scores on the fifth principal component when interpolated over the study area.
Data for samples within these clusters were extracted and principal component analysis conducted on the samples within each cluster, as was conducted in Sadeghi et al. [12], to determine their elemental association to determine if the clusters showed associations of elements that were influenced by mineralization or if they exhibited false associations. The principal components and the factor loadings of the individual elements for the analysis of each cluster are shown in Figure 10.
Cluster 2 showed a weaker spatial relationship to the positive factor loadings on PC5. The factor scores of the samples from cluster 2 (Figure 11) showed a similar relationship on the first PC loadings to cluster 6 with antipathic relationships between Y and La and the other elements typical of highly fractionated felsic intrusives. Cluster 2 differed in showing significantly stronger positive loadings on the second PC of a Cu-Ni-Cr association, likely indicating a more mafic component controlling the cluster.
Cluster 3 showed associations of elements likely related to highly fractionated felsic rock, Sn, Rb, Th, Zn, and U with negative loadings along the first PC (Figure 11), while positive loadings on the first PC demonstrated an association that was likely more mafic in composition, though with some potential control by mineralization, with a Ni-Se-Cd-Pb-V-Bi-Y-La association. An antipathetic association with the more felsic components and the REEs is observed in this cluster. A weak association of Co-As-Cr-Cu-Tl on the second PC likely represents a less-mixed mafic source rock for the till.
Cluster 6 showed similar trends of elemental associations to cluster 3. There was an Y-La-Cd-Pb-Bi-Ag association with positive loadings on the first PC. Positive loadings on PC2 show a Tl-Rb-Sr-Li association (Figure 11). These elements behave in a geochemically similar way to K, and may represent strong K alteration of the rock, rather than a felsic association based on their lack of association with the more felsic components (U-Sn-Th-Mo-Be) on the negative axis of the first PC. An antipathetic relationship is observed with Y and La when compared to other elements associated with highly fractionated granites, namely U, Sn, W, and Th. This relationship indicates that Y and La are more strongly associated here with the sulfide-associated elements than fractionation occurring within granites and cluster 6 can be interpreted as potentially showing an association with Bastnäs-type mineralization.

6.3. Alteration Index

The alteration index for each sample was calculated from molar data using the Ishikawa alteration index (Equation (1)) [27]. The results were interpolated as a continuous surface over the REE line, and all known mineralizations in the SGU database, both Fe and sulfide, were included to study the distribution (Figure 12). The areas of highest alteration appear to correspond well with known deposits, which typically demonstrate high levels of alteration [15]. Some areas, such as the Bastnäs deposits, are not captured by the model, which is possibly due to the inherent error in attempting to transform sampling points up-ice due to localized ice-stream dynamics.

7. Discussion

7.1. All-Element Data

The application of multivariate analysis to till data within this study has demonstrated mafic and felsic elemental associations are the strongest control, accounting for between 20% and 30% of the total variance within the system for both all- and trace-element data. As bedrock of volcanic origin comprises the bulk of the bedrock in the study area, this is not surprising. For under-explored areas, where bedrock maps may be less detailed and till samples are available, it appears that these first PCs may provide useful information for determining the composition of unexposed bedrock. This mafic–felsic leading to the most variance within the system has been observed in prior studies [14]. Similar to these studies, it was only the higher PCs that made it possible to identify associations within the till that point to other more subtle geochemical differences between rocks of more similar compositions or with different ages, or potentially mineralization.
When attempting to identify at the regional level areas with geochemistry that point to being favorable to mineralization using PCA in the REE line, it was these subtle variations within the higher PCs that enabled the recognition of potential anomalies related to mineralization. The third PC of the all-element data demonstrates positive scores showing a geochemical trend, pointing to potential control by REE mineralization-favorable processes, while also demonstrating a spatial association with known REE deposits. Of particular note is the strong positive anomaly of PC3 scores to the southwest of the Bastnäs mines, which demonstrates a dispersal pattern in the direction of the ice movement, indicating a point source for the anomaly and subsequent distribution in the down-ice tills. This location is within a few kilometers of the Cu-Co-Bi-Au mineralization at Håkansboda, where preliminary investigations are looking into potential REE mineralizations in the area, and some anomalous REE values were observed in drill-core sampling [47].
PC4 of the all-element data shows that positive loadings are associated with Fe, K, Y, and P. When the factor scores of the samples are interpolated, it is observed that the high scores associated with these elements occur in relation to the iron mines of the Norberg area, where K alteration, as well as HREE anomalies, occur more frequently when compared to the Bastnäs localities, with an REE-bearing mineralogy that is often dominated by britholite group minerals, which can be P bearing [21] and more enriched in heavy REEs and Y. The apparent distribution of high factor scores on PC4 appears to show some relationship to the distribution of higher alteration indices in the till, which supports the use of the interpolated map as a predictive layer for the representation of a geochemical trap for mineralizing fluids at the regional scale.

7.2. Trace-Element Data

The results from the trace-element data showed relatively similar results to the all-element data, with the lower PCs demonstrating a mafic versus felsic bedrock signal in the lower principal components. The only PC meeting the Kaiser rule (eigenvalue >1) to demonstrate a possible spatial relationship to REE mineralization was PC5. Additionally, several clusters resulting from K-means clustering showed a strong spatial association with high factor scores on PC5. The results of the PC analyses of the data points from each cluster demonstrated relationships between elements that relate to REEs and sulfide mineralization. The most important observation of the PC data of the individual clusters is the decoupling of the relationship between REEs and other felsic elements, indicating high degrees of fractionation REE were associated with both the all-element and full trace-element dataset. This indicates a separate process may be responsible for the relationship between the REE and other elements in these samples. In particular, the antipathetic associations between fractionated felsic-associated elements such as U, Th, Mo, W, and Sn [45] and the REEs as represented by Y and La appear to be a key factor in identifying potential signatures of REE mineralization in the till data. Additionally, Ba and Be show a strong association in PC5 with Y and La, and focus should be given to these elements in studies of REEs using high-density sampling at the district to camp scale.
The positive association of Cu-Cr-Ni on PC2 in cluster 6 is associated with mafic rock and may point to a spatial relationship with the heat source that drove mineralization, which has been proposed to have been mafic intrusions [26], and identifying potential heat sources at a regional scale may help identify areas for more targeted exploration.

7.3. Generation of Predictive Layers for Prospectivity Mapping

From the till data, a total of four predictive layers were generated that may prove useful as part of a detailed mineral prospectivity model in the Bergslagen region: the rasters of PC3 and PC4 from the all-element data, PC5 from the trace-element data, and the interpolated results of the alteration index. When used in conjunction with other predictive layers for creating predictive maps of the REE line, a better understanding of regional-scale factors leading to the formation of Bastnäs-type REE deposits may be gained.
The results of the K-means clustering cannot themselves be used to generate evidentiary layers for mapping; the use of K-means for the validation of PC scores is here useful in the interpretation of the PC scores. While interpretations of bedrock can be extracted from both the clustering data and the PC data, in the case of the REE line, the area is thoroughly mapped; however, the use of lower PCs in similar future studies may be beneficial in a more greenfield setting to aid in generating a rough picture of unexposed bedrock.
The use of till geochemical data has produced a reasonable approximation of areas of high alteration where correlations can be seen between the areas of high alteration and a high density of known mineralizations. The generation of an alteration map using till data is a novel approach for exploration. With the limited data for bedrock samples and their degree of alteration in the REE line, using till data to create a regional alteration map for MPM may provide a valuable method for identifying under-exposed areas where strong alteration may have led to the formation of mineral deposits.
While these predictive layers have been generated as part of the compilation of data within the EIS project [6] for creating full prospectivity models, the data show a few preliminary areas of interest: to the northwest of Nora, the southwesternmost corner, and the area near Håkansboda named earlier. However, due to the uncertainty introduced by dispersal patterns induced by transport, and uncertainty caused by the coordinate transformations, care should be taken when interpreting the results. While it is difficult to rectify a regional-scale till sampling program with the identification of unknown deposits at the district and camp scale [11], the spatial association of the patterns seen in this study to known deposits indicates that the maps produced here can provide targeting for prospecting efforts to smaller areas and may highlight areas of geochemical interest when used with other evidentiary layers in MPM using the mineral systems approach such as other studies conducted with till geochemistry as part of the evidentiary dataset [48].

8. Conclusions

The small size and often limited exposure of REE mineralizations in Bergslagen makes prospecting for these deposits difficult. Therefore, the use of MPM using machine learning-assisted methods may be key for identifying potential new deposits. Using statistical and clustering methods, we identified geochemical factors associated with REEs and sulfide minerals demonstrating a spatial association with known REE deposits in the study, demonstrating that till data can be used in generating evidentiary maps for use in GIS-based prospectivity mapping. Additionally, the use of till to recognize the extent of regional-scale alteration may aid in the understanding of patterns of alteration in Bergslagen and further guide geological exploration when used in the EIS.

Author Contributions

Conceptualization: M.S. and P.C. Methodology: G.M.; Writing—Original Draft Preparation: P.C., Writing—Review and Editing: P.C. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by European Union’s Horizon Europe research and innovation program under grant Agreement No. 101057357-HORIZON-CL-2021-RESILIENCE-01.

Data Availability Statement

Publicly available datasets were analyzed in this study. This data can be found here: [Markgeokemi (sgu.se)] (Site available only in Swedish).

Acknowledgments

The authors thank the Department of Mineral Resources at the Geological Survey of Sweden for providing the datasets and background knowledge used in this study, which were obtained from the ongoing Regional Till Geochemical Mapping Program. We also thank the four anonymous reviewers, whose help greatly improved this paper.

Conflicts of Interest

The authors declare no conflict of interest.

References

  1. Grohol, M.; Veeh, C. Study on the Critical Raw Materials for the EU 2023—Final Report. Publications Office of the European Union. 2023. Available online: https://data.europa.eu/doi/10.2873/725585 (accessed on 22 February 2024).
  2. Taylor, S.R.; McLennan, S.M. The Continental Crust: Its Composition and Evolution; Blackwell: Oxford, UK, 1985. [Google Scholar] [CrossRef]
  3. Nordenskiöld, I. Der pegmatit von Ytterby. Bull. Geol. Inst. Univ. Upps. 1908, 9, 183–228. [Google Scholar]
  4. Hisinger, W.; Berzelius, J. Cerium, en ny Metall, Funnen i Bastnäs Tungsten, från Riddarhyttan i Westmanland; Henrik A. Nordström förlag: Stockholm, Sweden, 1804. [Google Scholar]
  5. Sahlström, F.; Jonsson, E.; Högdahl, K.; Troll, V.R.; Harris, C.; Jolis, E.M.; Weis, F. Interaction between high-temperature magmatic fluids and limestone explains ‘Bastnäs-type’ REE deposits in central Sweden. Sci. Rep. 2019, 1, 15203. [Google Scholar] [CrossRef] [PubMed]
  6. Yousefi, M.; Carranza, E.J.M.; Kreuzer, O.P.; Nykänen, V.; Hronsky, M.A.; Mihalasky, M.J. Data analysis methods for prospectivity modeling as applied to mineral exploration targeting: State-of-the-art and outlook. J. Geochem. Explor. 2021, 229, 106839. [Google Scholar] [CrossRef]
  7. McCuaig, T.C.; Beresford, S.; Hronsky, J. Translating the mineral systems approach into an effective exploration targeting system. Ore Geol. Rev. 2010, 3, 128–138. [Google Scholar] [CrossRef]
  8. Sadeghi, M.; Andersson, S.S.; Jonsson, E. A Preliminary Approach to Target Generation for Bastnäs-Type REE Mineralisation in Bergslagen, Sweden; Open File Research Report 29/2023; Geological survey of Finland (GTK): Espoo, Finland, 2023. [Google Scholar]
  9. Canil, D.; Lacourse, T. An estimate for the bulk composition of juvenile upper continental crust derived from glacial till in the North American Cordillera. Chem. Geol. 2011, 3, 229–239. [Google Scholar] [CrossRef]
  10. Roberts, S.; Palmer, M.R.; Waller, L. Sm-Nd and REE characteristics of tourmaline and scheelite from the Bjorkdal gold deposit, northern Sweden: Evidence of an intrusion-related gold deposit? Eco. Geo. 2006, 7, 1415–1425. [Google Scholar] [CrossRef]
  11. McClenaghan, B.M.; Paulen, R.C.; Smith, R.C.; Rice, J.M.; Plouffe, A.; McMartin, I.; Campbell, J.E.; Parsasadr, M.; Lethonen, M. Review of till geochemistry, indicator mineral, and boulder tracing methods for mineral exploration in glaciated terrain. Geochem. Explor. Environ. Anal. 2023, 23, geochem2023-013. [Google Scholar] [CrossRef]
  12. Normandeau, P.X.; McMartin, I.; Corriveau, L. Till Geochemistry as a Vector to Metasomatic Iron and Alkali-Calcic Systems and Associated Deposits in the Great Bear Magmatic Zone, Northwest Territories, Canada. Minerals 2024, 6, 547. [Google Scholar] [CrossRef]
  13. Lax, K.; Selinus, O. Geochemical mapping at the Geological Survey of Sweden. Geochem. Explor. Environ. Anal. 2005, 5, 337–346. [Google Scholar] [CrossRef]
  14. Sadeghi, M.; Casey, P.; Carranza, J.; Lynch, E. Principal Components Analysis and K-Means Clustering of Till Geochemical Data: Mapping and Targeting of Prospective Areas for Lithium Exploration in Västernorrland Region, Sweden. Ore Geol. Rev. 2024, 167, 106002. [Google Scholar] [CrossRef]
  15. Allen, R.L.; Lundstrom, I.; Ripa, M.; Christofferson, H. Facies analysis of a 1.9 Ga, continental margin, back-arc, felsic caldera province with diverse Zn-Pb-Ag-(Cu-Au) sulfide and Fe oxide deposits, Bergslagen region, Sweden. Econ. Geol. 1996, 6, 979–1008. [Google Scholar] [CrossRef]
  16. Stephens, M.B.; Weihed, J.B. (Eds.) . Sweden: Lithotectonic Framework, Tectonic Evolution and Mineral Resources; Geological Society of London: London, UK, 2020. [Google Scholar]
  17. Jonsson, E.; Högdahl, K. New evidence for the timing of formation of Bastnäs-type REE mineralisation in Bergslagen, Sweden. In Proceedings of the Mineral Deposit Research for a High-Tech World–Proceedings of the 12th Biennial SGA Meeting, Uppsala, Sweden, 12–15 August 2013; pp. 1724–1727. [Google Scholar]
  18. Geijer, P. The geological significance of the cerium mineral occurrences of the Bastnäs type in central Sweden. Arkiv Min. Geol. 1961, 3, 99–105. [Google Scholar]
  19. Holtstam, D.; Andersson, U.B. The REE minerals of the Bastnas-type deposits, south-central Sweden. Can. Min. 2007, 5, 1073–1114. [Google Scholar] [CrossRef]
  20. Holtstam, D.; Andersson, U.B.; Broman, C.; Mansfeld, J. Origin of REE mineralization in the Bastnäs-type Fe-REE-(Cu-Mo-Bi-Au) deposits, Bergslagen, Sweden. Min. Dep. 2014, 49, 933–966. [Google Scholar] [CrossRef]
  21. Holtstam, D.; Casey, P.; Bindi, L.; Förster, H.J.; Karlsson, A.; Appelt, O. Fluorbritholite-(Nd), Ca2Nd3(SiO4)3F, a new and key mineral for neodymium sequestration in REE skarns. Min. Mag. 2023, 5, 731–737. [Google Scholar] [CrossRef]
  22. Robertsson, A.-M.; Lundqvist, J.; Brunnberg, L. Dark clayey till in central and northern Sweden—Microfossil content and stratigraphical importance. GFF 2005, 127, 169–178. [Google Scholar] [CrossRef]
  23. Björck, J.; Andrén, T.; Wastegård, S.; Possnert, G.; Schoning, K. An event stratigraphy for the Last Glacial-Holocene transition in eastern middle Sweden: Results from investigations of varved clay and terrestrial sequences. Quat. Sci. Rev. 2002, 1, 489–501. [Google Scholar] [CrossRef]
  24. Andersson, M.; Ripa, M.; Smith, C. Geokemisk kartering i Västra Bergslagen; Rapport 2020:01; Sveriges Geologiska Undersökning: Uppsala, Sweden, 2020; p. 38. [Google Scholar]
  25. Carranza, E.J.M. Geochemical mineral exploration: Should we use enrichment factors or log-ratios? Nat. Resour. Res. 2017, 26, 411–428. [Google Scholar] [CrossRef]
  26. Sadeghi, M.; Ripa, M. Conceptual models for rare earth element exploration in Sweden. In Rare Earth Elements Distribution, Mineralisation and Exploration Potential in Sweden; Sadeghi, M., Ed.; Sveriges Geologiska Undersökning Rapporter Och Meddelanden; Sveriges Geologiska Undersökning: Uppsala, Sweden, 2019; Volume 146, pp. 135–147. [Google Scholar]
  27. Ishikawa, Y.; Sawaguchi, T.; Iwaya, S.; Horiuchi, M. Delineation of prospecting targets for Kuroko deposits based on modes of volcanism of underlying dacite and alteration haloes. Min. Geol. 1976, 26, 105–117. [Google Scholar]
  28. Franklin, J.M.; Lydon, J.W.; Sangster, D.F. Volcanic-associated massive sulfide deposits. Eco. Geol. 1981, 75th Anniversary Volume. 485–627. [Google Scholar]
  29. Stephens, M.B.; Ripa, M.; Lundström, I.; Persson, L.; Bergman, T.; Ahl, M.; Wahlgren, C.-H.; Persson, P.-O.; Wickström, L. Synthesis of the Bedrock Geology in the Bergslagen Region, Fennoscandian Shield, South-Central Sweden; Sveriges Geologiska Undersökning: Uppsala, Sweden, 2009; Ba 58; p. 258. [Google Scholar]
  30. MacLean, W.H. Rare earth element mobility at constant inter-REE ratios in the alteration zone at the Phelps Dodge massive sulphide deposit, Matagami, Quebec. Min. Dep. 1988, 23, 231–238. [Google Scholar] [CrossRef]
  31. SS 28311; Soil Analysis—Determination of Trace Elements in Soils—Extraction with Nitric Acid. Swedish Standard: Stockholm, Sweden, 2007.
  32. Miguel, A.L.R.; Moreiraa, R.P.L.; Oliveira, A.F.D. ISO/IEC 17025: History and introduction of concepts. Química Nova 2021, 6, 792–796. [Google Scholar] [CrossRef]
  33. Morris, G.A.; Ladenberger, A. Till Geochemistry by Nitric Acid (HNO3) partial leaching at the Geological Survey of Sweden. Geol. Surv. Swed. 2017, 14, 135–147. [Google Scholar]
  34. Sadeghi, M.; Arvanitidis, N.; Ladenberger, A. Geochemistry of Rare Earth Elements in Bedrock and Till, Applied in the Context of Mineral Potential in Sweden. Minerals 2020, 10, 365. [Google Scholar] [CrossRef]
  35. Reimann, C.; Filzmoser, P.; Garrett, R.; Dutter, R. Statistical Data Analysis Explained: Applied Environmental Statistics with R; John Wiley Sons: Chichester, UK, 2008; ISBN 9780470987605. [Google Scholar]
  36. Kaiser, H.F. On Cliff’s formula, the Kaiser-Guttman rule, and the number of factors. Percept. Mot. Ski. 1992, 74, 595–598. [Google Scholar] [CrossRef]
  37. Aitchison, J. The statistical analysis of geochemical compositions. J. Intl. Assoc. Math. Geol. 1984, 16, 531–564. [Google Scholar] [CrossRef]
  38. Aitchison, J. The statistical analysis of compositional data. In Monographs on Statistics and Applied Probability; Chapman Hall Ltd.: London, UK, 1986; p. 486. [Google Scholar]
  39. Egozcue, J.J.; Pawlowsky-Glahn, V.; Mateu-Figueras, G.; Barcelo-Vidal, C. Isometric logratio transformation for compositional data analysis. Math. Geol. 2003, 35, 279–300. [Google Scholar] [CrossRef]
  40. Peres-Neto, P.R.; Jackson, D.A.; Somers, K.M. Giving meaningful interpretation to ordination axes: Assessing loading significance in principal component analysis. Ecology 2003, 9, 2347–2363. [Google Scholar] [CrossRef]
  41. Madhulatha, T.S. An overview on clustering methods. IOSR J. Eng. 2012, 4, 719–725. [Google Scholar] [CrossRef]
  42. Parsa, M.; Sadeghi, M.; Grunsky, E. Innovative methods applied to processing and interpreting geochemical data. J. Geochem. Explor. 2022, 237, 106983. [Google Scholar] [CrossRef]
  43. MacQueen, J. Some methods for classification and analysis of multivariate observations. In Proceedings of the Fifth Berkeley Symposium on Mathematical Statistics and Probability, Berkeley, CA, USA, 21 June–18 July 1965; pp. 281–297. [Google Scholar]
  44. Sadeghi, M. Regional metamorphism versus hydrothermal alteration/ metasomatism related to sulphide mineralisation in the Bergslagen district, Sweden. Mineral deposit research for a high-tech world. In Proceedings of the 12th Biennial SGA Meeting, Uppsala, Sweden, 12–15 August 2013; Volume 1, pp. 434–436. [Google Scholar]
  45. Large, R.R.; Gemmell, J.B.; Paulick, H.; Huston, D.L. The alteration box plot: A simple approach to understanding the relationship between alteration mineralogy and lithogeochemistry associated with volcanic-hosted massive sulfide deposits. Eco. Geol. 2001, 96, 957–971. [Google Scholar] [CrossRef]
  46. Lynch, E.; Berggren, R.; Bergman, S. Geochemical and Geophysical Assessment of c. 1.8 Ga Granites Associated with W-F ± Mo Mineralisation, Western Bergslagen; SGU Rapport 2023:13; Sveriges Geologiska Undersökning: Uppsala, Sweden, 2013; 75p. [Google Scholar]
  47. Månbro, C. The Geology and Geochemistry of the Håkansboda Cu-Co Deposit, Bergslagen, Sweden. Master’s Thesis, Stockholms University, Stockholm, Sweden, 2021; p. 90. [Google Scholar]
  48. Torppa, J.; Middleton, M.; Hyvönen, E.; Lerssi, J.; Fraser, S. A novel spatial analysis approach for assessing regional-scale mineral prospectivity in northern Finland. Geol. Surv. Finl. 2015, 57, 87–120. [Google Scholar]
Minerals 14 00753 g001
Figure 1. Bedrock map of the REE line taken from the SGU 1:1,000,000 map from SGUs database. Coordinates are based on Swedish SWEREFF-99TM system. The red points represent an up-ice coordinate transformation of original sample locations to roughly 16 km NNW based on inferred transport distance. Location the REE line is shown in the inset map of Sweden in a red rectangle.
Figure 1. Bedrock map of the REE line taken from the SGU 1:1,000,000 map from SGUs database. Coordinates are based on Swedish SWEREFF-99TM system. The red points represent an up-ice coordinate transformation of original sample locations to roughly 16 km NNW based on inferred transport distance. Location the REE line is shown in the inset map of Sweden in a red rectangle.
Minerals 14 00753 g001
Minerals 14 00753 g002
Figure 2. The study area with soil depth shown in meters. The thickest sediments are typically associated with fluvial or lacustrine areas. Arrows demonstrate ice-flow direction as measured from orientation of striations in bedrock.
Figure 2. The study area with soil depth shown in meters. The thickest sediments are typically associated with fluvial or lacustrine areas. Arrows demonstrate ice-flow direction as measured from orientation of striations in bedrock.
Minerals 14 00753 g002
Minerals 14 00753 g003
Figure 3. Flow chart defining mineral systems from critical processes through mappable proxies.
Figure 3. Flow chart defining mineral systems from critical processes through mappable proxies.
Minerals 14 00753 g003
Minerals 14 00753 g004
Figure 4. (A–C) Biplots of PC scores for PCs 1 through 4 for the ILR transformed all-element data. Individual points represent individual till samples, and colors represent their K-means cluster membership.
Figure 4. (A–C) Biplots of PC scores for PCs 1 through 4 for the ILR transformed all-element data. Individual points represent individual till samples, and colors represent their K-means cluster membership.
Minerals 14 00753 g004
Minerals 14 00753 g005
Figure 5. Interpolated factor scores from the all-element data. (A). PC1 shows spatial correlation to mafic (high factor scores) and felsic (low factor scores) bedrock. (B). High scores along PC2 demonstrate possible correlations to the more evolved 1.85–1.75 Ga granites and pegmatites. (C). High factor scores show special affinity to the Norberg REE mineralizations. Black arrows indicate ice direction.
Figure 5. Interpolated factor scores from the all-element data. (A). PC1 shows spatial correlation to mafic (high factor scores) and felsic (low factor scores) bedrock. (B). High scores along PC2 demonstrate possible correlations to the more evolved 1.85–1.75 Ga granites and pegmatites. (C). High factor scores show special affinity to the Norberg REE mineralizations. Black arrows indicate ice direction.
Minerals 14 00753 g005
Minerals 14 00753 g006
Figure 6. Clusters membership of individual till samples for all-element data plotted over the bedrock map of the REE line. REE mineralizations are shown in red crosses.
Figure 6. Clusters membership of individual till samples for all-element data plotted over the bedrock map of the REE line. REE mineralizations are shown in red crosses.
Minerals 14 00753 g006
Minerals 14 00753 g007
Figure 7. Biplots of trace-element data after ILR transformation for PCs 1 through 5. (A). PC1–PC2. (B). PC1–PC3. (C). PC1–PC5.
Figure 7. Biplots of trace-element data after ILR transformation for PCs 1 through 5. (A). PC1–PC2. (B). PC1–PC3. (C). PC1–PC5.
Minerals 14 00753 g007
Minerals 14 00753 g008
Figure 8. (A–C) Interpolated results for principal component analysis of the trace-element data. (A). PC1 demonstrates a rough mafic (low scoring) and felsic (high scoring) divide between till samples. (B). Positive PC2 scores demonstrate association with 1.85–1.75 Ga granites and pegmatites. (C). PC5 shows correlation with known REE deposits. Red crosses are known REE mineralizations. Arrows indicate ice direction.
Figure 8. (A–C) Interpolated results for principal component analysis of the trace-element data. (A). PC1 demonstrates a rough mafic (low scoring) and felsic (high scoring) divide between till samples. (B). Positive PC2 scores demonstrate association with 1.85–1.75 Ga granites and pegmatites. (C). PC5 shows correlation with known REE deposits. Red crosses are known REE mineralizations. Arrows indicate ice direction.
Minerals 14 00753 g008
Minerals 14 00753 g009
Figure 9. Clustering results from the trace-element data plotted over the bedrock map of the REE line.
Figure 9. Clustering results from the trace-element data plotted over the bedrock map of the REE line.
Minerals 14 00753 g009
Minerals 14 00753 g010
Figure 10. Clusters 2, 3, and 6 shown overlain on the interpolated factor scores of PC5 of the trace-element data.
Figure 10. Clusters 2, 3, and 6 shown overlain on the interpolated factor scores of PC5 of the trace-element data.
Minerals 14 00753 g010
Minerals 14 00753 g011
Figure 11. Principal components of till samples from each of the three clusters highlighted in Figure 11 showing association with mineralization and positive loadings on the fifth principal component of the trace-element data. (A). Cluster 2, (B). Cluster 3. (C). Cluster 6.
Figure 11. Principal components of till samples from each of the three clusters highlighted in Figure 11 showing association with mineralization and positive loadings on the fifth principal component of the trace-element data. (A). Cluster 2, (B). Cluster 3. (C). Cluster 6.
Minerals 14 00753 g011
Minerals 14 00753 g012
Figure 12. Interpolated results of the alteration index of till samples within the REE line with known non-REE bearing mineralizations shown as red points (sulfide bearing) or green triangles (Fe-oxide).
Figure 12. Interpolated results of the alteration index of till samples within the REE line with known non-REE bearing mineralizations shown as red points (sulfide bearing) or green triangles (Fe-oxide).
Minerals 14 00753 g012
Table 1. Potentially mappable proxies that can be extracted from till geochemical data.
Table 1. Potentially mappable proxies that can be extracted from till geochemical data.
FactorTarget EvidenceMappable Geochemical ProxySources of Uncertainty
SourceUltramafic to mafic magmasPrincipal component factor coupled with mafic chemistryUnclear elemental associations due to mixing of signals in till
TrapReactive horizons or potential trap rocksPrincipal component or cluster associated with carbonate rocks, positive HCl test of tillWeak reaction to acid, Ca from other Ca-rich minerals
Geochemical signatures of mineralizationMultivariate geochemical anomalies in till samplesSignatures obscured by other anomalous metal sources such as metal-rich granites
ModificationAlteration (Mg-K-Ca-Na) of bedrockIshikawa alteration index; chlorite–carbonate–pyrite IndexAnalytical uncertainty, poor leaching of elements from certain key minerals
Alkaline-type granitesPrincipal component factor coupled with felsic chemistryMixed geochemical signals
Table 2. Table showing results of exploratory data analysis of till sampled (n = 768). All values given in ppm.
Table 2. Table showing results of exploratory data analysis of till sampled (n = 768). All values given in ppm.
ElementMinMaxMeanMedianSTDSkew
Al250938,5929772916048251.3
Ag0.010.20.10.0570.01.8
As0.281215.110.0
Ba4.01622621182.2
Be0.12.70.60.50.32.1
Bi0.1140.20.150.619
Ca35191,78518851650329723.1
Cd0.00.90.10.050.09.9
Co0.419.43.22.82.12.5
Cr1.03219.98.412.417.9
Cu0.11256.44.57.88.4
Fe121348,8289685861554621.7
K2510,1536414796115.8
La4.313119.516.811.24.0
Li0.955.16.45.24.82.8
Mg199.023,0272685.1228819873.4
Mn14.012201441161163.3
Mo0.1200.40.21.014.0
Na24.0311.075.57027.02.2
Ni0.0119.05.04.25.910.9
P25.02091.0394.33682112.1
Pb2.3130.58.97.48.17.5
Rb1.3174.28.56.28.69.0
Se0.11.20.30.250.11.9
Sn0.11.70.40.380.23.0
Sr2.363.98.07.24.04.3
Th2.912911.19.58.55.3
Ti1602249642.06092441.4
Tl0.02.70.10.060.112.9
U0.533.02.62.12.35.5
V4.251.815.4147.41.4
W0.07.20.30.210.314.7
Y3.062.012.0115.93.0
Zn2.51520.216.116.22.8
Table 3. Eigenvalues and explained variance of each principal component of the all-element dataset.
Table 3. Eigenvalues and explained variance of each principal component of the all-element dataset.
PCEigenvalueVarianceCumulative Variance (%)
PC16.619.519.5
PC23.911.430.9
PC33.39.740.5
PC43.19.149.6
PC52.16.255.8
PC61.85.261.0
PC71.54.565.5
PC81.33.969.5
PC91.23.472.9
Table 4. Factor loadings of a factor analysis with Varimax rotation or orthogonal rotation carried out with all-element data for better understanding and interpreting geological and mineralized significance of the factors showing element associations. Scores considered to be significant in this study are bolded.
Table 4. Factor loadings of a factor analysis with Varimax rotation or orthogonal rotation carried out with all-element data for better understanding and interpreting geological and mineralized significance of the factors showing element associations. Scores considered to be significant in this study are bolded.
ElementPC1PC2PC3PC4PC5PC6PC7
Al−0.23−0.090.17−0.120.210.090.28
Ag0.300.080.100.150.110.100.06
As−0.020.15−0.19−0.060.370.260.16
Ba−0.090.090.36−0.050.200.190.17
Be0.00−0.10−0.330.02−0.11−0.36−0.14
Bi0.100.14−0.080.180.18−0.24−0.10
Ca0.21−0.16−0.09−0.04−0.02−0.020.20
Cd−0.17−0.17−0.050.020.010.020.14
Co0.100.27−0.040.090.270.12−0.08
Cr−0.150.100.22−0.250.010.000.06
Cu−0.12−0.28−0.130.170.160.310.10
Fe0.120.180.090.34−0.01−0.220.20
K0.05−0.07−0.070.440.010.160.19
La0.00−0.02−0.04−0.240.29−0.41−0.09
Li0.25−0.120.12−0.100.130.16−0.22
Mg0.04−0.19−0.220.010.420.03−0.19
Mn0.000.19−0.05−0.300.03−0.030.16
Mo0.00−0.36−0.080.100.040.19−0.12
Na0.03−0.24−0.090.02−0.03−0.040.15
Ni0.26−0.130.24−0.04−0.080.02−0.10
P−0.15−0.210.200.25−0.06−0.200.08
Pb0.170.06−0.220.100.28−0.070.20
Rb−0.27−0.16−0.03−0.240.02−0.010.00
Se−0.240.17−0.030.06−0.13−0.120.21
Sn−0.150.270.080.000.23−0.08−0.05
Sr−0.020.200.040.16−0.050.020.28
Th0.06−0.300.210.010.06−0.250.30
Ti0.260.05−0.22−0.17−0.010.010.18
Tl0.17−0.150.27−0.140.140.00−0.09
U−0.26−0.04−0.140.150.22−0.22−0.07
V−0.130.15−0.30−0.11−0.250.26−0.03
W−0.270.040.000.22−0.070.15−0.28
Y−0.140.100.270.230.14−0.02−0.34
Zn0.300.130.040.05−0.160.03−0.10
Table 5. Eigenvalues and explained variance of each principal component of the trace-element dataset.
Table 5. Eigenvalues and explained variance of each principal component of the trace-element dataset.
PCEigenvalueVarianceCumulative Variance (%)
PC15.421.721.7
PC23.012.133.8
PC32.610.344.1
PC42.39.453.5
PC51.76.760.2
PC61.45.665.8
PC71.35.070.8
PC81.14.375.0
PC91.04.179.1
Table 6. Factor scores for trace-element data for PC 1–7. Elements with factor scores > 0.2 or < −0.2 are bolded.
Table 6. Factor scores for trace-element data for PC 1–7. Elements with factor scores > 0.2 or < −0.2 are bolded.
ElementPC1PC2PC3PC4PC5PC6PC7
Ag0.290.200.160.210.110.000.13
As−0.17−0.010.31−0.330.110.30−0.02
Ba−0.12−0.090.02−0.190.52−0.230.20
Be−0.080.320.33−0.190.27−0.03−0.06
Bi−0.02−0.34−0.260.07−0.430.05−0.08
Cd0.25−0.200.01−0.040.07−0.02−0.02
Co−0.180.220.07−0.23−0.200.42−0.07
Cr0.01−0.06−0.020.220.120.260.42
Cu−0.16−0.350.180.210.170.28−0.13
La0.270.140.130.34−0.060.090.10
Li−0.01−0.050.01−0.30−0.27−0.400.45
Mo−0.060.29−0.34−0.26−0.040.010.03
Ni0.02−0.390.14−0.080.060.260.32
Pb0.24−0.11−0.18−0.260.090.170.01
Rb−0.190.20−0.11−0.03−0.170.240.41
Se0.33−0.04−0.06−0.160.090.21−0.04
Sn−0.260.10−0.100.04−0.110.15−0.20
Sr−0.040.21−0.060.170.07−0.06−0.25
Th−0.30−0.200.010.18−0.02−0.040.06
Tl0.12−0.110.40−0.06−0.23−0.24−0.22
U−0.180.15−0.110.340.12−0.170.24
V0.24−0.16−0.25−0.210.11−0.07−0.07
W−0.18−0.06−0.40−0.030.270.03−0.15
Y0.260.08−0.240.150.190.13−0.07
Zn−0.33−0.20−0.020.010.15−0.19−0.10
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Casey, P.; Morris, G.; Sadeghi, M. Derivation of Predictive Layers Using Regional Till Geochemistry Data for Mineral Potential Mapping of the REE Line of Bergslagen, Central Sweden. Minerals 2024, 14, 753. https://doi.org/10.3390/min14080753

AMA Style

Casey P, Morris G, Sadeghi M. Derivation of Predictive Layers Using Regional Till Geochemistry Data for Mineral Potential Mapping of the REE Line of Bergslagen, Central Sweden. Minerals. 2024; 14(8):753. https://doi.org/10.3390/min14080753

Chicago/Turabian Style

Casey, Patrick, George Morris, and Martiya Sadeghi. 2024. "Derivation of Predictive Layers Using Regional Till Geochemistry Data for Mineral Potential Mapping of the REE Line of Bergslagen, Central Sweden" Minerals 14, no. 8: 753. https://doi.org/10.3390/min14080753

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop