By Ryan Kressall, M.Sc.

Multi-element geochemical datasets for mineral exploration projects are commonly large and can contain thousands of analyses. Traditional methods of data analysis often do not fully assess all the multi-element variables and interpretation is usually done for only the main pathfinder elements using tables, charts and bivariate plots. Principal Component Analysis (PCA) is a powerful statistical technique that allows for efficient analysis and interpretation of large multi-variate datasets. It can effectively identify subtle elemental trends significant to exploration project targeting that may be missed if only the primary and pathfinder elements are addressed.

No alt text provided for this image

Figure 1: Principal Component Analysis reduces the number of variables within a dataset. 

PCA is a multivariate statistical transformation that is used across many scientific disciplines to reduce the number of variables and to describe the variability within a dataset (Figure 1). PCA transformations involve several linear algebra calculations, which become increasingly complex with increase in the number of variables. In basic terms, PCA transforms data into vectors that identify directions of greatest variability within a multi-element dataset (Figure 2). The PCA algorithm searches for the vector of greatest variability in the dataset, which becomes the first Principal Component (PC1), and then the algorithm searches for the next direction of greatest variability that is orthogonal to PC1, which becomes PC2. The PCA continues this process to define PC3, PC4, etc. until PCn, with n representing the number of variables within the dataset. 

No alt text provided for this image

Figure 2: Principal Component Analysis transforms data into vectors that represent directions of greatest variability within a dataset.

Mercator Geological Services (Mercator) has developed an expertise and understanding of the algorithms and machine learning codes necessary to help exploration and mining companies get the most out of their multi-element geochemical datasets using PCA. We have effectively applied PCA to identify and accentuate trends that were not apparent with traditional assessment methods. There is no limit to the type of datasets that can be assessed and this method has been successfully applied to bedrock, soil and till datasets as well as diamond drilling and RC drilling assay results, allowing us to map new trends in mineralization, alteration, structure and lithology. The results can be plotted as 2D GIS maps or modeled in 3D with programs such as Leapfrog® Geo.

Brief case studies of two applications that we have performed for clients are presented below. The first example covers application of PCA to a soil survey grid and the second example covers application of PCA to the results of a small reverse circulation drilling program.

Example 1: Soil Survey Grid

A common problem in mineral exploration is the lack of bedrock exposure at surface, which often is addressed in exploration programs through the use of soil, till, lake sediment or biological material geochemical surveys. This traditionally may have limited the assessment to a single component such as gold assay results or could have included analytical results for certain recognized pathfinder elements. Although raw survey results may successfully identify broad areas of anomalism, they commonly do not identify specific vectors to guide further exploration. 

No alt text provided for this image

Figure 3: The first Principal Component (PC1) correlates with the bedrock geology by mapping out an argillite-bearing bedrock sequence along the hinge of a regional anticline and indicating a 150 meter displacement of the anticline by a northwest trending fault. 

PCA is applied in the case study example to the entire multi-element dataset and extracts more detailed information from the soil geochemistry results. The images below show the gridded PCA results for the case study survey. These clearly identify features such as the location of an anticlinal hinge zone that is recognized from other information sources to be associated with nearby argillite-rich sequences that host orogenic gold deposits (Figure 3). A PCA trend offset identifies an independently mapped cross fault that displaces the anticlinal trend of the unit on the east side of the soil survey grid, and therefore contributes to the structural interpretation of the target area. Additional derivatives (i.e., additional PCs) provide further definition of soil-based anomalism and allow the client to get the maximum benefit from the multi-element dataset. Benefits include differentiating between clusters of gold anomalism (Figure 4) and the identification of a unique glacial landforms (drumlin) geochemical signature that locally masks geochemical trends of greater consequence (Figure 5), and enhancement of structural and bedrock lithological interpretations for the grid area.

No alt text provided for this image

Figure 4: The second Principal Component (PC2) distinguishes anomalous gold values in the north from anomalous gold values in the south where there is a close association between arsenic and gold anomalism.

No alt text provided for this image

Figure 5: The third Principal Component (PC3) correlates with the occurrence of glacial drumlins.

Example 2: Reverse Circulation Drilling Program

One of the most common PCA applications that we see involves the study of multi-element drill hole geochemical data. In this application of PCA, downhole multi-element geochemistry results can be used in a number of ways, including core logging verification, geological modelling, alteration mapping, interpolation of targets and geochemical fingerprinting.

In the example below the results of the PCA were used to guide the development of a 3D geological model in Leapfrog®. Our client was exploring a historic gold district that was previously mined for quartz vein hosted gold, but current exploration is focused on assessing potential for low-grade disseminated gold mineralization of orogenic association. In this instance, PC1 differentiates between bedrock intervals dominated by either argillite or greywacke and PC2 maps hydrothermal alteration in the host rocks, which is known to be associated with the targeted style of broadly dispersed low-grade gold mineralization. 3D modelling of the PCA results in Leapfrog® spatially distinguishes two different types of gold mineralization on the property based on the PCA defined alteration signature (Figure 6). One type of gold mineralization occurs in the area of the old mine workings where gold occurs within quartz veins that are accompanied by localized hydrothermal alteration with associated gold disseminations limited to the immediate host rocks. The second occurs primarily east of the old mine workings, where bedrock argillite and greywacke sequences are pervasively altered and associated with continuously distributed, very low-level gold (20 ppb to 100 ppb range). This important association is mapped by PC2, which defines an attractive target zone for future exploration to the east of past mining and most exploration (Figure 6).

No alt text provided for this image

Figure 6: Longitudinal section along anticlinal trend showing historical mine workings and recent RC drill holes. PCA has defined a coincident hydrothermal alteration and very low-grade gold target area (blue) associated with altered greywacke and argillite to the east. It is distinct from the quartz vein hosted gold mineralization in the historical mine workings area where low-grade disseminated gold associated with hydrothermal alteration is confined to the immediate contact areas of quartz veins.

Conclusions

Mercator’s expertise in PCA and the general application of this method has allowed us to unlock information within client datasets that is normally overlooked. It can provide a major advantage in guiding mineral exploration projects through geochemically fingerprinting important alteration and mineralization signatures that may not be recognized using other methods of assessment. Geochemical datasets amenable to PCA study include newly acquired assay lab results and/or portable-XRF results as well as data compiled from historical reporting. Common data sources include diamond or RC drilling programs, outcrop or trenching/mine workings sampling programs, and geochemical surveys of soil, till, lake bottom or biological materials. Using PCA in conjunction with surface mapping, drilling and 3D modeling can result in identification of important geochemical features that can guide exploration targeting. PCA is a cost effective analytical tool that has potential to greatly increase exploration value derived from commonly collected multi-element datasets. 

No alt text provided for this image

Ryan Kressall, M.Sc., is the Senior Geochemist/Geologist at Mercator and holds a B.Sc. (Hons.) and M.Sc. in Geology from the University of Manitoba. His experience includes application of geochemistry and statistical methods to mineral exploration, analytical instrumentation, and project management. He has also worked on a variety of commodities across Canada including gold, rare metals, rare earths, and diamonds. Joining Mercator in 2018 as our Senior Geochemist, Ryan is responsible for data analysis and database development to support prospectivity and targeting strategies. He is particularly focused on developing unique approaches to geochemical and mineralogical fingerprinting of specific deposit types