Service: Geology mapping that honors subsurface and surficial datasets

Geology mapping service post image

EXPLORERS facing imminent field reconnaissance work need a firm grasp of the area's geology, a good first-pass map for uncharted localities, and a means to help crystallize and focus their work plans. Auto-mapping can supply all of these, because modern survey data are embodied in the resulting maps. A case study in Western Australia's western Capricorn Orogen region shows how.

Flat, remote terrain that is devoid of geological outcrop can thwart even the most driven and best-funded exploration geologists.

In such frontier terrain, maps of any type can be a scarce commodity or older than some of the exploration geologists using them.

Maps tend to be prepared by clear-headed and knowledgable professionals. But if a map was last substantially updated in the 1960s or 70s, or largely charts the whereabouts of transported and in situ cover, it has probably had most of its usefulness wrung from it already by previous explorers.

Interpreted geology maps that draw on geophysical and geochemical data are becoming increasingly prominent within natural-resource exploration circles.

For instance:

  • Geological survey staff have been using geophysical data in their mapping projects in Australia. [1]
  • Remote predictive mapping of Canada's far north regions used an iterative revision process that tapped into geophysical and geochemical survey data to help update geological maps. [2]
  • NASA and the US Geological Survey have given automated geological mapping their blessing: they use automated orbital mapping to geologically map the surface of Mars and other planetary bodies based on data acquired in orbit. This is done largely because a mission's remote-sensed data can amass faster than human mappers can handle without the aid of such automated processing. [3] [4]

Reasons to use auto-mapping

Interpreted geology auto-mapping is a way of visually grouping and summarizing the variation in features that show up throughout explorers' tenements.

It does the same overall job of manually created interpreted geology maps — it simultaneously 'lumps' and 'splits' the area of interest according to important geological themes and trends.

Auto-mapping may not reproduce the map a human would come up with manually, but because auto-maps have strengths of their own, they are no less valid than a conventional map.

In fact, some of the other demonstrable benefits of using a geological map that incorporates computer-assisted interpretations of remotely-sensed and airborne data include:

  • new recognition of lithological or structural features that previously went unnoticed, or that were mapped wrongly or too coarsely,
  • an up-to-date best estimate of the geology that a given piece of ground represents,
  • rapid mapping of large swaths of land is feasible,
  • advance notice of the geological structures (folds, faults and contacts) and relationships likely to be encountered in a given area,
  • early assessment of which locations appear more geologically complex, and therefore seem to warrant special attention during ground-truthing fieldwork,
  • lithological predictions and inferences that honor subsurface and surficial data signatures in a given area, which may open up the area's standing geological model to a major re-think, and
  • creation of a robust, first-pass map that readily accommodates iterative revisions and upgrades as new ground-based and air-based data emerge, and as expert knowledge develops about local geological and ore-genesis models and is used to refine maps. This map-evolution process can be continued until the desired mapping confidence levels are reached.

This type of mapping can be completed and later refined by human experts in less time than it would take to build a map manually from scratch. [3]

Arguably, the most important benefit that comes with the use of data-augmented interpreted geology maps is the opportunity to make meaningful exploration-related extrapolations and decisions.

Geophysical data are powerful in exploring beneath cover, particularly when there is a close correspondence between the data and the region's ore-forming processes.

However, simply looking for exaggerated anomalies in geophysical and geochemical survey data without considering any subtleties present in the data will probably lead to wasted exploration effort and expense. [5] [6]

Instead, a more useful approach would involve, first up, ascertaining the physical, chemical, structural and physiographic signature of the host rock that contains known ore zones or mineralization shows. Understanding the geological milieu is paramount. [7]

The next task at hand would then be the extrapolation — looking at the available data and at the interpreted geology map for any signs that the host-rock signature is showing up under similar circumstances elsewhere in the tenements.

This data-augmented approach is now increasingly accessible to resource exploration companies because of improvements and greater efficiencies in auto-mapping.

Auto-mapping improvements

The list of automated mapping methods is something of an acronym soup that includes self-organizing maps (SOMs), support vector machine (SVM) routines, maximum likelihood estimation (MLE), artificial neural networks (ANN), and principal component analysis (PCA). [2] [8] [9] [10] [11] [12]

Quite often currently, a drawback in using some of these data analysis and partitioning methods is that resulting interpreted geology auto-maps often look speckled, due to miniscule domains of one interpreted lithology unit appearing dispersed throughout another interpreted unit. The avalanche of detail can be overwhelming, preventing an easy and intuitive understanding of the basic relationships among the interpreted units.

Another common hindrance is the need for a preliminary 'training' step, which establishes computer-assisted decision-making rules, before automated mapping algorithms can be launched. This kind of training often uses existing geology maps for double-checking and calibration purposes. But in frontier areas, reliable maps simply may not be available at the desired map-scale.

The procedure developed by Fathom Geophysics avoids creating distracting speckles and does not require any preliminary training step. The procedure produces a coherent map that honors the structural and other relationships seen in the input data and in the supporting data used in subsequent steps.

This can be demonstrated with a case study.

Western Australian case study

Fathom Geophysics recently assisted Eastern Goldfields Exploration with interpreted geology auto-mapping of survey data at the Cobra Project in Western Australia's central-west region. [13]

Geology mapping service figure 1FIGURE 1: Location map in Western Australia, with the Cobra Project case study area in the state's central west region defined by a red boundary.

Figures 1 to 5 show the location of the project, and the magnetic, radiometric and topographic data for the Cobra Syncline area.

Geology mapping service figure 2FIGURE 2: Reduced-to-the-pole (RTP) magnetic data for the Cobra Project case study area.

Geology mapping service figure 3FIGURE 3: Structural network (black lines) extracted from the Cobra Project case study area's RTP magnetic data using Fathom Geophysics' computer-vision routines. Structures shown are those with wavelengths of between about 25 meters and about 100 meters. The structural network is in vectorized form, superimposed on the area's RTP magnetic data.

Geology mapping service figure 4FIGURE 4: The same structural network (black lines) that appeared in the previous image, this time superimposed on the area's total-count radiometric data (potassium, thorium, uranium).

Geology mapping service figure 5FIGURE 5: The same structural network (black lines) that appeared in the previous images, this time superimposed on the area's Shuttle Radar Topography Mission (SRTM) topographic data.

The interpreted geology auto-map for the Cobra area encapsulates the following data processing results:

  • Total structure detection (see our May-June 2010 newsletter write-up), and
  • Analysis of structures detected, to define structure sets that are either parallel or perperpendicular to the predominant fabric of a geological belt (also described in the May-June 2010 newsletter write-up)

We carried out interpreted geology auto-mapping at the Cobra Project as follows:

  • Selection of the data that best reflects the geology across the entire area of interest, to be used as intial input for auto-mapping. We selected the area's magnetic data, having also tested the area's radiometric data, because the latter correlated poorly with geology in sub-areas possessing subdued topography.
  • Segmentation of the input data (see our March 2010 newsletter write-up). Figures 6 and 7 show the results of this step.
  • Grouping of segments via a K-means clustering technique (see the March-April 2012 newsletter news story on the topic).
  • Classification of segments according to inherent attributes observable as distinct signatures in the input data. We used segments' magnetic, potassium, thorium and topography characteristics to select which class a given segment belongs to. Figures 8 and 9 show the results of this classification step.
  • Inclusion of expert knowledge, to augment the metadata for segment classes with further details about their nature, such as any distinct lithological names. We assigned rock-unit names based on mapping information supplied by an expert familiar with the Cobra area's local geology.

Geology mapping service figure 6FIGURE 6: Results of automated interpretation of features (black lines) in the Cobra Project case study area using Fathom Geophysics' computer-vision routines, which are described in further detail in this article's text. The features' boundaries are superimposed on the area's RTP magnetic data.

Geology mapping service figure 7FIGURE 7: The same automated interpretation results (black outlines) that appeared in the previous image, this time superimposed on the area's total-count radiometric data (potassium, thorium, uranium).

Geology mapping service figure 8FIGURE 8: Results of grouping and classification (various colors) of features contained in the automated interpretation (black outlines). Each segment class was assigned a specific lithological name via consultation with a knowledgable expert familiar with the area's local geology. Colors were attributed as follows: yellow = Pooranoo Metamorphics pelitic schist, dark green = Pooranoo Metamorphics dolerite, light green = Pooranoo Metamorphics amphibolite, light spotted purple = high thorium Narimbunna Dolerite, light purple = highly magnetic Narimbunna Dolerite, dark purple = undifferentiated Narimbunna Dolerite, pink = Moorarie Supersuite granite, burnt orange = Leake Spring Metamorphics amphibolite, light gray = Edmund Group siltstone, dark gray = Edmund Group low-potassium sandstone, blue = Edmund Group dolostone.

Geology mapping service figure 9FIGURE 9: The area's structural network (black lines) superimposed on the lithologically grouped automatic interpretation results (colored areas).

In Figures 8-11 and Figures 14-16, segment classes were geologically attributed and colored as follows:

  • yellow: Pooranoo Metamorphics pelitic schist,
  • dark green: Pooranoo Metamorphics dolerite,
  • light green: Pooranoo Metamorphics amphibolite,
  • light spotted purple: high thorium Narimbunna Dolerite,
  • light purple: highly magnetic Narimbunna Dolerite,
  • dark purple: undifferentiated Narimbunna Dolerite,
  • pink: Moorarie Supersuite granite,
  • burnt orange: Leake Spring Metamorphics amphibolite,
  • light gray: Edmund Group siltstone,
  • dark gray: Edmund Group low-potassium sandstone, and
  • blue: Edmund Group dolostone.

The resulting interpreted geology auto-maps can be useful standing alone or in 'fused images', such as those appearing in Figure 10. Image fusion is a way to simultaneously view complementary information, [2] [4] such as rock-unit classes and magnetic amplitude.

Geology mapping service figure 10FIGURE 10: The area's structural network (black lines) superimposed on the lithologically grouped automatic interpretation results (colored areas), which have in turn been superimposed on a grayscale version of the area's RTP magnetic data. To view a larger version, click on either the image or this text link.

Auto-map acid test: Comparison with offical geology map

To help show how useful a well-constructed interpreted geology auto-map is, let's see how the Cobra area auto-map looks against an existing, high-quality map that has been prepared independently (Figure 11).

One such existing interpreted geology map of the Cobra area was published by the Geological Survey of Western Australia (GSWA) and appears in Johnson and others (2012). [14]

Geology mapping service figure 11aFIGURE 11a: The area's lithologically grouped automatic interpretation results (colored areas), superimposed on a grayscale version of the area's official geology map. To view a larger version, click on either the image or this text link.

Geology mapping service figure 11bFIGURE 11b: The area's official geology map (by the Geological Survey of Western Australia; in Johnson and others, 2012). To view a larger version, click on either the image or this text link.

Figure 12 illustrates the extent to which Fathom Geophysics' structural analysis of geological belt-parallel and belt-crossing structures successfully tie in with the structures and contacts officially mapped in the Cobra area.

Geology mapping service figure 12aFIGURE 12a: The area's belt-parallel structural network (magenta lines) superimposed the area's official geology map (by the Geological Survey of Western Australia; in Johnson and others, 2012). To view a larger version, click on either the image or this text link.

Geology mapping service figure 12bFIGURE 12b: The same image that appears in the previous figure, but this time with the area's belt-crossing structural network. To view a larger version, click on either the image or this text link.

Bear in mind, though, that our interpreted geology auto-map and the existing GSWA interpreted geology map aren't strict equivalents of each other. Because magnetic data were used in the auto-map's construction, the geological relationships seen within it contain a significant component of the area's subsurface configuration. In contrast, the GSWA map seems to closely conform to the area's physiography (this can be seen in the relatively close match in Figure 13).

Geology mapping service figure 13aFIGURE 13a: The area's official geology map (by the Geological Survey of Western Australia; in Johnson and others, 2012) superimposed on a grayscale version of the area's topography data. To view a larger version, click on either the image or this text link.

Geology mapping service figure 13bFIGURE 13b: Grayscale topography alone. To view a larger version, click on either the image or this text link.

Even so, it's still instructive to compare the two maps, because if both are considered to be accurate representations, then the auto-map should provide clues about what the geology is doing at depths greater than the surface and near-surface.

For instance, by comparing the auto-map with the GSWA map (Figure 11), we can see:

  • The GSWA map shows that the Cobra Synclinorium in the area's southeast contains a large central ellipse composed of Narimbunna dolerite (the green-colored unit 'Pnr-od') surrounded by a quite thin occurrence of the Edmund Group's upper formations (the orange-colored, vertically striped unit 'PMEP4-xs-k'), which include siltstone, mudstone and dolostone materials. The auto-map suggests that at that location, the subsurface is dominated more by the sedimentary rock material (mapped as gray), with only a relatively slim central sliver of the dolerite being evident (purple).
  • In the Cobra area's northwestern half, in places where the topography is rather subdued, the auto-map reveals that geological relationships are more complicated than what is indicated by the undifferentiated triangular sliver of Pooranoo Metamorphics material in the GSWA map (the salmon pink unit 'PPO-md').

Figures 14, 15 and 16 are 'fused' images and are examples of how an auto-map could be combined with other GIS map layers an exploration company may have at hand, such as known major faults running through the area. Depending on which combination of layers is chosen, the resulting overall image may prove more valuable than the simple sum of the information contained in the individual map layers.

Geology mapping service figure 14FIGURE 14: The area's major known faults as officially mapped by the Geological Survey of Western Australia (in Johnson and others, 2012), alongside the area's belt-parallel structural network (dark gray lines), all superimposed on the lithologically grouped automatic interpretation results (colored areas), which have been 'draped' over a grayscale version of the area's RTP magnetic data. To view a larger version, click on either the image or this text link.

Geology mapping service figure 15FIGURE 15: The area's major known faults as officially mapped by the Geological Survey of Western Australia (in Johnson and others, 2012), alongside the area's belt-crossing structural network (dark gray lines), all superimposed on the lithologically grouped automatic interpretation results (colored areas), which have been 'draped' over a grayscale version of the area's RTP magnetic data. To view a larger version, click on either the image or this text link.

Geology mapping service figure 16FIGURE 16: The area's major known faults as officially mapped by the Geological Survey of Western Australia (in Johnson and others, 2012), alongside the area's belt-parallel and belt-crossing structural network (dark gray lines), all superimposed on the lithologically grouped automatic interpretation results (colored areas), which have been 'draped' over a grayscale version of the area's RTP magnetic data. To view a larger version, click on either the image or this text link.

Invaluable exercise

We think the promise of interpreted geology auto-maps has been shown above in the Cobra case study.

But don't just take our word for it. Here is what Geological Survey of Canada researchers Mark Pilkington, Pierre Keating and Mike Thomas said in an exhaustive 2008 collaborative write-up on data-augmented geological mapping:

"Delineating the extent of units with similar magnetic properties follows the geological mapping approach of dividing the surface into rocks having similar properties. Whether or not the magnetization changes exactly mimic lithological ones, mapping such contrasts provides invaluable information on structural regimes, and deformation styles and trends." [15]

And:

"Geological mapping from magnetic and gravity data is essentially an exercise in pattern recognition. ... Different 'units' can be distinguished based on the strike of anomaly trends by grouping together regions having similar strike (assuming a constant direction is apparent). Similarly, regions can be defined on the basis of similar texture or pattern of anomalies. The resulting areas do not have a geological interpretation; they are merely areas defined by similarities in amplitude, trends, and patterns of anomalies. Broad geological inferences can, however, be made."

When it comes to automated methods of producing geological maps, the overall sense is that the minerals exploration industry is on the cusp of 'the future is now'.

Where once we may have asked ourselves Are auto-maps good enough to be of any real use in my exploration program?, now the question is more along the lines of What do I want my auto-map to highlight during this stage of exploration?

One clear sign that auto-mapping has come of age is that researchers now are beginning to attack the next tier of auto-map research topics, such as map generalization, which is the automated production of broad-scale, regional maps based on a series of more detailed maps. [16]

— Acknowledgement: Fathom Geophysics thanks Richard Cooke, Eastern Goldfields Exploration Pty Ltd, for permission to show and discuss data processing results for Cobra Project tenements.

References and notes

[1] M. Morse (2010) "Potential field methods prove effective for continental margin studies: Another option for offshore geology mapping", AusGeo News, 98, 4 pages. In this publication, Morse said: "The potential field methods described above [use of upward continuation results and the analytic signal] are just two of a number of methods used to interpret potential field geophysical data. They have proven useful in Geoscience Australia's frontier [offshore] basin studies and show how geological information can be extracted from geophysical datasets..."

[2] E.M. Schetselaar, J.R. Harris, T. Lynds and E.A. de Kemp (2007) "Remote predictive mapping (RPM): A strategy for geological mapping of Canada's north", Geoscience Canada, 34, 3/4, 93-111.

[3] NASA Applied Information Systems Research, http://aisrp.nasa.gov/projects/7bd92ada.html. The page states: "Planetary exploration has entered a new era in which our data-gathering capability has outpaced our capacity for timely analysis."

[4] R.O. Kuzmin, R. Greeley, R. Landheim, N.A. Cabrol and J.D. Farmer (2000) "Geologic map of the MTM-15182 and MTM-15187 Quadrangles, Gusev Crater-Ma'adim Vallis Region, Mars", US Geological Survey, prepared for the NASA Planetary Geology Program.

[5] P.G. Harman (2004) "Geophysical signatures of orebodies under cover", In: J. Muhling, R. Goldfarb, N. Vielreicher, F. Bierlein, E. Stumpfl, D.I. Groves and S. Kenworthy (Ed.) "SEG 2004: Predictive Mineral Discovery Under Cover", Extended Abstracts, University of Western Australia Centre for Global Metallogeny Publication No. 33, 85-89.

[6] A.M. Silva, A.C.B. Pires, A. McCafferty, R.A.V. De Moraes and H. Xia (2003) "Application of airborne geophysical data to mineral exploration in the uneven exposed terrains of the Rio Das Velhas Greenstone Belt", Revista Brasileira de Geociencias, 33, 17-28.

[7] See disscussion and citations in: F. Robert, R. Brommecker, B.T. Bourne, P.J. Dobak, C.J. McEwan, R.R. Rowe and X. Zhou (2007) "Models and exploration methods for major gold deposit types", In: B. Milkereit (ed.) Proceedings of Exploration 07: Fifth Decennial International Conference on Mineral Exploration, 691-711. In this paper, the authors said: "... geology should remain an important underpinning of future gold exploration programs. ... Another element will be the wise application of proven and emerging detection techniques, in close integration with geology."

[8] J.H. Hodgkinson, S.J. Fraser and P. Donchak (2012) "Using self-organising maps to derive lithological boundaries from geophysically-derived data in the Mt. Isa region, Queensland", ASEG Extended Abstracts, 1, 1-4.

[9] L. Yu, A. Porwal, E.-J. Holden and M.C. Dentith (August 2012) "Towards automatic lithological classification from remote sensing data using support vector machines", Computers and Geosciences, 45, 229-239.

[10] S. Grebby, J. Naden, D. Cunningham and K. Tansey (2011) "Integrating airborne multispectral imagery and airborne LiDAR data for enhanced lithological mapping in vegetated terrain", Remote Sensing of the Environment, 115, 214-226.

[11] S. Grebby, D. Cunningham, J. Naden and K. Tansey (2010) "Lithological mapping of the Troodos ophiolite, Cyprus, using airborne LiDAR topographic data", Remote Sensing of the Environment, 114, 4, 713-724.

[12] T. Villmann, E. Merenyi and B. Hammer (2003) "Neural maps in remote sensing image analysis", Neural Networks, 16, 3-4, 389-403.

[13] For background information on the Cobra area and the surrounding region, see, for example: (a) A.M. Thorne, S.P. Johnson, H.N. Cutten and O. Blay (2012) "Structural development and mineralization of the western Edmund and Collier Basins", Geological Survey of Western Australia 2012 extended abstracts, Record 2012/2, 20-23. (b) S. Sheppard, S.P. Johnson, M.T.D. Wingate, C.L. Kirkland and F. Pirajno (2010) "Explanatory notes for the Gascoyne Province", Geological Survey of Western Australia, 336 pages. (c) D. McB. Martin, K.N. Sircombe, A.M. Thorne, P.A. Cawood and A.A. Nemchin (2008) "Provenance history of the Bangemall Supergroup and implications for the Mesoproterozoic paleogeography of the West Australia Craton", Precambrian Research, 166, 93-110. (d) D. McB. Martin, S. Sheppard, A.M. Thorne, T.R. Farrell and P.B. Groenewald (2006) "Proterozoic geology of the western Capricorn Orogen: A field guide", Geological Survey of Western Australia, Record 2006/18, 43 pages.

[14] S.P. Johnson, A.M. Thorne, H.N. Cutten and O. Blay (compilers) (2012) "Geological interpretation of the western Capricorn Orogen", In: S.P. Johnson, A.M. Thorne and I.M. Tyler (eds.) "Capricorn Orogen seismic and magnetotelluric workshop 2011 extended abstracts", Geological Survey of Western Australia, Record 2011/25.

[15] M. Pilkington, P.B. Keating and M.D. Thomas (2008) "Geophysics", Chapter 3 Of: J.R. Harris (ed.) "Remote Predictive Mapping: An Aid for Northern Mapping", Geological Survey of Canada Open File 5643, 29-52.

[16] See, for example: A. Smirnoff, G. Huot-Vezina, S.J. Paradis and R. Boivin (March 2012) "Generalizing geological maps with the GeoScaler software: The case study approach", Computers and Geosciences, 40, 66-86.