Prospectivity of Western Australian iron ore from geophysical data using a reject option classifier

ArticleinOre Geology Reviews 71:761-776 · March 2015with 358 Reads
Cite this publication

Do you want to read the rest of this article?

Request full-text
Request Full-text Paper PDF
  • Article
    Big Data techniques have the potential to be paradigm-changing for applied geoscience if they are used widely. A significant number of such techniques, under the umbrella of Earth informatics, involve Machine Learning applied to high dimensional data to create new forms of value. This contribution presents two case studies of successful Earth informatics computation and the communication of the value of results, which provide insight into the uptake of ‘Big Data’ in geosciences. Machine Learning techniques split naturally into either supervised or unsupervised approaches. Supervised algorithms, such as Random ForestsTM (RF), support vector machines or neural networks, share the concept of training a classifier using an initial (training) dataset. They are generally applied to predictive tasks, such as our first case study, predicting lithology from remote sensing and airborne geophysical data. Unsupervised algorithms, such as Self-Organising Maps (SOM), allow patterns inherent in the data to emerge without the use of a training dataset. They are generally applied to tasks which seek to explore patterns in data, such as our second case study, which identifies new potentially prospective river catchments. We find that calculating and presenting explicitly the newly extracted value, of the result obtained through computation, is an essential component of the post-compute evaluation. As strong advocates for the use of a range of Big Data techniques in applied geosciences, we conclude that the benefits to be gained from the way that we ‘compute’ can be lost if we do not also take considerable care with the ways that we ‘communicate’.
  • Thesis
    Full-text available
    Exploration for mineral deposits is more challenging today as the target depth for new discoveries is increasing. Deep exploration often requires drilling through several hundreds metres of barren cover, which is expensive and time consuming. As deeper holes are drilled it is imperative that each hole delivers as much information as possible-preferably in real-time-to allow changes to be made such as geo-steering and altering or updating of the drill plan whilst the drill rig is still on site. Typically, most drilling campaigns are conducted to obtain rock samples for geochemical analysis to measure the metal content. Thus, various technologies that enable real-time elemental analysis are being developed; for example: DET CRC's Lab-at-Rig R or Boart Longyear's TruScan R. Geophysical wireline logging could assist the real-time decision making, but is not widely used in base metals and gold exploration despite many obvious advantages to those familiar with such data. One issue is the cost of standby for a dedicated and skilled operator waiting for the drilling to finish; another is that the hole can collapse before logging is performed. Thus, DET CRC has developed driller-deployable geophysical borehole tools that provide petrophysical data as part of the drilling process (e.g., AutoSonde TM and AutoShuttle). Both of these new measuring and sensing technologies offer a wealth of real-time data that will require timely analysis and interpretation to complement an effective exploration strategy. This study demonstrates several benefits of joint analysis of interdisciplinary data from three different ore deposits and offers methods and workflows that enable real-time data interpretation. The core objective was to identify the necessary data combinations to classify the rock mass in terms of lithological, mineralogical and textural features. It is not the single link between one measurement and a specific rock property that is the focus, but the importance of combining the 'right' data for useful classification. What is 'useful' and necessary depends on the purpose or reason of the classification; for example, lithological distinction, ore-waste delineation or to highlight textural differences or rock quality properties. In addition, an emphasis is placed upon the use of petrophysical data as it offers high resolution real-time data but is less understood amongst geoscientists. Examples are given where petrophysical downhole logs can substitute for elemental assay data, or where petrophysical measurements are indicative of rock mass properties that are not detectable with elemental or mineralogical analysis. Furthermore, this thesis shows that statistical analysis prior to classifying a data set can: firstly, give valuable insight into the interrelation of the data variables; secondly, provide a means for reducing dimensions of multivariate data sets; and thirdly, guide the optimal choice of variables for rock mass classification. This study also highlights the importance of necessary pre-processing steps for multivariate, cross-disciplinary data analysis and evaluates different clustering and prediction methods. The results indicate that a combination of two to three petrophysical downhole logs are sufficient for a range of possible classification purposes in various geological environments. For example, the natural gamma, resistivity and density logs sufficiently distinguish major lithology and identify the units of economic value at a sulphide deposit, whilst a combination that includes p-wave velocity data can separate quartz-rich, coarse grained from mica-rich, fine grained barren strata. A second example from an iron ore deposit demonstrates the limitations of clustering algorithms when a classification according to strict, industry standard cutoff grades is desired and how this can be overcome by applying adaptive learning algorithms for variable prediction. This second example also indicates that real-time iron ore classification is possible with data extracted from Lithodensity logging, which enables cost effective exploration and resource evaluation campaigns through near real-time interpretation and decision making. The final case study highlights the possibility of predicting and interpolating rock quality properties from petrophysical downhole measurements on data from a nickel-copper deposit. The results and findings of this research should encourage the wider application of geophysical downhole methods in future exploration campaigns and trigger a demand-pull from the industry for real-time measurement technologies and complementing data analytical workflows. These technologies and data analysis procedures are essential to ensure cost and time effective exploration for mineral deposits under deep cover in the future.
  • Article
    The circulation and accumulation of harmful elements in blast furnace were investigated. The results show that the maximum concentration of K and Na in the belly is about 50 times of that in the charge. The zinc increased by 80 times and the Pb is insignificant. The impact of the circulation and accumulation of K, Na, Pb and Zn on the fuel consumption was quantitatively analyzed using Rist diagram. The results show that the gas utilization decreased and the coke rate increased linearly with the increase of load and accumulation times of each harmful element. Given the accumulation times of each element, the loading of Na, K, Zn and Pb into BF leads to the increase of coke rate by 13.99 kg/tHM, 6.25 kg/tHM, 3.63 kg/tHM and 0.02 kg/tHM, respectively. The increased coke rate and the decreased gas utilization under various loads and accumulation times of each element was estimated.
  • Article
    Full-text available
    Machine learning algorithms (MLAs) are a powerful group of data-driven inference tools that offer an automated means of recognizing patterns in high-dimensional data. Hence, there is much scope for the application of MLAs to the rapidly increasing volumes of remotely sensed geophysical data for geological mapping problems. We carry out a rigorous comparison of five MLAs: Naive Bayes, k-Nearest Neighbors, Random Forests, Support Vector Machines, and Artificial Neural Networks, in the context of a supervised lithology classification task using widely available and spatially constrained remotely sensed geophysical data. We make a further comparison of MLAs based on their sensitivity to variations in the degree of spatial clustering of training data, and their response to the inclusion of explicit spatial information (spatial coordinates). Our work identifies Random Forests as a good first choice algorithm for the supervised classification of lithology using remotely sensed geophysical data. Random Forests is straightforward to train, computationally efficient, highly stable with respect to variations in classification model parameter values, and as accurate as, or substantially more accurate than the other MLAs trialed. The results of our study indicate that as training data becomes increasingly dispersed across the region under investigation, MLA predictive accuracy improves dramatically. The use of explicit spatial information generates accurate lithology predictions but should be used in conjunction with geophysical data in order to generate geologically plausible predictions. MLAs, such as Random Forests, are valuable tools for generating reliable first-pass predictions for practical geological mapping applications that combine widely available geophysical data.
  • Article
    Wisely used, mineral resources create wealth, employment, a vital social and natural environment, and peace. If the reverse of these conditions occurs only too often, illustrating the so-called “resource curse”, this should be attributed to the true perpetrators, namely irresponsible, weak or selfish leaders. This book, however, does not intend to provide rules for good governance. I wrote it as a broad overview on geoscientific aspects of mineral deposits, including their origin, geological characteristics, the principles of the search for ores and minerals, and the investigation of newly found deposits. In addition, practical and environmental aspects are adressed that arise during the life cycle of a mine and after its closure. I am convinced that in our time, economic geology cannot be tought, studied, or practiced without an understanding of environmental issues. The scientific core of the book is the attempt to present the extraordinary genetic variability of mineral deposits in the frame of fundamental geological process systems. The comprehensive approach of this book – covering materials from metal ores to minerals and hydrocarbons – is both an advantage and a loss. The second concerns the sacrifice of much detail, but I promote the first for its benefit of a panoramic view over the whole field of economic geology. Being aware that the specialist level of subjects presented in this book fills whole libraries, I do hope that even experienced practitioners, academic teachers and advanced students of particular subjects will find the synopsis useful. Over more than 50 years, several editions of this book were published in German language. Since the first edition (Wilhelm & Walther E. Petrascheck 1950), the book was intended to provide a concise introduction to the geological setting of mineral deposits, including its application to exploration and mining. The target audience has changed, however. Originally, it was written for students of mining engineering. Today, it is mainly directed to aspiring and practicing geologists. Each of the seven chapters of the book was developed with my own students as a university course, and should be useful to fellow academic teachers. After initially working in industry I never lost contact with the applications of economic geology which is my motive for the constant interweaving of practical aspects in the text and for dedicating one of the chapters to the practice of economic geology. For professional reference purposes, practitioners in geology and mining should appreciate this melange of science and application. Frequent explanations and references to environmental and health aspects of extraction and processing of ores and minerals should assist users involved in environmental work. To those with no background in geology, I recommend they acquire an introductory geoscience text for looking up terms that are employed but not explained in the book. Compared with the last German-language edition (Pohl 2005), this book has been rewritten for an international public. Although it retains a moderate European penchant by referring to examples from this region, important deposits worldwide are preferentially used to explain genetic types and practical aspects. I trust that this will be useful to both scholars and practitioners, wherever they work. Generally, it was my ambition to represent the state of the art in economic geology, by referring to and citing recent publications as well as earlier fundamental concepts. This should assist and motivate students to pursue topics to greater depth. Many people have supported me in my life-long pursuit of theory and practice of economic geology, and helped with this book, especially by donating photographs. I cannot name them all but in captions, donors are acknowledged. Here, just let me say thank you.
  • Article
    Geographic information systems (GISs) are defined as software systems. In this article, the relationships between GIS and other activities having to do with geographic information are reviewed. The use of GIS in social and behavioral sciences is discussed as an increasingly essential component of the research infrastructure and as a tool for acquiring and communicating geographic knowledge. Examples are used to discuss the importance of GIS across the social and behavioral sciences. Sources of data are reviewed, and GISs are discussed from the perspectives of client–server architectures, the Internet, Internet-based services, data archives, and digital libraries. GIS use is intimately related to the role of space in scientific explanation. The article ends with a discussion on the future of GIS.
  • Chapter
    What are ore deposits?Mining in the stress field between society and environmentThe mineral resources conundrum
  • Article
    The Hamersley iron ore province of Western Australia contains world-class high-purity hematite orebodies hosted within Lower Proterozoic banded iron formations at Mount Tom Price, Mount Whaleback, and Paraburdoo-Channar. New evidence indicates that the orebodies are structurally controlled along old normal fault systems that formed during a period of major uplift and extension in Proterozoic times. Hematite ores are always hosted by the Brockman Iron Formation, and ore formation resulted from a multistage, sequential removal of gangue minerals from the host, giving rise to residual concentration of iron. The first, hypogene, stage of ore formation removed silica only, leaving a thinned residue enriched in iron oxides, carbonates, magnesium silicates, and apatite, with no apparent change in the oxidation state of the iron minerals. In this stage of alteration, warm, highly saline bicarbonate-saturated fluids from the underlying carbonate-shale Wittenoom Formation leaked upward along fault zones into the lower part of the Brockman Iron Formation. During the second, deep meteoric, stage of ore formation a magnetite-siderite assemblage oxidized to hematite-ankerite, characteristic microplaty hematite developed, and magnetite converted to martite. The fluid responsible was moderately warm, of low salinity, and oxidized, and it most likely derived from the surface. A second stage of gangue removal followed this oxidation stage and stripped all carbonate from both magnetite and hematite zones, leaving highly porous and permeable iron ore bands with a high apatite content interbedded with magnesium-rich shale bands. The final, purely supergene, stage of upgrading is indistinguishable from modern weathering but penetrated deep below the present surface. Magnesium silicates were converted to a kaolinitic residue, greatly thinning the shale bands, apatite was destroyed, and both calcium and phosphorus were leached from the ore. The final product is a highly porous hematite ore of characteristic microplaty texture interbedded with kaolinitic shale containing a significant amount of aluminum and titanium, which retain their relative proportions throughout the upgrading process.
  • Article
    Iron-rich sediments have been deposited intermittently throughout earth history, but virtually all significant deposits of the cherty layered rock referred to as iron formation are of Precambrian age. Present evidence suggests three periods of peak deposition rate. This review deals only with those of the cherty oxide facies. Two distinct subtypes are recognized: well laminated cherty banded iron formation (BIF in this paper) and granule iron formation. Both are primary chemical precipitates, modified subsequently. The depositional basins of BIF showed unusual tectonic and environmental stability coupled with an absence of clastic contamination. The paleoenvironment of granule iron formation was, by contrast, a high-energy one in which the deposited material was subjected to para-depositional reworking. -from Author Precambrian chemical precipitates cherty oxide facies
  • Article
    A lithotectonic subdivision of the Yilgarn block is presented. The more ancient Western gneiss terrain is contrasted with three younger granite-greenstone provinces.J.A.H.
  • Chapter
    Principal component analysis has often been dealt with in textbooks as a special case of factor analysis, and this tendency has been continued by many computer packages which treat PCA as one option in a program for factor analysis—see Appendix A2. This view is misguided since PCA and factor analysis, as usually defined, are really quite distinct techniques. The confusion may have arisen, in part, because of Hotelling’s (1933) original paper, in which principal components were introduced in the context of providing a small number of ‘more fundamental’ variables which determine the values of the p original variables. This is very much in the spirit of the factor model introduced in Section 7.1, although Girschick (1936) indicates that there were soon criticisms of Hotelling’s method of PCs, as being inappropriate for factor analysis. Further confusion results from the fact that practitioners of ‘factor analysis’ do not always have the same definition of the technique (see Jackson, 1981). The definition adopted in this chapter is, however, fairly standard.
  • Chapter
    In the history of research of the learning problem one can extract four periods that can be characterized by four bright events: (i) Constructing the first learning machines, (ii) constructing the fundamentals of the theory, (iii) constructing neural networks, (iv) constructing the alternatives to neural networks.
  • Article
    The sedimentary iron-formations of Precambrian age in the Lake Superior region can be divided on the basis of the dominant original iron mineral into four principal facies: sulfide, carbonate, oxide, and silicate. As chemical sediments, these rocks reflect certain aspects of the chemistry of the depositional environments. The major control, at least for the sulfide, carbonate, and oxide types, probably was the oxidation potential. The evidence indicates that deposition took place in restricted basins, which were separated from the open sea by thresholds that inhibited free circulation and permitted development of abnormalities in oxidation potential and water composition. The sporadic distribution of metamorphism and of later oxidation permits description of the primary facies on the basis of unoxidized, essentially unmetamorphosed material. The sulfide facies is represented by black slates in which pyrite may make up as much as 40 percent of the rock. The free-carbon content of these rocks typically ranges from 5 to 15 percent, indicating that ultra-stagnant conditions prevailed during deposition. Locally, the pyritic rocks contain layers of iron-rich carbonate. The carbonate facies consists, in its purer form, of interbedded iron-rich carbonate and chert. It is a product of an environment in which oxygen concentration was sufficiently high to destroy most of the organic material but not high enough to permit formation of ferric compounds. The oxide facies is found as two principal types, one characterized by magnetite and the other by hematite. Both minerals appear to be of primary origin. The magnetite-banded rock is one of the dominant lithologies in the region; it consists typically of magnetite interlayered with chert, carbonate, or iron silicate, or combinations of the three. Its mineralogy and association suggest origin under weakly oxidizing to moderately reducing conditions, but the mode of precipitation of magnetite is not clearly understood. The hematite-banded rocks consist of finely crystalline hematite interlayered with chert or jasper. Oolitic structure is common. This facies doubtless accumulated in a strongly oxidizing, probably near-shore, environment similar to that in which younger hema-titic ironstones such as the Clinton oolite were deposited. The silicate facies contains one or more of the hydrous ferrous silicates (greenalite, minnesotaite, stilpnomelane, chlorite) as a major constituent. Granule structure, similar to that of glauconite, is typical of some varieties; others are nongranular and finely laminated. The most common association of the silicate rocks is with either carbonate- or magnetite-bearing rocks, which suggests that the optimum conditions for deposition ranged from slightly oxidizing to slightly reducing. The relationship between the iron-rich rocks and volcanism, stressed by many authors, is considered by the writer to be structural, not chemical: in the Lake Superior region both iron-deposition and volcanism are believed to be related to geosynclinal development during Huronian time. In Michigan, the lower Huronian rocks are iron-poor quartzite and dolomite-typical "stable-shelf" deposits; much of the upper Huronian consists of iron-poor graywacke and slate with associated volcanic rocks -a typical "geosynclinal" assemblage. Thus the iron-rich beds of the middle Huronian and lower part of the upper Huronian were deposited during a transitional stage in structural history. The major environmental requirement for deposition of iron-formation is the closed or restricted basin; this requirement coincides in time with what would be a normal stage in evolution of the geosyncline: namely, structural development of offshore buckles or swells that subsequently develop into island arcs characterized by volcanism.
  • Article
    The Hamersley iron ore province of Western Australia contains world-class high-purity hematite orebodies hosted within Lower Proterozoic banded iron formations at Mount Tom Price, Mount Whaleback, and Paraburdoo-Channar. New evidence indicates that the orebodies are structurally controlled along old fault systems that formed during a period a major uplift and extension in Proterozoic times. Hematite ores are always hosted by the Brockman Iron Formation, and ore formation resulted from a multistage, sequential removal of gangue minerals from the host, giving rise to residual concentration of iron. The first. hypogene, stage of ore formation removed silica only; leaving a thinned residue enriched in iron oxides, carbonates, magnesium silicates, and apatite. with no apparent change in the oxidation state of the iron minerals, ill this stage of alteration, warm, highly saline bicarbonate-saturated fluids from the underlying carbonate-shale Wittenoom Formation leaked upward along fault zones into the lower;er part of the Brockman Iron Formation. During thr second, deep meteoric, stage of ore formation a magnetite-siderite assemblage oxidized to hematite-ankerite. characteristic microplaty hematite developed. and magnetite converted to marmite. The fluid responsible was moderately warm, of low salinity; and oxidized, and it most likely derived from the surface. A second stage of gangue removal followed this oxidation stage and stripped all carbonate from both magnetite and hematite zones, leaving highly porous anti permeable iron ore bands with a high apatite content interbedded with magnesium-rich shale bands. The final, purely supergene, stage of upgrading is indistinguishable from modern weathering but penetrated deep below the present surface. Magnesium silicates were converted to a kaolinitic residue, greatly thinning the shale bands, apatite was destroyed, and both calcium and phosphorus were leached from the ore. The final product is a highly; porous hematite ore of characteristic microplaty texture interbedded with kaolinitic shale containing a significant amount of aluminum and titanium, which retain their relative proportions throughout the upgrading process.
  • Article
    Full-text available
    Previous prospectivity modelling for epithermal Au–Ag deposits in the Deseado Massif, southern Argentina, provided regional-scale prospectivity maps that were of limited help in guiding exploration activities within districts or smaller areas, because of their low level of detail. Because several districts in the Deseado Massif still need to be explored, prospectivity maps produced with higher detail would be more helpful for exploration in this region.
  • Article
    We present a mineral systems approach to predictive mapping of orogenic gold prospectivity in the Giyani greenstone belt (GGB) by using layers of spatial evidence representing district-scale processes that are critical to orogenic gold mineralization, namely (a) source of metals/fluids, (b) active pathways, (c) drivers of fluid flow and (d) metal deposition. To demonstrate that the quality of a predictive map of mineral prospectivity is a function of the quality of the maps used as sources of spatial evidence, we created two sets of prospectivity maps – one using an old lithologic map and another using an updated lithological map as two separate sources of spatial evidence for source of metals/fluids, drivers of fluid flow and metal deposition. We also demonstrate the importance of using spatially-coherent (or geologically-consistent) deposit occurrences in data-driven predictive mapping of mineral prospectivity. The best predictive orogenic gold prospectivity map obtained in this study is the one that made use of spatial evidence from the updated lithological map and spatially-coherent orogenic gold occurrences. This map predicts 20% of the GGB to be prospective for orogenic gold, with 89% goodness-of-fit between spatially-coherent inactive orogenic gold mines and individual layers of spatial evidence and 89% prediction-rate against spatially-coherent orogenic gold prospects. In comparison, the predictive gold prospectivity map obtained by using spatial evidence from the old lithological map and all gold occurrences has 80% goodness-of-fit but only 63% prediction-rate. These results mean that the prospectivity map based on spatially-coherent gold occurrences and spatial evidence from the updated lithological map predicts exploration targets better (i.e., 28% smaller prospective areas with 9% stronger fit to training gold mines and 26% higher prediction-rate with respect to validation gold prospects) than the prospectivity map based on all known gold occurrences and spatial evidence from the old lithological map.
  • Article
    E xisting geological models of the structure and evolution of the Yilgarn Craton are poorly constrained with extensive regolith (weathering and surficial cover) a significant factor in limiting understanding. Regional aeromagnetic data acquired with a ground clearance of 60 metres or greater are little effected by this regolith, and their interpretation provides a continuous model of the distribution of the Archaean rocks. Five distinct regionally distributed geophysical map units are defined (undivided gneiss-migmatite-granite, banded gneiss, granite plutons, sinuous gneiss, and greenstone) which in various proportions comprise eight large domains that may equate with geological provinces. The domain boundaries coincide with abrupt changes in magnetisation and or structural trends. Boundaries of geologically-based models commonly occur in areas where there is no geophysical evidence for changes in crustal structure or composition. Interlinking shears in the Norseman-Wiluna Belt define a craton scale shear zone with less deformed crust to the east and west.
  • Article
    Bayesian weight-of-evidence and logistic regression models are implemented in a GIS environment for regional-scale prospectivity modeling of greenstone belts in the Yilgarn Craton, Western Australia, for magmatic nickel sulfide deposits. The input variables for the models consisted of derivative GIS layers that were used as proxies for mappable exploration criteria for magmatic nickel sulfide deposits in the Yilgarn. About 70% of the 165 known deposits of the craton were used to train the models; the remaining 30% was used to validate the models and, therefore, had to be treated as if they had not been discovered. The weights-ofevidence and logistic regression models, respectively, classify 71.4% and 81.6% validation deposits in prospective zones that occupy about 9% of the total area occupied by the greenstone belts in the craton. The superior performance of the logistic regression model is attributed to its capability to accommodate conditional dependencies amongst the input predictor maps, and provide less biased estimates of prospectivity.
  • Conference Paper
    Full-text available
    Australia produces over 90% of the world's precious opal from highly weathered Cretaceous sedimentary rocks within the Great Artesian Basin. Since opal was first discovered around 1870 until the present day, opal mining has been carried out by private operators working a claim no larger than 50 × 50 m, usually in the direct vicinity of areas that have yielded precious opal in the past. Currently there is no formal exploration model for opal and its formation in the geological environment is poorly understood. Here we make the first systematic attempt to formulate a predictive model for opal exploration using a powerful data mining approach, which considers almost the entire Great Artesian Basin as a potential reservoir for precious opal. Our methodology uses all known locations where opal has been mined to date. Its formation and preservation in weathered Cretaceous host rocks is evaluated by a joint analysis of large digital data sets that include topography, regional geology, regolith and soil type, radiometric data and depositional environments through time. By combining these data sets as layers enabling spatio-temporal data mining using the GPlates PaleoGIS software, we produce the first opal prospectivity map for the Great Artesian Basin. Our approach reduces the entire area of the Great Artesian Basin to a mere 6% that is deemed to be prospective for opal exploration. It successfully identifies two known major opal fields (Mintabie and Lambina) that were not included as part of the classification dataset owing to lack of documentation regarding opal mine locations, and it significantly expands the prospective areas around known opal fields particularly in the vicinity of Coober Pedy in South Australia and in the northern and southern sectors of the Eromanga Basin in Queensland. The combined characteristics of these areas also provide a basis for future work aimed at improving our understanding of opal formation
  • Article
    Full-text available
    The Hellyer–Mt Charter region of western Tasmania includes three known and economically significant volcanic-hosted massive sulfide (VHMS) deposits. Thick vegetation and poor outcrop present a considerable challenge to ongoing detailed geological field mapping in this area. Numerous geophysical and soil geochemical datasets covering the Hellyer–Mt Charter region have been collected in recent years. These data provide a rich source of geological information that can assist in defining the spatial distribution of lithologies. The integration and analysis of many layers of data in order to derive meaningful geological interpretations is a non-trivial task; however, machine learning algorithms such as Random Forests and Self-Organising Maps offer geologists methods for indentifying patterns in high-dimensional (many layered) data. In this study, we validate an interpreted geological map of the Hellyer–Mt Charter region by employing Random Forests™ to classify geophysical and geochemical data into 21 discrete lithological units. Our comparison of Random Forests supervised classification predictions to the interpreted geological map highlights the efficacy of this algorithm to map complex geological terranes. Furthermore, Random Forests identifies new geological details regarding the spatial distributions of key lithologies within the economically important Que-Hellyer Volcanics (QHV). We then infer distinct but spatially contiguous sub-classes within footwall and hangingwall, basalts and andesites of the QHV using Self-Organising Maps, an unsupervised clustering algorithm. Insight into compositional variability within volcanic units is gained by visualising the spatial distributions of sub-classes and associated statistical distributions of key geochemical data. Compositional differences in volcanic units are interpreted to reflect contrasting primary composition and VHMS alteration styles. We conclude that combining supervised and unsupervised machine-learning algorithms provides a widely applicable, robust means, of analysing complex and disparate data for machine-assisted geological mapping in challenging terranes.
  • Article
    Full-text available
    Results from a multidisciplinary geoscience program since 1994 are summarized for the North Pilbara terrain of the Pilbara Craton. Major findings include the recognition of three separate terranes with unique stratigraphy, geochronological, and structural histories; the ca. 3.72 to 2.85 Ga East Pilbara granite-greenstone terrane, the ca. 3.27 to 2.92 Ga West Pilbara granite-greenstone terrane, and the ≤3.29 Ga Kuranna terrane in the southeast. These are separated by two late, dominantly elastic sedimentary basins deposited within tectonically active zones; the ca. 3.01 to 2.93 Ga Mallina basin in the west and the undated Mosquito Creek basin in the east. The oldest supracrustal rocks are the ca. 3.51 to 3.50 Ga Coonterunah and ca. 3.49 to 3.31 Ga Warrawoona Groups in the East Pilbara granite-greenstone terrane, deposited on fragments of older sialic crust to 3.72 Ga. The Warrawoona Group is subdivided into three main (ultra)mafic-felsic volcanic cycles including from base to top, the Talga (3.49-3.46 Ga), Salgasli (3.46-3.43 Ga), and newly defined Kelly (3.43-3.31 Ga) Sub-groups. These dominantly basaltic rocks include chert beds containing Earth's oldest stromatolites and are interbedded with significant felsic volcanics erupted intermittently from 3.49 to 3.43 Ga during emplacement of sheeted sodic granitoid sills. Estimates of autochthonous stratigraphic thickness range from 9 to 18 km. Deformation involved extensional growth faulting, local folding, and tilting of greenstones away from synvolcanic granitoid domes. Rapid partial convective overturn of upper and middle crust occurred at 3.32 Ga during voluminous potassic felsic magmatism, followed by deposition of the Budjan Creek Formation at 3.31 Ga. Granitoid plutonism at ca. 3.29 Ga in the Kuranna terrane preceded deposition of ultramafic through felsic volcanics and chert in the West Pilbara granite-greenstone terrane (3.27-3.25 Ga Roebourne Group) and western margin of the East Pilbara granite-greenstone terrane (3.26-3.24 Ga Sulphur Springs Group). Geochemical and isotopic data suggest that volcanism resulted from plume-related rifting of the East Pilbara granite-greenstone terrane, which was accompanied by granitoid plutonism and deformation. Following this was ca. 100 m.y. of relative quiescence during which locally economic concentrations of banded iron-formation and siliciclastics of the Gorge Creek Group were deposited in the East Pilbara granite-greenstone terrane. Thereafter, geologic events are more consistent with microplate tectonics, commencing with deformation at 3.15 Ga followed by deposition of 3.13 to 3.11 Ga bimodal volcanics in the West Pilbara granite-greenstone terrane (Whundo Group), which have juvenile Nd isotope signatures and thus may represent either a rift or island- are succession. Basaltic rocks and minor felsic tuff were deposited in the East Pilbara granite-greenstone terrane at 3.06 Ga and possibly in the West Pilbara granite-greenstone terrane (Regal Formation). At 3.02 Ga. the Whundo and Roebourne Groups share a common history of deposition of banded iron-formation and granitoid plutonism across the Sholl shear zone, suggesting accretion at, or immediately preceding, this time. This was followed by deposition in the Mallina basin of the volcanic Whim Creek Group at 3.01 Ga, possibly as an arc, and then the 2.97 to 2.93 Ga volcanic Bookingarra (west) and clastic De Grey (east) Groups during periods of intracontinental rifting interspersed with compression and granitoid intrusion. The geochemistry of 2.95 Ga high Mg diorites (sanukitoids) indicates a previous episode of subduction during either the Whundo or Whim Creek Groups or both. Final events include emplacement of ultramafic-mafic layered intrusions (2.925 Ga in the West Pilbara granite-greenstone terrane), local shearing and lode Au mineralization (2.92 Ga in the West Pilbara granite-greenstone terrane, 2.90 Ga in the Mosquito Creek basin, 2.89 Ga in the East Pilbara granite-greenstone terrane), and intrusion of fractionated, Sn-Ta-Li-bearing granites to 2.85 Ga (East Pilbara granite-greenstone terrane).
  • Article
    Full-text available
    The Madoonga iron ore body hosted by banded iron formation (BIF) in the Weld Range greenstone belt of Western Australia is a blend of four genetically and compositionally distinct types of high-grade (>55 wt% Fe) iron ore that includes: (1) hypogene magnetite–talc veins, (2) hypogene specular hematite–quartz veins, (3) supergene goethite–hematite, and (4) supergene-modified, goethite–hematite-rich detrital ores. The spatial coincidence of these different ore types is a major factor controlling the overall size of the Madoonga ore body, but results in a compositionally heterogeneous ore deposit. Hypogene magnetite–talc veins that are up to 3 m thick and 50 m long formed within mylonite and shear zones located along the limbs of isoclinal, recumbent F1 folds. Relative to least-altered BIF, the magnetite–talc veins are enriched in Fe2O3(total), P2O5, MgO, Sc, Ga, Al2O3, Cl, and Zr; and depleted in SiO2 and MnO2. Mafic igneous countryrocks located within 10 m of the northern contact of the mineralised BIF display the replacement of primary igneous amphibole and plagioclase, and metamorphic chlorite by hypogene ferroan chlorite, talc, and magnetite. Later-forming, hypogene specular hematite–quartz veins and their associated alteration halos partly replace magnetite–talc veins in BIF and formed during, to shortly after, the F2-folding and tilting of the Weld Range tectono-stratigraphy. Supergene goethite–hematite ore zones that are up to 150 m wide, 400 m long, and extend to depths of 300 m replace least-altered BIF and existing hypogene alteration zones. The supergene ore zones formed as a result of the circulation of surface oxidised fluids through late NNW- to NNE-trending, subvertical brittle faults. Flat-lying, supergene goethite–hematite-altered, detrital sediments are concentrated in a paleo-topographic depression along the southern side of the main ENE-trending ridge at Madoonga. Iron ore deposits of the Weld Range greenstone belt record remarkably similar deformation histories, overprinting hypogene alteration events, and high-grade Fe ore types to other Fe ore deposits in the wider Yilgarn Craton (e.g. Koolyanobbing and Windarling deposits) despite these Fe camps being presently located more than 400 km apart and in different tectono-stratigraphic domains. Rather than the existence of a synchronous, Yilgarn-wide, Fe mineralisation event affecting BIF throughout the Yilgarn, it is more likely that these geographically isolated Fe ore districts experienced similar tectonic histories, whereby hypogene fluids were sourced from commonly available fluid reservoirs (e.g. metamorphic, magmatic, or both) and channelled along evolving structures during progressive deformation, resulting in several generations of Fe ore.
  • Article
    Aeromagnetic data is important for the exploration of gold and other hydrothermal deposits because geologically favourable environments are associated with changes in rock magnetism. For example, Archean orogenic gold mineralisation is known to be present in areas of structural complexity near major shear-zones that form conduits for mineralising fluids. Potential fluid pathways such as shear zones and faults are often associated with magnetite destructive alteration resulting in linear negative anomalies in magnetic data. Here, we present a new image analysis method that identifies geological structural complexity using lineaments automatically mapped within magnetic data. This quantitative analysis is efficient and self consistent in dealing with large volumes of data, and is suitable as a first-pass ground selection tool for orogenic gold exploration in greenfield terrains.
  • Article
    Full-text available
    Banded iron formations (BIF) are the protolith to most of the world's largest iron ore deposits. Previous hypogene genetic models for Paleoproterozoic "Lake Superior" BIF-hosted deposits invoke upwards, down-temperature flow of basinal brines via complex silica and carbonate precipitation/dissolution processes. Such models are challenged by the necessary SiO2 removal. Thermodynamic and mass balance constraints are used to refine conceptual models of the formation of BIF-hosted iron ore. These constraints, plus existing isotope and halogen ratio evidence, are consistent with removal of silica by down- or up-directed infiltration of high-pH hypersaline brines, with or without a contribution from basinal brines. The proposed link to surface environments suggest that Paleoproterozoic BIF-ore upgrade may provide a record of a critical time in the evolution of the Earth's biosphere and hydrosphere.
  • Article
    Full-text available
    We have compiled a record of the geochronology of mantle plume activity between 3.8 and 1.6 Ga. Over this time period, the ages of komatiites, and those of global plumes, correlate strongly, at the 99% confidence level, with the ages of banded iron formations (BIFs). The ages of continental plumes correlate more weakly, at an overall 85% confidence level. Using the geochronological records of these events, we can define four periods characterized by mantle superplume activity. Three of these periods are also times of enhanced BIF deposition. The fourth mantle plume period may similarly be coeval with increased BIF accumulation, but the BIF chronostratigraphic resolution is not accurate enough to test this rigorously. Mantle superplume volcanism may promote BIF deposition by increasing the Fe flux to the global oceans through continental weathering and/or through submarine hydrothermal processes. It may also be enhanced by increasing the number of paleotectonic environments appropriate for BIF deposition (particularly plume-induced ocean plateaus, seamounts, and intracratonic rifts) and by promoting global anoxic, Fe-rich hydrothermal plumes in the shallow to intermediate marine water column.
  • Article
    Full-text available
    ■ Abstract This paper reviews the Precambrian history of atmospheric oxygen, beginning,with a brief discussion of the possible nature and magnitude,of life be- fore the evolution of oxygenic,photosynthesis. This is followed,by a summary,of the various lines of evidence constraining oxygen levels through time, resulting in a sug- gested history of atmospheric,oxygen,concentrations. Also reviewed,are the various processes regulating oxygen concentrations, and several models of Precambrian oxy- gen evolution are presented. A sparse geologic record, combined with uncertainties as to its interpretation, yields only a fragmentary and imprecise reading of atmospheric oxygen evolution. Nevertheless, oxygen levels have increased through time, but not monotonically, with major and fascinating swings to both lower and higher levels. DEDICATION This manuscript is dedicated to the memory of Robert M. Garrels, one of the fathers
  • Article
    Full-text available
    Similarities between Proterozoic (~ 1.8-2.5 Gyr) and Archean (> 2.5 Gyr) banded iron-formations are probably more significant than their differences. The contrasts largely reflect differences in the tectonic settings of Proterozoic and Archean terrains. Archean banded iron-formations are not as thick nor laterally as extensive as the major Proterozoic iron-formations. Nevertheless, some Archean iron-formations have strike lengths of over 150-200 km and may have been quite extensive prior to the deformation that has affected most Archean terrains. Stratigraphic sequences in which iron-formations occur are highly variable and indicate that iron-formations formed in many depositional environments. Sedimentary textures in the iron-formations are dominated either by granules and oolites or laminations (including microbanding) reflecting differences in their physical conditions of deposition. Granular and oolitic textures are abundant in only three Proterozoic depositional basins and most Precambrian iron-formations are laminated. Despite differences in associated lithologies and sedimentary textures Precambrian iron-formations have similar bulk compositions and mineral assemblages, implying that the chemical conditions of iron-formation deposition were similar through much of the Precambrian. The formation of banded iron-formation appears not to have reached a maximum around 1.8-2.0 Gyr but to have been an important process over a long period in the Precambrian.
  • Article
    Full-text available
    A multilayer feed‐forward neural network, trained with a gradient descent, back‐propagation algorithm, is used to estimate the favourability for gold deposits using a raster GIS database for the Tenterfield 1:100 000 sheet area, New South Wales. The database consists of solid geology, regional faults, airborne magnetic and gamma‐ray survey data (U, Th, K and total count channels), and 63 deposit and occurrence locations. Input to the neural network consists of feature vectors formed by combining the values from co‐registered grid cells in each GIS thematic layer. The network was trained using binary target values to indicate the presence or absence of deposits. Although the neural network was trained as a binary classifier, output values for the trained network are in the range [0.1, 0.9] and are interpreted to indicate the degree of similarity of each input vector to a composite of all the deposit vectors used in training. These values are rescaled to produce a multiclass prospectivity map. To validate and assess the effectiveness of the neural‐network method, mineral‐prospectivity maps are also prepared using the empirical weights of evidence and the conceptual fuzzy‐logic methods. The neural‐network method produces a geologically plausible mineral‐prospectivity map similar, but superior, to the fuzzy logic and weights of evidence maps. The results of this study indicate that the use of neural networks for the integration of large multisource datasets used in regional mineral exploration, and for prediction of mineral prospectivity, offers several advantages over existing methods. These include the ability of neural networks to: (i) respond to critical combinations of parameters rather than increase the estimated prospectivity in response to each individual favourable parameter; (ii) combine datasets without the loss of information inherent in existing methods; and (iii) produce results that are relatively unaffected by redundant data, spurious data and data containing multiple populations. Statistical measures of map quality indicate that the neural‐network method performs as well as, or better than, existing methods while using approximately one‐third less data than the weights of evidence method.
  • Article
    Full-text available
    Mineral exploration comprises three sequential steps: development of a business strategy, creation and application of a targeting model, and follow-up with direct detection in defined high-priority domains. The main geoscientific challenge is the conceptual targeting phase which can lower geological risk and ensure cost-effective direct-detection exploration. A fundamental tenet of conceptual targeting is that ore deposits are part of much more extensive systems, and hence that targeting must be carried out at global through province to district scales. The heterogeneous distribution of ore deposits and their power-law size frequency distribution in individual provinces leads to alternative ‘Elephant Country’ and ‘First Mover’ strategies, both of which employ conceptual targeting, but at different scales. The first stage of targeting science involves development of robust, multi-scale targeting models for ore-deposit types, particularly larger examples. The targeting models can then be applied to identify specific targets by interrogating databases compiled as layers of spatially referenced key themes or parameters. At larger scales in immature terrains, a Hierarchical approach is commonly used to progressively reduce terrains and identify targets, whereas a Venn-diagram approach, the basis of most GIS-based prospectivity analyses, is more commonly used in mature terrains where spatial databases are of higher, more homogenous quality. Target ranking is best achieved using a multiplicative probability approach in which it is required that all essential processes in a mineral system must have operated to form a significant ore deposit. In practice, one or more critical spatially referenced parameters are used as proxies for the essential processes to develop a target score, which is a semi-quantitative estimate of probability of the presence of a large ore deposit. Such target ranking can be used in both proactive ground acquisition and reactive submittal-based project acquisition. Once targets have been defined and explored, it is important that there is critical feedback on the robustness of the targeting exercise such that new information is used to build superior databases and/or targeting models for future area-selection programs.
  • Article
    The conceptual approach used in this study incorporates spatial analysis techniques for data integration and analysis to perform reconnaissance-scale mineral prospectivity mapping for iron oxide copper – gold (IOCG) mineralisation in Finland. The known IOCG occurrences in Finland are characterised by the following features: (i) an epigenetic magnetite-rich host-rock; (ii) an association of Fe – Cu – Au ± Co ± U; (iii) ore minerals comprising magnetite, chalcopyrite, pyrite or pyrrhotite, and native gold; (iv) a gangue dominated by Ca-amphibole ± diopside, albite and biotite; (v) enrichment in Ag, Au, Bi, Ca, CO2, Cu, Fe, S, Te ± As, Ba, Cl, Co, K, LREE, Mo, Na, Pb, Rb, Sb, Se, U; (vi) multi-stage alteration; (vii) formation in the P – T range of 400 – 600°C, 150 – 350 MPa; and (viii) a distinct structural control in regions that have experienced both extensive compression and extension. The datasets used for the prospectivity analysis include a 1:1 000 000 scale geological map, high-resolution airborne geophysics, regional-scale multi-element till-geochemistry data, and a mineral occurrence database. The derived parameters used in the conceptual analysis include: (i) proximity to the craton margin; (ii) intersecting fault structures; (iii) presence of granitic intrusions particularly those with compatible and incompatible element enrichment; (iv) Cu, Co and Fe concentrations in till samples; (v) presence of hematite; and (vi) airborne magnetic highs and radiometric U data. A conceptual fuzzy-logic model was used to predict and locate the most prospective or favourable areas for IOCG exploration in the study area using the above-mentioned data layers. The models identify several permissive and high-potential areas within a significantly reduced potential exploration area. Validation of the modelling was conducted by quantifying the spatial association between the predicted endowment as favourability classes on the prospectivity map and the known mineral deposit sites with IOCG affinities using the Bayesian weights-of-evidence method.
  • Article
    Geoscience Australia and the Australian State and Territory Geological Surveys have systematically surveyed most of the Australian continent over the past 40 years using airborne gamma-ray spectrometry to map potassium, uranium and thorium elemental concentrations at the Earth's surface. However, the individual surveys that comprise the national gamma-ray spectrometric radioelement database are not all registered to the same datum. This limits the usefulness of the database as it is not possible to easily combine surveys into regional compilations or make accurate comparisons between radiometric signatures in different survey areas. To solve these problems, Geoscience Australia has undertaken an Australia-Wide Airborne Geophysical Survey (AWAGS), funded under the Australian Government's Onshore Energy Security Program, to serve as a radioelement baseline for all current and future airborne gamma-ray spectrometric surveys in Australia. The AWAGS survey has been back-calibrated to the International Atomic Energy Agency's (IAEA) radioelement datum. We have used the AWAGS data to level the national radioelement database by estimating survey correction factors that, once applied, minimise both the differences in radioelement estimates between surveys (where these surveys overlap) and the differences between the surveys and the AWAGS traverses. The database is thus effectively levelled to the IAEA datum. The levelled database has been used to produce the first `Radiometric Map of Australia' - levelled and merged composite potassium (% K), uranium (ppm eU) and thorium (ppm eTh) grids over Australia at 100m resolution. Interpreters can use the map to reliably compare the radiometric signatures observed over different parts of Australia. This enables the assessment of key mineralogical and geochemical properties of bedrock and regolith materials from different geological provinces and regions with contrasting landscape histories.
  • Article
    The integrated use of airborne magnetic and radiometric data, and SPOT Panchromatic and LANDSAT Thematic Mapper satellite data has proved effective for regolith mapping in the Yilgarn Craton of Western Australia. They can form the basis for designing appropriate soil sampling and regional exploration drilling in gold exploration programs. Horizontal derivative filters have been applied to airborne magnetic line data to enhance short wavelength magnetic responses of maghemite-rich lateritic weathering products within the regolith profile. Magnetised maghemite-rich pisolites often occur in buried Tertiary palaeo-drainage channels or within residual laterite horizons, and their distribution can be delineated with these filters. A new filter for magnetic data (REGMAG) is described that maps the distribution of short-wavelength responses and has proved effective in mapping magnetic structures in the regolith. Ratios of airborne radiometric potassium and thorium channels, and ternary images of potassium, thorium and uranium, are useful in highlighting the radiometric signatures of various weathering products in the regolith. A new ratio normalisation algorithm improves the resolution of ratioed data. Combined SPOT Panchromatic and LANDSAT Thematic Mapper satellite data are useful for landform mapping. Ratios of LANDSAT Thematic Mapper bands provide discrimination of various weathering products such as saprolite, pisolitic/nodular goethite and hematite-rich laterite, quartz and kaolinite rich alluvial cover, and red-earth calcareous clays. Examples from the Archaean Yilgarn Craton of Western Australia show how these processing techniques can reveal valuable information from remotely sensed data which can assist with regolith mapping.
  • Article
    Full-text available
    Geophysical exploration in Western Australia is hindered by a mantle of conductive and magnetic weathered rocks that covers much of the State. This has required the adaptation of most geophysical methods for successful application in Western Australian conditions, and has led to the development and widespread use of, for instance, high-resolution aeromagnetics and time-domain electromagnetic methods. However, these difficulties have not prevented geophysics from being an integral part of exploration for base metal, diamond, gold, iron ore, manganese, nickel and uranium deposits in Western Australia. Mississippi Valley-type base-metal deposits are difficult geophysical targets and direct detection of the ore is not usually possible. However, gravity and magnetic data can be used to locate basement highs associated with such deposits and, on a semi-regional scale, induced polarisation surveys have been used to locate marcasite halos associated with the orebodies. Volcanic-hosted massive sulphide base-metal deposits have variable geophysical responses. Physical property contrasts with their host are highly variable and thus methods such as magnetics, induced poiarisation and electromagnetics may fail to generate recognisable responses. Mise-a-la-masse surveys have proved highly successful for mapping such mineralisation on a prospect scale, once it has been intersected by drilling. The only example of a sedimentary exhalative deposit in the State for which data are available has distinct gravity, magnetic and time-domain electromagnetic anomalies. Diamonds in Western Australia mainly occur in lamproite pipes. These pipes have variable magnetisations but can usually be detected using high-resolution aeromagnetic surveys. The pipes can also be conductive and mapped using electromagnetic techniques if the host rocks are suitably resistive. The major geophysical method utilised in gold exploration is high-resolution aeromagnetics which is used to map favourable structures and rock types. Electrical and electromagnetic methods can also be used where gold is associated with sulphides. Geophysics has been comparatively little used in exploration for iron ore. Exploration for supergene-enriched deposits mainly uses aeromagnetics, to map favourable structures and to detect magnetite destruction and replacement associated with mineralisation, and gamma-ray logging for stratigraphic correlation purposes. The major technique used in manganese exploration is the gravity method, taking advantage of the positive density contrast between ore and host rocks. The mineral sands industry uses aeromagnetic data to map placer deposits containing ilmenite, but the relatively low cost of drilling limits the use of geophysical exploration methods. Nickel sulphide mineralisation can be directly detected using induced polarisation and electromagnetic techniques. Gravity and magnetic surveys are also used, but mainly in a mapping role. Carbonatitic intrusions associated with rare-earth-element mineralisation give rise to large magnetic anomalies. Radiometric and gravity anomalies can also occur. Uranium mineralisation has been directly detected using radiometric data, but some deposits are concealed below cover. Magnetic, electromagnetic, electrical and gravity surveys can be used to locate the rocks and structures which host mineralisation.
  • Article
    The Bureau of Mineral Resources has been routinely acquiring airborne magnetic surveys over the land area of Australia since 1951 to record and map anomalies in the earth's magnetic field attributable to geological structures and lithologies. In forty years, over four million line kilometres of survey data have been flown, while the technology of survey practice has passed through various stages of development. About 83 per cent of the land area has now been covered with so-called reconnaissance surveys flown 150 m above terrain at line spacing between 1.5 and 3.2 km. Located profile data for these surveys have been gridded using a minimum curvature technique, to 15 second of arc (approximately 400 m) and, where necessary, micro-levelled. Data for much of the remaining areas ? particularly the inland sedimentary basins covered by surveys of lower specifications ? were obtained from digital data on an approximately 2-km grid (72 seconds of arc) published in 1976; these have also been interpolated to 15 second of arc. The data were first assembled for each of over five hundred 1:250 000 map sheets. The 1:250 000 sheets were linked by minimizing the discrepancies along their common boundaries (which were often also survey boundaries) and reducing remaining mis-ties through Laplacian smoothing to minimize the visibility of boundaries between surveys acquired separately. While the data quality varies with instrumentation and survey parameters, it is almost everywhere good enough to provide a useful synoptic view of rnagnetic anomaly patterns, which can be expected to give important new insights into geology and tectonics at a continental scale, and to provide a regional framework within which to interpret more local magnetic anomalies. The purpose of this short paper is to report the latest progress on compilation of the Magnetic Anomaly Map of Australia, which is scheduled for publication late in 1992.
  • Article
    The Mount Bruce Megasequence Set (formerly the Mount Bruce Supergroup) was deposited on the Pilbara Craton during the late Archaean and early Proterozoic. It comprises two metasequences, the Chichester Range Megasequence and the overlying Hamersley Range Megasequence. Each megasequence comprises three supersequences or supersequence packages whose tectonic history can be explained in terms of Phanerozoic-style tectonics. The Mount Bruce Megasequence Set and its contained iron formations are preserved in the relatively distal hinterland of this collisional orogen. -from Authors
  • Article
    Full-text available
    The support vector machine (SVM) is a group of theoretically superior machine learning algorithms. It was found competitive with the best available machine learning algorithms in classifying high-dimensional data sets. This paper gives an introduction to the theoretical development of the SVM and an experi- mental evaluation of its accuracy, stability and training speed in deriving land cover classié cations from satellite images. The SVM was compared to three other popular classié ers, including the maximum likelihood classié er (MLC), neural network classié ers (NNC) and decision tree classié ers (DTC). The impacts of kernel coné guration on the performance of the SVM and of the selection of training data and input variables on the four classié ers were also evaluated in this experiment.
  • Article
    A prima facie comparison is made between diagenetic, ‘sedimentary’ boudinage structures at outcrop scales (scales of centimetres to tens of centimetres), and zones of localised stratigraphic thinning (on scales of tens of metres) in beds of the Marra Mamba and Brockman Iron Formations of the Hamersley Iron Province of Western Australia. If the comparison is valid, it suggests that some of the hematite enrichment ores of the province may be diagenetic ores located in necks of diagenetic boudinage structures related to extensional disturbance of the basin when the sequence was only partly consolidated. This interpretation is seen as similar to the consensual supergene metasomatic replacement hypothesis for the origin of the ores in respect of mineral solution-precipitation mechanisms, but differs in respect of important aspects of bulk process, and in their implications for iron ore exploration. A prima facie comparison is also made with the structure locating some ores of the Krivoi Rog region of Ukraine for which a boudinage control has been explicitly described, and with the structure controlling the Nimba Range deposit, Liberia. If such a comparison is valid, then boudinage could account simultaneously for the Proterozoic age of the deposits, the localised stratigraphic thinning, the influx of iron, and the ‘removal’ of silica. Further, on the basis of self-similarity of boudinage structure across scale, region and tectonic regime, and in conjunction with the recognition by others on different grounds that the examples described in the paper may be extrapolated world-wide, boudinage may provide a partial framework within which existing models for the formation of enriched hematite ores of Proterozoic banded iron formations can be adapted. The paper is conceptual and provides no new data.
  • Article
    Full-text available
    Banded iron-formations (BIFs) were comparatively abundant and widespread marine sedimentary rocks in the Archean and Lower Proterozoic eras, but thereafter they appear to be restricted to the Neoproterozoic and Paleozoic eras, although there are indications of similar rocks forming at present. BIFs are important as the major source of iron ore for industry and have also been used to support hypotheses regarding the evolution of life, oceans, and the atmosphere in the Archean and Proterozoic. They apparently formed in deep water and consisted of a semi-regular alternation of quartz (chert) and iron-rich minerals with little or no terrigenous sediment and low (
  • Article
    Iron formations are economically important sedimentary rocks that are most common in Precambrian sedimentary successions. Although many aspects of their origin remain unresolved, it is widely accepted that secular changes in the style of their deposition are linked to environmental and geochemical evolution of Earth. Two types of Precambrian iron formations have been recognized with respect to their depositional setting. Al-goma-type iron formations are interlayered with or stratigraphically linked to submarine-emplaced volcanic rocks in greenstone belts and, in some cases, with volcanogenic massive sulfide (VMS) deposits. In contrast, larger Superior-type iron formations are developed in passive-margin sedimentary rock successions and generally lack direct relationships with volcanic rocks. The early distinction made between these two iron-formation types, although mimimized by later studies, remains a valid first approximation. Texturally, iron formations were also divided into two groups. Banded iron formation (BEE) is dominant in Archean to earliest Paleoproterozoic successions, whereas granular iron formation (GIF) is much more common in Paleoproterozoic successions. Secular changes in the style of iron-formation deposition, identified more than 20 years ago, have been linked to diverse environmental changes. Geochronologic studies emphasize the episodic nature of the deposition of giant iron formations, as they are coeval with, and genetically linked to, time periods when large igneous provinces (LIPs) were emplaced. Superior-type iron formation first appeared at ca. 2.6 Ga, when construction of large continents changed the heat flux at the core-mantle boundary. From ca. 2.6 to ca. 2.4 Ga, global mafic magmatism culminated in the deposition of giant Superior-type BIF in South Africa, Australia, Brazil, Russia, and Ukraine. The younger BIFs in this age range were deposited during the early stage of a shift from reducing to oxidizing conditions in the ocean-atmosphere system. Counterintuitively, enhanced magmatism at 2.50 to 2.45 Ga may have triggered atmospheric oxidation. After the rise of atmospheric oxygen dining the GOE at ca. 2.4 Ga, CIF became abundant in the rock record, compared to the predominance of BEE prior to the Great Oxidation Event (GOE). Iron formations generally disappeared at ca. 1.85 Ga, reappearing at the end of the Neoproterozoic, again tied to periods of intense magmatic activity and also, in this case, to global glaciations, the so-called Snowball Earth events. By the Phanerozoic, marine iron deposition was restricted to local areas of closed to semiclosed basins, where volcanic and hydrothermal activity was extensive (e.g., back-arc basins), with ironstones additionally, being linked to periods of intense magmatic activity and ocean anoxia. Late Paleoproterozoic iron formations and Paleozoic ironstones were deposited at the redoxcline where biological and nonbiological oxidation occurred. In contrast, older iron formations were deposited in anoxic oceans, where ferrous iron oxidation by anoxygenic photosynthetic bacteria was likely an important process. Endogenic and exo-genic factors contributed to produce the conditions necessary for deposition of iron formation. Mantle plume events that led to the formation of LIPs also enhanced spreading rates of midocean ridges and produced higher growth rates of oceanic plateaus, both processes thus having contributed to a higher hydrothermal flux to the ocean. Oceanic and atmosplieric redox states determined the fate of this flux. When the hydrothermal flux overwhemed the oceanic oxidation state, iron was transported and deposited distally from hydrothermal vents. Where the hydrothermal flux was insufficient to overwhelm the oceanic redox state, iron was deposited only proximally, generally, as oxides or sulfides. Manganese, in contrast, was more mobile. We conclude that occurrences of BIF, CIF, Phanerozoic ironstones, and exhalites surrounding VMS systems record a complex interplay involving mantle heat, tectonics, and surface redox conditions throughout Earth history, in which mantle heat unidirectionally declined and the surface oxidation state mainly unidirectionally increased, accompanied by superimposed shorter term fluctuations.
  • Article
    Full-text available
    The Mount Gibson banded iron formation lies within the Windanning Formation of the Luke Creek Group, which is found in almost all greenstone belts throughout the Murchison province. The banded iron formation (BIF) typically consists of alternating bands of magnetite and microcrystalline quartz (chert) with rare carbonaceous and iron silicate-rich shale partings and layers and rare, thin fine-grained tuff bands. Owing to repetition by isoclinal folding and attenuation by faulting, the true thickness of the Mount Gibson BIF is unknown but appears to be on the order of 100 m. Although the Mount Gibson BIF is typical of many Algoma-type iron formations, situated within a greenstone belt on an Archean craton, it is also similar to Hamersley-type BIF in petrology, areal extent, and ore genesis. High-grade hematite deposits formed within BIF were thought to have formed by the supergene leaching of chert from typical cherty BIF. Recent evidence suggests that at least some of these deposits are formed by hypogene replacement of chert by carbonates with subsequent supergene leaching of the carbonate and accessory minerals and oxidation of magnetite to hematite. Magnetite-carbonate BIF, in which there is clear evidence of hydrothermal replacement of chert by carbonate, forms distinctive magnetite-goethite ore with magnetite locally persisting to the surface. Mount Gibson shows clear evidence of the formation of high-grade ore by this process but also contains high-grade hematite ore and chert-free BIF that show no evidence of the hypogene replacement of chert. High-grade hematite occurrences, up to 1 km in strike length, are found within the weathered zone overlying the magnetite BIF at Mount Gibson and continue into unweathered chert-free BIF at depth that show no evidence of hydrothermal carbonate or supergene enrichment. The cherty BIF shows sharp contacts against chert-free BIF and high-grade ore, even when strongly weathered. This suggests that deep saprolitic in situ high-grade ore may be produced by different processes, including hydrothermal replacement of chert mesobands by carbonates with subsequent supergene leaching of the carbonate and by the oxidation of chertfree BIF, in which chert bands either never developed or were apparently removed during diagenesis. Neither model requires supergene selective leaching of quartz (chert) during deep weathering.
  • Article
    The ores derived from banded iron-formation in situ in the Hamersley Iron Province represent a series of events probably related to the time of emergence of their parent rocks by the slow process of erosion of the overlying cover. The simplest and least mature ores consist essentially of residual oxides in a matrix of goethite, the latter derived from the supergene replacement of part of the chert and other components of the original banded iron-formation. The iron necessary for this enrichment logically comes from the now-eroded extension of banded iron-formation outcrop.
  • Article
    Iron-formation of Archean age (older than 2,500 m.y.) is widely distributed in supra-crustal assemblages of the Canadian Shield. Although individual iron-formations are comparatively small, the total ore reserve is 35,000 million tons or 25 percent of the established iron ore resources in Canada. Archean iron-formations form an intimate part of volcanic-rich greenstone belts or segments. Individual iron-formations are commonly associated with upper pyroclastic phases of predominantly tholeiitic to calc-alkaline, mafic to felsic volcanic sequences and nearby turbidite assemblages. The iron-formations are readily attributed to volcanic processes in terms of the source of the chemical components. The distribution of Archean iron facies conforms in the main to the common worldwide depositional pattern of iron. Thus oxide, carbonate, and sulfide facies are present in that shallow-to-deep order upon Archean paleoslopes which are components of Archean basins. Archean iron facies thereby provide a powerful clue to basin analysis. In general, oxide facies is the most common and readily recognized form of Archean iron-formation. Sulfide and carbonate facies iron-formations, although widely distributed, are comparatively thin, discontinuous, and inconspicuous. Most if not all Archean iron-formation is attributed to a volcanic exhalative source (exhalite). A plot of known Archean iron-formation by facies within the greenstone belts of the Canadian Shield reveals the presence of a number of large Archean basins. Each basin, although presently elliptical in outline as a result of structural deformation with the major axis 250 to 400 miles long, may originally have been more nearly circular in outline and 450 to 600 miles in diameter. The margin of a basin features a triple lithofacies association of 1) oxide-carbonate-sulfide transition, 2) arc-type felsic volcanic rocks, and 3) proximal conglomerates. Interior parts feature sulfide facies iron-formation, tholeiitic volcanic rocks and fine-grained sediments. A typical basin encompasses several greenstone belts and intervening batholithic complexes. Archean basins of this type represent first order crustal features and are particularly significant to an understanding of early crustal evolution. A meteorite impact scar hypothesis is considered unlikely. Rather the basins are considered to be mantle thermal plume or "hot spot' derivates, a manifestation of particularly active vertical thermal streams resulting from prevailing high geothermal gradients and a thin mobile crust in Archean time.
  • Article
    Banded iron-formations (BIFs) occur in the Precambrian geologic record over a wide time span. Beginning at 3.8 Ga (Isua, West Greenland), they are part of Archean cratons and range in age from about 3.5 until 2.5 Ga. Their overall volume reaches a maximum at about 2.5 Ga (iron-formations in the Hamersley Basin of Western Australia) and they disappear from the geologic record at about 1.8 Ga, only to reappear between 0.8 and 0.6 Ga. The stratigraphic sequences in which BIFs occur are highly variable. Most Archean iron-formations are part of greenstone belts that have been deformed, metamorphosed, and dismembered. This makes reconstruction of the basinal setting of such BIFs very difficult. The general lack of metamorphism and deformation of extensive BIFs of the Hamersley Range of Western Australia and the Transvaal Supergroup of South Africa allow for much better evaluations of original basinal settings. Most Archean iron-formations show fine laminations and/or microbanding. Such microbanding is especially well developed in the Brockman Iron Formation of Western Australia, where it has been interpreted as chemical varves, or annual layers of sedimentation. BIFs ranging in age from 2.2 Ga to about 1.8 Ga (e.g., those of the Lake Superior region, U.S.A., Labrador Trough, Canada, and the Nabberu Basin of Western Australia) commonly exhibit granular textures and lack microbanding. The mineralogy of the least metamorphosed BIFs consists of combinations of the following minerals: chert, magnetite, hematite, carbonates (most commonly siderite and members of the dolomite-ankerite series), greenalite, stilpnomelane, and riebeckite, and locally pyrite. Minnesotaite is a common, very low-grade metamorphic reaction product. The Eh-pH stability fields of the above minerals (and/or their precursors) indicate anoxic conditions for the original depositional environment. The average bulk chemistry of BIFs, from 3.8 through 1.8 Ga in age, is very similar. They are rich in total Fe (ranging from about 20 to 40 wt%) and SiO2 (ranging from 43 to 56 wt%). CaO and MgO, contents range from 1.75 to 9.0 and from 1.20 to 6.7 wt%, respectively. Al2O3 contents are very low, ranging from 0.09 to 1.8 wt%. These chemical values show that they are clean chemical sediments devoid of detrital input. Only the Neoproterozoic iron-formations (of 0.8 to 0.6 Ga in age) have very different mineralogical and chemical make-ups. They consist mainly of chert and hematite, with minor carbonates. The rare-earth element profiles of almost all BIFs,with generally pronounced positive Eu anomalies, indicate that the source of Fe and Si was the result of deep ocean hydrothermal activity admixed with sea water. The prograde metamorphism of iron-formations produces sequentially Fe-amphiboles, then Fe-pyroxenes, and finally (at highest grade) Fe-olivine-containing assemblages. Such metamorphic reactions are. isochemical except for decarbonation and dehydration. The common fine lamination (and/or microbanding) as well as the lack of detrital components in most BIFs suggest that such are the result of deposition, below wave base, in the deeper parts of ocean basins. Those with granular textures are regarded as the result of deposition in shallow water, platformal areas. Carbon isotope data suggest that for a long period of time (from Archean to Early Proterozoic) the ocean basins were stratified with respect to delta(13)C (in carbonates) as well as organic carbon content. In Middle Proterozoic time (when granular BIFs appear) this stratification diminishes and is lost. The Neoproterozoic BIFs occur in stratigraphic sequences with glaciomarine deposits. These BIFs are the result of anoxic conditions that resulted from the stagnation in the oceans beneath a near-global ice cover, referred to as "Snowball Earth." All of the most "primary" mineral assemblages appear to be the result of chemical precipitation under anoxic conditions. There are, as yet, no data to support that BIF precipitation was linked directly to microbial, activity. The relative abundance of BIF throughout the Precambrian record is correlated with a possible curve for the evolution of the 02 content in the Precambrian atmosphere.
  • Article
    Full-text available
    The Koolyanobbing banded iron formation (BIF)-hosted iron ore deposits (total premining resources ∼150 million metric tons (Mt), indicated reserves ∼32 Mt) are located in the Mesoarchean lower succession BIF of the Koolyanobbing greenstone belt, Younami terrane, Yilgarn craton in Western Australia. In the Koolyanobbing greenstone belt a deformation sequence that broadly correlates with the proposed deformation history of most greenstones belts within the Southern Cross domain includes: D 1 structures (mainly small-scale F 1a and F 1b folds, formed in a north-south to northwest-southeast compressional regime), a ductile to brittle deformation sequence, D 2 to D 4 (generated during east-west compression) and, a late-stage brittle segmentation of BIF and reactivation of faults, attributed to D 5. The formation of the seven known medium- (45-58 wt % Fe) to high-grade (58-68 wt % Fe) magnetite-,martite-, specularite-, and goethite-bearing orebodies can be subdivided into four Archean stages and one weathering-related upgrade from the Permian and/or Mesozoic to recent times. The Archean ore-forming stages comprise: (1) early Fe-Mg ± Ca metasomatism causing local ferroan carbonate and ferroan talc alteration of the metamorphosed quartz-magnetite BIF protolith; (2) sequential syn-D 2a (coaxial) to syn-D 4 (transpressional) tight folding-driven removal of carbonate, quartz and minor ferroan talc by solution and mechanical transfer, producing residual enrichment of medium- to high-grade magnetite ore; (3) magnetite mineralization in syn-D 2b and syn-D 4 breccias and fractures, forming medium-grade ore zones, or overprint magnetite in BIF and first-stage magnetite ore; and (4) mineralization of hydrothermal specularite and locally associated ferroan dolomite-quartz alteration, and local oxidation of magnetite in and near brittle D 4 faults, fractures, and reactivated F 1 and F 2a fold cores. Modern weathering-related leaching of carbonate (and minor quartz), pseudomorphic goethite replacement of existing iron oxides and gangue, and coeval or subsequent to oxidation in the vadose zone formed goethitemartite ore with local relics of specularite or magnetite and/or kenomagnetite. The intensity and localization of this supergene modification is, in most deposits at Koolyanobbing, controlled by existing hypogene magnetite, specularite-rich medium- to high-grade ore zones and/or carbonate-altered BIF at depth. The existence of high-grade ore below the weathering horizons suggests the possibility of further concealed magnetite- and/or specularite-rich orebodies within the deposits and region.