Graeme F. Bonham-Carter

Numerical procedures and computer program for fitting an inverted Gaussian model to vegetation reflectance data

Abstract Four alternative procedures are outlined for fitting an Inverted Gaussian Model to vegetation spectra in the red-edge region, 680–800 nm. Two procedures use a linearized equation, resulting in computational efficiency. Two procedures use nonlinear optimization which is demanding computationally but always gives a superior fit to the data. The wavelength parameters, λo and σ, that define the position and shape of the model, are used to measure spectral shifts of the red-edge. The red-edge is reported to differ in position in vegetation under stress, so images of λo and σ derived by fitting the model to spectra measured from airborne sensors may be used to map the spatial extent of vegetation stress. A FORTRAN 77 program for making the fitting calculations is included in the Appendix.

Review of methods to determine gold, platinum and palladium in production-oriented geochemical laboratories, with application of a statistical procedure to test for bias

Abstract Current methods for determining Au, Pt and Pd in geological materials in Canadian commercial laboratories are reviewed. The relative merits of fire assay (Pb and NiS) and wet-chemical attack as methods of decomposition of rocks, soils and sediments are discussed. Amongst various analytical techniques, neutron activation is compared to atomic absorption spectrometry (AAS), and to three types of analysis employing the inductively coupled plasma as a source-emission spectrometry, mass spectrometry and atomic fluorescence. Clearly, the sensitivity and flexibility of ICP-mass spectrometry, introduced commercially in 1983, ensure a dominant role for this technique in the determination of Au and the platinum-group elements. A new statistical method for detecting bias, applied to the determination of Au in 157 rock samples, shows that instrumental neutron activation analysis (INAA) yields values which are 30% greater than those by graphite furnace AAS with an aqua regia attack; this same bias occurs in a second data set. Although INAA is an excellent technique for the determination of Au in vegetation, the low background levels of Pt and Pd and poor sensitivity of INAA for these elements make analysis by GF-AAS or ICP-MS more attractive. Further research is required in wet-chemical dissolution procedures for Pt and Pd in dried and ashed vegetation. Fire assay is an alternative choice of decomposition for ashed vegetation but, currently, low-level detection (to 1 ppb for 1 g samples) is hampered by contamination during fusion and by contributions to blank levels from flux constituents. Gold, Pt and Pd in waters can now be determined to levels of 1, 4 and 2 ppt, respectively, in a 1-L water sample using ICP-MS or graphite furnace AAS following preconcentration by adsorption on to activated charcoal. The detection limits for Pt and Pd must be lowered further by a factor of ten for application to exploration goechemistry.

Catchment basin analysis applied to surficial geochemical data, Cobequid Highlands, Nova Scotia

Abstract Areal proportions of mapped lithologic units are used as independent variables in a regression equation to predict the element content of stream sediments. Geochemical variation due to average differences in chemical composition between lithologic units is removed by linear regression from the data, highlighting anomalies due to mineralization and to secondary processes active in the surficial environment. For the Cobequid Highlands lithologic effects explain between 30% and 65% of the variance, depending on the element examined. The effects of scavenging metal ions by Fe and Mn in the secondary environment are also removed using the same technique. Co, Ni and As show particularly strong associations with Fe and Mn. Regression calculations also show that if geologic map units are recorded as present or absent (1 or 0) in each basin, instead of using area measurements, the proportion of variance explained is almost as large. However, using only the most areally abundant unit in each basin results in about 25% less efficient prediction, while similar results are obtained using only the presence of the map unit closest to each sample point. Co-occurrence of units is represented in the regression equation by using cross-products of lithologic variables in presence/absence form. Co-occurrence of certain map units in a catchment basin appears to be a relatively important factor for predicting Zn. A method for obtaining a weighted mean and weighted dispersion for the element contents characterizing each map unit is proposed. These values are more easily interpreted than those from regression coefficients, which suffer from large variances for map units which cover only small areas. A catchment basin map showing positive Zn anomalies after removing geological background and secondary effects due to Fe and Mn scavenging is shown as a grey-tone image, based on a percentile scale. Digitization of geologic and catchment basin maps and their merging with geographically located geochemical data, allows routine production of colour or grey-tone images depicting measured or derived geochemical parameters and provides a powerful tool for geochemical exploration.

Evaluation of portable X-ray fluorescence (pXRF) in exploration and mining: Phase 1, control reference materials

This paper describes a project sponsored by the Canadian Mining Industry Research Organisation (CAMIRO) to evaluate the strengths and weaknesses of portable XRF for use in mineral exploration and mining and to develop best-practice protocols in the analysis of rocks, soils, sediments and drill-core. Phase I focussed on the analysis of pulp control reference materials (CRMs) to determine the figures of merit, principally accuracy and precision, of the technique before introducing the confounding parameters associated with in-situ analysis such as heterogeneity, particle size and moisture. Five instruments (three handheld and two portable benchtop) from three manufacturers were used to carry out replicate analyses (n = 10) of a diverse suite of 41 CRMs, from barren granites, through soils and sediments, to ores. Standard factory calibration, in mining and soil modes, was used. The performance of the instruments was evaluated using x-y plots of results versus established element concentrations for the CRMs and the values of goodness of fit (r2), slope and intercept documented for both full and restricted concentration ranges. For many elements, the performance across the instruments varied markedly, as did the ability to correct for spectral interferences. Numerous interferences were encountered, particularly from the rare-earth elements (REEs) on transition elements, but also for well-known interference pairs such as Pb on As, Zn on Au, U on Mo, and Th on Bi. In general, major elements with the exception of the light element Mg were well determined, as were Mn and Ti. Sensitivity was inadequate for Cl and P; however, S could be measured with acceptable precision to c. 0.05% S. Performance for the trace elements was categorized as follows: very good for As, Cu, Nb, Pb, Rb, Sr, and Y (r2>0.9 for more than 1 instrument); good for Ba, Mo, Sn, Zn and Zr (r2>0.9 for 1 instrument); moderate for Cr, Sb, Se, Th and U (r2 = 0.6–0.8); poor for Ag, Cd, Co, Ni and V; and very poor for Au, Bi, Cs, Hf, Hg, Pd, Sc, Ta, Te and W. Of the REEs determined (La, Ce, Nd, Sm, using the standard calibration by the manufacturer, i.e. not REE-specific), only La showed adequate sensitivity and precision (3–5% RSD), however, only at concentrations approaching c. 1000 ppm (50–100% RSD at La <100 ppm). Slopes of the best-fit lines, where r2 ≥0.6, ranged from 0.5 to 5.0, indicating that calibration is required by the user for both the soil and mining modes. The precision, shown by 10 replicate readings, was excellent and usually better than 10% RSD except where close to detection limit or where major interferences were present. The beam time study showed that, in most situations, 60 s was a good compromise between productivity and precision but also highlighted cases of significant drift and a lack of improvement in precision with longer beam time. A study of the thin-film sample cover used for cups demonstrated that 4.0-µm Prolene® is superior to the same thickness of Mylar® in both transmittance (especially for Mg, Al, Si) and contamination properties.

Background corrections to stream geochemical data using digitized drainage and geological maps: application to selwyn basin, yukon and northwest territories

Abstract The effect of lithology has been evaluated for considering the spatial variation of trace elements measured in stream sediment and water. Geochemical data from about 1000 samples collected in the Nahanni River area (NTS 1051) were merged with a digital file of areal proportions of 34 bedrock map units occurring in the sampled catchment basins. Multiple-regression experiments showed that for many ore-associated elements, areal proportions of map units explain over 50% of the total variation. The effects of iron, manganese, loss on ignition and pH are subordinate to lithology for all elements examined, except for cobalt whose distribution is strongly correlated to manganese. Zinc (water), zinc (sediment) and lead (sediment) residuals, after removing the effects of lithology, and correcting for differential dilution due to variation in catchment area, provide a better prediction of known lead-zinc deposits than raw data. Combined zinc (water and sediment) anomalies appear to be as good as lead in predicting known lead-zinc occurrences. Combined zinc in sediment and water anomalies have a further advantage over lead in that they are usually more extensive and are therefore easier to detect. The reduction of all data to a machine-processable form enables automatic production of digital maps of drainage basins, coloured or shaded according to a geochemical abundance scale.

Modelling lake sediment geochemical distribution using principal component, indicator kriging and multifractal power-spectrum analysis: a case study from Gowganda, Ontario

Combined geostatistical and multifractal power-spectrum modelling of geochemical distributions can provide suitable indicators of metal dispersion, and is capable of analysing complex problems for targeting potential areas for mineral exploration. A case study analysing lake sediment geochemical data for the Gowganda area is presented and development of the methodology for spatial analysis of the data is described. The Gowganda-Cobalt area of northeastern Ontario is a textbook example of Co, Ag-Co vein-type deposit, which by 1984 had yielded one-half billion ounces of Ag. The area is also known for shear-zone-hosted Au mineralization. This paper uses the spatial and geometric distribution of lake sediment data to discriminate geochemical anomalies from background values. The application of two geostatistical techniques (spatial principal component analysis and indicator kriging) allows the estimation of geochemical distributions by utilizing their statistical and spatial properties. The newly developed multifractal power-spectrum method additionally allows for the geochemical distributions to be modelled by their multifractal Fourier-transformed power-spectrum characteristics. Verification of the estimates produced by these techniques has been enabled through spatial analysis of bedrock geology and mineral deposit occurrences in the area.

Statistical study of trace elements in the soluble organic and amorphous Fe-Mn phases of surficial sediments, Sudbury Basin. 1. Multivariate and spatial analysis

Abstract Samples of humus and B-and C-horizon till collected from the north and east margins of the Sudbury Basin were analyzed for about 30 trace elements. The data consist of two sets of analyses, one being a conventional “near total” analysis following an aqua regia digestion, the other being a “partia” analysis from a selective leach designed to extract the metals bound to the soluble organic fraction of humus and to the amorphous Fe and Mn oxides of B- and C-horizon till. Statistical processing, using both standard approaches and some newly developed methods of spatial analysis, were applied to evaluate spatial geochemical patterns and their relationship to geological factors such as proximity to known mineral occurrences. In particular, the information content of the partial analyses was evaluated to determine whether the partial analyses reveal interpretable patterns not detected in the “near total” analyses. From a visual appraisal of dotplots of Ni and Cu in till, it is difficult to determine whether the partial data add anything not already seen in the “total” data. However, when residuals obtained by removing that part of the partial variation explained by the “total” variation are mapped, distinctive patterns are produced. These patterns are enhanced by applying an adaptive filtering technique using a spatial U-statistic. Multielement groups (“total” data, 30 elements in till) determined by principal components analysis show that the first two components are due to Ni-Co-Cr-Cu-Zn and Pd-Au-Pt-Sb associations. A multifractal area-concentration model fitted to component scores allows multielement anomalies to be mapped. The first component is shown to be spatially associated with Ni-Cu deposits, the second with Au and Ni-Cu-Au deposits. Analysis of variance, with “total”, partial and residual data taken in turn as the dependent (“response”) variable, shows that both Ni and Cu in till are significantly “explained” by geological factors (underlying rock type, proximity to contacts and to mineralization). Neither the “total” nor the partial analyses on their own show significant associations with these geological factors for either element in C-horizon till. This demonstrates the value of selective leach analyses, where they are carried out in addition to the “total” analyses, for this dataset. A multivariate analysis of variance shows that if all three surfical media (humus, B- and C-horizon till) are used, the selective extraction data are not so advantageous, although still adding some predictive capability, because additional “explanatory” information comes from multimedia rather than multiphase sources.

Lithodiversity and its spatial association with metallic mineral sites, Great Basin of Nevada

Geographical information system (GIS) techniques were used to investigate the spatial association between metallic mineral sites and lithodiversity in Nevada. Mineral site data sets include various size and type subsets of about 5,500 metal-bearing occurrences and deposits. Lithodiversity was calculated by counting the number of unique geological map units within four sizes of square-shaped sample neighborhoods (2.5-by-2.5, 5-by-5, 10-by-10, and 20-by-20 km) on three different scales of geological maps (national, 1:2,500,000; state, 1:500,000; county, 1:250,000). The spatial association between mineral sites and lithodiversity was observed to increase with increasing lithodiversity. This relationship is consistent for (1) both basin-range and range-only regions, (2) four sizes of sample neighborhoods, (3) various mineral site subsets, (4) the three scales of geological maps, and (5) areas not covered by large-scale maps. A map scale of 1:500,000 and lithodiversity sampling neighborhood of 5-by-5 km was determined to best describe the association. Positive associations occurred for areas having >3 geological map units per neighborhood, with the strongest observed at approximately >7 units. Areas in Nevada with more than three geological map units per 5-by-5 km neighborhood contain more mineral sites than would be expected resulting from chance. High lithodiversity likely reflects the occurrence of complex structural, stratigraphic, and intrusive relationships that are thought to control, focus, localize, or expose mineralization. The application of lithodiversity measurements to areas that are not well explored may help delineate regional-scale exploration targets and provide GIS-supported mineral resource assessment and exploration activity another method that makes use of widely available geological map data.

Comparison of metal distributions in snow, peat, lakes and humus around a Cu smelter in western Québec, Canada

Concentrations of metals around the Cu smelter at Rouyn-Noranda, Québec, in snow, peat, the humus horizon of soil, lake water and sediments, show generally similar spatial distribution patterns. The contours of smelter-related metal concentrations in all these media are approximately circular in shape, centred on the smelter, and elongated somewhat towards the east due to the prevailing wind direction. Close to the smelter, the concentrations are up to 1000 times greater than regional background levels. The distances at which concentrations of smelter-related metals decrease to background levels is in the range 65 ± 5 km. The ratio of (Cu+As)/(Al+Fe) is a useful indicator of source apportionment for comparing the results between sampling media because it is independent of measurement units and maximizes the contrast between smelter and non-smelter materials. The value of the ratio, which changes as a function of distance from the smelter, reflects the mixing and dilution of emissions with non-smelter dust and aerosols. Deposition rates of metal calculated for snow and peat integrated over the area of the smelter-centred anomaly allow estimates to be made of the amounts of metal deposited from the smelter, after correcting for deposition of background metal. The proportions of reported metal emissions accounted for by deposition within 150 km of the smelter are: Cu, 25–50%; Pb, 10–20%; Zn, 25%; As, 5%; and Cd 20–25%. Metals in excess of these proportions are probably transported outside the immediate smelter footprint, but deposition of smelter-emitted metals becomes so small at these distances, and so mixed with airborne metals from other sources, that it is difficult to discriminate between smelter and non-smelter sources in samples taken on the ground.

2006 Special issue: Time dependent neural network models for detecting changes of state in complex processes: Applications in earth sciences and astronomy

A computational intelligence approach is used to explore the problem of detecting internal state changes in time dependent processes; described by heterogeneous, multivariate time series with imprecise data and missing values. Such processes are approximated by collections of time dependent non-linear autoregressive models represented by a special kind of neuro-fuzzy neural network. Grid and high throughput computing model mining procedures based on neuro-fuzzy networks and genetic algorithms, generate: (i) collections of models composed of sets of time lag terms from the time series, and (ii) prediction functions represented by neuro-fuzzy networks. The composition of the models and their prediction capabilities, allows the identification of changes in the internal structure of the process. These changes are associated with the alternation of steady and transient states, zones with abnormal behavior, instability, and other situations. This approach is general, and its sensitivity for detecting subtle changes of state is revealed by simulation experiments. Its potential in the study of complex processes in earth sciences and astrophysics is illustrated with applications using paleoclimate and solar data.