A spatial statistical approach for sedimentary gold exploration – a Portuguese case study

A spatial statistical approach for sedimentary gold exploration – a Portuguese case study Pierre Goovaerts1, Teresa Albuquerque2, Margarida Antunes2 1 BioMedware, Inc., 121 W. Washington St., 4th Floor-TBC, Ann Arbor, MI 48104 CIGAR – Geo-Environmental and Resources Research Center, FEUP, Oporto and Polytechnic Institute of Castelo Branco, 6001-909 Castelo Branco, Portugal

2

Abstract. This paper describes the mapping of gold content in the surroundings of abandoned gold mines located in central Portugal. In 1988, 376 samples were collected and analyzed for 22 elements. Gold (Au) was measured only inside the gold mines and its value was predicted at other locations using linear regression (R2=0.46) and four metals (Fe, As, Mn, and W) which are known to be mostly associated with the local gold's paragenesis. One hundred realizations of the spatial distribution of gold content were generated using sequential Gaussian simulation. Each simulated map then underwent a local cluster analysis to identify areas of significantly low or high values. The one hundred classified maps were processed to derive the most likely classification of each simulated node and the associated likelihood. The distribution of the hot-spots and cold-spots shows a clear enrichment in Au along the Erges River. Keywords: Gold prospection, Linear regression, Gaussian simulation, cluster analysis, Erges River.

1 Introduction In Portugal, ore extraction and processing has been an important economic activity and mines were actively developed until the early 1970s. The recent spike in gold prices and technologic development make extraction and processing more effective, leading to a regain interest in abandoned gold mining areas with a few experimental explorations in Alentejo, southern Portugal. Geochemical cartography is a goal in mining prospection since the late 1920’s. Recent development of analytical methods and computational resources facilitates the implementation of geochemical mapping and its use in natural resources management [1]. Geochemical modeling in environmental applications is mostly

oriented to the recognition and quantification of anthropogenic impacts. An accurate characterization of the natural background values is an essential and unavoidable step to evaluate the influence of mining activities on the environment. The aim of this manuscript is the development and application of a spatial statistical approach for sedimentary gold exploration in an old abandoned mining region – Monfortinho region. The analysis focuses on heavy metals (Fe, Ba, Cu, Cr, B, Zn, Sb, Pb, Sn, Ni, Mn, Be, Mo, Co, Y, Cd, Ag, V, As, W, Nb and U) and is based on a point support dataset of 376 stream sediment samples.

2 Methodology 2.1. Sampling Around mine tailing sites, the mineralogical content of the material exploited consists of inert materials from the gangue constituent’s mineralization or mineral constituents of rocks [2]. The presence of anomalies in some chemical elements in the surrounding areas of tailings or mineralized areas indicate the action of dominant wind and transport of fine dust from the superficial layers of the heap [3]. The stream sediments resulting from the alteration of rocks by various physical and chemical processes are mobilized, transported and deposited along the water lines. The geochemical composition of stream sediments and their spatial distribution in the study area, were characterized using a total of 376 representative samples, collected in a narrow region ranging from 50 m upstream to 100 m downstream the streams’ confluences [4] (Fig. 1). Almost all water lines correspond to open valleys, so our point-support stream sediments samples correspond to incipient and not evolved soils. All the samples were collected on schist and were prepared through reduction, drying and grinding. Total concentration of As, B, Cu, Ba, Pb, Zn, Ni, Sb, Mn, Be, Mo, V, Co, Y, Cd, Nb, Fe, Cr, Ag, Au, W and U were analyzed by ICPAES, with a precision of 20 ppm for As and 10 ppm for the other elements [4]. Tin and W were analyzed by X-ray fluorescence spectrometry and plasma emission with a precision of 10 ppm [4]. Gold was determined only for the 12 samples collected inside the old mine area. Its value was predicted at other locations using linear regression (R 2=0.46) and four metals (Fe, As, Mn and W) which are known to be mostly associated with the local gold's paragenesis. Fig. 1 shows the location of samples with the gold content that was either measured directly or estimated by regression. In the subsequent analysis, these values were assumed to be known with certainty. Indeed, the main objective of the present study is the visualization and delineation of potential zones of low and high values for future prospections instead of the accurate estimation of gold content.

Fig.1. Location map of sediment samples. The size of yellow dots is proportional to the gold concentration.

2.2 Geostatistical methodology The delineation of zones of high and low contents in gold was conducted through the application of local cluster analysis (LCA) [5]. The basic idea is to compute at each grid node a local indicator of spatial autocorrelation (LISA) and test whether this statistic is significantly positive, indicating the existence of an aggregate of grid nodes with similar gold content. Since the data are not gridded, the first step is the derivation of gold content at all grid nodes. Kriging is not recommended for this step because its smoothing effect will cause the detection of artificial clusters. Following Goovaerts [6], gold content was first simulated using sequential Gaussian simulation and each of the 100 realizations underwent a LCA to identify each grid node that belongs to a cluster of small gold content (low value surrounded by low values) or a cluster of large gold content (high value surrounded by high values). The one hundred classified maps were then processed to derive the most likely classification of each node and the associated likelihood (i.e. frequency of occurrence of that class). This approach has the advantage of incorporating the uncertainty attached to the gold map through the local cluster analysis.

3 Results and conclusions Fig. 2b shows the average of 100 simulated maps of gold content which indicates higher values along the Erges River and downstream the abandoned gold mines. Each simulated map underwent a LCA, leading to the allocation of some grid nodes

to clusters of low gold content (Low-low) or high gold content (High-high), whereas the LISA statistic was not significantly different from zero at other grid nodes. Fig. 2b shows the frequency of the most likely class that is displayed in Fig. 2c. Lower frequencies indicate zones of changes (i.e. boundaries) or transitions between classes of values. The location of hot and cold spots shows a clear Au enrichment along the Erges River downstream the old abandoned sedimentary mineralization.

Fig.2. a) Average simulated map of gold content; b) Likelihood map for the classification into clusters of low or high values; c) Classes with the highest frequency of occurrence. Acknowledgments We are grateful to Instituto Geológico e Mineiro (Portugal) for the data on sediments.

References 1. Antunes, I.M.H.R., Albuquerque, M.T.D.: Using indicator kriging for the evaluation of arsenic potential contamination in an abandoned mining area (Portugal). Sci. Total Envir. 442, 545-552 (2013). 2. Maroto, A.G., Navarrete, J., Jimenez, R.A.: Concentraciones de metales pesados en la vegetación autoctona desarrollada sobre los suelos del entorno de una mina abandonada. Bol. Geol. Minero 108-1, 67-74 (1997). 3. Santos Oliveira, J. M., Pedrosa, M.Y, Canto Machado, M.J, Rochas Silva, J.: Impacte ambiental provocado pela actividade mineira. Caracterização da situação junto da Mina de Jales, avaliação dos riscos e medidas de reabilitação. Actas do V Cong. Nac. Geol. 84/2, E74-E77 (1998). 4. Instituto Geológico e Mineiro: Reports from a Prospecting project for tungsten, tin and associated minerals from Góis-Segura. Metalic Mineral Prospecting Section. Porto, Portugal, 10 pp (1988). 5. Anselin, L.: Local indicators of spatial association-LISA. Geogr Anal, 27, 93-115 (1995) 6. Goovaerts, P.: Geostatistical analysis of disease data: visualization and propagation of spatial uncertainty in cancer mortality risk using Poisson kriging and p-field simulation. Int J of Health Geogr, 5:7 (2006)