Impact of nickel mining in New Caledonia assessed by compositional data analysis of lichens

Nickel exploitation began in the archipelago of New Caledonia (South Pacific Ocean), soon after the discovery by Jules Garnier in 1864 of nickel silicates, called garnierite. Ever since, the economic life of the island has beaten to the rhythm of nickel demand. Nickel mining is now becoming the largest employer in New Caledonia (more than 7500 employees in 2015). When exploitation first began, high-grade veins of garnierite (6–7% Ni), identifiable by their green colour, were extracted. Miners were soon interested by other lateritic horizons that developed on ultramafic weathered rocks. These nickel-enriched soils are located near the surface, so opencast mining was preferred. For such an approach, vegetation and top soil are first removed, and ore is extracted and transported by trucks on unpaved roads or by belt conveyors to loading zones or to ore treatment plants. All these operations are expected to release into the atmosphere huge amounts of dust enriched in trace metals (Chakraborty et al. 2002; Huertas et al. 2012), known to affect human health: it can lead to respiratory dysfunction, cardiovascular disease and cancer (International Agency for Research on Cancer 1990; Andersen et al. 1996; Harrison and Yin 2000; Menvielle et al. 2003). Nickel mining and ore treatments are known to have deleterious effects on the quality of the New Caledonian environment by degrading natural ecosystems (Bramwell 2011; Kettle et al. 2007; Losfeld et al. 2015), forests (Jaffré et al. 1977), and the lagoon (Gunkel-Grillon et al. 2014; Hédouin et al. 2007). Furthermore, the smallest dust particles can be spread over very long distances, far from emission sources. The impact of mining on the surroundings and the geographical dispersion of metals are often estimated using atmospheric filtration units covering the territory of interest. The filters are frequently changed because meteorological conditions (humidity, wind strength and direction variability, etc.) drastically affect the quality and quantity of dust material collected. This procedure is commonly used as a reference method, but remains costly and time consuming. Yet, at least for preliminary screening, some alternatives do exist. They are based on the biomonitoring capabilities of some organisms or organic material, such as tree rings, peat deposit, mosses or epiphytic lichens (Aslan et al. 2012; Balabanova et al. 2013a; Boamponsem et al. 2010; Mihaljevič et al. 2008; Monna et al. 2012; Richter et al. 2007; Spiro et al. 2004). Epiphytic lichens are symbiotic organisms, composed of fungi and algae. They have the ability to absorb and accumulate metals, through wet and dry deposition, without symptoms, at least up to a certain level (see the seminal work from Nylander 1866, and for reviews of recent studies, Conti and Cecchetti 2001; Szczepaniak and Biziuk 2003). Their nutrient uptake therefore relies exclusively on air constituents, since they have neither roots nor cuticles. Concerning absorption of metals, three main mechanisms have been invoked: intercellular absorption by an exchange process, intercellular accumulation, and entrapment of metal-rich particles (Richardson 1995; Szczepaniak and Biziuk 2003). The respective role of these processes is, however, not fully understood, all the more since metal contents in thalli seem to experience periods both of accumulation and release, related to several environmental factors. In any case, it has been demonstrated that lichens are good biomonitors of trace elements, as their thalli concentrations are correlated with those in their surrounding environment (Conti 2008). The sampling of these organisms is inexpensive and easy (when thalli are abundant), in any case free from heavy logistics. Furthermore, it can be problematic to compare concentrations from different individuals, even within the same species, because biomonitors may be affected by several internal factors, such as body morphology, age, and exposure (Carignan et al. 2002; Doucet and Carignan 2001; Kinalioglu et al. 2010; Senhou et al. 2002). That is why, instead of examining absolute concentration levels, some authors have proposed normalizing concentrations of potentially anthropogenic elements by a single chosen reference element, exclusively of crustal origin, such as titanium, aluminium or a lanthanide (i.e. all metal concentrations are divided, for instance, by those of Ti, the normalizing element used in the following, Goix et al. 2013; Monna et al. 2012). This can be viewed as the first step towards the so-called enrichment factor (EF) calculation, which aims at detecting, in combination with other tools, which elements are enriched in relation to local soils or earth crust (Agnan et al. 2013; Aničić et al. 2009; Aslan et al. 2012; Cloquet et al. 2015). The only difference is that there is no final normalization by local reference values or, more simply, by those of the upper continental crust (e.g. Wedepohl 1995).

Although convenient, and most of the time efficient (Bergamaschi et al. 2002; Nyarko et al. 2006), the normalization of concentrations of potentially anthropogenic elements by a single chosen reference element (e.g. Metals/Ti or Metals/Al), or possibly by the sum of reference elements (e.g. Metals/(Ti + Al)), presents an unavoidable drawback: the use of new ratio-based variables with the same element(s) as denominator is mathematically inadequate for all statistics based on correlations (see Pearson 1896; Kim 1999, for a complete discussion, and Monna et al. 2006 for a debatable use of multivariate statistics on such ratio data). The use of [metal/Ti] or [metal/(Ti + Al)] ratios are still mathematically correct when correlations are not computed (only sites are compared), but the geochemical signals carried by all other pairwise ratios (e.g. Ni/Cr, Co/Zn, etc.) are not really taken into account, although they may contain subtle information about the origin of the metals.

The compositional data are by nature closed because the sum of all components, including those not measured, is constrained (to 1, or to 100%). The covariance structure of such a dataset is necessarily biased (Filzmoser et al. 2009a, b), so that most multivariate techniques become doubtful without a proper transformation. Undeniable progress has, however, been made since the 1980s to open the dataset and to remove problems related to spurious inter-element correlations (Aitchison 1982; Van der Weijden 2002). The data are, with these new techniques, treated as a whole, allowing each inter-elemental ratio to be examined (Van den Boogaart and Tolosana-Delgado 2013).

In the present study, elemental concentrations in Co, Cr, Ni, Fe, Cu, Zn and Ti were analysed in more than 30 epiphytic lichens collected in New Caledonia, at various distances from modern mining activities, and from an ore treatment plant in Noumea, from within the city, and from areas free from any known historical mining. Ti is supposedly of crustal origin and used as reference, Cu and Zn have potentially an industrial/domestic origin, while Ni, Co, Cr and, to a lesser extent, Fe are especially targeted, because they are expected to be emitted by mining and metallurgical activity. The objective of this study is therefore to evaluate the potential of lichens as biomonitors of air quality in the context of Ni mining and ore treatment. Compositional data analyses were used to compute a synthetic index summarising the local degree of contamination related to mining and metallurgy.

Our results demonstrate that epiphytic lichens, at least those belonging to the Parmeliaceae family, can be used as biomonitors for air quality assessment in a Ni mining context. As expected, raw Ni, Co, Cr concentrations in lichen bodies are high, even extremely high, close to the mine and to the ore treatment plant. Concentrations decrease as distance from these infrastructures increases. In our case, the sample scores for the first principal component, computed after appropriate transformation, can be used as an overall environmental indicator. The PC1 includes all the main elements emitted into the atmosphere during Ni extraction and treatment. This approach may provide valuable information for the management and protection of ecosystems, highlighting those areas most contaminated by mining activity, and identifying the best sites for the installation of costly filtration units, for continued air quality monitoring. However, as the synthetic index drawn from PC1 drastically simplifies the problem by providing a broad picture, it may also preclude the differentiation of sources, when their compositional differences are subtle. Applying compositional data analyses in combination with close examination of raw data from lichens is therefore recommended, because contact with their absolute values is lost during the statistical analysis. Although the results of this study indicate that lichens can be used beneficially for assessing atmospheric dispersion of Ni in the surroundings of mining and metallurgical activities, further study in an area where air quality or atmospheric deposition data are also available would be useful to evaluate more precisely the performances of the approach described here.