Multivariate Statistical Analyses and Potentially Toxic Elements Pollution Assessment of Pyroclastic Products from Mt. Etna, Sicily, Southern Italy

Abstract

Potentially toxic elements contamination represents a universal problem of major concern, due to several adverse health effects on human beings when permissible concentration levels are overcome. In this sense, the assessment of potentially risky elements content in different environmental matrices plays a key role in the safeguarding of the quality of the environment, and thus of the strictly correlated public health. In this article, measurements of the average potentially toxic elements concentrations in pyroclastic products from Mt. Etna, Eastern Sicily and Southern Italy were performed together with a comparison with the allowable levels set by Italian legislation, with the aim to evaluate the level of toxicity imposed on the ecosystem. For this purpose, Inductively-Coupled Plasma Mass Spectrometry (ICP-MS) measurements were performed to investigate any possible chemical pollution by potentially risky elements, by applying different indices such as Enrichment Factor (EF), Geo-accumulation Index (I_geo), Contamination Factor (CF) and Pollution Load Index (PLI). Finally, the multivariate statistical analysis was performed by processing potentially toxic elements content and pollution indices. It is worth noting that the used approach could be applied, in principle, for the evaluation of the chemical risk due to the presence of potentially toxic elements in a large variety of samples of particular environmental interest, and can constitute a guideline for investigations focused on the monitoring of the environmental quality.

1. Introduction

The rapid industrialization and the uncontrolled urbanization in many cities and coastal areas led to an alarming level of chemical contaminants around these environments [1,2]. Among these pollutants, heavy metals are of major concern, being characterized by a persistent and bio-accumulative character [3,4]. In order to address the extent of contamination, the knowledge of the potentially risky elements’ sources is of particular importance, together with the contamination mechanisms of systems where toxicity levels of concentration are reached [5,6,7].

The environmental pollution by potentially toxic elements (including heavy metals) is by now a universal problem, especially for the toxic effects on living organisms when permissible concentration levels are exceeded [4]. In fact, under such circumstances, a decline in the mental, cognitive and physical health of the individual could appear [8,9].

Moreover, the potentially toxic elements abundance in soils, waters and biota can be indicative of the presence of natural/anthropogenic sources, as reported in the literature [10,11]. In this sense, geochemical and mineralogical studies of soils can be, in particular, helpful in order to understand, on one side, the elemental distribution patterns and the environmental conditions existing in a specific area [12], and on the other side, the geological history of the transport and sorting process [13].

In the view of a sustainable development, pyroclastic products are considered as natural resources for the production of building materials [14,15]. In addition, in countries where active volcanoes exist, pyroclastic products could be employed to supply nutrients and reduce CO₂ from the atmosphere [16].

In the present study, Inductively-Coupled Plasma Mass Spectrometry (ICP-MS) was employed to measure the potentially toxic elements content, i.e., arsenic (As), cadmium (Cd), copper (Cu), mercury (Hg), nickel (Ni), lead (Pb), antimony (Sb), thallium (Tl), and zinc (Zn), of pyroclastic products sampled in a surrounding area of the Mt. Etna (Eastern Sicily, Southern Italy) [17,18,19]. With the aim of estimating the level of toxicity imposed on the ecosystem by the potentially risky elements, different pollution indices, such as Enrichment Factor (EF), Geo-accumulation Index (I_geo), Contamination Factor (CF) and Pollution Load Index (PLI) were also calculated. Finally, with the aim to analyze the chemical pollution, Pearson correlation, Principal Component Analysis (PCA), Hierarchical Cluster Analysis (HCA), and Multidimensional Data Analysis (MDA), were conducted with the aim of finding out any possible relationship among the variables [20,21].

2. Materials and Methods

2.1. Samples Collection

Table 1. The Identification (IDs) and the Global Positioning System (GPS) latitude and longitude coordinates of the investigated pyroclastic products sampling sites.

Site ID	GPS Coordinates
	Latitude	Longitude
1	37.68864	15.09499
2	37.64975	15.05663
3	37.71456	15.11615
4	37.68809	15.13123
5	37.61574	15.09846
6	37.57984	15.09326
7	37.61873	15.05600
8	37.69899	15.00120
9	37.65249	15.09540
10	37.69900	15.05389

reports the Identification (IDs) and the Global Positioning System (GPS) coordinates of the sampling sites (fifty samples of pyroclastic products in total, five for each site).

Figure 1 shows the map of the sampling area, with the site IDs (1–10) indicated.

Pyroclastic products were collected according to [22]. In particular, in the ID 1, 4, 5, 7 and 9 sites, freshly erupted air-fall pyroclastic products were sampled, while in the remaining ID 2, 3, 6, 8 and 10 sites, samples from heaps of pyroclastic wastes from previous volcanic activities were collected [22]. The samples were stored into labeled plastic vials, with adequate caution taken in order to prevent their contamination.

2.2. Inductively Coupled Plasma Mass Spectrometry (ICP-MS) Measurements

For the ICP-MS analysis, approximately 0.5–1.0 g of sample, together with 3 mL of ultrapure (for trace analysis) HNO₃ (67–69%) and 9 mL of ultrapure (for trace analysis) HCl (32–35%) (aqua regia), were directly introduced into a 100 mL TFM vessel, according to [23].

For the measurements a Thermo Scientific (Waltham, MA, USA) iCAP Qc ICP-MS was used [24,25], according to [26]. The sample introduction system consisted of a Peltier cooled (3 °C), baffled cyclonic spray chamber, PFA nebulizer and quartz torch with a 2.5 mm i.e., removable quartz injector. The instrument was operated in a single collision cell mode, with kinetic energy discrimination (KED), using pure Helium as the collision gas. All samples were presented for analysis using a Cetac ASX-520 (Teledyne Cetac Technologies, Omaha, NE, USA) and for each one, data were recorded in duplicate.

The quality of the ICP-MS experimental results was certified by the Italian Accreditation Body (ACCREDIA) [27]. This implies the continued verification (with annual periodicity) of the maintenance of the ICP-MS method performance characteristics.

2.3. Level of Potentially Toxic Elements Pollution Assessment

The level of potentially toxic elements pollution was evaluated by calculating the pollution indices reported in the following.

2.3.1. The Enrichment Factor

The Enrichment Factor is

$$\mathrm{EF}=\frac{\left\{{\mathrm{C}}_{\mathrm{x}}/{\mathrm{C}}_{\mathrm{Fe}}\right\}\mathrm{sample}}{\left\{{\mathrm{C}}_{\mathrm{x}}/{\mathrm{C}}_{\mathrm{Fe}}\right\}\mathrm{reference}}$$

(1)

where C_x is the concentration of the potentially enrichment element and C_Fe is the concentration of the normalizing element, usually Fe [28]. Considering that the regional geochemical background values for these elements are not available, the world average elemental concentrations reported by [28] in the Earth’s crust were used as reference.

2.3.2. The Geo-Accumulation Index

The Geo-accumulation Index is

$${\mathrm{I}}_{\mathrm{geo}}={\mathrm{Log}}_2\left[{\mathrm{C}}_{\mathrm{n}}/\left({\mathrm{kB}}_{\mathrm{n}}\right)\right]$$

(2)

where C_n is the concentration of the potentially hazardous trace element in the sample, B_n is the geochemical background value in average shale of the element n, and k is the background matrix correction factor [28,29].

2.3.3. The Contamination Factor

The Contamination Factor is

$$\mathrm{CF}={\mathrm{C}}_{\mathrm{metal}}/{\mathrm{C}}_{\mathrm{background}}$$

(3)

where C_metal and C_background are the concentration and the background values for each heavy metal, respectively [30].

2.3.4. The Pollution Load Index

The Pollution Load Index is defined as the n-th root of the product of the Contamination Factor of potentially risky elements

$$\mathrm{PLI}={\left({\mathrm{CF}}_1\times {\mathrm{CF}}_2\times {\mathrm{CF}}_3\times \dots \times {\mathrm{CF}}_{\mathrm{n}}\right)}^{\raisebox{1ex}{$1$}\!\left/ \!\raisebox{-1ex}{$\mathrm{n}$}\right.}$$

(4)

where n is the number of metals [31,32].

2.4. Statistical Analysis

The XLSTAT statistical software for Windows was used for all statistical calculations [33].

With the aim to individuate the presence of the relationships among the original variables (Pearson correlation analysis), an exploratory method (PCA) was performed. The PCA elaboration ensures the reduction of the data dimensionality, whereas the combinations of variables identified by the Principal Components (PCs) provide the greatest contribution to sample variability [34]. Moreover, in order to add a further degree of detail to the implemented statistical approaches, the Hierarchical Clusters Analysis (HCA) [35] and a new innovative statistical technique called Multidimensional Data Analysis (MDA) [36], which merges the calculation algorithms of PCA with those of HCA, were also employed.

3. Results and Discussion

3.1. Potentially Toxic Elements Analysis

Table 2. Potentially toxic elements content (mg kg⁻¹ dry weight, d.w.), as obtained through ICP-MS analysis for the analyzed samples.

ICP-MS Analysis
	Site ID
	1	2	3	4	5	6	7	8	9	10	Threshold Limit
CAs	1.73	1.74	1.72	2.09	1.85	2.11	1.98	2.32	1.97	2.14	20
CCd	0.10	0.15	0.10	0.11	0.10	0.13	0.10	0.11	0.12	0.12	2
CCu	76.91	76.83	76.04	74.02	82.61	87.23	82.45	85.23	85.08	92.91	120
CHg	0.03	0.03	0.03	0.03	0.03	0.03	0.03	0.03	0.04	0.04	1
CNi	16.61	14.95	14.32	15.06	15.91	17.45	17.81	21.95	18.82	18.32	120
CPb	4.93	5.02	4.67	5.17	5.18	5.55	5.23	5.16	5.02	5.46	100
CSb	0.08	0.08	0.08	0.15	0.09	0.12	0.09	0.11	0.35	0.10	10
CTl	0.11	0.09	0.08	0.07	0.08	0.09	0.09	0.08	0.09	0.09	1
CZn	62.15	58.45	56.56	58.41	62.32	69.32	63.39	80.23	68.65	69.05	150

reports the potentially toxic elements content (mg kg⁻¹ dry weight, d.w.) for the analyzed pyroclastic products, as obtained through ICP-MS analysis.

Worth noting is that the obtained concentrations are lower than the threshold limits [37], also reported in Table 2, for all detected metals. Therefore, they cannot be considered as pollutants, and thus do not compromise the welfare of the environment. For this reason, they do not constitute a risk to human health.

3.2. Estimation of the Level of Potentially Toxic Elements Pollution

3.2.1. EF

Table 3. Calculated values of the Enrichment Factor (EF), Geo-accumulation Index (Igeo), Contamination Factor (CF) and Pollution Load Index (PLI) for all the investigated sites.

Site ID	Metal	Index of contamination				Site ID	Metal	Index of contamination				Site ID	Metal	Index of contamination
		EF	Igeo	CF	PLI			EF	Igeo	CF	PLI			EF	Igeo	CF	PLI
1	As	0.14	−3.49	0.13	0.21	2	As	0.15	−3.49	0.13	0.22	3	As	0.15	−3.50	0.13	0.20
	Cd	0.34	−2.19	0.33			Cd	0.54	−1.58	0.50			Cd	0.37	−2.17	0.33
	Cu	1.80	0.19	1.71			Cu	1.85	0.19	1.71			Cu	1.89	0.17	1.69
	Hg	0.07	−4.48	0.07			Hg	0.07	−4.48	0.07			Hg	0.08	−4.48	0.07
	Ni	0.26	−2.62	0.24			Ni	0.24	−2.78	0.22			Ni	0.24	−2.83	0.21
	Pb	0.26	−2.61	0.25			Pb	0.27	−2.58	0.25			Pb	0.26	−2.68	0.23
	Sb	0.06	−4.79	0.05			Sb	0.06	−4.81	0.05			Sb	0.06	−4.81	0.05
	Tl	0.08	−4.25	0.08			Tl	0.07	−4.54	0.06			Tl	0.06	−4.71	0.06
	Zn	0.69	−1.20	0.65			Zn	0.67	−1.29	0.61			Zn	0.67	−1.33	0.59
Site ID	Metal	Index of contamination				Site ID	Metal	Index of contamination				Site ID	Metal	Index of contamination
		EF	Igeo	CF	PLI			EF	Igeo	CF	PLI			EF	Igeo	CF	PLI
4	As	0.19	−3.22	0.16	0.22	5	As	0.15	−3.40	0.14	0.21	6	As	0.16	−3.21	0.16	0.24
	Cd	0.42	−2.03	0.37			Cd	0.36	−2.17	0.33			Cd	0.43	−1.79	0.43
	Cu	1.89	0.13	1.64			Cu	1.97	0.29	1.84			Cu	1.90	0.37	1.94
	Hg	0.08	−4.48	0.07			Hg	0.07	−4.48	0.07			Hg	0.07	−4.48	0.07
	Ni	0.25	−2.77	0.22			Ni	0.25	−2.68	0.23			Ni	0.25	−2.55	0.26
	Pb	0.30	−2.54	0.26			Pb	0.28	−2.53	0.26			Pb	0.27	−2.43	0.28
	Sb	0.12	−3.91	0.10			Sb	0.06	−4.64	0.06			Sb	0.08	−4.23	0.08
	Tl	0.06	−4.91	0.05			Tl	0.06	−4.71	0.06			Tl	0.06	−4.54	0.06
	Zn	0.71	−1.29	0.61			Zn	0.70	−1.19	0.66			Zn	0.72	−1.04	0.73
Site ID	Metal	Index of contamination				Site ID	Metal	Index of contamination				Site ID	Metal	Index of contamination
		EF	Igeo	CF	PLI			EF	Igeo	CF	PLI			EF	Igeo	CF	PLI
7	As	0.16	−3.30	0.15	0.22	8	As	0.18	−3.07	0.18	0.24	9	As	0.16	−3.31	0.15	0.28
	Cd	0.35	−2.17	0.33			Cd	0.36	−2.03	0.37			Cd	0.42	−1.91	0.40
	Cu	1.94	0.29	1.83			Cu	1.86	0.34	1.89			Cu	1.98	0.33	1.89
	Hg	0.07	−4.48	0.07			Hg	0.07	−4.48	0.07			Hg	0.11	−3.91	0.10
	Ni	0.28	−2.52	0.26			Ni	0.32	−2.22	0.32			Ni	0.29	−2.44	0.28
	Pb	0.28	−2.52	0.26			Pb	0.25	−2.54	0.26			Pb	0.27	−2.58	0.25
	Sb	0.06	−4.64	0.06			Sb	0.07	−4.35	0.07			Sb	0.25	−2.68	0.23
	Tl	0.07	−4.54	0.06			Tl	0.06	−4.71	0.06			Tl	0.07	−4.54	0.06
	Zn	0.71	−1.17	0.67			Zn	0.83	−0.83	0.84			Zn	0.76	−1.05	0.72
						Site ID	Metal	Index of contamination
								EF	Igeo	CF	PLI
						10	As	0.17	−3.19	0.16	0.25
							Cd	0.40	−1.91	0.40
							Cu	2.00	0.46	2.06
							Hg	0.10	−3.91	0.10
							Ni	0.27	−2.48	0.27
							Pb	0.27	−2.46	0.27
							Sb	0.07	−4.49	0.07
							Tl	0.06	−4.54	0.06
							Zn	0.73	−1.05	0.73

reports the calculated EF values.

According to the literature [38], EF values < 2 indicate deficient to minimal enrichment. In particular, values between 0.5 and 1.5 indicate an entirely crustal material/natural origin metal, while EF > 1.5 suggests a more likely anthropogenic one [39]. Moreover, 2 ≤ EF < 5 means moderate enrichment; 5 ≤ EF < 20 significant enrichment; 20 ≤ EF ≤ 40 high enrichment and EF > 40 extremely high enrichment.

In our case, the obtained EF values are lower than 2 for all investigated sites, indicating no or minimal enrichment.

3.2.2. I_geo

Obtained I_geo values are presented in Table 3.

According to previous literature [40], the I_geo values were interpreted as follows:

I_geo ≤ 0 indicates no contamination; 0 < I_geo ≤ 1 no/moderate contamination; 1 < I_geo ≤ 2 moderate contamination; 2 < I_geo ≤ 3 moderate/strong contamination; 3 < I_geo ≤ 4 strong contamination; 4 < I_geo ≤ 5 strong/extreme contamination; and I_geo > 5 extreme contamination.

In our case, all values were found to be lower than 0, with the only exception of copper (Cu), indicating a very moderate contamination by Cu for all the investigated samples in all sampling sites. This occurrence can be justified, in agreement with [41,42], considering that this metal is typically employed as a soil defense product, the soil texture and high pH—all factors that can contribute to Cu accumulation.

3.2.3. CF

Table 3 reports the assessed CF values.

According to [43], CF ≤ 1 implies no contamination, 1 < CF ≤ 3 low or moderate contamination, 3 < CF ≤ 6 high contamination and CF > 6 very high contamination.

In our case, all values are <1 in all cases except of Cu, once again indicating a moderate contamination in all the sampling sites.

3.2.4. PLI

PLI values are presented in Table 3.

According to [44], the PLI value > 1 indicates the presence of pollution, whereas PLI value < 1 indicates no pollution. PLI turned out to be lower than 1 in all cases, thus revealing no pollution by the assessed potentially risky elements.

3.3. Statistical Features

The sphericity (or Bartlett) test was preliminarily conducted with the aim of making multivariate treatments more reliable and statistically significant, furnishing a p-value equal to 0.04. Thus, considering that a high statistical significance of the data is ensured for p-values < 0.05, it follows that starting from reliable data, the results of the calculation algorithms will also be trustworthy [45]. Furthermore, in order to understand any possible relationship between the different variables, expressed as concentrations of the detected potentially toxic elements, the Pearson correlation matrix was calculated.

The results of the Pearson’s correlation coefficients are presented in

Table 4. Pearson correlation matrix among the considered variables. Values in bold are different from 0 at the significance level α = 0.05.

Variables	CAs	CCd	CCu	CHg	CNi	CPb	CSb	CTl	CZn
CAs	1.00	0.03	0.61	0.43	0.76	0.67	0.19	−0.15	0.80
CCd	0.03	1.00	0.15	0.13	−0.05	0.24	0.16	−0.04	0.04
CCu	0.61	0.15	1.00	0.68	0.68	0.71	0.18	0.27	0.72
CHg	0.43	0.13	0.68	1.00	0.44	0.30	0.55	0.01	0.45
CNi	0.76	−0.05	0.68	0.44	1.00	0.39	0.28	0.23	0.97
CPb	0.67	0.24	0.71	0.30	0.39	1.00	−0.03	0.16	0.48
CSb	0.19	0.16	0.18	0.55	0.28	−0.03	1.00	−0.08	0.22
CTl	−0.15	−0.04	0.27	0.01	0.23	0.16	−0.08	1.00	0.18
CZn	0.80	0.04	0.72	0.45	0.97	0.48	0.22	0.18	1.00

.

All positive correlations were detected, with the exception of C_Sb-C_Tl (−0.08), C_Cd-C_Ni (−0.05), C_Cd-C_Tl (−0.04), C_Pb-C_Sb (−0.03).

As far as the PCA algorithm is concerned, ten variables (Site IDs, C_As, C_cd, C_Cu, C_Hg, C_Ni, C_Pb, C_Sb, C_Tl and C_Zn) were processed, also performing the Varimax rotation [46]. The obtained results are reported in Figure 2.

As can be seen from a first inspection of the figure, the PCA and Pearson matrix results appear to be coherent with each other. In particular, PCA elaboration allowed us to clearly discriminate three groups of samples. The first is formed by pyroclastic products from ID 6, 8 and 10 sites (yellow area in Figure 2), and it is clearly separated from the others on the first main component (PC1). The second group contains pyroclastic products from ID 1, 4, 5, 7 and 9 sites (salmon area in Figure 2) whose variance is more justified by the PC2. The last ideal group seems to be formed by pyroclastic wastes from ID 2 and 3 sites (light-blue area in Figure 2).

It is worth noting that the difference in the behavior of samples of the same type, i.e., pyroclastic wastes from previous volcanic activities, may depend on the different residence time of the pyroclastic products in the soil, and on the possible differential pollution that the pyroclastic waste samples have undergone [47].

Figure 3 shows the outcome dendrogram of the HCA. The dotted line represents the cut automatically made by the calculation algorithm, which allows us to identify the formation of three clusters (C1, C2 and C3).

The C1 cluster (green color in Figure 3) is made up of 60% of the samples (ID 1, 2, 3, 4, 5 and 7 sites), while 30% of the pyroclastic products fall within the C2 cluster (blue color in Figure 3) and come from the ID 6, 9 and 10 sites. The last cluster (C3) is instead formed by samples from the ID 8 site.

Noteworthy, the similar behavior of samples collected as “fresh eruption” and “pyroclastic wastes” could be probably due to the fact that pyroclastic wastes have not—or have negligibly—undergone contamination processes.

With reference to the MDA, the outcome graph is shown in Figure 4. The graphical interface consists of the results previously obtained from the PCA, with the superimposition of a contour map comprising different colors in relation to the degree of similarity that the analyzed samples highlight. The interpretation of the results is made clearer by the presence of circular, elliptical or square vectors capable of explaining the differences observed in multidimensional space. Obviously, a greater degree of difference observed between the considered samples will correspond to a lower degree of similarity [48].

The results of the MDA approach confirm what was suggested by the previous analyses (PCA and HCA). In fact, it is possible to identify three groups, which respectively include:

• Samples from the ID 8 and 10 sites, with a degree of similarity equal to 100%, and samples from the ID 6 and 9 sites, with a degree of similarity equal to 78%. These pyroclastic products can be ideally grouped within the same cluster. However, while samples from the ID 6, 8 and 10 sites are pyroclastic wastes, those from the ID 9 site were collected as a fresh eruption sample;
• All other samples are placed in the quadrants on the left side of the MDA graph. In particular, samples from ID 7 site have a degree of similarity of 56% (green color), and belong to the category of freshly erupted air-fall pyroclastic products;
• The remaining samples show heterogeneous degrees of similarity, but commonly less than 40% for all of them. This last hypothetical cluster seems to consist of the remaining fresh eruption samples (ID 1, 4 and 5 sites) and the pyroclastic waste samples (ID 2 and 3 sites), which probably have undergone pollution phenomena capable of determining potentially toxic elements concentrations similar to those shown by the fresh eruption pyroclastic samples.

4. Conclusions

The concentration levels of potentially toxic elements in pyroclastic products from Mt. Etna, Sicily, Southern Italy, were investigated through ICP-MS. The obtained values were found to be lower than the threshold limits set by the Italian legislation, and for this reason they do not constitute a risk to human health.

Moreover, the calculation of different pollution indices was performed to evaluate the ecological risk imposed on the ecosystem by potentially risky elements. In particular, EF values indicate no or minimal enrichment in all cases, suggesting an anthropogenic origin for Cu only. Furthermore, the obtained I_geo and CF data indicate a moderate Cu contamination for all the investigated samples, probably due to the use of this metal as a soil defense product, to the soil texture and high pH. Finally, the PLI values indicated no pollution by the assessed potentially risky elements.

Statistical analyses, i.e., Pearson correlation, PCA, HCA and MDA, were also performed by processing measured potentially toxic elements contents with the aim to determine their correlation with the sampling locations. In detail, the calculated PCA biplot identifies three different clusters, into which the analyzed pyroclastic products can be grouped. Furthermore, results of HCA and MDA turned out to be in good agreement with those produced by the PCA with a high degree of accuracy, validating the use of such statical approaches for the analysis of the relationship between chemical contaminants and sampling sites.

Finally, the approach reported in this paper could be applied, in principle, for the evaluation of the chemical risk due to the presence of potentially toxic elements in a large variety of environmental samples, by constituting a guideline for investigations focused on the monitoring of the environmental quality.