**4. Discussion**

The pseudo-total PTE concentrations (Table 1) were generally higher than the average values for the world's soils: 58% of the samples presented pseudo-total As concentrations higher than the average concentrations for world's soils reported by Kabata-Pendias [27]; this percentage was also 58% for Cd, 33% for Co, 31% for Cr, 87% for Cu, 65% for Mn, 58% for Ni, 70% for Pb, 71% for Zn, and 84% for Hg. Furthermore, the average V concentration for world's soils was exceeded by all samples, while the average Ti concentration for world's soils was never exceeded. However, Ti and V pseudo-total concentrations were highly correlated.

Conversely, the measured pseudo-total metal concentrations were similar to those published by other authors for urban soils [6,11–15,28–30], except for vanadium, which presented very high concentrations in the soils studied.

The pseudo-total concentrations for most PTEs did not exceed the maximum concentrations allowed in agricultural soils. The maximum allowable concentrations (MAC) in agricultural soils reported by Kabata-Pendias [27] for Cr and Ni were exceeded by 12% and 22% of samples, respectively. As discussed below, these elements have a mainly lithogenic character, associated with ultramafic rocks. The MAC in agricultural soils for Cu, Pb, and Zn were exceeded by 4% (4 samples), 1% (one sample), and 3% (3 samples) of samples, respectively. The V concentrations were very high, always exceeding the MAC for agricultural soils. This element, as discussed below, can have a lithogenic character, although anthropogenic sources might contribute to increase its concentrations.

The measured pseudo-total Co, Cu, Ni, and Pb are in accordance with those reported by Diaz-Rizo et al. [31] for surface soils (0–10 cm) of Havana, measured by X-ray fluorescence. The Zn concentrations were mostly lower in the present study, rather closer to the values for nonurbanized areas in [31].

Cuban soils have naturally high concentrations of some heavy metals, exceeding usual quality standards [26]. This is true especially for Ni, Cr, Cu, and Co, abundant in ultramafic rocks, which are common in the island. Consequently, Rodriguez Alfaro et al. [26] established quality reference values (QRVs) for PTEs in Cuban soils, based in the analysis of representative natural soils (Table 8).

The pseudo-total concentrations of all the elements determined by Rodriguez Alfaro et al. [26] in Cuban natural soils correlated significantly with the concentration of the lithogenic element Fe. Similarly, in the present study this happens to be true for all the elements studied except Cd, Pb, Zn, and Hg, which is a clear indication of the human origin of these elements.


**Table 8.** Quality reference values (QRVs) established by Rodriguez Alfaro et al. (2015) for Cuban soils, enrichment factors (EF) and geoaccumulation indices (Igeo) calculated for the urban soils studied by comparison with these QRVs (median and interval in parentheses).

\* In the case of Ti, the average concentrations of Ti and Fe in the earth's crust were used as a reference.

These QRVs for Cuban soils were used as background concentrations to calculate the EF and Igeo for PTEs in urban soils.

The EFs for Ti were always lower than 1, indicating no enrichment, while the calculated Igeo were always negative, indicating the absence of pollution. Titanium is a lithogenic element, and a common constituent of several rocks [27].

Except for Ti, the lowest enrichment factors (Table 8) were calculated for As, Co, Cr, Mn, and Ni. These EFs were lower than 2 (minimum enrichment) for most samples, although there are a few samples with values between 2 and 5 (moderate enrichment) for Co, Cr, Mn, and Ni and 19% of samples (in all districts) with EFs between 2 and 5 for As. Co, Cr, Mn, and Ni are lithogenic elements, Co, Cr, and Ni being associated with ultramafic rocks and Mn with both acid and basic rocks [27]. However, soil enrichment can occur through anthropogenic emissions. The application to soils of composts and organic amendments can contribute with several PTEs, such as As, Co, Cr, Cu, Mn, Ni, Pb, and Zn [32].

According to the EF values, there was no or minimum enrichment in Co and Mn for most samples, with a few outliers showing moderate enrichment (Figure 3). There was minimum enrichment in As, Cd, Cr, Cu, and Ni for most samples, but outliers showed moderate (EF between 2 and 5) to significant (EF between 5 and 20) enrichment. These elements can be of lithological origin, but the presence of outliers suggest the contribution of anthropogenic sources in some samples. These possible anthropogenic sources include the traffic of vehicles [13], metal smelting, waste disposal and incineration, burning of fossil fuels [33,34], and the application of organic amendments to soils [35].

The enrichment factors were high or very high for Pb, V, Zn, and Hg (Table 8, Figure 3), which points to a human origin of these elements. This origin is unquestionable for Pb and Hg. The past use of leaded gasoline, Pb-containing paints, metal smelting, and fuel combustion, among other anthropogenic activities, can be sources of lead [27,36]. Mercury can originate from combustion of coal and oil, smelting industry, or waste incineration [27,36]. Zn can be lithogenic, but it can also be brought to the soil with organic amendments [32,35] or as a result of various human activities, such as metal smelting or traffic of vehicles. V can have a lithogenic character, mainly associated to ultramafic rocks, but its enrichment may be linked to anthropogenic activities, such as combustion of fuel oil, metal smelting, or oil refining [27,36]. There was significant enrichment in V, while for Pb, Zn, and Hg the median values denote moderate enrichment, but outliers reached significant enrichment in Zn and extremely high enrichment in Pb and Hg.

The values of the geoaccumulation index (Igeo) were always negative, indicating the absence of pollution, for As, Mn, and Ti, even in cases of moderate enrichment in As and Mn. For Co and Cu, the Igeo were negative for all but one (Co) or six (Cu) samples, which had Igeo values between 0 and 1 (unpolluted to moderately polluted). In the cases of Cd, Cr, and Ni, the Igeo values ranged from negative to 2.00, 1.84, and 1.87, respectively, indicating situations between practically unpolluted and moderately polluted. The Igeo values reached 2.49, 2.82, and 2.37 for Pb, V, and Zn, respectively, indicating some situations of moderate to strong pollution (2 < Igeo < 3). The worse situation is that of Hg, with 13 samples (14%) with Igeo values between 3 and 4 (strongly polluted) and 1 sample with Igeo = 4.28 (strongly to extremely polluted). In general, the interpretation of Igeo values is in accordance with that of EF values.

**Figure 3.** Enrichment factors for the studied elements relative to Cuban natural soils.

The first principal component in the PCA, accounting for 25% of the variance in the rotated matrix (Tables 3 and 4), groups the elements Ti, V, Fe, As, and Hg. Ti, V, and Fe, correlating positively to each other (Table 5), must be mostly lithogenic. Conversely, As and Hg, correlating negatively with Ti, V, and Fe, seem to be associated with soil organic fraction and have an anthropogenic origin. Both elements can originate from metal processing industry, combustion of fuel, waste incineration, and application of organic amendments to soil [27,36].

The vanadium enrichment could be attributed to smelting industries and to generators located in all the Havana districts, since V can be present as an impurity in fuels. The combustion of oil and its derivatives is considered the main source of vanadium contamination. V concentrations in crudes from Venezuela are particularly high, ranging from 282 to 1180 mg kg−<sup>1</sup> [37].

As referred to before, the titanium and vanadium concentrations showed a very high correlation (Table 5), which points to a common origin. Vanadium appears as an impurity in titanium minerals, such as ilmenite and rutile. Both elements are present in pyroxenes and amphiboles, being, therefore, relatively more abundant in basic and ultrabasic rocks [27]. Furthermore, both elements are used in the steel industry. Therefore, its origin seems to be mainly lithological, although there seems to be an anthropogenic enrichment of vanadium.

The second principal component includes the elements Ni, Cr, and Co and accounts for 22% of the variance in the rotated matrix (Tables 3 and 4). These elements, highly correlated to each other (Table 5), are abundant in ultramafic rocks. They are therefore considered as lithogenic. Anthropogenic activities (fuel combustion, smelting industry, and soil organic amendments) may contribute to increasing the concentrations of these elements in some samples. Accordingly, Díaz-Rizo et al. [38] report high bivariate correlations among Co, Cr, and Ni in soils of Las Tunas City (NE Cuba), which they interpret as a result of a common natural (nonanthropogenic) origin.

The third principal component, which explains 19% of the variance, includes Pb, Zn, Cd, and Cu (Tables 3 and 4). Cd, Pb, and Zn did not significantly correlate with Fe, which indicates a nonlithogenic origin. The association in the same PC points to an anthropogenic origin of the four elements. The higher concentrations of Zn and Cd (both pseudo-total and bioavailable) and bioavailable Cu in surface horizons (Table 6) are in accordance with this anthropogenic origin. The surface soil is usually more likely affected by anthropogenic sources than the subsurface horizons, closer to the parent material. However, urban soils in general, and the soils studied here in particular, are often allochthonous and, therefore, the enrichment of the surface soil in a certain element does not necessarily indicate a nonlithogenic character, since different horizons may have different origins. Our results are in accordance with those reported by Díaz-Rizo et al. [31], which, based on the bivariate correlations among Cu, Zn, and Pb, suggest a common origin for these three elements in Havana urban soils. However, by comparing soils from urban and nonurbanized areas, these authors conclude that Pb and Zn have a human origin, while Cu, the same as Co and Ni, may have a lithogenic character. Furthermore, for Las Tunas soils, Díaz-Rizo et al. [38] report a high correlation between Pb and Zn, which are considered anthropogenic elements. Because Cu is an abundant element in ultramafic rocks, its presence in Havana soils may have in part a geological origin, but the PCA points to an important anthropogenic contribution. The absence of significant correlations between Cu and any of Co, Cr, Ni, Pb, and Zn and their different spatial distribution in the aforementioned study from Las Tunas [38] would be in accordance with the double lithogenic and anthropogenic character of Cu.

Similarly to our results, in a study of urban soils of Hong Kong [16], Cd, Cu, Ni, Pb, and Zn were associated in the same principal component and considered anthropogenic. Massas et al. [11] and Gasparatos et al. [12] inferred from their data for urban soils of Greece a common anthropogenic origin for Zn, Cu and Pb. Also in Vigo (Spain), Pb, Cu and Zn were identified as anthropogenic [13]. Sun et al. [6] reported an association among the anthropogenic Cd, Cu, Pb and Zn in urban soils of Shenyang, China. In topsoils of urban parks in Beijing, China [33], Pb, Zn, and Cu were strongly correlated and were considered of anthropogenic origin, mainly related to vehicular traffic. In summary, profuse research shows that Cd, Cu, Pb, and Zn are ubiquitous in urban soils and arise from human activity.

The last component (PC4) explains 11% of the variance and includes only Mn (Tables 3 and 4). This is usually a lithogenic element. However, its isolation in a unique component and its high concentration in some samples (up to 1230 mg kg<sup>−</sup>1) could indicate a partially human origin. The major anthropogenic sources of Mn are municipal wastewaters, sewage sludge, coal combustion, and metal smelting processes [27,36].

Our results are broadly consistent with those of Manta et al. [14], which for urban soils of Palermo (Italy) conclude that Pb, Zn, Cu, Sb, and Hg can be considered tracers of anthropogenic pollution, whereas Mn, Ni, Co, Cr, V, and Cd were interpreted to be mainly inherited from parent materials.

Ti and V, having a common origin, showed the same spatial distribution (Figure 2). Similarly, Ni, Cr, and Co presented an analogous spatial distribution. On the contrary, the anthropogenic As and Hg showed different spatial distribution patterns. The spatial distribution of Mn was different from any other element. Cd, Cu, Fe, Pb, and Zn were not affected by the district; this is interpreted as a result of lithological origin for Fe and ubiquitous contamination with Cd, Cu, Pb, and Zn.

The high levels of Ti and V can be linked to the oil refinery Ñico López and the electric power plant Antonio Maceo in Regla and to the smelting industry FUNALCO in San Miguel del Padrón, as well as to various small sources like electricity generating facilities. The highest Hg concentrations in Habana Vieja, the historical centre of the city, are in accordance with reports of a relationship between the accumulation of Hg and the duration of urbanization [33]. High concentrations of Co, Cr, and Ni (Figure 2) can be related to lithological factors (ultramafic rocks). They could also be associated with the oil refinery Ñico López, in Regla, with the traffic of vehicles and with the application of organic amendments to soils. All the enriched elements can originate from local application of different organic amendments.

The high concentrations of PTEs, mainly Cr, Ni, and V, in some soils pose an environmental risk, particularly in soils for agricultural use. However, it should be noted that these elements are expected to be scarcely soluble in moderately alkaline soils, such as those in this study. Moreover, to the extent that they are lithogenic elements, they must be part of primary minerals and be released only upon weathering of these minerals. Chromium is solubilized above pH 8 as CrO4 <sup>2</sup><sup>−</sup> [27]. Therefore, an increase of soil pH would be risky, taking into account that Cr (VI) is carcinogenic to humans. The high content of organic matter in soil favors the reduction of Cr (VI) to Cr (III), less mobile and less toxic to human. Vanadium is found in neutral and alkaline soils in anionic forms, which are adsorbed by iron oxides or clay minerals through Fe cation bridges [27]. Nickel is rather mobile in horizons rich in organic matter, where it is solubilized in the form of chelates, although the complexation by organic ligands reduces the toxicity of Ni for soil organisms.

Pb, Zn, and Hg, having high enrichment factors and geoaccumulation indices, do not exceed the upper limits of the maximum allowable concentrations for agricultural soils (Table 1).

To evaluate the risk associated to the presence of PTEs in soils, it is necessary not only to know the total metal concentrations, but also the concentrations of mobile or bioavailable forms. The soil pH, the presence of carbonates, the organic matter content, the redox potential, and the presence of potential adsorbents, such as iron, aluminum, and manganese oxyhydroxides, are, among others, factors that influence the retention/mobility of PTEs.

The Mehlich 3 reagent extracts bioavailable PTEs [24,39,40]. The low proportions of Mehlich-3 extractable Ti (0.10%), V (0.07%), Fe (0.66%), As (1.02%), Hg (1.44%), and Cr (1.60%) indicate a low mobility and a low risk of exportation of these elements to food chain or water bodies. This is particularly relevant for vanadium and mercury: even though the pseudo-total V concentrations were very high, the bioavailable V concentrations (Table 1) were very low; even though the EFs of Hg were very high, the bioavailable Hg concentrations were also very low, suggesting that there is no risk of V or Hg being exported to vegetation or water bodies. This risk is also very low for Cr, whose pseudo-total concentrations were high in certain samples. According to Baken et al. [41], vanadium is strongly adsorbed onto poorly crystalline oxyhydroxides. This is in accordance with the fact that V extracted by acid oxalate (associated with poorly crystalline oxyhydroxides) is considerably higher than V extracted by Mehlich-3 reagent (bioavailable, Table 1). For soils derived from fluvioglacial sands in Poland, Jeske and Gworek [42] reported low mobilities of Cr, Ni, and V, the latter being the least mobile among them. According to Larsson et al. [43], V toxicity to plants and soil microorganisms is controlled by the vanadium sorption capacity of soils and, therefore, V concentration in the soil solution is a better predictor of V toxicity than total V in soil.

On the contrary, high proportions of metals in available form (observed for Mn, Cd, Pb, Zn, and Cu) indicate a recent enrichment, so that there has been no time for these metals to be immobilized in organic or inorganic forms [11,15]. In accordance with our results, Massas et al. [11] reported very low availability ratios (available metal/pseudo-total metal) for Cr and high availability ratios for Pb, Cu, and Zn in urban soils in Greece. Similarly, Pb was the most mobile and Cr the least mobile metal, among Pb, Zn, Cu, Ni, and Cr, in urban soils of the same area [12].

For Ni, whose pseudo-total concentrations exceeded the MAC for agricultural soils in 22% of samples, some values of the concentrations in bioavailable form are still high. The critical toxicity limit according to Monterroso et al. [39] (7 mg kg−<sup>1</sup> at pH 8) was exceeded by three soils of the district Regla, one of them being an agricultural soil. However, the determined Mehlich-3 extractable concentrations were all within the range reported by Caridad-Cancela et al. [44] for natural soils of Galicia, NW Spain (0.22–42.65 mg kg<sup>−</sup>1).

The toxicity limit for bioavailable Cd (0.25 mg kg−1) was exceeded in some horizon of five agricultural soils and three park or vacant soils. Nevertheless, it is worth mentioning that the determined concentrations were very similar to those reported by Caridad-Cancela et al. [44] for natural soils of Galicia, NW Spain (<0.01–0.41 mg kg−1). The toxicity limits for Mn (140 mg kg−1) and Pb (12 mg kg<sup>−</sup>1) were often (39% and 56% of samples, respectively) exceeded by soils of various uses. One sample had a concentration of bioavailable Cu above the toxicity limit (60 mg kg<sup>−</sup>1), and two samples had a concentration of bioavailable Zn above the toxicity limit (100 mg kg<sup>−</sup>1).

According to Rodriguez Alfaro et al. [45], high concentrations of PTEs can be present in Cuban agricultural soils as a result of the application of municipal-solid-waste compost, which contain elevated concentrations of Cd, Pb, Hg, As, Se, and Ni. According to these authors, as a result of their study, the use of municipal-solid-waste composts for food production has been forbidden by Cuban authorities. Upon discontinuation of the application of contaminated municipal-solid-waste compost, the PTEs will expectedly evolve into immobile forms.

The extraction with acid oxalate allows estimation of the elements occluded in noncrystalline iron oxides, which can be made available in the medium to long term [46], particularly if a change in soil conditions (e.g., pH and redox potential) favors the dissolution of oxides. This would be of particular concern for Hg, although the pseudo-total concentrations for this metal were always well below the reported upper limit of the MAC for agricultural soils.

The existence of gleyic properties had a significant influence on the concentrations of PTEs, both pseudo-total and available. The pseudo-total Fe, Co, Cr, and Ni concentrations were significantly (p < 0.05) lower in the soils with gleyic properties. Mobility and subsequent loss of iron are well-known facts in soils with reducing conditions. Co, Cr, and Ni can coprecipitate with iron or be adsorbed or occluded in iron oxyhydroxides, so that they are mobilized upon solubilization. It may also happen that these elements are mobilized in the form of complexes with soluble organic matter, more abundant under reducing conditions. Similarly, Shaheen et al. [47] reported a release of Cd, Cu, Co, Fe, Mn, Ni, and Zn under flood-dry cycles in a contaminated floodplain soil, related to variations of redox potential, pH, and dissolved organic carbon. In laboratory experiments, waterlogging was observed to affect the mobility of heavy metals (Cu, Pb, and Zn) retained or occluded in Fe or Mn compounds in urban soils [34,48].

The soils with gleyic properties showed lower concentrations (p < 0.05) of Mehlich-3 extractable Co, Ni, and Ti. In the cases of Co and Ni, this behavior was the same as that observed in pseudo-total concentrations. For these elements, pseudo-total and Mehlich-3 extractable concentrations were highly correlated. In the case of Ti, there were no significant differences in pseudo-total concentrations between gleyic and nongleyic soils. The extractable Ti/pseudo-total Ti ratio was significantly lower in soils with gleyic properties (p < 0.01). This appears to be a loss of Ti solubility under reducing conditions, which could be due to Ti precipitation as sulfide. This precipitation as sulfide may also occur for Co and Ni; in these cases, the precipitation, added to the decrease in total contents of these elements, could intensify the decline of available Co and Ni. Therefore, a change in the redox conditions of soils (for example, a variation of the water table) can result in mobilization or immobilization of some PTEs, increasing or decreasing the risk for these metals being exported to the food chain or water bodies. According to Ajmone-Marsan et al. [34], waterlogging of urban soils contaminated with PTEs may pose a serious environmental risk, particularly in the context of climate change.
