Factors of Variation of Soil Chemical Properties in Metalliferous Ecosystems of Tenke-Fungurume, Katanga, Democratic Republic of the Congo

Our study aimed at deepen our understanding of relationships between soil properties and vegetation distribution in metalliferous ecosystems of Tenke-Fungurume in the Democratic Republic of Congo. The first question concerned the differences and similarities between soils of the main vegetation units and four variation factors of soil properties were summarized by multivariate analysis. They were all linked to lithology and significantly contributed to explain the distribution of vegetation units. Our result suggest that the variation of soil properties which is observed within the various vegetation units (rocky steppe savanna, sward, and steppe savannas on slope or on Dembo) should partially be attributed to differences of geochemical composition of rocks between sites but the main source of variability is to be found inside each hill. The soil contamination in Cu and Co originates from rock weathering and besides site effect and topographic distribution of the rocks, the variability of soil properties within one vegetation unit may be due to variability of soil parent material and not only to erosion. The second question dealt with the changes of soil properties at small distances. Metric variation was studied from transects between adjacent vegetation units. Our results showed that the abrupt changes of vegetation units which were clearly identified on the field were all truly explained by the variations of one or more properties linked to lithology. The key point being the Cu bioavailability.


Introduction
Metal-rich soils provide very restrictive habitats for plants due to phytotoxicity and resulting severe selection pressure (1). They can host a unique flora (12), such as copper flora from which plant species contribute highly to global biodiversity (20) and are priceless related to their properties (44,49). Primary calamine and serpentinic sites are other examples of sites on which metal-specific vegetation develops (23,50).
Soil enrichment in copper (Cu) and cobalt (Co) may result from natural anomalies or human activities (21). In soils, metals from natural origin are generally less mobile than anthropogenic one (16,21,37).
Factors of Variation of Soil Chemical Properties in Metalliferous Ecosystems ... Indeed, copper and cobalt are nutrients to living organisms when they are at low concentrations (26,38,47) and become toxic at high concentrations (11,47). Excess of Cu induces injuries to plants by generating oxidative stress and reactive oxygen species while Co adversely affects shoot growth and biomass (36). However, some plants are able to tolerate high concentrations of Cu and Co in soils (2, 6). Tolerance mechanisms to Cu and Co were found on some cuprophytes from Katanga (4,7,8,10,34,35,39).
In Katanga province, mineralized rocks rich in Cu and Co outcrop at the summit of hills which they protect against erosion. The concentrations of these two elements can reach up to a few tens of thousands of mg.kg -1 in soils (9,24,25,28). Saad et al. (43) reported total Cu concentrations between 100 and more than 35,000 mg/kg. On these metalliferous outcrops, grows an original flora composed of at least 600 species of plants, of which 33 were recognized as strictly endemic to this environment (13). These species are distributed in the landscape within plant communities, called further as vegetation units. Mineralized particles are redistributed along the slope by erosion.
These phenomena generate a gradient of Cu and Co concentrations in the topsoil that directly affects the distribution of native vegetation (9,28).
Mining activities lead to the destruction of the primary plant communities covering the outcrops and the surrounding soils and contribute to the total or partial loss of the species composing them. The protection of plant biodiversity in this specific context relies on ex-situ conservation of threatened species and requires knowledge of their biotic and abiotic requirements for growing (12,18).
The importance of soil properties to plant growth in reclaimed soils was reviewed by Sheoran et al. (45). Some key soil properties (acidity-basicity and redox potential) and soil constituents (clays, oxides and hydroxides of Fe, Mn and Al; carbonates, phosphates, organic matter) govern the behavior of trace elements in soils (21,29,42). The availability of nutrients as well as changes in soil physical properties can both contribute to the differential distribution of plants within ecosystems.
The change in plants communities in copper hills of Katanga was for a long time attributed only to Cu and Co in topsoils (5,9,28,30). However, recent studies suggested that this variation would be also explained by the combination of edaphic factors other than trace metals concentration, such as nutrient and water availability or physical constraints (12,19,43,44). This combination of factors constitutes an edaphic gradient which influences the vegetation structure and would be at the origin of ecosystem complexity (44) We intended to examine the diversity of edaphic conditions for given plant communities within and between sites in order to gain objective elements for restoration strategies. Especially, the missing scale of investigation in previous studies is the metric variation between two adjacent vegetation units. To achieve this, two questions were developed: 1. What are the differences and similarities between soils of the four main vegetation units encountered from top to bottom of the hills? 2. Are the scales of variation of soil properties and vegetation units congruent for small distances?

Study area
The study area is located in the region between the cities of Tenke The region hosts more than 40 copper outcrops (44). The geology is largely influenced by the RAT and the Mines Series, the latter being the most mineralized zone of the Roan Group (15,22).
The rocks within these series include, from youngest to oldest, calcareous rock with dark minerals (CMN), dolomitic shales and schists (SDS, SDB), cellular or foliated siliceous rocks (RSC, RSF), stratified dolomites (D-Strat), and talcose argillaceous rocks (RAT) (24,25,44). The siliceous rocks make up the backbone of the hilly landscape due to higher resistance to erosion processes.

Soil sampling
To answer the first question, fifty-seven samples of surface soils (0-10 cm) from the main vegetation units of Fungurume copper-cobalt deposits have been sampled within a list of 300 floristic 1-square meter observation plots in the 13 metalliferous hills of Tenke-Fungurume complex (18).  To answer the second question, five short-distance transects across neighbouring vegetation units as identified on the field have been sampled. This part of the study aimed to assess whether the physiognomic sudden change observed on vegetation was parallel to similar sudden changes of soil physicochemical properties (Table 1)

Soil analysis
All soil samples were dried at open air inside a room for 8 days and then passed through a 2 mm sieve. The pH was determined by mixing 2 g of soil with 50 ml of distilled water and/or 1 N KCl.
The mixture was stirred for 2 hours on a rotary device and centrifuged for 10 minutes at 3,000 rev/ min. Measurement was performed with a pH meter. Total organic carbon was measured by titration after wet oxidation with K 2 Cr 2 O 7 , according to the Walkley & Black method (48).
Available cations (K, Mg, Ca, P, Cu, Co, and Mn) were extracted with EDTA + CH 3 COONH 4 at pH 4.65 (27). Total concentrations of elements were obtained by a digestion of 0.5 mg of soil with a mixture of three acids namely: 2 ml HNO 3 + 1 ml HClO 4 + 5 ml HF according to AFNOR 1996:NF X31-147.
Total contents (Cu, Co, Al, Fe, Mn) were only measured on transect samples. The determination of Al, Fe and Mn aimed at characterizing the general soil properties, especially they express the mineralogical signature of rocks (15,22) and Al and Fe may be used as proxies of soil texture (29).
Soluble metals (Cu, Co) were obtained by extraction with 0.01 M CaCl 2 (17). This is considered as labile or mobile fraction (33,46) in soils. Measurement of total, available and soluble metals have been made by flame atomic absorption spectrometry (VARIAN model 220).

Statistics
Factorial analysis was performed from Principal Component Analysis (PCA) with varimax rotation in order to identify the underlying factors of variability among the studied soil properties. Analysis of variance was used in order to test the significance of "Site" and "Vegetation" factors in the first analysis, and of differences between vegetation units in the study of transects. Soil characteristics were transformed, except for pH, in order to approach normality and homoscedasticity. In the semivariance analysis, the "within-unit" estimates the variance with the closest point within the same unit while the "between-unit" relates to differences across the limit between two vegetation units. Semivariance was also compared to variance within vegetation units, which was estimated by the residual mean square in the ANOVA.
The software used for statistical analysis were Minitab 17 and R.   Figure 2). In particular, SHC, Fu1 and Fu3 show higher mean

Soil properties under main vegetation units of the studied hills
Cu content than the three other sites, while Fu9 is clearly the less contaminated of the study sites.
Regarding Co content, Fu3 and Fu8 show the highest levels and SHC and SHW the lowest, which means that mineralization of rocks with Cu and Co might have differed from one site to another.
Regarding vegetation units (Table 2, Figure 2), eight of eleven parameters considered showed a significant difference (p < 0.05). Among them, pH KCl , Mg, P, and Cu contents were the most significant (p < 0.001). The analysis of variance was performed on log10-transformed data excepted for TOC (square root) and pH (no transformation). Interactions between factors were not significant. Means that do not share a letter are significantly different after Tukey at 95%. The soil properties in metalliferous ecosystems of Katanga are usually significantly correlated and PCA analysis was used by several authors to identify edaphic factors (14,43,44). We performed a factorial analysis from a PCA with varimax rotation on soil chemical properties. The figure 3 shows the results of the PCA before rotation. Four factors were kept as they make up more than 85% of total variance. These factors should be identified as: 1. a Cu-contamination factor, 2. the richness in major nutrients, 3. a Co-contamination factor different from the first one, and finally 4. an acidification factor.
The first factor, not only reflects the direct effect of contamination in Cu due to mineralized rock but it also shows lithological origin of P and indirect effect on the accumulation of organic matter probably due to a decrease of biological activity and decomposition processes. The second factor is clearly under the influence of major nutrients, P excepted, and Mn. Soils downslope developped on RAT are clearly richer in these elements and lithology seems to be a predominant factor of spatial distribution, even if downward redistributions with soil water fluxes cannot be discarded at this stage. The factor 3 constitutes another factor linked to contamination by the parent material, which also indicates differences of rock elemental composition between sites. Finally, the fourth factor is driven by pH KCl , Ca and Mg content, which separates the rocky steppic savannas on siliceous rocks from the three other vegetation units, or SHW from the other hills, as they are more acidic.

Short-distance transitions between vegetation units
The chemical properties of soils sampled in the various transects are summarized in table 3. Each transect should be analyzed for itself first.
The transect Fu3T on the small flat summit of Fungurume 3 concerned the transition between a rocky steppic savanna and a natural sward with Xerophyta sp. (Table 1). Each vegetation unit is associated to a different rock outcrop, RSC and RSF respectively. The results (Table 3) show that average soil properties are clearly different between these two units as pH, TOC and every element content are higher on the RSF. Only the Co content difference, when expressed in log is at the limit of the significance (p = 0.051). The most significant differences between the two vegetation units are due to Cu content but at this stage none of the other elements/properties could be dismissed of being a factor of differenciation.
In the transect Fu5T1, across the slope of Fungurume 5, three vegetation units were sampled from the natural sward on RAT, contaminated by colluviating particles from the upslope RSF, to a steppic savanna and a grove with small trees of Uapaca robynsii. Both savanna and grove were on slopes over SDB shales. Excepted K and Fe contents, every soil properties showed significant differences beween at least two vegetation units. The C and I units were developped on the same type of rocks, that is SDB, and the B unit on RAT. The total Al content confirmed the influence of lithology on soil properties (Table 3) as Al in soil over RAT is almost 1/3 lower than over SDB. The B unit is clearly differing from the other units by chemical properties as TOC, P, Cu and Co were far higher than in the two other vegetation units (Figure 4). Regarding the differences between C and I units, it appeared that they were significant for pH, Mg and Ca higher in the I unit, as well as for Cu CaCl2 ,  (Table 3) (Table 3). Regarding pH, it is not possible to evaluate whether the differences are due to effect of fire but the observations are in contradiction with the usually admitted rise of pH after burning. Regarding the other chemical properties, they showed intermediate levels between sward and grove and the transition with sward appeared more abrupt than with the grove (Figure 4).   (Table 4), which expresses significant differences between sites.
Regarding transitions, semivariances between two neighbour vegetation units are 2 to 6 times higher than semivariances within vegetation units, to the exception of Co content for which both semivariances are similar. The less pronounced differences concern the total Al, Fe and Mn which reflect the nature of the soil parent material and variations occur mainly between siliceous rocks (RSC) and rocks with clay minerals (RSF, RAT, SDB, colluviums). The pH and nutrient status vary Tropicultura 2295-8010 Volume 37 (2019) Numéro 1, 250 strongly between two vegetation units, mainly between the Uapaca groves (I) and steppic savannas (C, D) and with less strength between swards (B, E) and neighbours. The most abrupt transitions between vegetation units is due to Cu content and organic matter and they concern dominantly the swards as can be seen by comparing the specific g-swards given at table 4.

Metric variations of soil chemical characteristics in transects
As a general rule, the swards (B, E) show higher levels of TOC, P, Cu and Co contents and lower levels in nutrients than the steppic savannas. At the opposite, the groves (I) are characterized by more favourable pH and nutrient conditions. The steppic savannas (A, C, D) present intermediate soil chemical properties. Regarding the transitions between the vegetation units, the transects perpendicular to the slopes show that they were abrupt between swards and steppic savannas and more gradual between the latter and the Uapaca groves. Soil properties can be affected by burning of vegetation but effects are not completely understood (40). We found no effect (Fu8T1) or lower pH in burned steppic savanna compared to unburned, which does not seem to be an expected result of burning (40). Moreover, it is not realistic that the burning of the vegetation could affect the soil iron content. Hence, we cannot consider that burning is a real factor of variation in the studied transects.
Significant differences of soil properties from successive vegetation units located are observed along the mini transects. The transitions are abrupt between swards and steppic savannas for TOC, Cu and nutrient (K, P, Mg) content (Table 4). However, among these elements only the Cu content appears to be a limitation factor for vegetation due to phytotoxicity. Nutrients are clearly linked to geochemical composition of soil parent material and swards present higher content in P due to presence of phosphates (pseudomalachite, Cu 5 (PO 4 ) 2 (OH) 4 ) in RSF and SDB (41). Higher TOC content can be associated to organic matter accumulation through reduced microbial activity or increased root development. Soil P and TOC levels cannot however be considered as limiting factors but as correlated variables. The transition between steppic savannas and Uapaca groves were gradual ( Figure 4). However, the only common factor between both studied transitions (Fu5T1 and Fu8T2) was the increase of pH and decrease of Cu CaCl2 from steppic savanna to grove. The levels of Mg and Ca also tend to be higher under the grove. At this point, we don't know if the pH and nutrient status are the result, the factor or only correlated variables of the vegetation differentiation but the reduction of toxicity seems to be a crucial factor (7,10,11,32,41,42,50). Most studies so far used total Cu or Cu EDTA to analyse soil-vegetation relationships in the metalliferousecosystems of Katanga. However, it seems from our results that the use of Cu CaCl2 might be better to discriminate vegetation units because it is linked to a potential reserve (total Cu) and effective conditions of solubility such as acido-basic status (32). Assessment of chemical fractionation by geochemical modelling is another alternative (41).

Conclusion
Our study aimed at deepen our understanding of relationships between soil properties and vegetation a strong selection pressure for plants due to phytotoxicity, other properties, such as topographic position, soil parent material, soil nutrient status, soil depth… also vary within the landscape.
Four factors of variation of soil properties were summarized by multivariate analysis, two are linked to Cu or Co contamination, one to nutrient status and one to pH KCl . The four of them can be linked to lithology and they contribute to explain a significant part of the distribution of vegetation units.
However, the residual variability of soil properties within each vegetation unit remains significant.
The lithological factor is important in hilly landscapes even under tropical climate because soils are rejuvenated by erosion processes. The distribution of swards and various steppic savannas in the landscape is clearly the result of an adaptation of species to the phytotoxic effect of metals originating from rocks. Our result suggest that the variation of soil properties which is observed within the various vegetation units should partially be attributed to differences of geochemical composition of rocks between sites for Cu and Co contents. These differences however do not concern the pH nor the nutrient status for which the main source of variability is to be found inside each metalliferous hill. The distribution of pH KCl and nutrients in the hill follows the mineralogical composition of rocks: acidic reaction and low nutrient content over siliceous rocks at the top, less acidic reaction and enrichment in P over mineralized outcrops, intermediate soil reaction, lower P content and higher K and Mg content over RAT. The soil contamination in Cu and Co also originates from rock weathering and we think that besides the above-mentioned site effect and topographic distribution of the rocks, the variability of soil properties within one vegetation unit may be due to spatial variability of soil parent material and not only due to erosion processes.
A deeper insight was put on the transition between vegetation units at metric scale, which had never been done so far in the copper ecosystems of Katanga. The abrupt changes of vegetation units which were clearly identified on the field were all truly explained by the variations of one or more properties linked to lithology. The key point seem to be the Cu-phytotoxicity which depends on total reserve in Cu and acidity level and was estimated by 0,01 N CaCl 2 extraction in our study.