While SiO₂ and K₂O contents in Allanche bedrock agree with the range of basanite and feldspathic basalt (De Goër De Hervé and Tempier, 1988; Nehlig et al., 2001), sediments and soils show a depletion in both SiO₂ and K₂O as weathering processes alter the primary phases of the bedrock, the lowest contents being observed in soil samples (Négrel and Deschamps, 1996). One way to understand the chemical compositions of riverbed sediment and soil developed over the same parent rocks is by normalizing each chemical species X by the average value of the same species in the parent rock (Cullers et al., 1988; Dennen and Anderson, 1962). Some of the chemical elements found in suspended particulate matter, riverbed sediments and soil, like Zr, Ti or Th, can be considered as non-reactive species in the weathering process (Cullers et al., 1988). Zirconium displays mostly insoluble behavior in hydrosystems and, in sediments, is largely concentrated in detrital minerals (Aiuppa et al., 2000). In order to evaluate the extent of chemical mobility during bedrock weathering, the percentage change relative to Zr (Nesbitt, 1979) was applied to lead following Eq. (1):

The mean values of Pb and Zr for the bedrock are respectively 35 and 309 μg·g⁻¹. Results of the percentage change for Pb are expressed vs. the SiO₂ content on Fig. 2. Depletion reaches −46% in sediments and −54% in soil. In soil, the depletion is consistently negative, except for sample A10-3/2, with a positive value of over 20%. In sediment, the samples from the headwaters of the main river and from its tributaries have negative values, and the depletion is less marked for the rest of the samples collected downstream along the Allanche River with the percentage changes being positive. This Pb-enrichment, with values between 40 and 80%, might correspond to an external Pb source.

Fig. 3 illustrates the relationships between the 1000Pb/Zr vs. K/Zr ratios for bedrock, river sediment and soil. This graph combines element ratios used when studying weathering profiles or weathering suites of materials rather than contents alone, with the approach using an immobile element (Zr) as the reference. We used also K, based on its solubility and its presence in one of the two phases in basalt (feldspar phenocrysts or in the ground mass). We postulated that K is mobile during weathering as the percentage changes relative to Zr for K according to Eq. (1) ranged between 26 and 79% in sediment and between −30 and −60% in soil.

Bedrock constitutes an extreme set of ratios and the two relationships defined between the 1000Pb/Zr and K/Zr ratios can be interpreted in two different ways. First, the linear relationship with a large variation in K/Zr ratios and a lower one for 1000Pb/Zr ratios may reflect the evolution between two extreme end-members, one being fresh basalt (the unaltered end-member) and the other clay and altered phases, whereby the position of each sample along these linear relationships reflects its degree of weathering (Négrel and Deschamps, 1996). However, the deviation of the data points from this alteration line must be related to lead enrichment in sediments and soil. This agrees well with the calculation of the percentage changes showing lead enrichment for some samples (Fig. 2). As for lake sediments where lead enrichment was detected (Renberg et al., 1994), atmospheric deposition was suspected to be the major contributor. This will be explored below through lead isotope fingerprinting.

The lead content, the ²⁰⁶Pb/²⁰⁴Pb ratios, the 1000Pb/Zr ratios and the Zn content (Fig. 4a–d) in bedrock, soil and sediment are reported versus the distance from the river source. Lead content (Fig. 4a), is depleted compared to bedrock in sample A1 collected 6 km downstream from the river source, reflecting the effect of weathering processes. Therefore, two types of behavior can exist, with the lead content in soil and sediment:

• • having a similar range to that of bedrock;
• • having a clearly higher range than that of bedrock.

The first sample with a lead content higher than that of bedrock was A4, 12 km downstream from the source, and increasing lead contents were found in samples A8, 9 and 10. Regarding the evolution of lead content in the mainstream sediment, a log-relationship better fits with an equation Pb = 20.63 × ln (D)–11.75 (R² = 0.87, n = 8). A similar plot for the ²⁰⁶Pb/²⁰⁴Pb ratios vs. distance is given on Fig. 4b. Except one sample (A2, 7.5 km downstream from the source) with a ²⁰⁶Pb/²⁰⁴Pb ratio in agreement with those of the bedrock (as well as for Pb concentration), all other soil and sediment samples had ²⁰⁶Pb/²⁰⁴Pb ratios clearly lower than those of bedrock. As for the lead content, the plot of the 1000Pb/Zr ratios versus the distance from the river source (Fig. 4c), reflecting again the role of weathering processes with values lower than bedrock ones in sample A1–A2 and then increasing downstream. Regarding the evolution of this ratio in the mainstream sediment, a log-relationship better fits with an equation 1000Pb/Zr 1000Pb/Zr = 53.28 × ln (D)–18.57 (R² = 0.71, n = 8) that reflects lead addition. A plot displaying the Zn content vs. the distance is given on Fig. 4d. The zinc content is enriched compared to bedrock in all sediment samples. There is no correlation between the Zn content and the distance, as illustrated by the R² value ∼ 0.09 for a log-relationship. Coupling the downstream increase of lead content in sediment with the lowering of lead isotope ratios may help constraining the additional end-members represented by the anthropogenic input, as illustrated in the graph. However, the multiplicity of anthropogenic sources (gasoline, industry) with low lead isotope ratios precludes any further interpretation of this graph.

Fig. 6c is a plot of the 1000Pb/Zr ratios vs. the distance and will be explored further.

The relative importance of sediment sources in catchment for fluvial sediment supply depends on whether these sources are connected with the channel. Large catchments have more opportunities than smaller catchments for sediment storage, because of floodplains that act as a buffer between hillslope sediment sources and the river, and as storage areas for sediment deposited by overbank flow, thereby reducing the volume of sediment reaching the catchment outlet (Bouchez et al., 2012). Because small catchments generally have non-existent floodplains, hillslope sources have a greater connectivity with the river and overbank, and are therefore more susceptible to erosion and less capable of sediment storage. The rate of sediment delivery from catchment to river depends on the rate of sediment production in the catchment and on the level of connectivity between sediment source and channel (Owens et al., 2005). There are several mechanisms of sediment production in upland areas; generally, cultivated areas are most sensitive to erosion. High grazing intensity has been shown to initiate soil erosion and to exacerbate existing soil erosion; pasture next to a channel thus may significantly increase sediment input into the river (Owens et al., 2005).

Erosion processes and sediment transfer in the catchment can be represented by a box model (Fig. 5) with inputs and outputs that can be summarized by the mass balance equation:

Q_s is the production of soil on hillslopes due to bedrock weathering, and anthropogenic input through atmospheric deposition (Q_a). The inflow from upstream (Q_i) is counterbalanced by outflow (Q_o) plus Q_s and Q_a. This equation can be solved assuming Mtx as the amount of an element X at time t in the box model, X being Pb in this study.

Then it comes:

If a steady state is reached (i.e. input being counterbalanced by output), the residence time τ of X is MtxQiorMtxQ0 and dMdt can be solved. Here, we postulate that Q_s is constant over recent time as less land-use changes have occurred, nor climatic or physical changes (runoff, temperature…, Cerdan et al., 2010), Q_a cannot be considered as constant and a steady state is thus not fully reached. In this case, the dynamic box model must be considered (Boyer et al., 2005). We therefore assume that the initial amount M at t = 0 is M₀ and the initial mass balance equation is:

The next stage of this dynamic box model follows an increase of M₀ to a new value M₁, which is equal to [Q_i + Q_s + Q_a]₁ and thus leads to:

This equation is solved according to:

We have considered the residence time τ of the sediments in the catchment as being of about the order of the century, as demonstrated by Gautier et al. (2009) in mountainous rivers.

In the model, t = 0 is the present day and Q_a is considered as inactive, whereas only Q_s produces lead in the sediment through bedrock weathering (Cerdan et al., 2012). The content M₀ then corresponds to the most depleted sample (A1, Pb = 19 μg·g⁻¹), with the highest percentage change compared to Zr (e.g., −46%, Fig. 4), and represents the most weathered stage. We will consider hereafter three cases for the lead amount M_t (at time t) varying from the present period up to 200 years in the past, which corresponds to the time when the atmospheric deposition of lead (Q_a) was active neither through gasoline nor from mining activities. Within this time frame, we cover twice the admitted residence time of the sediment in the catchment (Gautier et al., 2009). The three cases are chosen along the course of the river and according to the evolution of the content vs. the distance (Fig. 4). The first case M₁ is represented by sample A2 (Pb = 35 μg·g⁻¹), the second case M₂ by the samples A4 and A8 (Pb = 48–49 μg·g⁻¹) and the last case M₃ by sample A10 (Pb = 56 μg·g⁻¹). The variations with time of the lead content M_t^Pb* for the three cases M₁, M₂ and M₃ are calculated using Eq. (6) with M₀ being around 19 μg·g⁻¹. The variation of M_t^Pb* demonstrates first that the lead content observed in samples A2, A4–A8, and A10 was yet obtained at t = 200 yr ago, suggesting that Q_s allows the lead content to increases along the course of the river. The M_t^Pb* remains constant during the mining period, between 1830 and 1930 and after that during the gasoline lead influence. Similar results were obtained in an experimental site by Semlali et al. (2004) and in selected sites in Switzerland (Hansmann and Köppel, 2000).

Determining the source of Pb in soil and sediment requires knowledge of the different possible Pb reservoirs. Because there is no relationship between lead isotopes and the lead content in soils and sediment, more than two end-members must be suspected. In addition to the nature of the bedrock and its weathering, the other inputs may be of natural or anthropogenic origin, but originating outside the catchment. Among these inputs, that from tetraethyl lead when it was extensively used as an additive to gasoline is probably one of the main sources of anthropogenic lead. The Pb isotope composition of gas additives depends on the source of the Pb ore used. In France, Roy (1996) and Monna et al. (1997) reported the range of lead isotope composition in gasoline as 17.12–17.29 for ²⁰⁶Pb/²⁰⁴Pb and 36.81–36.97 for ²⁰⁸Pb/²⁰⁴Pb. In the ²⁰⁸Pb/²⁰⁴Pb vs. ²⁰⁶Pb/²⁰⁴Pb plot (Fig. 6), the positions of Pb isotopic compositions of bedrock and those of the gasoline end-member define the two extreme end-members regarding the lead cycling in soils and sediments (i.e. lead originating from gasoline inputs and lead originating from bedrock weathering). When comparing the lead isotopes in soil and sediment with this relationship, all samples plot with a shift towards lower ²⁰⁶Pb/²⁰⁴Pb ratios. This implies the existence of a third end-member with characteristics that help explaining the shift in the Pb isotope composition. This high lead isotope cannot be due to the application of fertilizers, themselves with a variable lead isotope composition (²⁰⁶Pb/²⁰⁴Pb > 19.21 and ²⁰⁸Pb/²⁰⁴Pb > 33.30; Roy and Négrel, 2001). Hence, the other lead input must be related to natural input, like the impact of granite weathering near the catchment or the presence of ore mineralization (Nehlig et al., 2001). Both may affect the catchment via atmospheric circulation and the transfer of dust to the Allanche catchment.

Even though the lead isotope analyses of the granite population in the Massif Central, made on acid-leached feldspars, plagioclases and whole rocks (Downes et al., 1997; Michard-Vitrac et al., 1981), yielded ²⁰⁶Pb/²⁰⁴Pb, ²⁰⁷Pb/²⁰⁴Pb and ²⁰⁸Pb/²⁰⁴Pb isotope ratios in agreement with the high lead isotope composition in sediment and soil (respectively 18.07–18.45, 15.59–15.72, and 38.15–38.60), two arguments refute this origin for lead. First, the amount of dust needed to shift the lead isotope signature would be too large compared to the lead content in both granite and basalt (Négrel and Roy, 2002). The second argument is based on the Sr isotope ratios in bedrock and soil/sediment in the catchment (Négrel et al., 2013), which does not support such input in view of the values in soil and sediment developed on granite (Négrel, 1999, 2006). Thus, the influence of wind-blown dust from mining and mineral-processing activities remains a major probable source of lead (Carignan et al., 2002; Guéguen et al., 2012) as they are rich in lead. The role of the different lead sources is detailed below.

Input of lead from mining activities can thus be considered as an important source of lead in soil and sediment of the Allanche catchment. In the Massif Central, the small (600 km²) Brioude–Massiac mining district (Marcoux and Brill, 1986) is known primarily for its Sb mineralization, in veins with Sn–W, W–Au, Sb–quartz, and Pb–Zn barite or fluorite, which chronologically belong to the Late Hercynian and Early Jurassic periods. France was the world's largest producer of antimony between 1890 and 1910 thanks to extraction in the Massif Central (Massiac) and the Mayenne region (Laval), with a total production of 120,000 tons, of which the Brioude–Massiac district provided one-third (Fig. 1c). In 1870, a smelter was established at Blesle (some 20 km northeast of the Allanche catchment), supplied mainly by mining in Conche and Ouche. The better days lasted sixty years, but economic development stopped during the crisis of the 1930s, the Ouche mine closing in 1932. Activity resumed in 1945, lasted until 1953 and continued sporadically until 1971. In all, the Ouche deposit has produced some 9000 tons of antimony (1830–1970) of the 40,000 tons produced by the whole district.

In addition to the smelter locations near the Allanche catchment, dominant winds blow from two main directions (Négrel and Roy, 1998). The first is westerly and mainly of marine origin from the Atlantic Ocean and crosses part of the Aquitaine basin. The second is northerly, being both marine (North Atlantic and North Sea) and continental (Great Britain); this crosses the smelter area and thus can have moved dust during the time of mining and ore processing. According to the studies of mineralization in the Massif Central, the Pb isotope ratios resulting from mining activity (²⁰⁶Pb/²⁰⁴Pb, ²⁰⁷Pb/²⁰⁴Pb and ²⁰⁸Pb/²⁰⁴Pb) varied between 18.10–18.47, 15.62–15.66, and 38.20–38.76, respectively, i.e. in the same range as that found in granites from the Massif Central (Marcoux, 1986; Marcoux and Brill, 1986). These data, with mean values of around ²⁰⁶Pb/²⁰⁴Pb = 18.299 ± 0.162 and ²⁰⁸Pb/²⁰⁴Pb = 38.461 ± 0.160 are plotted on Fig. 6 and perfectly agree with the soil and sediment data, partly reflecting the origin of lead from ore mineralization (along the dashed lines). Using the variation in lead content and lead isotopes in sediment vs. the distance (Fig. 4) and the mixing diagram between ²⁰⁸Pb/²⁰⁴Pb and ²⁰⁶Pb/²⁰⁴Pb ratios, the existence of three end-members, i.e. bedrock and two atmospheric components—the influence of mining activities and the deposition of lead from gasoline—, is required to explain the variation in the soil and sediment of the Allanche catchment, as not at all samples align along the bedrock-mining end-member. Fig. 7 shows the relationship between ²⁰⁶Pb/²⁰⁴Pb vs. 1/Pb ratios, the latter shown as a broken axis for easier viewing. The mixture of two or more components can be modeled with this type of standard diagram (Bird et al., 2010; Hansmann and Köppel, 2000; Négrel et al., 2010). With the Pb isotope ratio and 1/Pb coordinates, the mixing curve of two components with different isotope ratios and lead contents results in a straight line. In the case of a multi-end-member mixture, several linear mixing curves can be calculated, meaning that such diagrams require the characterization of end-members as Pb sources. In such a representation, the isotope systematics of two-component mixtures [Eq. (7)] described in detail by Bird (2011) can be expressed as follows:

In that frame, on Fig. 7 the samples A2 and A6 plot in the bedrock field, A4-2 and A10-3/2 plot along the mixing line between the bedrock and mining and processing end-members, reflecting that lead is only derived from these two end-members. The rest of the sediment samples display an additional influence of lead from gasoline input, but A8-3 and A9b plot more closely to the mixing line between the mining and processing and gasoline input end-members, while A1-2 and A2-2 plot more closely to the bedrock–gasoline input end-members mixing line. All soil samples show an influence of the three end-members.