The stream sediments sampling was carried along a grid which consider one sample per square kilometer (1 sample/km²) between February and March ( Figure 3 ). Each sample is a composite of at least 5 sampling points in the fine fraction of sediments in the stream bed, sampled within an interval of several tens meters around the ideal or theoretical sampling point. The average weight of the final sample is around one kilogram (1 kg). Each sample was taken using a small, tough, non-contaminating plastic scoop.

Soil sampling was carried out on a grid of 500 × 200 m over the defined study area, i.e. 10 samples per km² on the anomalies highlighted after the processing and interpretation of the results of the analysis of stream sediments. Depending on the pedo-morpho-regolitic profile encountered, different specific treatments were used for sampling:

1. Lateritic profile present

These are the cases where the lateral armour is flush or a gravelly floor covers the breastplate. Under these conditions, the 5 to 10 cm relatively rich in organic matter is stripped with a shovel and the ochre-beige soil rich in sub-centimeter-sized gravel is removed. The biggest fragments (>2 cm) generally quartz is removed manually. If the lateritic crust is flush over a large area, scales of the lateritic crust (chipping with a hammer or crowbar) are taken at least 5 points 10 to 20 m from the central point.

2. Eroded lateritic profile, saprock or sub-outcropping rock

This profile is encountered in particular towards the northwest of the prospecting area, whose geology is characterized by sub-outcropping schists, crosscut by quartz veins. At surface, the soil is mainly composed of quartz fragments compared to the shales. The highly altered rock (saprock), clayed, is observed after 3 to 10 cm of stripping. It is this material that is sampled on at least 5 points within a radius of 15 to 20 m around the theoretical point.

3. Silty plain area and granitic zone

On lowland areas, it is always difficult to establish whether the cover is real soil or alluvium. In these contexts, the vein is sampled below the humus-bearing horizon at a depth of around 35 cm, taking care to specify on the sampling sheet that it is probably sediment. On the other hand, in certain granitic areas, the sandy horizon encountered at depth (between 30 and 80 cm) is either a granitic arena (i.e. saprolite) or transported sandy deposits (i.e. sedimentary in nature). The water beneath the slightly organic zone, at a depth of around 15 cm, is sampled and the material (presence of lateritic granules) is carefully described.

The stream sediment samples were spread out in aluminum panels, dried away from sunlight and dust at a temperature not exceeding 40˚C. They were then crushed with porcelain mortar and sieved to 63 μm using a sieve column composed from the bottom to the top, of 63 μm sieve, 125 μm sieve, 150 μm sieve and a lid. The fraction less than 63 μm was quartered by means of a riffle divider so as to obtain 2 sachets of 100 g (+/− 10 g) each.

Samples from the Kangounadéni area were analyzed at the Bureau-Veritas laboratory in Vancouver, Canada using their MA250 method. This multi-acid etching method (hydrochloric, nitric, perchloric and hydrofluoric) is followed by the determination of multi-elements (57 chemical elements) by Inductively Coupled Plasma Mass Spectrometry (ICP-MS).

The soil samples were spread out in aluminum panels, dried away from sunlight and dust at a temperature not exceeding 40˚C. In order to reduce the particle size, the samples were crushed with porcelain mortar and sieved to 63 μm using a sieve column composed from the bottom to the top, of 63 μm sieve, 125 μm sieve, 150 μm sieve, 2 mm sieve and a lid. The fraction less than 63 μm was quartered by means of a riffle divider so as to obtain 2 sachets of 100 g (+/− 10 g) each. The samples were analyzed at the Bureau-Veritas laboratory in Vancouver, Canada using the MA250 method. The multi-acid etching method (hydrochloric, nitric, perchloric and hydrofluoric) followed by the determination of multi-elements (57 chemical elements) by Inductively Coupled Plasma Mass Spectrometry (ICP-MS).

For all samples, both in stream sediments and in soils, a proportion of 10% of QA/QC samples was integrated to ensure quality control of the analyses. These control samples include blanks, standard and duplicates:

This QA/QC procedure ensures the reliability and accuracy of the geochemical data used for anomaly mapping and subsequent investigations.

A total of 281 samples were analyzed and processed. The treatment consisted in the examination of the spatial distribution of the chemical elements; the definition of anomaly thresholds for some key elements (Li, Pb, Sn) and the definition of anomalous areas ( Figure 4 ). The zone located in the continuity of the two-micas

Ferké leucogranite, presents geochemical disparities compared to the zone located in the western flank (Niarébama-Labola). The axis of the batholith contains the highest concentrations of Li, Be, Cs, Ti, Bi, Nb, Ta, W and Sn ( Figure 4 and Figure 5 ). The extreme northern part of the axis, located in volcanosediments, has samples more enriched in siderophile elements (Fe, Mn, V, Cu, Cr, Mo and Sc), as well as in Cs and Bi. On the other hand, the southern part (Sakora) is richer in K, Rb and U than the northern part. In the northwest, samples from the Niarébama-Labola zone are more enriched in As and Sb compared to those from the batholith axis. As one moves northwest from the central axis of the batholith, there is a gradual enrichment of Zr ( Figure 5(A) ), Sb, Mo, and REE concentrations (especially marked for light REEs). In contrast, the concentrations of W, Ta and Nb decrease in the same direction ( Figure 5(B) and Figure 5(C) ).

1. Lithium (Li)

As a criterion for discriminating abnormal values in a normal distribution, we use the indicator given by the formula: threshold = mean + 2 * standard deviation [33] . The average concentration of Li in stream sediments is 25.7 ppm ( Figure 6(A) ). This value is significantly higher than the Clarke of Li (21 ppm) [34] . Its geochemical background threshold is calculated at 64.6 ppm. Based on this threshold, five (5) samples are outlying and can be divided into two (2) anomalies, the first is located at the southwest of the village of Sakora (up to 133 ppm). It is located on the two-micas Ferké leucogranite and its orientation is poorly constrained. This area contains points with relatively high concentrations of Rb (up to 226 ppm), U (up to 9.8 ppm), Be (up to 4 ppm), Sn (up to 2.9 ppm) and Ta (up to 1.8 ppm). The second is located at the northeast of the village of Kangounaba. This zone also has points with relatively high concentrations of La (up to 38.7 ppm), Be (up to 5 ppm), Cs (up to 42.9 ppm), and medium concentrations of Sn (up to 1.8 ppm). It is located on the metavolcanosedimentary schists outside the two-micas Ferké leucogranite.

2. Lead (Pb)

The average concentration of Pb is 11.60 ppm, which is lower than its Clarke value (14 ppm) [34] ( Figure 6(B) ). Its geochemical background threshold is calculated at 17.24 ppm. As a result, five (5) sampling points are anomalous: 3 from the south of Sakora (up to 26.7 ppm) on the two-micas Ferké leucogranite and one from the northwestern part of the sampling area (Naniagara-Tangora, up to 17.74 ppm) in the volcanosedimentary shales.

3. Tin (Sn)

The average concentration of Sn is 1.2 ppm. This value is much lower than its Clarke (9.8 ppm) [34] ( Figure 6 (C)). Its geochemical background threshold is calculated at 1.93. This results in five (5) points being anomalous, near Sakora, on the leucogranite with two-micas Ferké leucogranite as well as in the surrounding volcanosedimentary units to the west of the village.

Regarding to the chemical elements analyzed in the stream sediments, the following anomaly areas were identified ( Figure 7 ):

A total of 640 samples were chemically analyzed and processed. The treatment was first carried out by describing the spatial distribution of the chemical elements. Then, the identification of abnormal samples, the reduction of variables, principal component analysis and hierarchical ascending classification. At the end of the classification, the dominant chemical behaviours and the areas of anomalies were identified.

Analysis of the distribution of chemical elements in soil allows for the identification of general distribution trends. The samples coinciding with the two-micas Ferké leucogranite are relatively richer in alkaline elements (K, Li, Rb), in U and are very depleted in Sc. The K, Li, Rb and Cs elements are particularly enriched along multi-kilometer corridors-oriented NNE-SSW at the south of Sakora and the west of Sakora, ENE-WSW trend at the south of Kangounadéni, but also in less extensive clusters scattered at the north of Kangounadéni. The samples coinciding with the units of the volcanosedimentary belt are relatively more enriched in Sc, W, Cr and Ni.

The analysis of the box-plot plots identified a list of samples with very high positive anomalous values in concentrations of Be, Mn, Ba, Ta, W, Pb, Sn and Nb and rare earth elements ( Table 1 ). These samples correspond for the most part to point anomalies, with the exception of two zones at Ta, Be and S, and at Mo and Bi where these anomalies are respectively observed on two neighboring samples.

1. Filtering for non-significant and repetitive variables

For data reduction, our approach is to discard non-significant and repetitive variables. Variables S, Re and Cd were removed because very few samples had concentrations above the detection limits (640, 626, 505, respectively) for a total of 640 analyses. The elements Tm, Lu and Tb have respectively 104, 107 and 186 values above the detection limit. However, they have been removed, but their chemical behaviour can be inferred from the elements Gd and Yb. To identify repetitive variables, we calculated Pearson correlation coefficients that qualify the linear relationship between a pair of variables. Based on a correlation coefficient threshold of 0.75, we grouped the chemical elements into groups, each characterized by a representative element. The variables selected at the end of this treatment are presented in Table 2 . Despite the correlations, we have kept some elements because of their geochemical or metallogenic significance. The element Al is kept for the characterization of the aluminosity. The element Th is kept for its metallogenic interest and for the characterization of geochemical mobility compared to U. The elements Cs and Rb are kept despite their correlation with Li and for their use as a marker of enrichment in incompatible elements during magmatic differentiation. Arsenic is kept despite its correlation with Fe, for its metallogenic interest (association with chalcophile elements).

After the non-significant and repetitive variables have been excluded, the variables retained for the rest of the analysis are as follows: Li, Be, Na, Mg, Al, P, Ca, Fe, Cu, Co, Zn, As, Se, Rb, Sr, Zr, Nb, Mo, Ag, Sn, Cs, Ba, La, Gd, Yb, Ta, W, Pb, Bi, Th and U. The elements of this selection with a strong correlation (>0.7) are presented in the Table 3 .

2. Principal component analysis

A principal component analysis (PCA) was carried out on the 31 elements retained with all 640 samples. The first three factors express a cumulative variance of 53.11% ( Table 4 ).

Factor 1 shows an opposition between a positive pole dominated by Th, immobile elements (Al, Fe, Zr) and heavy REE, and an alkaline negative pole represented by the elements Li, Na and Rb ( Table 5 ). The F2 factor opposes a positive pole dominated by alkaline earth elements (Mg, Ca, Sr, Ba), and a negative pole dominated by organophile elements (P, As, Se, U), and siderophile. In the factor diagram, the distribution of chemical elements indicates an opposition between alkaline and alkaline earth elements associated with Li (negative F1 and positive F2), and organophile and semimetal elements (positive F1 and negative F2). The F3 factor is controlled by the elements Sn, U, P, Li in opposition to certain incompatible elements of high load (medium and heavy REE, Zr, W), and elements with calcium tendency (Ba, Mg, Ca) and chalcophile (Co, Cu, Zn) ( Table 5 ). The Ta element does not have a noticeable influence in the three (3) main factors, which is characterized by its scores close to zero in them ( Figure 8 ).

We performed a hierarchical ascending classification using all the analysis results, transformed into a centered log-ratio for the same elements selected for the principal component analysis. Agglomerative hierarchical clustering was applied using Ward’s linkage with Euclidean distance. The clustering quality indices gave a low silhouette score (0.158) and a moderate cophenetic correlation (0.573) indicating weak separation between groups. However, the choice of five (5) classes is justified by the observed consistency between the clusters ( Table 6 ), the identified anomalies, and the lithologies of the sampling areas ( Table 7 ). The dendrogram is shown in Figure 9 . Classes C1 and C5 coincide spatially with volcanosediments. They also contain the samples with the highest Cu values (μ = 56 ppm). Class C5 is characterized by high values in Mg but also in Li (μ = 60 ppm) and is exclusively observed in the North. Classes C2, C3 and C4 coincide with two-micas Ferké leucogranite. Class C3 particularly the two previous ones by relatively high concentrations of U (μ = 6 ppm). The other two classes C2 and C4 have relatively low values of Fe, Al, As and Th. The class C2 corresponds to the zone of Li anomaly (μ = 69 ppm) and contains samples rich in Na and K, but depleted in REE. These two trends allow us to group these classes into two (2) superclasses, the CVS class (C1 + C5) which contains the analyses with a volcanosedimentary type signature and the CGR class (C2 + C3 + C4) which contains the granitic type signatures. Li-rich samples are preferentially found in classes C2 and C5, each corresponding respectively to the anomalous class in Li in its superclass CGR and CVS ( Figure 10 ).

1. Lithium (Li)

As a criterion for discriminating abnormal values in a normal distribution, we use the indicator given by the formula: threshold = mean + 2 * standard deviation [33] . The concentration of Li in soil samples averages 37 ppm, which is higher than the Clarke of Li (20 ppm) [34] . In the CVS superclass, this average is 25 ppm, while it is 47 ppm in the CGR superclass ( Figure 11 ). The granitic batholith is

therefore naturally richer in Li than the volcanosedimentary host. The geochemical background threshold of Li for the dataset is 120 ppm. About 4.5% of the samples have values above this threshold. These samples are located south of Tiéfora (Kougounaba, up to 170 ppm), south of Kangounadéni (up to 175 ppm) and south of Sakora (up to 194 ppm). The area at the southernmost point of the Kangounadéni anomaly has a Li ± Sn signature with some high Sn grades at 6.4 ppm.

Thus, four anomalies are highlighted in the study area ( Figure 12 ):

2. Lead-Zinc (Pb-Zn)

The average concentration of Pb ( Figure 13(A) ) is 34 ppm, which is much higher than Clarke (14 ppm) [34] . In addition, nearly 82% of the analyses are superior to the Clarke. The calculated geochemical background threshold is 109 ppm, which leaves 25 samples (approximately 4%) in anomalous concentration. These samples are found exclusively west of Kangounaba, and in Kangounadéni. The average concentration of Zn ( Figure 13(B) ) is 33 ppm. This average is much lower than the Clarke of Zn which is (70 - 80 ppm) [34] . The calculated geochemical background is 93 ppm, which leaves 25 samples (about 4%) in anomalous concentrations. With the exception of a point anomaly observed near Sakora, anomalous points with respect to the threshold are located at Kangounadéni (up to 523 ppm). We have retained as an anomalous zone a corridor-oriented NNE-SSW, west of Kangounadéni, which coincides with the edge of the batholith. This corridor is 8 km long, and its widest area is 1.8 km ( Figure 14 ).

3. Rare Earth Elements (La-Ce-Nd)

The element La ( Figure 15(A) ) has an average in soil samples of 15 ppm, which is lower than Clarke (39 ppm) [34] . The geochemical background threshold (42 ppm) is close to the Clarke, leaving only 9 anomalous samples rich in La. These two samples are located exclusively in the anomalous Li-zone at Kangounaba (up to 247 ppm). The element Ce ( Figure 15(B) ) has an average in soil samples of 39 ppm, lower than its Clarke (66.5 ppm) [34] . On the other hand, its geochemical background threshold is high (109 ppm), which leaves less than 5% anomalous samples. These samples are all distributed in Li-anomalous zones with concentrations especially high in the Kangounaba anomaly (up to 439 ppm). The element Nd ( Figure 15(C) ) has an average in soil samples of 12 ppm, which is lower than its Clarke (41.5 ppm) (Lide, 2005). Its geochemical background threshold is calculated at 28 ppm, which leaves less than 3% anomalous samples. These samples are located on the Kangounaba (up to 162 ppm) and Kangounadéni (up to 66 ppm) Li anomalies. The element Dy has an average in soil samples of 1.37 ppm, which is lower than its Clarke (5.2 ppm) (Lide, 2005). Its geochemical background threshold is calculated at 2.59 ppm, which remains lower than the Clarke and leaves less than 4.6% anomalous samples. These samples are located in the Li anomalous zones of Kangounaba and Kangounadéni. Ultimately, the REE anomaly zones coincide with those of the Li in the Kangounaba (volcanosedimentary zone) and Kangounadéni (granitic zone) parts.

4. Sn, W, Ta and U anomalies

Tin has an average content of 1.8 ppm in soil samples, lower than its Clarke (9.8 ppm). Its geochemical background threshold is calculated at 4.1 ppm, which leaves less than 2.7% of anomalous samples, which are located in the Li-anomalous zones of Kangounadéni (up to 8 ppm) and Sakora (up to 19.2 ppm). Tungsten has an average content of 0.9 ppm in soil samples, lower than its Clarke (1.25). Its geochemical background threshold is calculated at 2.3 ppm, which leaves about 2% of anomalous samples that are located in the Li-anomalous zones of Kangounadéni (up to 6.2 ppm) and Kangounaba (up to 6.7 ppm). A point anomaly is also observed North of Sakora (6.8 ppm). Tantalum has an average content of 0.9 ppm in soil samples, lower than its Clarke (2.0 ppm). Its geochemical background threshold is calculated at 6.6 ppm, which leaves only 6 anomalous samples. These samples are scattered around the Kangounadéni (up to 56 ppm) and Sakora (up to 37 ppm) Li anomalies. Uranium has an average content of 4.0 ppm in soil samples, higher than its Clarke (2.7 ppm). Its geochemical background threshold is set at 9.1 ppm, which leaves 5.5% anomalous samples (35 samples). These samples are located in the anomalous zones of Kangounadéni (up to 18.2 ppm) and Sakora (up to 14.5 ppm).

The anomalous zones of Kangounaba and Sakora observed in stream sediments are also found in soils. However, the enrichments observed in each type of rock may show slight variations. The sampling density allowed the Sakora level to distinguish two sub-enrichment zones, one to the north of the locality, and the other to the south ( Figure 16 ). The F1 component of the PCA shows the trend in

Li-Rb enrichment (F1 < 0). However, this alone is not sufficient to constrain the anomaly observed at Kangounaba. The Pb-Zn anomalies observed east of Kangounadéni correspond to samples of class C1, identified as representing volcanosedimentary units. This association suggests that the anomaly is hosted in volcanosedimentary units, and not in the two-micas Ferké leucogranite ( Table 8 ).

In stream sediments, a decrease in the concentration of the elements W, Nb and Ta is observed from the edge of the batholith to the northeast. These elements are among high field strength elements (HFSEs), usually behaving as immobile elements during meteoric weathering [35] [36] . They are likely to be reconcentrated on the surface on the one hand by the relative accumulation during meteoric weathering, and on the other hand by the trapping of oxides carrying these elements. In addition, the refractory minerals zircon, rutile and columbo-tantalite could be carriers of Nb and Ta in sediments [37] . The observed Pearson correlations between the elements W and Ti (0.609), Nb and Ti (0.804), and Ta and Ti (0.697) suggest that the presence of titanium oxide primarily controls W, Nb, and Ta concentrations. In soils, W tungsten is depleted from granitic zones to volcanosedimentary zones, but the elements Nb and Ta do not show a consistent distribution with lithology. The distribution of uranium U in soils is due to several different processes. The highest concentrations of U are observed in soil samples around Kangounadéni and Sakora (19 samples at U > 11 ppm). They are higher than the average concentrations found worldwide in soils [38] . Specifically, the most extreme values are found near Kangounadéni, on covers corresponding to the lateritic surfaces of high glacis. It is therefore very likely that the concentration of U in soils is partly linked to supergene accumulation (preferably in clays) during episodes of lateritic cover formation. The element U is found to be very weakly correlated with P (Pearson coefficient = 0.549), Sb (Pearson coefficient = 0.509), V (Pearson coefficient = 0.461) and Sn (Pearson coefficient = 0.407). On the one hand, the association with the organophile elements P and Sb suggests that the U concentration is controlled by organic matter. On the other hand, it is also associated with soils of hierarchical class C3, which corresponds to less evolved granitic units (low concentration of Ba, Li) and high concentration of Fe compared to the other granitic classes. This distribution is not consistent with the natural tendency of uranium enrichment with differentiation in peraluminous magmas [39] , and rather suggests a depletion of the U during the magmatic differentiation of the two-micas Ferké leucogranite, coupled with a fractionation induced by meteoric oxidation.

In the Kangounaba area, high concentrations of Cu, Mo and Bi are observed in soil samples. Although these concentrations are found sporadically, they may reflect geochemical trends of Cu-Mo association in the form of a “nugget effect” due to the presence of bearing minerals (molybdenite, or secondary oxides and clays) that have maintained or enhanced these concentrations. The Cu-Mo association is usually observed in deposits in a magmatic-hydrothermal transition context, e.g. skarns and porphyries [40] - [42] , while Bi can also be associated with Cu-Pb or Pb-Zn sulfide mineralization resulting from hydrothermal episodes [43] . The abundance of Mo in pegmatites is mentioned in the literature, but is generally associated with Sn-W mineralization (cassiterite, wolframite) [44] [45] . Since Cu, Mo are Bi are not associated with Li enrichment in the Kangounaba area, the hypothesis of an enrichment related to magmatic evolution of the two-micas Ferké leucogranite can be ruled out. In view of the difference in the geochemical trend of Cu-Mo-Bi enrichment between the Kangounaba anomaly and the Kangounadéni-Sakora anomalies, we can assume the existence of a not unearthed intrusion or a distant intrusion), whose exsolution of late-magmatic fluids is at the origin of hydrothermal alteration in the Kangounaba zone.

The Dougoulogo massif in the North, has significant Cu concentrations (10 - 50 ppm) but lacks Bi and Mo, (Bi < 1 ppm, Mo below the detection limit) [46] , reflecting remobilization from deep source. Cu is concentrated in the oxidized horizons by hydrothermal or meteoric transport, while Bi and Mo, less mobile, remain at depth, explaining their low concentration at the surface [47] .

A pole of Pb-Zn enrichment is clearly identifiable to the west of Kangounadéni, and seems to correspond to a kilometric lens at the edge of the two-micas Ferké leucogranite, enriched in Zn (up to 286 ppm), Pb (up to 523 ppm), Ag (up to 182 ppm) and depleted in Bi (<1 ppm). This type of association is typical of polymetallic sulfide deposits, generally associated with the circulation of hydrothermal fluids of metamorphic origin such as Sedimentary Exhalative Deposit (SEDEX) and the formation of sulphide associated mineralization [48] . As the geological context of the Banfora belt does not exhibit evidence of a paleo unconformity (birimian sedimentary of a pre/early birimian basement), it is more likely that the anomalies observed derive from a context similar to that of the VMS, which are also observed in other Birimian belts of the Baoulé-Mossi domain [49] - [51] .

The Li-Sn geochemical anomalies are commonly observed in rare metal granite/pegmatite deposits of the Lithium-Cesium-Tantalum (LCT) petrogenetic family [52] [53] . These pegmatites are generally associated with peraluminous granites, and are most of the time interpreted as resulting from fractional crystallization, resulting in a concentration of alkaline elements (Li, K, Rb, ...), and fluxing elements (B, F, P, ...) and H₂O [54] - [56] . These processes are accompanied by a more or less notable enrichment in Nb, Ta and Sn. The geological setting of the Banfora belt presents an ideal configuration to host this type of mineralization. Indeed, the two-micas Ferké leucogranite is a peraluminous granite and therefore a potential parent source for Li-Rb enrichments. Its geochemistry exhibits stronger REE depletion, compared to calc-alkaline plutonic rocks of the BB (Dougoulogo and Folonzo granite), with Li concentration up to 320 ppm, but low content in Sn (<6 ppm) and Cs (<44 ppm) [46] . In general, peraluminous granites tend to deplete Zr and REE during differentiation [57] . This feature is observed in soil geochemistry samples, which are depleted in Zr and REE in granitic zones compared to volcanosediments, suggesting that the underlying granite is depleted of REE. The granitic zone specifically exhibits an enrichment in Li, Be, Cs, Nb, Ta, W and Sn in stream sediment geochemistry and Li, K, Rb, U and Nb in soils geochemistry, consistent with granitic fractionated signature with increasing content in Nb, Ta, W and Sn. Then, the Li ± Sn soils anomalies are located in the northeastern margin of the batholith, which according to an inverse zonation model with differentiated leucogranitic margins [58] could derive from localized late magmatic facies or from late pegmatite dikes intrusive in volcanosediments. A similar setting is observed in the batholith of Kédougou-Kéniéba inliner, where spodumene-bearing pegmatites, dikes are described in the extreme northeastern part of the granitic batholith of Saraya [59] . According to the typical zonation model of LCT pegmatites around their source pluton [60] , the proximal pegmatites are enriched in Be, then Nb and Ta, while the distal pegmatites show the most pronounced enrichments in Li, Cs and Ta. This setting is not perceptible in our context, where the Li and Be elements are enriched in both the granitic massif and the surrounding area, while the Sn and Rb elements seem to be mainly concentrated in the anomalies within the batholith (especially in the granitic corridors defined from class C2). The highest Cs content (with also Ca and Mg) is specifically found in the north out of the granitic body, suggesting a geochemical enrichment trend different from the Li-Rb-trend in granitic facies, and controlled by alkaline earth elements. As in the case of the Saraya batholith, this model is not suitable for the elongated structure of the batholith, because the zones of Li enrichment correspond more to lateral margin of a granite massif rather than the roof of the intrusion. The samples classified C2 (granitic trend, Li-rich) are distributed in the form of clusters often extended over 2 to 3 km in NE-SW, N-S and NW-SE orientations in the Kangounadéni zone, and NESW in the Sakora zone. These oriented clusters suggest the presence of linear anomalies that could be attributed to enriched pegmatite veins. Apart from this, the C2 point clusters are distributed on the hillsides of a ridge line extending from Kangounadéni to Sakora, around high lateritic glacis, [61] . The areas corresponding to the glacis are covered mainly by C3 points. This is consistent with the observed association of C3 points with enrichment in Fe, Al, and suggests that in these areas, Li and Ba concentrations are depleted by lateralization ( Table 7 ).

Apart from samples of very anomalous values in Ta concentration (mainly class C3 and of class C2, <19 ppm < Ta < 57 ppm), Ta concentrations are generally homogeneous throughout the surveyed area, with a slight depletion in the extreme northern part. Niobium also follows this same trend, with many more anomalous points. Although these Ta concentrations are significantly high for soil samples (overall average at 1 ppm, [36] , principal component analysis shows that Ta does not contribute significantly to the data set. It therefore does not follow any of the geochemical trends identified for the elements (alkaline, alkaline-earth, organophile or siderophile elements). Moreover, the highest values in Ta and Nb do not coincide with the enrichments observed in Li. As suggested above, the high concentrations of Ta could result from the accumulation of Fe/Ti oxides, and this tendency of gravitational accumulation is therefore not perceptible through the multi-criteria analysis of the data. Either the potential for Nb/Ta enrichment of the identified anomalies is low to the detriment of Sn, or these elements constitute poor indicators for the search for differentiated and enriched zones in our study area. In addition, interpretation biases can be induced by the choice of variables for the analysis of anomalies. The Li anomaly observed in the Kangounaba zone coincides with high values in Fe, Co, Cu, Mu, which we attribute to a signature deriving from the host volcanosedimentary units. However, the component analysis shows an opposition between the alkaline chemical elements (Li, K, Rb) and the siderophile elements, particularly according to the B component. An interpretation based solely on the principal component analysis could lead to the elimination of units rich in siderophile elements, whereas in our context Li enrichment is superimposed on rocks rich in siderophile elements. In addition, the principal component analysis highlights more the overall geochemical affinities of the lithologies rather than the enrichment trends of the targeted elements (Li, Nb, Ta, ...).