Inﬂuence of Sediment Cycling on the Rare-Earth Element Geochemistry of Fluvial Deposits (Caculuvar–Mucope, Cunene River Basin, Angola)

: The rare-earth element (REE) geochemistry of sedimentary deposits has been used in provenance investigations despite the transformation that this group of elements may suffer during a depositional cycle. In the present investigation, we used the geochemistry and XRD mineralogy of a set of sand and mud ﬂuvial deposits to evaluate the ability of REE parameters in provenance tracing, and the changes in REE geochemistry associated with weathering and sorting. The analyzed deposits were generated in a subtropical drainage basin where maﬁc and felsic units are evenly represented, and these crystalline rocks are covered by sedimentary successions in a wide portion of the basin. A few element ratios appear to hold robust information about primary sources (Eu/Y, Eu/Eu*, La N /Yb N , La N /Sm N , and Gd N /Yb N ), and the provenance signal is best preserved in sand than in mud deposits. Sediment cycles, however, change the REE geochemistry, affecting mud and sand deposits differently. They are responsible for signiﬁcant REE depletion through quartz dilution in sands and may promote discernible changes in REE patterns in muds (e.g., increase in Ce content and some light REE depletion relative to heavy REE). Ce, Pr, Nd, and Sm), Eu, and heavy REE (HREE; Gd, Tb, Dy, Ho, Er, Tm, Yb, and Lu). In the analysis and graphical representations, Y and Sc were separated from the REE. Univariate and multivariate statistical analyses were performed using the software JMP Pro 14.0. To better evaluate the associations between compositional parameters, principal component analysis (PCA) was performed on different sets of geochemical results.


Introduction
Rare-earth elements (REE) have received major attention due to their economic importance for new technologies, with special emphasis on green energies, electronics, informatics, and innovative agricultural practices. An increasing demand for REE has thus been observed worldwide during recent decades [1,2]. REE deposits can be formed through magmatic, metamorphic or hydrothermal primary processes or be associated with secondary processes involving weathering and erosion of primary sources [1,3]. As REE tend to be insoluble in surface environments, their concentrations in a sedimentary deposit reflect the respective source area geology. Hence, the REE patterns of river sediments, besides being used to identify commercially valuable deposits, are frequently applied in provenance investigations without a direct economic focus [4][5][6][7][8][9]. However, the REE patterns of river sediments can be affected by exogenous processes, such as weathering [10][11][12][13][14] and sorting [5,[11][12][13][14][15][16][17]. For example, fine-grained deposits are expected to be particularly influenced by weathering-related transformations [18,19], but REE geochemistry is still widely used to trace the provenance of mud deposits [6,7,20,21]. The cumulative effects of exogenous changes associated with multiple depositional cycles are likely to have a major influence on REE geochemistry where the recycled component is abundant.
The present investigation is focused on the geochemistry of a set of present-day sand and mud deposits from the Caculuvar River basin, a sub-basin of the Cunene River in The present investigation is focused on the geochemistry of a set of present-day sand and mud deposits from the Caculuvar River basin, a sub-basin of the Cunene River in the subtropical SW of Angola ( Figure 1). It aims to better understand (1) the source rock con trol on the REE contents of the produced sediment and (2) the exogenous processes tha influence the REE geochemistry. In the study region, two trunk rivers are separated by a ridgeline striking broadly north-south, leaving the Caculuvar sub-basin to the west and the Mucope sub-basin river to the east ( Figure 2). The Caculuvar trunk river flows through an upstream sector, where felsic igneous rocks largely prevail, and a downstream sector mainly with mafic igneous rocks, before entering the Kalahari Basin; the Mucope is en tirely placed in the Kalahari Basin, thus draining exclusively a recent sedimentary succes sion. Several geological features of the investigated area, such as the similar representa tion of mafic and felsic crystalline rocks, and the good separation of regions with sharply different geologies, provide unique conditions to assess how the recycling of sediments with different primary sources controls REE geochemistry.

Geology
The Caculuvar River, with a length of 273 km and a drainage basin of 25,322 km 2 , is the biggest tributary of the Cunene River. The Caculuvar River basin integrates two trunk rivers, the Caculuvar and the Mucope. The catchment areas of these trunk rivers (hereafter referred to as the Caculuvar and Mucope sub-basins) are approximately the same size. The two rivers only join together less than 10 km from the confluence with the Cunene. The geologies of the Caculuvar and Mucope sub-basins are substantially different. The Mucope River is entirely placed in the sedimentary units of the Kalahari Basin, while the Caculuvar mostly drains the Angolan Block of the Congo Craton, entering the Kalahari only in its downstream path ( Figure 2). Different units of the Angolan Block are drained by the Caculuvar River. In the NW tip of the catchment belonging to the Humpata-Bimbe Plateau, its headwaters drain the Proterozoic Chela Group and Leba Formation [23]. The Chela Group comprises a succession of siliciclastic rocks of different grain sizes and intercalated volcanoclastic rocks that, as a general rule, are Si-rich. Zircon grains retrieved from rhyolitic beds allow dating the Chela Group at the Paleoproterozoic [24]. Occasionally, younger igneous rocks (~1.5 Ga) enriched in the mafic component also occur [25]. The Leba Formation unconformably overlies the Chela Group in the higher elevation areas of the Humpata-Bimbe Plateau. It consists mostly of dark dolomitic limestones, frequently with well-preserved stromatolites.
Southward, the Caculuvar flows on the pre-Chela basement, which is dominated by intrusive rocks. In northern locations, the so-called "regional granite" occurs [26]. In general, it is a deformed peraluminous leucocratic granite associated with the Eburnean orogeny (~2.0 Ga) [24,27]. Further downstream, the river drains Mesoproterozoic mafic rocks of the Kunene Complex of SW Angola, previously called the gabbro-anorthosite complex

Geology
The Caculuvar River, with a length of 273 km and a drainage basin of 25,322 km 2 , is the biggest tributary of the Cunene River. The Caculuvar River basin integrates two trunk rivers, the Caculuvar and the Mucope. The catchment areas of these trunk rivers (hereafter referred to as the Caculuvar and Mucope sub-basins) are approximately the same size. The two rivers only join together less than 10 km from the confluence with the Cunene. The geologies of the Caculuvar and Mucope sub-basins are substantially different. The Mucope River is entirely placed in the sedimentary units of the Kalahari Basin, while the Caculuvar mostly drains the Angolan Block of the Congo Craton, entering the Kalahari only in its downstream path ( Figure 2). Different units of the Angolan Block are drained by the Caculuvar River. In the NW tip of the catchment belonging to the Humpata-Bimbe Plateau, its headwaters drain the Proterozoic Chela Group and Leba Formation [23]. The Chela Group comprises a succession of siliciclastic rocks of different grain sizes and intercalated volcanoclastic rocks that, as a general rule, are Si-rich. Zircon grains retrieved from rhyolitic beds allow dating the Chela Group at the Paleoproterozoic [24]. Occasionally, younger igneous rocks (~1.5 Ga) enriched in the mafic component also occur [25]. The Leba Formation unconformably overlies the Chela Group in the higher elevation areas of the Humpata-Bimbe Plateau. It consists mostly of dark dolomitic limestones, frequently with well-preserved stromatolites.
Southward, the Caculuvar flows on the pre-Chela basement, which is dominated by intrusive rocks. In northern locations, the so-called "regional granite" occurs [26]. In general, it is a deformed peraluminous leucocratic granite associated with the Eburnean orogeny (~2.0 Ga) [24,27]. Further downstream, the river drains Mesoproterozoic mafic rocks of the Kunene Complex of SW Angola, previously called the gabbro-anorthosite complex [27][28][29]. These units, coupled with their continuation in Namibia, constitute the largest mafic complex of Africa. In places, between outliers of the Kunene Complex of SW Angola, coeval A-type red granites that usually strike SW-NE occur [30].
After crossing the eastern border of the Kunene Complex of SW Angola, the Caculuvar River enters the Kalahari Basin approximately 85 km upstream of its confluence with the Mucope. Its sedimentary infill includes the Kalahari Group with aeolian and fluvial deposits dated at the late Cretaceous to Cainozoic [31] and a series of Quaternary loose sands associated with the recycling of the Kalahari Group.

Climate
The climate of the Caculuvar catchment presents substantial seasonal variation, with a wet season, from October to April, in which heavy rainfall and high temperatures occur, and a dry season, from May to September, with lower temperatures [32]. At these latitudes, the influence that some processes such as the El Niño Southern Oscillation (ENSO) and the Angola Low (AL) play on rainfall significantly varies. At the interannual scale, ENSO is primarily responsible for rainfall variations in southern Africa [33], giving rise to seasonal below-normal rainfall [34]. In contrast, the AL is a low-pressure system occurring from October to March, favoring the wet periods, bringing the summer rains [35].
According to the Koppen-Geiger climate classification (http://koeppen-geiger.vuwien.ac.at/present.htm, accessed on 26 March 2021), in the Caculuvar catchment, the mountainous sector of the basin to the NW has a subtropical highland climate (Cwb), transitioning as we move southwards to a hot semi-arid climate (Bsh). In this way, despite registering a tropical thermal regime softened by the catchment's mountainous distribution (monthly average close to 22 • C), the average temperature increases to the east and south. In contrast, the annual precipitation varies in the opposite direction. The southern sectors of the Caculuvar and Mucope catchments record about 450-500 mm per year, while in the north, particularly in the mountainous sector of the basin, about 900-950 mm per year is recorded ( Figure 2).

Methods
For this study, the main lithological data came from the Geological Map of Angola, at a scale of 1/1,000,000, sheet 3, published in 1980 [22]. The delimitation of the basins, sub-basins, and hydrographic network was carried out in GIS and based on the altimetric data from SRTM DEM V4, with a 30 m pixel size. The selection of the sampling sites was carried out according to geomorphological and geological criteria, considering two main objectives: (1) the choice of sampling points along the main river, individualized by major lithological sectors to capture the sediments of the sector, and those coming from the upstream sub-catchments; and (2) the choice of sampling points in exclusive subcatchments of each of the major lithological sectors to obtain only sediments of that sector. Following these criteria, field locations were first defined on the cabinet by observing imagery from the Google Earth and Terra Incognita platforms (to evaluate the terrestrial access and sediment availability) and then transferred to a database to use in the field. In total, 25 present-day sediment samples were selected for compositional analyses, including 15 sands and 10 muds (Table 1; Figure 2).
The mineralogical composition was determined, for both sand and mud deposits, by X-ray diffraction (XRD) using an Aeris instrument (PanAlytical) with a Cu tube, at 15 kV, 40 mA. Diffractograms were obtained on the ground (to <15 µm) with randomly oriented grains in the range 2-60 • 2θ. Semi-quantitative estimations of mineral proportions were based on the areas of characteristic reflections identified in the diffractograms after extracting the background. Estimations were performed for quartz (peaks at 3.34 Å, after correcting for mica reflection, and 4.26 Å), K-feldspar (peak at~3.24 Å), plagioclase (~3.18 Å), calcite (~3.03 Å), pyroxene (~2.9 and 2.95 Å), hematite (~2.7 Å), and phyllosilicates (summed reflections of~7 Å, 10 Å, 12-13 Å, and 14-15 Å). Chemical element concentrations were determined with the same aliquots (~5 gr) used for bulk mineralogy. Analyses were performed at the laboratories of Bureau Veritas (Vancouver; group 4A-4B and code LF200) (Supplementary Material Table S1). Most major oxides were determined by ICP-AES (using a Spectro Ciros/Arcos instrument) and trace elements by ICP-MS (using an ICPMS ELAN 9000 instrument), following a lithium metaborate/tetraborate fusion and nitric acid digestion. For quality control, accuracy was obtained through the standard STD SO-19. The estimated errors for the elements considered in this research were, in general, below 2% and, except for Sm (~6%), always below 5%. Blanks were almost always below the detection level.
The composition of the Upper Continental Crust (UCC; [36,37]) was applied for the normalization of the geochemical data. Lanthanides were also normalized to the chondrite composition [38], with the fractionation parameters being computed after this normalization. For simplicity, REE are grouped here as light REE (LREE; La, Ce, Pr, Nd, and Sm), Eu, and heavy REE (HREE; Gd, Tb, Dy, Ho, Er, Tm, Yb, and Lu). In the analysis and graphical representations, Y and Sc were separated from the REE. Univariate and multivariate statistical analyses were performed using the software JMP Pro 14.0. To better evaluate the associations between compositional parameters, principal component analysis (PCA) was performed on different sets of geochemical results.

Sediment Classification Based on Source Geology
Sampled sediments can be classified into four types according to the geology of their catchment areas (Table 1). Where the drainage area includes mafic and felsic crystalline rocks and each represents at least~1/3 of the crystalline component (i.e., after excluding sedimentary and low-grade meta-sedimentary terranes), the source is classified as mixed. Felsic and mafic sources are considered where one of these rock types is prevalent, and the (meta-)sedimentary cover does not occupy more than 75% of the drainage area. Otherwise, the sampling sites are considered to have a recycled source. With one exception, collected in the upstream portion of the Caculuvar sub-basin, recycled sediments came from the Mucope sub-basin. Mafic-derived sediments were collected within the Kunene Complex of SW Angola, while felsic-derived sediments came from upstream locations in streams draining the Eburnean granitoids. Sampling sites in the Cunene and at the middle to lower courses of the Caculuvar are mixed source.

Mineralogy
The XRD mineralogy is presented in Table 1. Sand samples yield variable amounts of quartz and feldspar, with phyllosilicates in secondary to minor amounts. Carbonates and Fe oxides can be present in minor amounts. Sands with a sedimentary to meta-sedimentary source are strongly enriched in quartz (>92%). Mafic-derived sediments tend to hold more feldspar and less quartz than felsic-derived sediments. The composition of sands with a mixed source are widely variable, revealing either a clear enrichment in quartz, especially where sedimentary units occupy extensive areas of the respective catchments (12S), or abundant feldspar, namely, near the downstream border of the Kunene Complex of SW Angola (37S).
Except for some sediments with a dominant recycled component, which contain very low feldspar, sampled muds usually yield similar proportions of quartz, phyllosilicates, and feldspar. Recycled muds (i.e., mainly sourced by previous sedimentary and metasedimentary units) can yield the highest contents of quartz (73%) or phyllosilicates (57%). The proportions of feldspar in felsic-derived and mixed-source muds are not substantially different (24−44%). Carbonates were detected in some felsic and recycled muds, but always in secondary to minor amounts.

Overall Geochemistry
Sand composition is widely variable, particularly with regards to the concentrations of mobile elements Na, Ca, K, Mg, Sr, Rb, and Ba ( Figure 3). Sands yield more SiO 2 than muds. Conversely, the contents of Fe 2 O 3 , MgO, REE (including Y and Sc), Cs, U, V, Zr, and Hf tend to be higher in muds. Lost on ignition is minor in sands (0.9-4.4%) and moderately high in muds (11.9-18.5%).
All sand deposits are dominated by SiO 2 (60.1-97.4%), containing Al 2 O 3 (0.9-20.3%), CaO (non-detected to 9.3%), Fe 2 O 3 (0.2-4.3%), K 2 O (0.02-4.56%), Na 2 O (non-detected to 3.1%), and TiO 2 (0.1-2.9%) in more variable percentages. The abundances of other elements are always below 1%. As expected, sands mainly derived from mafic-dominated source areas are enriched in CaO, Fe 2 O 3 , TiO 2 , MgO, and Sr ( Figure 3; Table 2). They are also distinguished by enrichment in Eu relative to the other sand samples. Felsic-derived sands are distinguished by relatively high contents of K 2 O and Rb. Sand derived from sedimentary units yields the lowest contents of mobile elements such as Na, Ca, and K and are the most enriched in silica (>93%).
Muds are still dominated by SiO 2 (47.1-65.9%) and, compared to sands, yield higher and more homogenous contents of Al 2 O 3 (13.1-25.5%) and Fe 2 O 3 (3.2-8.0%). Except for Na 2 O, which occurs in lesser amounts in recycled sediments, the element abundances measured in muds with different source geologies broadly overlap ( Figure 3).

Fingerprints of Primary Sources
The total contents of REE in the studied samples with mafic and felsic sources are very similar. However, some features of the REE patterns, such as Eu/Eu* and the REE fractionation parameters LaN/YbN and GdN/YbN, are higher in mafic-than felsic-derived deposits ( Figure 6). The links between these parameters and primary source rocks become particularly evident in the PCA conducted with REE data and major chemical elements (Figure 7). In the PCA map, Na and Ca concentrations are plotted with Eu/Eu*, reflecting the presence of Eu in plagioclase [4], which is abundant in mafic-derived deposits. These variables also appear linked with LaN/YbN and GdN/YbN, demonstrating that REE parameters are good proxies of the mafic contribution. The studied mafic-derived sediments appear to display slightly steeper HREE patterns than sediments enriched in the felsic component. The biplot Eu/Y also provides a good discrimination of sands with predominantly mafic and felsic sources, with higher Eu/Y in the former (Figure 5), confirming the ability of this ratio in the assessment of primary sources.
Sands with a mixed provenance can display REE patterns resembling either maficor felsic-dominated deposits ( Figure 6). The dominant primary source can be estimated with the aforementioned REE parameters mentioned before (Eu/Y, Eu/Eu*, LaN/YbN, LaN/SmN, and GdN/YbN). At least concerning the REE contents, the contributions from different rock types do not directly mimic the spatial representation of these rock types in the drainage areas. In other words, some areas appear to be supplying more sediment than others. Note, for example, sample 37S, for which the REE patterns display a typical mafic fingerprint, but in its catchment, felsic rocks are even more abundant than mafic rocks (Table 1). In this case, the proximity to potential source rocks is crucial, as the sample

Fingerprints of Primary Sources
The total contents of REE in the studied samples with mafic and felsic sources are very similar. However, some features of the REE patterns, such as Eu/Eu* and the REE fractionation parameters La N /Yb N and Gd N /Yb N , are higher in mafic-than felsic-derived deposits ( Figure 6). The links between these parameters and primary source rocks become particularly evident in the PCA conducted with REE data and major chemical elements (Figure 7). In the PCA map, Na and Ca concentrations are plotted with Eu/Eu*, reflecting the presence of Eu in plagioclase [4], which is abundant in mafic-derived deposits. These variables also appear linked with La N /Yb N and Gd N /Yb N , demonstrating that REE parameters are good proxies of the mafic contribution. The studied mafic-derived sediments appear to display slightly steeper HREE patterns than sediments enriched in the felsic component. The biplot Eu/Y also provides a good discrimination of sands with predominantly mafic and felsic sources, with higher Eu/Y in the former (Figure 5), confirming the ability of this ratio in the assessment of primary sources.
Sands with a mixed provenance can display REE patterns resembling either maficor felsic-dominated deposits ( Figure 6). The dominant primary source can be estimated with the aforementioned REE parameters mentioned before (Eu/Y, Eu/Eu*, La N /Yb N , La N /Sm N , and Gd N /Yb N ). At least concerning the REE contents, the contributions from different rock types do not directly mimic the spatial representation of these rock types in the drainage areas. In other words, some areas appear to be supplying more sediment than others. Note, for example, sample 37S, for which the REE patterns display a typical mafic fingerprint, but in its catchment, felsic rocks are even more abundant than mafic rocks (Table 1). In this case, the proximity to potential source rocks is crucial, as the sample was collected at the downstream end of the area with mafic outcrops, while felsic rocks are dominant only approximately 75 km upstream (Figure 2). A sample collected slightly upstream, closer to outcrops of felsic crystalline units and covering sedimentary successions (20aS), is much less enriched in the mafic component. The other sands, either with a mixed or recycled source, are dominated by the felsic component.  The discrimination of source areas is not as obvious for muds. The highest mafic contributions appear to occur in 15L (Figure 7), but the composition is still not substantially different from some muds with a distinct provenance (e.g., 9L, 40L, 43L; Table 1). Although the geochemistry of fine-grained sediment fractions, frequently with an emphasis on REE, is an important tool in provenance investigations [6,7,20,21,39,40], the present results show that the interference of other exogenous processes renders the recognition of source areas more complicated.

Influence of Sediment Cycling on REE Geochemistry
Comparisons of fluvial sediments with their source rocks indicated that depositional cycles are responsible for discernible changes in REE [14,[41][42][43][44][45]. These transformations can be associated with a different process. In the following sections, we discuss the effects of weathering and sorting.

Weathering
REE are mobile within weathering profiles, particularly during the early stages of alteration [46,47]. Still, because of the decomposition of labile minerals that host minor REE and the retention of these elements in secondary minerals, weathering can promote an REE enrichment relative to the parent materials. To better understand the effect of The discrimination of source areas is not as obvious for muds. The highest mafic contributions appear to occur in 15L (Figure 7), but the composition is still not substantially different from some muds with a distinct provenance (e.g., 9L, 40L, 43L; Table 1). Although the geochemistry of fine-grained sediment fractions, frequently with an emphasis on REE, is an important tool in provenance investigations [6,7,20,21,39,40], the present results show that the interference of other exogenous processes renders the recognition of source areas more complicated.

Influence of Sediment Cycling on REE Geochemistry
Comparisons of fluvial sediments with their source rocks indicated that depositional cycles are responsible for discernible changes in REE [14,[41][42][43][44][45]. These transformations can be associated with a different process. In the following sections, we discuss the effects of weathering and sorting.

Weathering
REE are mobile within weathering profiles, particularly during the early stages of alteration [46,47]. Still, because of the decomposition of labile minerals that host minor REE and the retention of these elements in secondary minerals, weathering can promote an REE enrichment relative to the parent materials. To better understand the effect of weathering on REE, PCA was performed with parameters that characterize the REE geochemistry (∑LREE, ∑HREE, Eu, Eu/Eu*, Ce/Ce*, La N /Yb N , La N /Sm N , and Gd N /Yb N ) and a selection of compositional parameters that can be regarded as weathering proxies (Figure 8). weathering on REE, PCA was performed with parameters that characterize the REE geochemistry (∑LREE, ∑HREE, Eu, Eu/Eu*, Ce/Ce*, LaN/YbN, LaN/SmN, and GdN/YbN) and a selection of compositional parameters that can be regarded as weathering proxies (Figure 8). The extent of weathering transformations can be estimated through many different indices based on the concentration of major elements. Their formulations, advantages, and weaknesses were discussed in more detail elsewhere [48][49][50]. The parameters adopted in the PCA include the Chemical Index of Alteration (CIA; [51]), a modified CIA that does not consider CaO (CIX; [52]), and the Mafic Index of Alteration that also considers the fate of Fe/Mg silicates (MIA; [53]). The values obtained for all these weathering proxies result from ratios between the concentrations of one or more non-mobile elements and the concentrations of sets of mobile elements in the sample (Table 3). In addition, the weathering The extent of weathering transformations can be estimated through many different indices based on the concentration of major elements. Their formulations, advantages, and weaknesses were discussed in more detail elsewhere [48][49][50]. The parameters adopted in the PCA include the Chemical Index of Alteration (CIA; [51]), a modified CIA that does not consider CaO (CIX; [52]), and the Mafic Index of Alteration that also considers the fate of Fe/Mg silicates (MIA; [53]). The values obtained for all these weathering proxies result from ratios between the concentrations of one or more non-mobile elements and the concentrations of sets of mobile elements in the sample (Table 3). In addition, the weathering intensity was assessed through the parameters α Al E that measure the depletion of different mobile elements (E) using a ratio of their concentrations to the concentration of non-mobile element Al and applying a normalization to a reference material (e.g., UCC, as adopted in the present research) [54]. Using both sand and mud samples, the PCA with weathering indices and REE features separates these two sets of deposits ( Figure 8A). However, one sand sample collected in the upper reaches of the Caculuvar sub-basin that was sourced from meta-sedimentary units is plotted along with muds. This PCA also clusters most variables indicative of the weathering intensity (CIA, CIX, MIA, α Al Na, and α Al Ca) with LREE, HREE, and Eu in the region of the PCA where the mud deposits are plotted. Only α Al Mg and α Al K, which are strongly influenced by the source rock composition [18,56], are not grouped with the other weathering parameters, being plotted on opposite sides of the PCA map.
To remove the grain size effect on the sediment composition, two additional PCAs were performed for mud and sand deposits in separate. Due to the limited number of samples, some variables found to be redundant were not included (Gd N /Yb N and MIA in both PCAs; CIX, α Al K, and α Al Ca in the PCA for muds) and only ∑REE was considered.
In the PCA map for sand deposits, Eu/Eu* and La N /Yb N appear linked and in opposition to Ce/Ce*, while weathering indices are plotted orthogonally, indicating that surface alteration is not a major factor influencing REE patterns ( Figure 8B). Regarding muds, weathering indices appear to correlate with Ce/Ce* in opposition to fractionation indices ( Figure 8C), suggesting a concentration of Ce during surface alteration and a preferential release of La relative to Sm and HREE. The limited number of mud samples does not allow safe conclusions about weathering effects. However, they are compatible with the reported preferential mobilization of LREE [14,57] and the concentration of Ce in secondary oxides formed during weathering [58,59]. The PCA for muds also shows REE contents in opposition to Eu/Eu*. As secondary REE-bearing minerals tend to show stronger negative Eu anomalies (lower Eu/Eu*) than the primary minerals they result from (e.g., feldspar, mica, and ferromagnesian silicates [60,61]), surface decomposition could be promoting REE enrichment.

Sorting
The segregation of sediment materials with different sizes, shapes, and densities plays a fundamental role in sediment geochemistry [14,17,42,62,63]. Significant amounts of REE in river sediments are hosted by particles whose concentrations are strongly influenced by sorting processes, such as heavy minerals [64][65][66] and clay minerals [65,67,68]. For the studied deposits, the PCA performed with geochemical data suggests that sorting is the principal factor determining the compositional variability (Figure 7).
Sorting accounts for the separation of fine-grained particles (clay to fine sand, depending on flow conditions) that are preferentially transported as suspended loads from coarser particles mainly transported as a bedload [69]. Muds are characterized by enrichment in elements that tend to be hosted in phyllosilicates, such as Al (clay minerals and micas), Fe and Mg (e.g., biotite, chlorite, and some illite), and K (muscovite and some illite). Titanium, along with very fine-grained REE minerals also tends to be concentrated with the finer fractions. On the other hand, sands are enriched in quartz and contain variable amounts of feldspars and minor heavy minerals. REE can also be frequently adsorbed onto clay minerals and Fe-Mn oxides, which may hold the majority of the leachable REE compo-nent [14]. Except for some mafic-derived sands, these deposits contain minor amounts of heavy mineral grains. Hence, sorting and quartz dilution are partially responsible for the low contents of most REE minerals in sands. Occasional exceptions are Eu, hosted in feldspars, and Ce, probably associated with coatings ( Figure 9).
Geosciences 2021, 11, 384 14 of 18 micas), Fe and Mg (e.g., biotite, chlorite, and some illite), and K (muscovite and some illite). Titanium, along with very fine-grained REE minerals also tends to be concentrated with the finer fractions. On the other hand, sands are enriched in quartz and contain variable amounts of feldspars and minor heavy minerals. REE can also be frequently adsorbed onto clay minerals and Fe-Mn oxides, which may hold the majority of the leachable REE component [14]. Except for some mafic-derived sands, these deposits contain minor amounts of heavy mineral grains. Hence, sorting and quartz dilution are partially responsible for the low contents of most REE minerals in sands. Occasional exceptions are Eu, hosted in feldspars, and Ce, probably associated with coatings ( Figure 9). Figure 9. Plots of CIA-WIP applied to assess weathering, recycling, and grain size control on REE geochemistry. Bubble sizes represent ∑REE (A), REE fractionation determined as LaN/GdN (B), Eu anomaly (C), and Ce anomaly (D). Eu/Eu* is Eu anomaly; Ce/Ce* is Ce anomaly.

Effect of Multiple Depositional Cycles
Intense quartz dilution can be accomplished through recycling, which promotes a progressive enrichment in chemically stable and mechanically resistant minerals [70]. Feldspar is frequently the most abundant mineral of crystalline rocks that does not endure a sediment cycle as quartz. With the decrease in the feldspar contents, the negative Eu anomaly becomes progressively more evident, reflecting the release of Eu. Recycling is also responsible for the depletion of fine-grained detritus, which is frequently dominated by secondary clay minerals [50]. The quartz dilution effects on REE concentrations are particularly obvious for sands sourced from the Kalahari Group (Figure 7). These sediments are strongly enriched in quartz (92-100%) and yield the lowest REE contents, with Figure 9. Plots of CIA-WIP applied to assess weathering, recycling, and grain size control on REE geochemistry. Bubble sizes represent ∑REE (A), REE fractionation determined as La N /Gd N (B), Eu anomaly (C), and Ce anomaly (D). Eu/Eu* is Eu anomaly; Ce/Ce* is Ce anomaly.

Effect of Multiple Depositional Cycles
Intense quartz dilution can be accomplished through recycling, which promotes a progressive enrichment in chemically stable and mechanically resistant minerals [70]. Feldspar is frequently the most abundant mineral of crystalline rocks that does not endure a sediment cycle as quartz. With the decrease in the feldspar contents, the negative Eu anomaly becomes progressively more evident, reflecting the release of Eu. Recycling is also responsible for the depletion of fine-grained detritus, which is frequently dominated by secondary clay minerals [50]. The quartz dilution effects on REE concentrations are particularly obvious for sands sourced from the Kalahari Group (Figure 7). These sediments are strongly enriched in quartz (92-100%) and yield the lowest REE contents, with their geochemistry indicating that they are associated with the most intense surface decomposition ( Figure 8). Different REE are not affected in the same way during sediment cycling. For example, LREE appears to be preferentially leached relative to HREE, leading to a decrease in La N /Yb N , but the higher values of Ce/Ce* show that Ce is not hosted by the same minerals as the remaining LREE. Part of it is likely retained in coatings that evolved sand grains instead of fine-grained particles, which tend to be preferentially lost during sediment cycles.
Recycling effects can be assessed with a combination of compositional parameters. Unlike CIA and other indices used in this paper to assess the weathering intensity, the Weathering Index of Parker (WIP; [55]) is strongly influenced by the addition of recycled quartz. Hence, plots of CIA-WIP have been used to understand how recycling controls sediment composition [52,71]. Here, we used this biplot with a third dimension, the bubble size, to represent a feature of the REE geochemistry ( Figure 9). The diagram clearly shows how quartz dilution in recycled sands is responsible for REE impoverishment. This set of samples is also characterized by particularly low Eu/Eu*, responding to feldspar decomposition and breakdown, and flatter REE profiles due to the solubility of LREE. As expected, the composition of muds is not as affected by quartz dilution as observed with sands.

Conclusions
Sediment cycles, through weathering and sorting, influence the REE geochemistry of sands and muds differently. The decomposition of labile components and the segregation of particles according to size, shape, and density culminate with significant quartz enrichment in sand deposits, while mud deposits become enriched in secondary minerals. Quartz dilution explains the low REE content in sands, particularly in those most affected by recycling. Sand deposits with evidence of the most intense surface decomposition hold the lowest amounts of REE. As weathering transformations are accumulated during successive sediment cycles, the composition of recycled sands also points to more intense weathering, but the REE depletion should be ascribed to quartz dilution. REE concentrations in muds are influenced by two processes with opposite effects. Quartz dilution can be responsible for some REE depletion, while the decomposition of labile components and the retention of REE in secondary minerals account for some REE enrichment. The present research confirms that several parameters of REE geochemistry (e.g., Eu/Y, Eu/Eu*, La N /Yb N , La N /Sm N , and Gd N /Yb N ), particularly for sands, provide excellent information about the source geology. Surface decomposition seems to be responsible for an increase in Ce and some LREE depletion relative to HREE in muds but has no obvious consistent effects on the REE patterns of sands.