Sediment generation in humid Mediterranean setting: Grain-size and source-rock control on sediment geochemistry and mineralogy (Sila Massif, Calabria)

    Sediment Generation in humid Mediterranean setting: Grain-size and Source-rock Control on Sediment Geochemistry and Mineralogy (Sila Massif, Calabria) Hilmar von Eynatten, Raimon Tolosana-Delgado, Volker Karius, Kai Bachmann, Luca Caracciolo PII: DOI: Reference:

S0037-0738(15)00226-2 doi: 10.1016/j.sedgeo.2015.10.008 SEDGEO 4926

To appear in:

Sedimentary Geology

Received date: Revised date: Accepted date:

3 July 2015 15 October 2015 19 October 2015

Please cite this article as: von Eynatten, Hilmar, Tolosana-Delgado, Raimon, Karius, Volker, Bachmann, Kai, Caracciolo, Luca, Sediment Generation in humid Mediterranean setting: Grain-size and Source-rock Control on Sediment Geochemistry and Mineralogy (Sila Massif, Calabria), Sedimentary Geology (2015), doi: 10.1016/j.sedgeo.2015.10.008

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT Sediment Generation in humid Mediterranean setting: Grain-size and Source-rock Control on Sediment Geochemistry and Mineralogy (Sila Massif, Calabria)

PT

Hilmar von Eynatten1, Raimon Tolosana-Delgado2, Volker Karius1, Kai Bachmann2, and Luca

Geowissenschaftliches Zentrum der Georg-August-Universität Göttingen, Abteilung

SC

1

RI

Caracciolo3

NU

Sedimentologie/ Umweltgeologie, Göttingen, Germany

Helmholtz Institute Freiberg for Resource Technology, Freiberg, Germany

3

Chemostrat Ltd, Welshpool, UK

MA

2

TE

D

Corresponding author: [email protected]

AC CE P

Keywords: geochemistry; mineralogy; chemical weathering; comminution; provenance; compositional linear regression

Abstract

Grain-size control on sediment composition is investigated in modern proximal sediment from the Sila Massif, where basic to felsic intrusive rocks are exposed in a Mediterranean humidtemperate upland climate. Samples were taken from small creeks and weathering profiles from three areas reflecting different bed rock composition. Samples were separated into eleven grain size fractions from very coarse sand to clay and analyzed by (i) X-ray fluorescence for chemical composition, and (ii) X-ray diffraction and Mineral Liberation Analysis for mineralogical composition. The chemical composition vs. grain size relations were modelled by compositional linear regression. Mineralogical composition of selected samples is used to substantiate the interpretations based on geochemistry.

1

ACCEPTED MANUSCRIPT Results reveal a high degree of chemical weathering with chemical index of alteration (CIA) up to 92. High CIA values are restricted to the fine-grained fractions, while sand-sized sediment average at low to moderate CIA values (~ 60). Although strongly weathered, the

PT

three sample suites reflecting basic to felsic plutonic bed rock can be effectively

RI

discriminated across all grain-size classes using trace elements such as V, Rb, and Sr. Linear trend modelling and mineralogical data reflect similar patterns for all sample suites

SC

implying similar processes independent of source rock composition. This includes overall

NU

decrease of quartz and K-feldspar over the full grain-size range from very coarse sand to clay, which is contrasted by overall increase of sheet silicates from coarse to fine. Among the

MA

latter, increase of clay minerals strongly outpaces the increase of micas in silt to clay fractions. A more complex behaviour is shown by plagioclase, which is most abundant in intermediate grain-size fractions for all sample suites. This is likely caused by initial

TE

D

hydrolysis along cleavage and twinning planes and subsequent breakage of plagioclase crystals into smaller fragments. Towards finer grain size, intense hydrolysis has destroyed

AC CE P

most feldspars.

1. Introduction

The chemical and mineralogical composition of sediments and sedimentary rocks is strongly controlled by grain size (e.g. Blatt et al., 1972; Basu, 1976). Other factors include source rock composition, weathering, and physical as well as chemical processes that modify the sediment while being transported from source to sink (e.g. Johnsson, 1993). The effect of grain size is generally thought to exceed the impact of all other control factors on the variability of sediment composition (e.g. Garzanti et al., 2011; Bloemsma et al., 2012). Nevertheless, many studies inferring geologic or paleoclimatic conditions from sediment characteristics do not consider grain size influence adequately. The high relevance of grain-size for analyzing sediment composition is contrasted by the fact that analyzing the entire grain-size spectrum is often impossible, due to limitations of

2

ACCEPTED MANUSCRIPT the accessible rock record as well as specific methodological constraints (e.g., microscopic determination of framework grains or heavy minerals is restricted to certain grain-size ranges). Therefore, classic approaches to infer geological conditions from sediment

PT

composition typically rely on data obtained from a given narrow grain size range (e.g. Basu

RI

et al., 1975; Ingersoll et al., 1984; Morton and Hallsworth, 1994). The shortcomings of such approaches have been recently outlined by, for instance, Garzanti et al. (2009). Therefore,

SC

comparing data obtained from different grain size fractions or from sediment samples with

NU

contrasting grain size distributions inevitably requires an understanding of the compositional relations across grain size grades. If these relations are controlled by purely physical

MA

processes such as hydrodynamics or mechanical comminution, relatively simple empirical or numerical models can be used to correct for the grain-size dependent control on composition (e.g. Garzanti et al., 2009; von Eynatten et al., 2012). For complex interactions of chemical

TE

D

and physical processes, which must be considered the rule rather than the exception in natural systems, Bloemsma et al. (2012) developed a statistical method to separate the

AC CE P

grain-size dependent part of compositional variation from a residual part, which reflects other controlling factors such as provenance or diagenesis. Application of this tool revealed that trends of grain size vs. composition strongly depend on the specific geological setting, implying that more empirical studies and models are needed to better understand individual settings.

The aim of the paper is to describe and quantify the effects of grain size and sourcerock lithology on geochemical and mineralogical composition of sediments in a specific case study. The samples were taken from three areas of the humid-temperate Sila Massif, southern Italy, composed of contrasting basic to felsic plutonic source rocks. A broad grain size spectrum covering more than three orders of magnitude from very coarse sand to clay is considered. Previous studies have documented significant chemical weathering of soils and sediment in the Sila Massif (e.g. Scarciglia et al., 2007). In a first step, weathering and provenance effects on sediment composition, as well as their variations across the grain size grades from coarse to fine is evaluated. In a second step, the results are compared to a

3

ACCEPTED MANUSCRIPT previous study (von Eynatten et al., 2012) from a highly contrasting climatic setting where chemical weathering is negligible and thus mechanical comminution controls sediment composition across grain-size grades.

PT

The study is essentially based on geochemical data because they can be obtained

RI

easily and precisely for the full grain size spectrum from very coarse sand to clay, which covers >95% of the global clastic sediment budget given that conglomerates and breccias

SC

comprise only a few percent of the total clastic sediment record (e.g. Pettijohn, 1975).

NU

Mineralogical composition is obtained by X-ray diffraction for selected samples and grainsize fractions to verify interpretations based on geochemical data. Moreover, Mineral

MA

Liberation Analysis (MLA) has been tested on some samples because it potentially provides a tool to analyze mineral composition over a large grain size range.

TE

D

2. Study area and sampling strategy

AC CE P

The study area is located in the uplands of the Sila Massif in northern Calabria (southern Italy) at altitudes between approx. 1150 and 1600 m a.s.l., characterized by a smooth topography with gentle hills and slightly incised streams in between. It has a typical Mediterranean humid-temperate upland climate, characterized by mean annual precipitation of 1400 to 1800 mm and mean annual temperature of 10-12 °C, with mean monthly temperatures ranging between -1°C and 18°C (cf. Sca rciglia et al., 2012). The climatic and geomorphological conditions have caused deep and intense weathering of the respective bed rocks, which has been investigated in numerous weathering profiles (e.g. Le Pera and Sorriso-Valvo, 2000; Scarciglia et al. 2007), and is reflected in the sediments exported from the Sila Massif through the main drainage systems, i.e. the Crati and Neto River basins (e.g. Le Pera et al., 2001). The Calabrian Massif constitutes an allochthonous crustal segment of the western Variscan belt in Europe, exposed in a structurally high position between the Alpine Mesozoic to Cenozoic sedimentary nappe piles of the Apennines to the North and the Maghrebides in

4

ACCEPTED MANUSCRIPT Sicily to the Southwest (Ayuso et al., 1994; Graessner et al., 2000). The Sila Massif forms the northern part of the Calabrian Massif and is mainly composed of high-grade (granulite facies) metapelites in the West and Southwest, low to medium-grade metasedimentary rocks

PT

in the East, and the Sila Batholith comprising most of the central part of the Sila Massif (Fig.

RI

1). Peak metamorphism is dated at around 300 Ma, roughly coeval to the Late Variscan intrusion ages of the batholith (Graessner et al., 2000). The Sila Batholith is a complex

SC

intrusive body composed of variable plutonic rocks that range from granites and

NU

granodiorites to tonalite, diorite, and even gabbroic rocks (Messina et al., 1991; Caggianelli et al., 1994; Ayuso et al., 1994). After intrusion the plutons cooled and were exhumed to mid-

MA

crustal levels in Permian to Triassic time (Ayuso et al., 1994; Graessner et al., 2000). Final exhumation of the Sila Massif is constrained by zircon and apatite fission-track analysis to Oligocene to Mid-Miocene time (35 to 15 Ma; Thomson, 1994). Late Pleistocene to Holocene

TE

D

catchment-averaged erosion rates for the flat uplands are low, scattering around 0.1 mm/a as inferred from cosmogenic nuclide data (Olivetti et al. 2012).

AC CE P

The samples were collected in June 2010 (RT3) and September 2013 (RT4). They derive from three different areas of the Sila Massif, which were selected to represent the diverse range of the Sila Batholith from felsic to basic plutonic rocks. From north to south, the first area is located to the northeast of Lago Cecita and exposes felsic rocks of granodiorite to monzogranite composition (i.e. cordierite-bearing biotite-muscovite granodiorite to monzogranite according to Messina et al. 1991; peraluminous granite according to Graessner et al. 2000). These rocks range in SiO2 content from 68 to 74 wt.-% (Ayuso et al. 1994) and constitute the most felsic sample suite, termed F2 (Fig. 1). The second area is located around Carlo Magno and Silvana Mansio north of Lago Arvo. It belongs to the widely exposed hornblende and biotite-bearing tonalites to granodiorites, which range in SiO2 content from 60 to 71 wt.-% (Messina et al. 1991; Ayuso et al. 1994). The selected area constitutes the felsic to intermediate sample suite, termed F1 (Fig. 1). Some bedrock samples of this area indicate a dominant composition of F1 bedrocks in the less felsic part of the tonalites to granodiorites with SiO2 content around 61 to 62%. The third area is located

5

ACCEPTED MANUSCRIPT directly northeast of Lago Arvo around the village of Rovale and exposes basic to intermediate rocks of the so-called Rovale Zone or Complex. This small area (~15 km2) consists of gabbroic rocks (norites, amphibole gabbros; 42-49 wt.-% SiO2, 18-26 wt.-%

PT

Al2O3) in its western part close to Rovale, while in its eastern part plagioclase and amphibole-

RI

rich diorites and tonalites prevail (52-57 wt.-% SiO2, 17-18 wt.-% Al2O3), with small local

sampling area of the study, termed B (Fig. 1).

SC

gabbroic bodies (Caggianelli et al., 1994). This area constitutes the basic to intermediate

NU

Two types of samples have been collected: modern sediment from small proximal creeks (Sed, N=10) as well as material from weathering profiles or grus exposed in roadcuts

MA

(WP, N=6). Small creek means channel width less than 2 meters. Sampling was intended to cover a broad grain-size spectrum from very coarse sand to clay from each sampling site. Therefore, Sed-samples may represent a mixture of up to three subsamples taken from a

TE

D

small area of no more than approx. 20m2. Grus samples were taken from the debris cones developed along roadcuts through deeply disintegrated granitoid bed rock. While sample

AC CE P

suites F1 and F2 comprise both Sed and WP samples (altogether seven samples for each suite), only two sediment samples are available from sampling suite B (Table 1).

3. Methods

Samples were separated in up to eleven grain-size fractions ranging from very coarse sand (1 to 2 mm, -1 < Φ < 0) to clay (< 2 µm, Φ > 9), with Φ being the negative logarithm of the grain diameter d to the basis of 2 (Φ = - log2 d). Grain size separation was achieved in oneΦ-unit steps by wet sieving of the sand-sized fractions (-1 < Φ < 4), and by gravity settling for all finer fractions (Φ >4). Each separation step of the fine-grained fractions in Atterbergcylinders was repeated 8 to 14 times until quantitative separation of the respective grain-size fraction was achieved. The remaining suspension was vacuum-filtered through 0.2 µm cellulose acetate filters. Therefore, the finest fraction actually represents the grain size fraction 9 < Φ < 12.

6

ACCEPTED MANUSCRIPT Each sediment grain-size fraction was powdered, fused using lithium metaborate, and subsequently analysed by X-ray fluorescence (XRF) method using a PANalytical AXIOS Advanced sequential X-ray spectrometer at the Geoscience Center Göttingen. All samples

PT

and grain size fractions were analyzed for ten major element oxides (SiO2, Al2O3, TiO2,

RI

Fe2O3tot = total iron calculated as Fe2O3, MnO, MgO, CaO, Na2O, K2O, P2O5), the loss on ignition (LOI), and 16 trace elements (Ba, Co, Cr, Cu, Ga, Nb, Nd, Ni, Pb, Rb, Sc, Sr, V, Y,

SC

Zn, Zr). The RT4 samples were additionally analyzed for Ce, Hf, La, Mo, Sm, Th, U, and Yb.

NU

The elemental concentrations of all samples and grain-size fractions are available in the supplementary data file.

MA

Analytical precision is typically in the range of 1% for major elements and 5% for trace elements. Accuracy of measurements was tested against internationally certified reference materials with values for JB-3, JA-2, and JR-1 taken from Imai et al. (1995) and values for

TE

D

JG-2 taken from Guevara et al. (2001). Accuracy is better than 3% for all major oxides except for TiO2 (JG-2, 6%), MnO (JA-2, 4%, JG-2, 14%), MgO (JA-2, 4%, JR-1, 8%), P2O5

AC CE P

(JA-2, 5%, JG-2, 46%, JR-1, 14%) and Fe2O3tot (JG-2, 7%), and better than 8% for all trace elements except for Ba (JR-1, 15%), Ce (JB-3, 12%), Ga (JR-1, 12%), Nb (JG-2, 31%,), Rb (JB-3, 14%), Sr (JG-2, 75%), Zn (JB-3, 9%) and Zr (JG-2, 45%). Observed deviations higher than 10% are mostly related to concentrations near the detection limit and/or below or at the lower end of the data range of our samples. The only exception is Zr: although all four reference materials have comparable Zr concentrations (96-180 ppm), the deviations for JA2, JB-3 and JR-1 are only 2% and thus strongly contrast the high deviation for JG-2. Selected samples were analysed by X-ray diffraction (XRD) for identification and quantification of mineral phases. Individual grain size fractions were wet milled with a McCrone micronising mill in distilled water, evaporated to dryness at 60°C, and pulverised using agate mortar. 10% ZnO was admixed as internal standard. X-ray diffraction analyses were performed on a Philips X’Pert MPD, equipped with a PW3050 Goniometer, and CuKα radiation. Qualitative phase analysis was achieved using a search match routine of the program X’Pert Highscore (Version 2.2a) (Panalytical B.V., 2006). Quantitative phase

7

ACCEPTED MANUSCRIPT analysis was performed with Rietveld refinements based on the diffraction data using the program AutoQuan (Version 2.80) (GE Inspection Technologies, 2014). All Mineral Liberation Analysis (MLA) measurements were run on a scanning electron

PT

microscope FEI Quanta 650F equipped with two Bruker Quantax X-Flash 5030 Silicon Drift

RI

Detectors (SDD) and the specific software MLA Suite 3.1.4 for automated data acquisition. The MLA measurements were carried out at the Geometallurgy Laboratory at the Helmholtz

SC

Institute Freiberg. More detailed information about the functionality of the MLA system can be

NU

found in Fandrich et al. (2007). The operating conditions used for this study are listed in Table 2. All measured datasets were processed with the MLA Image Processing software,

classification of the measurements.

MA

version 3.1.4. A coincidence threshold of 90% with the standard spectra was used for the

The statistical techniques applied are fundamentally based on the log-ratio approach

TE

D

to compositional data, as introduced by Aitchison (1986). This includes the evaluation of the covariance structure via biplot analysis, the calculation of geometric means and predictive

AC CE P

regions, and the linear regression model. All this is performed on transformed data, i.e. the data are transferred from the constrained compositional space to the Euclidean space by centred log-ratio transformation (clr; Aitchison, 1986). The transformation frees the data from the constant-sum constraint implying, for instance, that incremental changes in concentrations in clr-scale are independent of the respective level of concentration, i.e. the clr-scale provides a translation-invariant measure of the degree of change (e.g. von Eynatten et al., 2003a). For linear regression, the transformation is followed by standard multiple regression techniques to estimate the coefficients of the compositional linear trend. The trend may include steps if distinct breaks are observed at certain grain-size thresholds. Methodological details are described elsewhere (Tolosana-Delgado and von Eynatten, 2009; Tolosana-Delgado and van den Boogart, 2011; von Eynatten et al. 2012; and references therein).

4. Results

8

ACCEPTED MANUSCRIPT

4.1 Major and trace element concentrations and patterns The variability of the major element oxide concentrations for the three sample suites

PT

B, F1, and F2 against the eleven separated grain-size fractions from very coarse sand to clay

RI

is illustrated in Figure 2. Contrasts between the three suites are clearly visible for some elements (e.g. Si, Ca, K, Fe) but typically converge or even disappear towards finer grain

SC

size. Some major element oxides show general trends such as overall decrease of SiO2 and

NU

increase of Al2O3 and loss on ignition (LOI) with decreasing grain size. Other elements display specific patterns for individual parts of the grain size spectrum for some or all of the

MA

three sample suites (e.g. Na, Ca, P, Ti, Fe). For instance, CaO displays a more or less pronounced increase in the coarser fractions from very coarse sand to very fine sand and a decrease in the fine fractions from very coarse silt to clay. The latter is very pronounced for

D

suites B and F1. Maximum CaO concentrations are thus measured in the very fine sand to

TE

very coarse silt fractions (3 < Φ < 5) for all sample suites (Fig. 2). A similar observation is

AC CE P

made for Na2O but is restricted to the intermediate to felsic sample suites F1 and F2. In the basic sample suite Fe2O3, TiO2, and MnO are peaking in the fine to very fine sand range (2 < Φ < 4), accompanied by a small trough in SiO2 and MgO in this range. P2O5 is peaking in very coarse silt (4 < Φ < 5) for all sample suites, decreases in the coarse to fine silt fractions and then again notably increases in the clay fraction (Φ > 9) (Fig. 2). Upon visual inspection of the raw percentage data K2O appears to discriminate best between all suites across the full grain size range, while MgO separates best the basic suite B from the rest, and CaO separates best F1 from F2. Trace element concentrations typically vary by an order of magnitude across sample suites and grain-size range (Fig. 3). Some elements show general increase with decreasing grain size for all sample suites (Th, Pb, U, Ga) while three more elements show this trend for F1 and F2 only (Zn, Cr, Nb). Zr and Hf peak in the very coarse silt fraction for all sample suites and Zr/Hf ratios scatter around 40 without grain-size dependence. Y and the light (LREE: La, Ce) and middle rare earth elements (MREE: Sm, Nd) all peak in the very fine

9

ACCEPTED MANUSCRIPT sand to very coarse silt fractions (3 < Φ < 5) for all sample suites, however, in F2 the concentrations of these elements stay high over the entire fine range (Fig. 3). V and Nb peak at 3 < Φ < 5 in the basic sample suite only. Upon visual inspection Th, U, and Rb appear to

PT

discriminate best between the three sample suites across the full grain size range. V nicely

RI

separates the basic suite from the rest, while Sr and Ba nicely separate F1 from F2 across

SC

the full grain size range (Fig. 3).

NU

4.2 Covariance structure

Evaluation of the covariance structure of the dataset is performed using compositional

MA

biplots, which provide a 2D-representation of multivariate compositional data based on principal components (e.g. Aitchison 1990, von Eynatten et al. 2003b). The data set has

D

been transformed using centered logratio transformation (clr) to account for the specific

TE

compositional nature of geochemical data (Aitchison, 1986). This implies that the center of the biplot corresponds to the geometric mean of the dataset.

AC CE P

The major element biplot of the entire data set (Fig. 4A) reveals a strong grouping depending on sample suite and grain size. The first principal component (PC1, x-axis) captures ~49% of the total variability and is characterized by strong positive loadings of K and Na and negative loadings of Mg,Ti, Fe, P, Ca, and LOI. The second principal component (PC2, y-axis) captures ~22% of the total variability and is characterized by strong positive loading of LOI and strong negative loading of CaO. PC1 effectively separates sediments from basic source rocks from intermediate to felsic ones (i.e. sample suites B vs. F1 vs. F2; Fig. 4A). This, however, holds only for grain size fractions coarser than medium silt (i.e. Φ ≤ 6). Towards finer grain size the contrast between the sample suites decreases with decreasing scores of PC1 and increasing scores of PC2. PC2 groups the very coarse silt to sand fractions for each sample suite, and effectively orders the finer grain size fractions by increasing PC2 scores (brown arrows in Fig. 4). The mixed biplot of major elements and selected trace elements (Fig. 4B) reveals a roughly similar picture. While V behaves like the mafic elements Mg and Ti (strong negative

10

ACCEPTED MANUSCRIPT loading on PC1), Rb, Y and Zr have positive loadings on PC1. Zr also has strong negative loading on PC2. Zn appears to be strongly related to the high LOI of the finest fractions (positive loading on PC2). The separation of the three sample suites appears even more

PT

pronounced in the mixed biplot. Moreover, grain size separation is enhanced. Fine-grained

RI

fractions still have highest scores on PC2, while the coarser fractions are now better differentiated by PC2 (Fig. 4B): lowest scores in each suite are restricted to very fine sand

SC

and very coarse silt (3 ≤ Φ < 5), while the coarser sand fractions yield intermediate scores on

NU

PC2.

Comparing only the two granitoid suites F1 and F2 allows for focussing the biplot

MA

analysis on features controlled by parameters other than source rock, such as grain-size and/or sample type. A threefold structure becomes visible that is predominantly controlled by

D

grain size (Fig. 5): (i) very coarse to medium sand (Φ ≤ 2) is mainly characterized by positive

TE

scores on PC1 and negative scores on PC2 and PC3 translating to high proportions of Si, Na, K, and Rb, (ii) very fine sand and very coarse silt (3 ≤ Φ < 5) is mainly characterized by

AC CE P

positive scores on PC1, PC2 and PC3 translating to high proportions of Ca, Zr, Y and P, and (iii) medium silt to clay (Φ > 6) characterized by negative scores on PC1 while PC2 and PC 3 are rather unspecific. The latter translates to high proportions of Zn, V, Ti, Mg, Fe, and P. In contrast, the sample type (i.e. sediment from creek vs. weathering profile) does not reveal systematic differences in the biplots. However, the fine-grained fractions from the weathering profile of suite F2 (WP, triangles upside down) appear to be separated from their sediment equivalents (Sed, rhombs) by high proportions of Y relative to e.g. Ca. This contrast explains the high variance for F2 samples regarding Y and REE concentrations (Fig. 3, third row).

4.3 Linear regression model The modelling has been restricted to the major elements (including LOI) and the intermediate to felsic suites F1 and F2, as only these provide reasonable numbers of samples. Constraints on the linear regression model like (i) similar slope over the full grainsize range and (ii) distinct steps at Φ = 4 and Φ = 8, as applied in a previous study (von

11

ACCEPTED MANUSCRIPT Eynatten et al. 2012), resulted in a poor fitting of the data especially for Ca, Na, P as well as Si and K. The raw data already point at such misfit because some elements show distinct breaks in increase/decrease patterns at around Φ = 4 (Fig. 2). Moreover, the three-pole

PT

structure governed by grain size as revealed in Figure 5A calls for modifications from a

RI

simple uniform linear trend over the full grain-size range. We, therefore, chose model

SC

constraints which account for the observed breaks in slope, i.e. the slopes are allowed to be different for the sand fraction (Φ < 4) compared to the silt-clay fraction (Φ > 4). Further, a

NU

possible step has been allocated at Φ = 4 but no further steps appear necessary. The regressions are calculated separately for F1 and F2 with intercepts as well as the step at Φ =

MA

4 being independent, while the slopes are forced to be equal for F1 and F2. The resulting linear trend indicates a reasonable fit to the compositional range of

D

sample suites F1 and F2 for all major elements except for P2O5 (Fig. 6). The relatively poor

TE

linear fit for P is due to its double peaking in very coarse silt and clay with a significant trough in between. Well adjusted negative slopes for the full grain-size range (i.e. decreasing

AC CE P

relative concentration with decreasing grain size) are obtained for Si, Na, and K, while general positive slopes are obtained for Fe, Mg, Ti, and LOI (Fig. 7). In contrast, Ca and Al reveal opposing slopes for sand and silt-clay: Ca show distinct increase from coarse sand to very fine sand and very coarse silt, and then markedly decreases towards clay, while Al behaves vice versa (Fig. 6). However, Al exhibits very small slopes in both parts of the grainsize spectrum implying relatively small changes over the full grain-size spectrum.

4.4 Mineralogical composition The XRD analysis of three selected samples covers the grain-size spectrum from coarse sand to clay as well as all three sample suites B, F1, and F2. Inferred mineral compositions include quartz, K-feldspar (orthoclase, microcline), plagioclase (almost pure albite to Ca-rich varieties), amphibole (hornblende), pyroxene (enstatite, augite), Ti-phases (ilmenite, rutile), muscovite, biotite, chlorite, clay minerals (illite, smectite, kaolinite), and gibbsite. Quantitative phase models are well adjusted if the quality parameter 1-ρ of the

12

ACCEPTED MANUSCRIPT Rietveld refinement is approx. 1%. This criterion is reasonably fulfilled for the sample suites F1 and F2 (1-ρ: 0.8%-3.5%) but less well for sample suite B (1-ρ: 2.5%-6.6%). For this sample qualitative analysis revealed a Mg-rich amphibole phase (e.g. cummingtonite) which

PT

could not be quantified by AutoQuan due to the lack of an appropriate structural model. As a

RI

consequence, the amorphous content in this sample is calculated to ~ 10% within the sand and coarse silt fractions which is likely related to the missing amphibole phase. Because

SC

structural phase models do not necessarily reflect the real structure of the mineral in the

NU

sample, especially in case of highly weathered materials, quantitative Rietveld modelling delivered partly ambiguous results for these highly weathered materials, especially in the

MA

finer-grained fractions. We therefore present, besides quartz, quantitative data for the summed values of the clearly determined and quantified phase groups, i.e. quartz, total Kfeldspar, total plagioclase, total chain silicates, total mica including chlorite, and total clay

TE

D

minerals (Fig. 8). Moreover, gibbsite and Ti-phases are listed, if appropriate. In general, mineral proportions are highly variable with respect to both sample suites

AC CE P

and grain size. As expected, clay minerals and mica increase from coarse sand to clay. Increase of clay minerals is strongest in the finest fraction (~ 60-75%). Only in the F2 samples micas (mainly muscovite) decrease relatively from fine silt to clay, which is, however, more than compensated by the strong increase of clay minerals (~ 60%) and gibbsite (~ 10%). Quartz content decreases continuously from coarse (~ 22% in B, ~ 40% in F1, F2) to fine (<5%). The same is observed for K-feldspar, starting from ~ 20%, except for the B-sample where K-feldspar is pretty low. Plagioclase content is highest (~ 20-30%) in the very fine sand and/or coarse silt fractions for all samples, and then decreases dramatically towards finer fractions. Amphibole and pyroxene decrease strongly from coarse (~ 25%) to fine (<5%) for the B-sample, is relatively low in F1 (max. ~5%, mostly hornblende) and almost lacking in F2. Ti-phases are significant in the B-sample only (max. ~5%), and peak in very fine sand (Fig. 8). Results for Mineral Liberation Analysis (MLA) are restricted to the sand-sized fractions of samples RT4-3 and RT4-4 (F1) as well as RT4-7 (B). The mineral phases (or

13

ACCEPTED MANUSCRIPT groups) identified and quantified include quartz, K-feldspar, albite, plagioclase, amphibole (mainly hornblende), pyroxene (augite, pigeonite), biotite, chlorite, muscovite, clay minerals, Ti-minerals (e.g. ilmenite, titanite, TiO2-polymorphs), apatite, zircon, and other accessories

PT

(e.g. monazite, magnetite, carbonate). The mineral concentrations for each sample and the

RI

trends from coarse sand to very fine sand are listed in Table 3. Separation between suites B and F1 is straightforward due to strongly contrasting proportions of quartz, K-feldspar, and

SC

chain silicates (amphiboles + pyroxenes). Very good discrimination within the sand fractions

NU

is thus obtained by the quartz to chain silicates and K-feldspar to plagioclase ratios (Table 3). Very fine sand typically shows decrease of quartz and K-feldspar and increase of plagioclase

MA

and amphibole compared to coarse sand. Heavy minerals such as apatite, zircon, and Timinerals increase by factors of ~2 to 17 in very fine sand compared to coarse sand. Comparison of XRD vs. MLA data for samples RT4-3 (suite F1) and RT4-7 (suite B),

TE

D

for which both methods were applied on two grain-size fractions each, yields an overall good agreement of the respective mineral phases (Fig. 9). RT4-3 yields good agreement for all

AC CE P

phases except for muscovite, which yields lower values with MLA. RT4-7 yields very good agreement for quartz, amphibole, K-feldspar, and clay minerals, good agreement for plagioclase, biotite + chlorite, and muscovite, but some contrast in case of pyroxene and Timinerals (especially for coarse sand; Fig. 9). The higher values for pyroxene in the MLA dataset may derive from some problems in discriminating Al-poor amphiboles from pyroxene with this technique. Alternatively, Mg-rich chain silicates have been observed in the XRDspectra but could not be quantified and may be as high as ~ 10% (see above). Taking together, the data indicate reasonable consistency between the two methods.

5. Discussion

5.1 Chemical weathering Chemical weathering of silicate minerals such as feldspar or chain silicates leads to exchange of cations of alkaline and alkaline earth elements for H+ via hydrolysis (e.g.

14

ACCEPTED MANUSCRIPT Bahlburg and Dobrzinski, 2011). Mineralogically, this process is characterized by the transformation of the original minerals into clay minerals, with kaolinite commonly reflecting intense weathering and complete removal of mobile elements including potassium. Chemical

PT

weathering may also mobilize silica, including desilicification of clay minerals and formation

RI

of gibbsite (Vazques, 1981). Among the vast number of proxies for chemical weathering the Chemical Index of Alteration (CIA; Nesbitt and Young, 1982) is by far the most popular. It

SC

essentially describes the degree of removal of Ca2+, Na+, and K+, normalised to Al, through

NU

transformation of feldspars to clay minerals. Unweathered source rocks typically have CIA of 40-50 while extremely weathered materials entirely composed of, for instance, kaolinite,

MA

gibbsite, and Fe-(hydr)oxides have CIA close to 100. Average pelitic rocks that typically experienced significant weathering yield CIA of ~70 (e.g. CIA 70 for post-Archean Australian shale, Taylor and McLennan, 1985; CIA 69 for Phanerozoic average cratonic shale, Condie,

TE

D

1993). It must be noted that CIA values or other weathering indices does not exclusively depend on the degree of chemical weathering. Other factors such as initial source rock

AC CE P

composition (Nesbitt and Young, 1984), grain size (von Eynatten et al., 2012), or provenance settings including e.g. sheet-silicate rich units or sediment recycling (Garzanti and Resentini, 2015) are prone to add significant variation to the calculated CIA values. Calculated CIA values for the Sila samples show a very wide range from 55 to 92, i.e. from almost “fresh” to heavily weathered materials (Fig. 10). There is no clear distinction between the three sample suites. Nevertheless, the most felsic suite F2 tends to slightly lower CIA at the coarse and fine tails of the grain-size range and the basic suite shows slightly higher CIA from coarse silt to very fine silt (5 < Φ < 9; Fig. 10). In general, CIA values show a pronounced increase with decreasing grain size. However, within the sand fraction there is only little increase from CIA 55-61 for very coarse sand to CIA 56-68 for very fine sand. The average CIA for all sand fractions from all samples is 60. Very coarse silt (4 < Φ < 5) is still relatively low in CIA (57-68; medians still around 60) while all finer fractions mark a continuous increase up to the clay fraction, which range in CIA from 77 to 92 with mean

15

ACCEPTED MANUSCRIPT values between 84 and 88 for the three suites (Fig. 10). Average CIA for all the fine fractions from coarse silt to clay (Φ > 5) is 75.4, which is higher than average shale (see above). The strong chemical weathering in the Sila Massif has long been known and is

PT

described elsewhere. It is characterized by the development of deep weathering profiles on

RI

granitoid bedrock (e.g. Le Pera and Sorriso-Valvo, 2000), dissolution microtextures including intensely etched quartz grains (e.g. Scarciglia et al., 2007), and the occurrence of various

SC

neo-formed clay minerals such as illite-smectite mixed layers, kaolinite, and halloysite (e.g.

NU

Borelli et al., 2014). Our data support this through (i) high CIA up to 92 and (ii) high proportions of minerals indicating strong chemical weathering such as kaolinite and gibbsite

MA

in the finest sediment fractions.

Grain size dependence of chemical weathering indices was described earlier, but has

D

been typically restricted to broader categories such as sand(stone) vs. mud(stone) (e.g.

TE

Ohta, 2008). The detailed grain-size resolution obtained here fully confirms that comparing weathering proxies from samples with different grain size is likely to produce biased results.

AC CE P

The effect appears less relevant for different grades of sand(stone) but highly relevant for comparison of sand vs. mud and for comparison of different grades and mixtures of silt to clay. For the latter, our data suggest strong contrast in CIA although the source area is similar in terms of both bedrock type (for each sample suite) and weathering conditions. This statement is crucial given that many case studies have based their interpretations on relative subtle changes in weathering proxies like, for instance, 5 to 15 units in CIA (e.g. Nesbitt and Young, 1982; Passchier and Krissek, 2008; Bahlburg and Dobrzinski, 2011).

5.2 Source-rock discrimination The three sample suites B, F1 and F2 reflect marked contrasts in bedrock mineralogy and geochemistry. The question is whether the sediments reflect these contrasts across the full grain size range despite intense chemical alteration? Raw data concentrations and covariance structure suggest high discriminative potential for K2O, MgO, and CaO, as well as Rb, Th, U, V, Sc, Co, Ba, and Sr (Figs. 2 to 5). Because some of these elements are

16

ACCEPTED MANUSCRIPT sensitive to chemical weathering (e.g. Ca, Sr, Rb, K, U), this observation implies that chemical weathering was not extreme enough to fully destroy source rock signatures. Typical element ratios used to discriminate sediments derived from basic vs. sediments derived from

PT

felsic rocks such as Ba/Co, La/Co, La/Sc, or Th/Sc (e.g. Bhatia and Crook, 1986, Cullers et

RI

al. 1988) almost completely separate sample suite B from the more felsic suites, like does Rb/Sc and Rb/Co. The best separation between suite B and all other samples based on

SC

single ratios is obtained by Rb/V. All these ratios, however, do not sufficiently separate F1

NU

from F2 across all grain size grades. This can be achieved by using Rb/Sr as additional discriminator (Fig. 5A).

MA

Consequently, a very good discrimination of the three sample suites is achieved in the ternary diagram with trace elements Rb, Sr and V, where relative V concentration effectively discriminates sample suite B from all the other samples (probability >90%) and

TE

D

Rb/Sr ratio is the main discriminator between F1 and F2 (Fig. 11). The latter discrimination is less strict due to minor overlap between F1 and F2, which mainly derive from the finer

AC CE P

fractions (Φ > 5) of one single sample at the western boundary of suite F2 (RT3-6) that shows relatively low Rb/Sr ratios compared to the rest of F2. Replacing the trace elements Rb, Sr and V by the major element oxides K2O, CaO, and MgO, respectively, leads to almost the same level of discrimination.

5.3 Grain-size control on composition Biplot analysis of major and trace elements as well as the linear regression model reveals strong grain size control on sediment composition. In mineralogical terms the general trend from coarse to fine is accompanied by the increase of sheet silicates at the expense of quartz and the silicates. The increase of sheet silicates is twofold: (i) newly formed clay minerals are preferentially concentrated in the finest fractions due to intense chemical weathering of feldspars and other silicates, and (ii) micas are enriched in the finer fractions due to enhanced chemical alteration and mechanical comminution along cleavage planes, as well as hydrodynamic sorting. This is reflected in the ratios of quartz and feldspar over both

17

ACCEPTED MANUSCRIPT micas and clay minerals, which strongly decrease from sand to silt-clay (Fig. 12). Decrease of the quartz/clay ratio from coarse sand to clay may exceed two orders of magnitude. In the finest fractions, clay minerals even increase relative to mica (i.e. stronger decrease of

PT

quartz/clay compared to quartz/mica; Fig. 12).

RI

Chemically, sheet silicate enrichment is reflected by an overall decrease of SiO2 and increase of Al2O3 and LOI from very coarse sand to clay (Fig. 2). All other major elements

SC

have a more complex pattern, especially in relative terms (clr-scale, Fig. 6) where even Al

NU

increases in the Φ > 4 fractions only. Considering the sand-fractions separately, the main trend is controlled by strong depletion of SiO2 and K2O while CaO and P2O5 increase. All

MA

other major elements show only minor changes (slopes <0.05) except for LOI showing moderate increase (Fig. 7). This pattern mostly reflects depletion of quartz and K-feldspar

D

and increase of plagioclase (and partly amphibole, RT4-3) in (very) fine sand compared to

TE

(very) coarse sand, as corroborated by XRD data (Fig. 8). Chemical weathering of feldspar grains preferentially evolves at the surfaces as well as along cleavage and twinning planes,

AC CE P

the latter in turn accelerate mechanical comminution of mineral grains (e.g. Berner and Holdren, 1979; Banfield and Eggleton, 1990). Among the feldspars, plagioclase weathers much faster than K-feldspar (e.g. White et al., 2001), implying that comminution of plagioclase is more effective than K-feldspar at the initial stages of feldspar weathering. Therefore, plagioclase is enriched over K-feldspar in fine sand to coarse silt fractions, where Ca peaks in both raw data and the linear model (Figs. 2 and 6). This decrease in Kfeldspar/plagioclase ratio from (very) coarse to very fine sand is strongly corroborated by both XRD data (Fig.12) and MLA data (Table 3). P2O5 reflects enrichment of apatite and/or other phosphate minerals such as monazite in very fine sand/very coarse silt, consistent with general enrichment of heavy mineral in these size fractions (e.g. zircon, Ti-minerals; Table 3). In the silt to clay range, the linear model is characterized by strong increase in LOI and strong decrease of CaO and especially Na2O. Moreover, a moderate increase of Fe2O3, and MgO (slopes around 0.1) is observed, and only minor changes for TiO2, P2O5, Al2O3,

18

ACCEPTED MANUSCRIPT SiO2 and K2O (slopes at around ± 0.03-0.07, Fig. 7). This observation reflects strong increase of clay minerals and mica compared to all other minerals (Figs. 8 and 12) and is well in line with the strong increase of CIA obtained for Φ > 5 (Fig. 10). Because some of

PT

these sheet silicates still keep potassium (illite, biotite, muscovite), the decrease of K2O is

RI

only minor compared to CaO and Na2O, although most K-feldspar and plagioclase have

SC

disappeared in the finest fractions (Fig. 8). Relative increase of Fe2O3 and MgO also reflects increase of the respective sheet silicates (biotite, chlorite, smectite). Fe-(hydr-)oxides appear

NU

negligible because Fe2O3 and MgO are strongly coupled (Fig. 7). Interestingly, the least changes in the linear regression model over the full grain size

MA

range (i.e. smallest slopes regardless of sign; Fig. 7) refer to Al2O3 (Figs. 6 and 7). Obviously, Al increase in raw data concentrations towards fine fractions (Fig. 2) but has low overall

D

variability (Figs. 4 and 5) and only small (F1) or almost no (F2) increase from coarse to fine

TE

in relative clr-scale (Fig. 6). Because Al is largely immobile during weathering processes and is commonly enriched in most clay minerals, Al typically increases towards fine fractions,

AC CE P

even in clr-scale (e.g. Bloemsma et al., 2012). This effect is quite small here, likely because we included the loss on ignition (reflecting mainly the water content of the sheet silicates), which captures most of the relative increase towards the fine grain-size classes. Regarding trace elements, the Zr/Zn ratio provides a straightforward proxy for grain size, as already observed for basic and granitoid rocks in glacial setting (von Eynatten et al., 2012). The new data corroborate this finding, and extend its use to intensely weathered settings, where Zn is similarly enriched in the finest fractions, regardless of source-rock lithology (Fig. 13). The source rocks, i.e. basic versus more felsic lithology, are best discriminated by the Rb/V ratio (Figs. 3, 11, 13).

5.4 Comparison to non-weathered sediments In an analogous study, von Eynatten et al. (2012) investigated the influence of grain size and source rocks on sediment composition in glacial settings where chemical weathering is negligible and mechanical comminution is the main process that governs the

19

ACCEPTED MANUSCRIPT separation of mineral phases and, hence, their distribution in different grain-size fractions. Study area was granitic bedrock of the Aar Massif in the Central Alps and basic amphibolites from the Silvretta Massif in the eastern Alps. The Aar example is quite similar in bedrock

PT

composition to sample suite F2 from the Sila Massif. Given the similar starting point (i.e.

RI

granitoid bedrock composition) the comparison of the two cases is actually a comparison of the effects of strong vs. negligible chemical weathering. For a reasonable comparison of the

SC

linear regression, we recalculated the model adjusted to the Aar Massif case study, applying

NU

the same model constraints as for the Sila Massif, except for the step at Φ = 8 which is obviously necessary to describe the Aar Massif data set (von Eynatten et al., 2012).

MA

The comparison shows striking contrasts in the modelled linear regression trends of the major elements against grain size (Fig. 14). The main observations on contrasting trends

D

for the Sila case vs. the Aar case include (i) strong decrease of SiO2 already in the sand

TE

fraction from very coarse to very fine sand, (ii) strong increase of CaO in the sand fraction from coarse to fine, (iii) continuous and pronounced decrease for CaO and Na2O in the silt-

AC CE P

clay fraction, (iv) continuous increase of Al2O3 and Fe2O3 for the silt-clay fraction (i.e. no significant steps at Φ = 8), and (v) no increase of K (i.e. no step) for the finest fractions at Φ > 8. These observations mainly reflect the following processes, which are prominent in the Sila case compared to the Aar case: (1) Initial plagioclase weathering through hydrolysis along cleavage and/or twinning planes causes breakage and preferential enrichment of plagioclase over quartz and K-feldspar from coarse sand to very fine sand, as reflected in Ca increase (observation ii) and some Si decrease (observation i). The effect may be supported by initial amphibole weathering in case of F1. (2) Intense chemical weathering mainly through hydrolysis leads to the almost complete removal of feldspar and chain silicates from the intermediate grain sizes (3 < Φ < 6) to the finest fractions (observation iii).

20

ACCEPTED MANUSCRIPT (3) Enrichment of various Al- and/or Fe-rich sheet silicates (biotite, chlorite, smectite, illite, kaolinite) towards the finer and finest fractions as a consequence of process no.2 (observation iv)

PT

(4) Enrichment of K-poor clay minerals (kaolinite, smectite) relative to micas in the finest

RI

fractions (observation v). This reflects ongoing removal of potassium, which is essential to obtain the high CIA values.

SC

The combination of processes (1) and (2) causes the smoothed Ca (Figs. 2 and 6)

NU

and plagioclase (Fig. 8) peaks in the Sila case instead of the abrupt break (i.e. step) in Ca concentration of the glacial Aar case (Fig. 14), which is caused by mechanical comminution

MA

in combination with inherited mineral-specific grain size distributions (von Eynatten et al., 2012). The latter combination was also found to be responsible for quartz concentrations

D

peaking in fine to very fine sand in the glacial case study (2 < Φ < 4), which is not the case in

TE

the Sila example. There, SiO2 continuously decreases over the full grain-size range from coarse to fine (observation i), which is caused by intense hydrolysis supported by quartz

AC CE P

leaching (Scarciglia et al., 2007) and, finally, desilicification of clay minerals and formation of gibbsite as observed in some samples (Fig. 8).Despite all these differences between the highly weathered and the glacial case study, some elements converge towards the finest fractions to very similar relative concentrations independent of similar (e.g. SiO2) or contrasting starting points (e.g. Fe2O3). For the fine fractions, the elements showing most contrast between the highly weathered and the glacial case study are Al2O3 and Na2O (Fig. 14), suggesting that the ratio of these two elements represents a simple and straightforward proxy for chemical weathering, which avoids potential bias through K-metasomatism (Fedo et al. 1995) and/or the need for corrections for Ca associated with non-silicate phases (von Eynatten et al., 2003a).

6. Conclusions

21

ACCEPTED MANUSCRIPT Analysis and modelling of composition vs. grain size relations of proximal sediments in a highly weathered Mediterranean setting (Sila Massif) and its comparison to a previous study in glacial settings in the European Alps (von Eynatten et al., 2012) reveal the following

PT

main conclusions:

RI

(1) The chemical index of alteration (CIA) corroborates strong chemical weathering by individual values as high as 92 and average values of 84 to 88 for the finest fractions.

SC

Fractions coarser than coarse silt (i.e. Φ < 5), however, reveal low to moderate values in

NU

CIA (55 to 68) and an average of 60 for all sand fractions from all sample suites. The CIA value of a specific sample thus strongly depends on its grain-size distribution.

MA

(2) Although strongly weathered, the three sample suites reflecting basic to intermediate (B), intermediate to felsic (F1), and felsic (F2) plutonic bed rock can be effectively

D

discriminated across all grain-size classes using trace elements such as V, Rb, and Sr. At

TE

least in the coarser fractions, major elements Mg, Ca, and K do so, too. (3) The chemical composition vs. grain-size relations can be accurately modelled by linear

AC CE P

regression only if different slopes are allowed for the sand (Φ < 4) and the silt to clay (Φ > 4) fractions. Similar slopes in the respective grain-size range for different sample suites (i.e. intermediate and felsic source rocks F1 and F2, respectively) underline similar processes independent of source rock composition. (4) The mineralogy behind the geochemical data reflect overall (i.e. valid for all three sample suites) decrease of quartz and K-feldspar over the full grain-size range from coarse to fine. This is compensated by overall increase of sheet silicates from coarse to fine, where the increase of clay minerals strongly outpaces the increase of micas in the silt to clay fractions (i.e. Φ > 5). A more complex behaviour is shown by plagioclase, whichpeaks in intermediate grain-size fractions (i.e. 3 < Φ < 6) in all sample suites. The latter is caused by initial hydrolysis along cleavage planes and subsequent breakage of plagioclase crystals into smaller fragments, preferentially in the 125 µm to 16µm range. (5) Mechanical comminution, mineral durability, and inherited grain-size distribution were found to be relevant parameters in the glacial case study, responsible for peaking of, for

22

ACCEPTED MANUSCRIPT instance, quartz and total SiO2 concentration in fine to very fine sand (2 < Φ < 4). This is not observed in the highly weathered Sila case. Instead, both quartz and total SiO2 concentration continuously decrease from coarse to fine, reflecting the cumulative effects

PT

of minor mechanical forces, quartz leaching, and hydrolysis.

RI

(6) Given the constantly contrasting behaviour of SiO2 and Al2O3 the ratio Si/Al provides a reasonable proxy for grain size in the Sila case. This is not valid for the glacial case (Fig.

SC

14). However, Zr/Zn provides a good proxy for grain size valid for both settings: high

NU

ratios indicate zircon concentration in the intermediate fractions (mostly 3 < Φ < 5) while low ratios indicate sheet silicate enrichment in the fine fractions (e.g. Φ > 7).

MA

The application of Mineral Liberation Analysis (MLA) yields some encouraging results, namely (i) the apparent compatibility with X-ray diffraction data for the main source-rock

D

forming minerals quartz, K-feldspar, plagioclase, amphibole, and most sheet silicates, and (ii)

TE

the abundance and grain-size distribution of certain heavy minerals. The potential of the method with respect to the fine grain size ranges and especially highly weathered materials

AC CE P

needs further exploration.

Acknowledgements

Salvatore Critelli greatly supported us during field work in the Sila Massif and subsequent discussions. Sample preparation and grain-size fractionation was performed by Cornelia Friedrich. Journal reviewers Abhijit Basu and Gert Jan Weltje and guest editor Eduardo Garzanti are thanked for their careful and constructive handling of the manuscript.

References Aitchison, J., 1986. The Statistical Analysis of Compositional Data. Monographs on Statistics and Applied Probability. Chapman & Hall Ltd., London (UK) (Reprinted in 2003 with additional material by The Blackburn Press), 416 pp.

23

ACCEPTED MANUSCRIPT Aitchison, J., 1990. Relative variation diagrams for describing patterns of compositional variability. Mathematical Geology, 22, 487–511.

PT

Ayuso, R. A., Messina, A., De Vivo, B., Russo, S., Woodruff, L.G., Sutter, J. F., Belkin, H. E., 1994. Geochemistry and argon thermochronology of the Variscan Sila batholith,

RI

southern Italy: source rocks and magma evolution. Contributions to Mineralogy and

SC

Petrology, 117, 87–109.

Bahlburg, H., Dobrzinski, N., 2011. A review of the chemical index of Alteration (CIA) and its

NU

application to the study of Neoproterozoic glacial deposits and climate transitions. In: Arnaud, E., Halverson, G.P., Shields-Zhou, G. (eds.) The Geological Record of

MA

Neoproterozoic Glaciations. Geological Society, London, Memoirs, 36, 81-92. Banfield, J.F., Eggleton, R.A., 1990. Analytical transmission electron microscope studies of

D

plagioclase, muscovite, and K-feldspar weathering. Clays and Clay Minerals, 38, 77-

TE

89.

AC CE P

Basu, A., 1976. Petrology of Holocene fluvial sand derived from plutonic source rocks: implications to paleoclimatic interpretation. Journal of Sedimentary Petrology, 46, 694709.

Basu, A., Young, S.W., Suttner, L.J., James, W.C., Mack, G.H., 1975. Re-evaluation of the use of undulatory extinction and polycrystallinity in detrital quartz for provenance interpretation. Journal of Sedimentary Petrology, 45, 873-882. Berner, R.A., Holdren, G.R., 1979. Mechanism of feldspar weathering – II. Observations of feldspars from soils. Geochimica et Cosmochimica Acta, 43, 1173-1186. Bhatia, M., Crook, K.A.W., 1986. Trace element characteristics of greywackes and tectonic setting discrimination of sedimentary basins. Contribution to Mineralogy and Petrology, 92, 181–193. Blatt, H., Middleton, G.V., Murray, R.C., 1972. Origin of sedimentary rocks. Prentice-Hall, Englewood Cliffs, NJ (US), 634 p.

24

ACCEPTED MANUSCRIPT Bloemsma, M.R., Zabel, M., Stuut, J.B.W., Tjallingii, R., Collins, J.A., Weltje, G.J., 2012. Modelling the joint variability of grain size and chemical composition in sediments.

PT

Sedimentary Geology 280, 135-148. Borelli, L., Perri, F., Critelli, S., Gullà, G., 2014. Characterization of granitoid and gneiss

RI

weathering profiles of the Mucone River basin (Calabria, southern Italy). Catena, 113,

SC

325-340.

Caggianelli, A., Del Moro, A., Piccarreta, G.,1994. Petrology of basic and intermediate

NU

orogenic granitoids from the Sila Massif (Calabria, southern Italy). Geological Journal, 29, 11-28.

MA

Condie, K. C., 1993. Chemical composition and evolution of the upper continental crust: Contrasting results from surface samples and shales. Chemical Geology 104, 1-37.

D

Cullers, R.L., Basu, A., Suttner, L.J., 1988. Geochemical signature of provenance in sand-

TE

size material in soils and stream sediments near the Tobacco Root Batholith, Montana,

AC CE P

USA. Chemical Geology, 70, 335-348. Fandrich, R., Gu, Y., Burrows, D., Moeller, K., 2007. Modern SEM-based mineral liberation analysis. International Journal of Mineral Processing, 84, 310-320. Fedo, C.M., Nesbitt, H.W., Young, G.M., 1995. Unravelling the effects of potassium metasomatism in sedimentary rocks and paleosoils, with implications for paleoweathering and provenance. Geology 23, 921-924. Garzanti,E., Andò, S., Vezzoli, G., 2009. Grain-size dependance of sediment composition and environmental bias in provenance studies. Earth and Planetary Science Letters, 277, 422–432. Garzanti, E., Andò, S., France-Lanord, C., Censi, P., Vignola, P., Galy, V., Lupker, M., 2011. Mineralogical and chemical variability of fluvial sediments. 2. Suspended-load silt (Ganga–Brahmaputra, Bangladesh). Earth and Planetary Science Letters, 302, 107– 120.

25

ACCEPTED MANUSCRIPT Garzanti, E., Resentini, A., 2015. Provenance control on chemical indices of weathering (Taiwan river sands). Sedimentary Geology.

PT

http://dx.doi.org/10.1016/j.sedgeo.2015.06.013 (this volume). GE Inspection Technologies GmbH, 2014. Rayflex – X-ray diffraction software, AutoQuan

RI

Version 2.80, Ahrensburg, Germany.

SC

Graessner, T., Schenk, V., Bröcker, M., Mezger, K., 2000. Geochronological constraints on the timing of granitoid magmatism, metamorphism and post-metamorphic cooling in the

NU

Hercynian crustal cross-section of Calabria. Journal of Metamorphic Geology, 18, 409421.

MA

Guevara, M., Verma, S.P., Velasco-Tapia, F., 2001. Evaluation of GSJ intrusive rocks JG1, JG2, JG3, JG1a, and JGb1 by an objective outlier rejection statistical procedure.

D

Revista Mexicana de Ciencias Geologicas, 2001, 18, 74-88.

TE

Ingersoll, R.V., Bullard, T.F., Ford, R.L., Grimm, J.P., Pickle, J.D., Sares, S.W., 1984. The

AC CE P

effect of grain size on detrital modes: a test of the Gazzi-Dickinson point-counting method: Journal of Sedimentary Petrology, 54, 103-116. Imai, N., Terashima, S., Itoh, S., Ando, A., 1995. 1994 Compilation of analytical data for minor and trace elements in sebenteen GSJ geochemical reference samples, “IGNEOUS ROCK SERIES”. Geostandards Newsletter, 1995, 19, 135-213. Johnsson, M.J., 1993. The system controlling the composition of clastic sediments. Geological Society of America Special Paper, 284, 1-19. Le Pera, E., Sorriso-Valvo, M., 2000. Weathering and morphogenesis in a Mediterranean climate, Calabria, Italy. Geomorphology, 34, 251-270. Le Pera, E., Arribas, J., Critelli, S., Tortosa, A., 2001. The effects of source rocks and chemical weathering on the petrogenesis of siliciclastic sand from the Neto River (Calabria, Italy). Sedimentology, 48, 357-378.

26

ACCEPTED MANUSCRIPT Messina, A., Compagnoni, R., De Vivo, B., Perrone, V., Russo, S., 1991. Geological and petrochemical study of the Sila Massif plutonic rocks (northern Calabria, Italy).

PT

Bollettino della Società Geologica Italiana, 110, 165-206. Morton, A.C., Hallsworth, C.R., 1994. Identifying provenance-specific features of detrital

RI

heavy mineral assemblages in sandstones. Sedimentary Geology, 90, 241-256.

SC

Nesbitt, H.W. and Young, G.M. (1982) Early Proterozoic climates and plate motions inferred from major element chemistry of lutites. Nature, 299, 715-717.

NU

Nesbitt, H.W., Young, G.M., 1984. Prediction of some weathering trends of plutonic and volcanic rocks based on thermodynamic and kinetic considerations. Geochimica et

MA

Cosmochimica Acta, 48, 1523-1534.

Ohta, T., 2008. Measuring and adjusting the weathering and hydraulic sorting effects for

D

rigorous provenance analysis of sedimentary rocks: a case study from the Jurassic

TE

Ashikita Group, south-west Japan. Sedimentology, 55, 1687-1701.

AC CE P

Olivetti, V., Cyr, A.J., Molin, P., Faccenna, C., Franger, D.E., 2012. Uplift history of the Sila Massif, southern Italy, deciphered from cosmogenic 10Be erosion rates and river longitudinal profile analysis. Tectonics, 31, TC3007. Panalytical B.V., 2006. X’pert Highscore 2.2a. Software for X-ray diffraction analysis. Panalytical B.V., Almelo, The Netherlands. Passchier, S., Krissek, L. A., 2008. Oligocene-Miocene Antarctic continental weathering record and paleoclimatic implications, Cape Roberts drilling project, Ross Seal, Antarctica. Palaeogeography Palaeoclimatology Palaeoecology, 260, 30-40. Pettijohn, F., 1975. Sedimentary rocks. Harper & Row, New York, 3rd edition, 628 p. Scarciglia, F., Le Pera, E.,Critelli, S., 2007. The onset of the sedimentary cycle in a midlatitude upland environment: Weathering, pedogenesis, and geomorphic processes on plutonic rocks (Sila Massif, Calabria). Geological Society of America Special Paper, 420,149-166.

27

ACCEPTED MANUSCRIPT Scarciglia, F., Saporito, N., La Russa, M.F., Le Pera, E., Macchione, M., Puntillo, D., Crisci, G.M., Pezzino, A., 2012. Role of lichens in weathering of granodiorite in the Sila

PT

uplands (Calabria, southern Italy). Sedimentary Geology, 280, 119-134. Taylor, S.R., McLennan, S.M., 1985. The continental crust: its composition and evolution.

RI

Blackwell Science, Oxford. 315pp.

SC

Thomson, S. N., 1994. Fission track analysis of the crystalline basement rocks of the Calabrian Arc, southern Italy: evidence of Oligo-Miocene late-orogenic extension and

NU

erosion. Tectonophysics, 238, 331-352.

Tolosana-Delgado, R., van den Boogart, K.G., 2011. Linear models with compositions in R.

MA

In: V. Pawlowsky-Glahn, A. Buccianti (Eds.) Compositional Data Analysis: Theory and Applications. John Wiley & Sons, Ltd., Chapter 26, p. 356-371.

D

Tolosana-Delgado, R., von Eynatten, H., 2009. Grain-size control on petrographic

TE

composition of sediments: compositional regression and rounded zeroes.

AC CE P

Mathematical Geosciences, 41, 869-886. Vazques, F.M., 1981. Formation of gibbsite in soils and saprolites of temperate-humid zones. Clay Minerals, 16, 43-52.

von Eynatten, H., Tolosana-Delgado, R., Karius, V., 2012. Sediment generation in modern glacial settings: source-rock and grain-size control on sediment composition. Sedimentary Geology, 280, 80-92. von Eynatten, H., Barceló-Vidal, C., Pawlowsky-Glahn, V., 2003a. Modelling compositional change: the example of chemical weathering of granitoid rocks. Mathematical Geology, 35, 231-251. von Eynatten, H., Barceló-Vidal, C., Pawlowsky-Glahn, V., 2003b. Sandstone composition and discrimination: a statistical evaluation of different analytical methods. Journal of Sedimentary Research, 73, 47-57.

28

ACCEPTED MANUSCRIPT White, A.F., Bullen, T.D., Schulz, M.S., Blum, A.E., Huntington, T.G., Peters, N.E., 2001. Differential rates of feldspar weathering in granitic regoliths. Geochimica et

RI

PT

Cosmochimica Acta, 65, 847-869.

SC

Figure captions

NU

Fig.1: Simplied geological map of the study area, modified from Messina et al. (1991) and

MA

Borelli et al. (2014).

Fig. 2: Major element raw data concentrations in oxide wt.-% for all three sample suites of

D

the Sila Massif (blue = B, gabbro to diorite; red = F1, tonalites to granodiorites; brown = F2,

TE

granodiorite to monzogranite) versus grain size in Φ-grades from very coarse sand (-1 < Φ <

AC CE P

0) to clay (Φ > 9). The variability of individual samples of a grain-size class is illustrated in box plots for suites F1 and F2 (N=7 each), while sample suite B (N=2) is represented by stippled lines with solid line reflecting the mean.

Fig. 3: Trace element raw data concentrations in ppm (note log-scale) for all three sample suites of the Sila Massif (blue = B, gabbro to diorite; red = F1, tonalites to granodiorites; brown = F2, granodiorite to monzogranite) versus grain size in Φ-grades from very coarse sand (-1 < Φ < 0) to clay (Φ > 9). The variability of individual samples of a grain-size class is illustrated in box plots for suites F1 and F2 (N=7 each), while sample suite B (N=2) is represented by stippled lines with solid line reflecting the mean. For key to boxplots see Fig. 2.

Fig. 4: Compositional biplots of the clr-transformed geochemical data (Aitchison 1990). Symbols reflect sample suite (B, F1, F2) and sample type (Sed = sediment, WP = weathering

29

ACCEPTED MANUSCRIPT profile), colour coding refers to grain size class in Φ-units from very coarse sand (yellow; -1 ≤ Φ < 0) to clay (black; Φ > 9). A given value x thus represents the grain-size class of x ≤ Φ < x+1. Both biplots indicate sample position in the two dimensional projection based on the first

PT

(x-axis, PC-1) and second (y-axis, PC-2) principal components, as well as the variables and

RI

their loadings on the respective principal component. Left side (A) shows major element

SC

oxide data including loss on ignition (LOI). Right side (B) shows the same major elements along with selected trace elements (Rb, Sr, V, Y, Zn, Zr). Numbers in italics refer to

NU

percentages of the total variability explained by PC1 and PC2.

MA

Fig. 5: Compositional biplots of major element oxides (excluding LOI) and selected trace elements (Rb, Sr, V, Y, Zn, Zr) for the intermediate to felsic suites F1 and F2. The

D

compositional contrast that still exists between F1 and F2 has been minimized by matching

TE

the means of each suite (i.e. by perturbing one data set by the inverse geometric mean of the other). (A) shows first (x-axis) vs. second (y-axis) principal component, (B) shows first (x-

AC CE P

axis) vs. third (y-axis) principal component. Numbers in italics refer to percentages of the total variability explained by PC1 and PC2 (A) or by PC1 and PC3 (B). For key to symbols and colours see Fig. 4.

Fig. 6: Linear regression model (solid lines) adjusted for the F1 (red) and F2 (brown) sample suites against grain size classes, displayed in clr-scale for each of the 10 variables included in the model (9 major elements plus LOI) separately. For key to boxplots see Fig. 2.

Fig. 7: Graphical representation of the model parameters slope in the sand fraction (Φ < 4) vs. slope in the silt to clay fraction (Φ > 4). Dotted line reflects 1:1 relation of the slopes.

Fig. 8: Quantitative X-ray diffraction (XRD) results of three samples, one from each of the sample suites B (RT4-7), F1 (RT4-3), and F2 (RT4-11). For each sample five grain-size fractions from coarse sand (0 < Φ < 1) to clay (Φ > 9) were analyzed and quantified. Mica

30

ACCEPTED MANUSCRIPT includes biotite, muscovite, and chlorite; claymin includes illite, smectite, and kaolinite; gib = gibbsite; ti-phases include ilmenite and rutile. Error bars include 3 sigma errors of Rietveld

PT

refinements. Note different scale for concentrations >32%.

RI

Fig. 9: Comparison of results on mineral composition from X-ray diffraction(XRD) vs. Mineral

SC

Liberation Analysis (MLA) for two samples from suite F1 (RT4-3) and suite B (RT4-7). Two grain-size fractions from each sample are available for comparison (qtz = quartz, pla =

NU

plagioclase incl. albite, kf = K-feldspar, am = amphibole, px = pyroxene, bio = biotite, chl =

MA

chlorite, clm = Σ clay minerals). For explanation see text.

Fig. 10: Chemical index of alteration (CIA) for each grain-size fraction and each sample suite

D

B, F1, and F2. The variability of individual samples of a grain-size class is illustrated in box

TE

plots for suites F1 and F2 (N=7 each), while sample suite B (N=2) is represented by stippled

AC CE P

lines with solid line reflecting the mean. For key to boxplots see Fig. 2.

Fig. 11: V-Rb-Sr ternary diagram for discrimination of the three sample suites B, F1, and F2. Fields represent predictive regions for each suite with 67% (solid line) and 90% (stippled line) probability.

Fig. 12: Selected mineral ratios from coarse sand to clay for all three sample suites based on X-ray diffraction data. Note logarithmic scale implying that, for instance, quartz to clay mineral ratio (qtz/clay) decreases by more than two orders of magnitude from coarse sand to clay.

Fig. 13: Rb/V vs. Zr/Zn diagram showing all samples and all grain-size fractions. For explanation see text.

31

ACCEPTED MANUSCRIPT Fig. 14: Comparison of the adjusted linear regression trends for the weathered case (Sila Massif, sample suites F1 and F2) against the glacial case (Aar Massif) for the most important major elements. Minor differences in F1 and F2 values compared to Fig. 6 are caused by

PT

excluding loss on ignition (LOI) because these data are not available for the Aar Massif data

RI

set.

NU

SC

Table captions

Table 1: List of samples including bed rock suite, sample type, coordinates and

MA

approximate elevation (Sed = sediment from small creeks; WP = weathering profile

D

or grus from roadcuts)

TE

Table 2: Summary of MLA parameters for samples RT4-3, RT4-4, and RT4-7.

AC CE P

Table 3: Results of MLA Analysis

32

Fig. 1

AC CE P

TE

D

MA

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

33

AC CE P

Fig. 2

TE

D

MA

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

34

MA

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

AC CE P

TE

D

Fig. 3

35

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

AC CE P

TE

D

MA

Fig. 4

Fig. 5

36

SC

RI

PT

ACCEPTED MANUSCRIPT

Fig. 7

AC CE P

TE

D

MA

NU

Fig. 6

37

SC

RI

PT

ACCEPTED MANUSCRIPT

AC CE P

TE

D

MA

NU

Fig. 8

38

AC CE P

TE

D

MA

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

Fig. 9

39

SC

RI

PT

ACCEPTED MANUSCRIPT

AC CE P

TE

D

MA

NU

Fig. 10

Fig. 11

40

MA

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

AC CE P

TE

D

Fig. 12

Fig. 13

41

MA

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

AC CE P

TE

D

Fig. 14

42

ACCEPTED MANUSCRIPT

suite

type

RT3-1

F1

WP

39°16'06.17''

16°32'01.42''

RT3-2

F1

Sed

39°16'06.61''

16°31'59.36''

RT3-3A

F1

WP

39°16'49.60''

16°32'18.91''

~1570 m a.s.l.

RT3-3B

F1

WP

39°16'49.60''

16°32'18.91''

~1570 m a.s.l.

RT3-5A

F2

Sed

39°23'14.85''

RT3-6

F2

Sed

39°22'12.59''

RT4-3

F1

Sed

39°16'48.28''

RT4-4

F1

Sed

RT4-5

F1

RT4-6

elevation

NU

SC

RI

~1480 m a.s .l.

~1480 m a. s.l.

~1170 m a.s.l.

16°31'30.34''

~1150 m a. s.l.

16°34'36,48''

~1580 m a. s.l.

39°17'34.23''

16°31'16.66''

~1530 m a. s.l.

Sed

39°17'36.39''

16°31'23.51''

~1520 m a. s.l.

B

Sed

39°15'10.33''

16°32'10.70''

~1330 m a.s .l.

RT4-7

B

Sed

39°15'17.02''

16°33'39.36''

~1380 m a.s .l.

RT4-10

F2

WP

39°23'21.79''

16°33'44.29''

~1270 m a.s.l.

RT4-12 RT4-13 RT4-15

TE

AC CE P

RT4-11

MA

16°33'12.04''

D

sample

N - coordinates - E

PT

Table 1: List of samples including bed rock suite, sample type, coordinates and approximate elevation (Sed = sediment from small creeks; WP = weathering profile or grus from roadcuts)

F2

Sed

39°24'17.18''

16°36'26.30''

~1430 m a.s.l.

F2

Sed

39°23'03.87''

16°36'51.97''

~1390 m a.s.l.

F2

WP

39°23'42.77''

16°36'03.80''

~1320 m a.s.l.

F2

WP

39°23'22.18''

16°33'43.33''

~1270 m a.s.l.

Table 2. Summary of MLA parameters for samples RT4-3, RT4-4, and RT4-7. SEM parameters Mode Voltage (kV) Working dist. (mm) Probe current (nA) Spot size HFW (µm) Brightness Contrast BSE calib.

GXMAP 25 12 10 5.84 800 83.5 26.2 Au 252

MLA parameters Scan speed 16 Resolution (pixels) 500 x 500 Pixel size (µm/px) 1.6 Acq. time (ms) 6 Min. EDX-count 2000 Step size (px) 6x6 GXMAP trigger 25 - 255 Min. particle size (px) 10 Min. grain size (px) 4

43

ACCEPTED MANUSCRIPT Table 3: Results of MLA Analysis RT4-4

RT4-7

cs - vfs *

cs - vfs *

cs - vfs *

quartz

40.5 - 20.2

38.2 - 19.2

22.7 - 11.3

K-feldspar

23.3 - 14.1

13.7 - 17.4

4.3 - 2.4

albite

7.4 - 9.2

6.8 - 10.6

4.5 - 3.2

plagioclase

5.1 - 11.9

9.6 - 16.1 (6.6)**

amphibole

0.4 - 5.1

0.03 - 6.0

5.3 - 16.1 21.2 - 21.7 (26.3)**

pyroxene

0.0 - 0.0

0.0 - 0.0

14.9 - 9.9

biotite

7.4 - 11.4

10.8 - 7.1 (12.4)**

8.2 - 6.9

chlorite

0.5 - 1.7

0.6 - 1.3

1.6 - 8.1 2.8 - 1.7 11.6 - 13.5

RI

SC

clay minerals

13.6 - 23.7

2.5 - 1.7 17.1 - 17.4 (14.1)**

Ti-minerals

0.10 - 0.39

0.25 - 0.86

2.2 - 4.5 (5.9)**

apatite

0.14 - 0.29 0.05 - 0.12 (0.03)**

0.20 - 1.25

0.44 - 0.63

0.03 - 0.54

0.02 - 0.06

1.1 - 0.6

1.3 - 0.8

1.6 - 0.5

1.9 - 0.7

0.8 - 0.7

0.4 - 0.1

~100 - 4.0

>100 - 3.2

0.6 - 0.4

MA

1.4 - 1.6 (2.4)**

zircon

AC CE P

quartz / Σ feldspar K-feldspar / plagioclase ° quartz / chain silicates °°

TE

D

muscovite

NU

grain-size

PT

RT4-3

sample

quartz / clay minerals 3.0 - 0.9 2.2 - 1.1 2.0 - 0.8 * the values [wt.-%] describe the trend from coarse sand (cs) to very fine sand (vfs) ** values in brackets refer to exceptions from the trend observed in medium or fine sand ° albite included in plagioclase °° chain silicates = amphibole + pyroxene

44