Chapter 3 Application of a confluence-based sediment fingerprinting approach to a dynamic
3.5 Material and methods
3.5.1 Sample Collection
The research was conducted through an exploratory approach purposed to provide an overall appreciation on the effectiveness of sediment fingerprinting to differentiate source material at a very broad-scale before the outlay of further resources. The most direct way was to break the catchment down into sub-catchments and key confluences. This simplifies the sediment
sources by isolating upstream sediment from a given confluence and allows the assumption to be made that the sediment flowing out of a given confluence is a proportionally representative upstream signal of the entire sub-catchment at a given point in time. The sub-catchment tributaries utilized are the Upper Manawatu, Tiraumea, Mangatainoka (C1-2), Mangahao (C3), Pohangina (C4) and Oroua sub-catchments (C5) (Fig. 15).). The sub-catchments that drain to the five confluences cover a significant area of the Manawatu Catchment, Upper Manawatu (1301 Km2), Tiraumea (876 Km2), Mangatainoka (481 Km2), Mangahao (339 Km2), Pohangina (551 Km2) and Oroua sub-catchment (903 Km2) with each sub-catchment containing a unique composition of geology, albeit all sedimentary in nature.
The confluences of the six major sub-catchments were selected as sample site locations. For each confluence, upstream and downstream samples were collected from fine drapes over bars within the channel during a period of low flow. Upstream samples were collected as close to the confluence point as possible. Downstream of the confluence, attempts were made to collect samples approximately 500 m to 1 km downstream to provide distance for mixing to occur but limiting the introduction of new sediment sources. Five samples were initially collected for each location on one occasion. A trowel was used to collect multiple (typically 5) point grabs/scrapings from the immediate area (≈ 10 m2). Deposits that clearly contained fine sediment were targeted over larger clast material and samples near the bank edge or where the sediment was clearly a direct input from the channel bank were avoided. Fine sediment was found in three key positions: around the fringes of point bars, particularly on the lee side and after receding waters; trapped in semi-dried-out pools following receding waters; and caught within the interstices of larger clasts and objects (Fig. 16). Sampling occurred at the culmination of the winter season, following a period of increased rainfall across the catchment, common for the time of year. Repeated sampling was not undertaken due to initial resource constraints and was beyond the purpose of the exploration which poses important limitations to the study and the ability to evaluate temporal information.
Fig. 16: Typical in-channel fine sediment deposits: point bar deposition (left); trapped in the interstices of larger clasts on a bar (right); and deposited in semi-dry to dried-out pools on bars (middle)
3.5.2 Sample Preparation and Analysis
Samples were washed through a plastic sieve stack on a sieve shaker; fractions < 63 μm were retained and larger fractions were discarded. Deionized water was rinsed through when needed but used sparingly. The samples were collected in plastic containers and dried at 40°C. XRF preparation consisted of crushing the dried samples with a tungsten carbide grinder and weighing into crucibles. The samples were and left overnight at 105°C, re-weighed and left in a furnace at 850°C overnight to combust any organics. 2 g of sample was mixed with 6 g of lithium tetraborate and fused into glass discs. The glass discs were analysed using Siemens SRS 3000 Sequential XRF Spectrometer for elements SiO2, TiO2, Al2O3, Fe2O3, MnO, MgO, CaO, Na2O, K2O, P2O5, Ba, Rb, Sr, Pb, Th, Zr, Nb, Y, V, Cr, Ni, Cu, Zn, La, and Ce. Concentrations were all above the Limit of Quantitation (LoQ) with several trace elements displaying values close to their limits (Table 7).
Table 7: Limits of Quantification (LoQ) from XRF analysis for the major and minor elements
Element Mean (%) LoQ (ppm) Element Mean (ppm) LoQ (ppm)
SiO2 63.5 295.4 Ba 478.4 27.3 TiO2 0.7 42.6 Rb 108.1 4.2 Al2O3 15.3 193.5 Sr 234.2 3.2 Fe2O3 4.8 23.1 Pb 20.4 10.3 MnO 0.1 13.8 Th 9.0 6.1 MgO 1.4 102.2 Zr 260.2 3.4 CaO 1.4 70.4 Nb 11.2 3.2 Na2O 2.8 121.5 Y 23.9 3.3 K2O 2.5 23.8 V 94.1 10.2 P2O5 0.2 26.8 Cr 54.1 11.6 Ni 17.9 8.2 Cu 14.2 9.2 Zn 77.4 6.2 La 31.8 15.9 Ce 50.9 43.1
The glass discs were retained and analysed on a Agilent Technologies 7700 Series LA-ICP-MS for Sc, V, Cr, Co, Ni, Cu, Zn, Ga, Rb, Sr,Y, Zr, Nb, Cs, Ba, La, Ce, Pr, Nd, Sm, Eu, Gd, Tb, Dy, Ho, Er, Tm, Yb, Lu, Hf, Ta, Pb, Th, U. The LA-ICP-MS used BCR-2G chemical standard every 10 samples to check accuracy and precision. Since the glass discs are what are being ablated and measured, Ca is used as an internal standard to calibrate the LA-ICP-MS values against the XRF values.
3.5.3 Data Analysis
Recent sediment fingerprinting approaches have encouraged the use of a combination of independent statistical techniques to explore sediment source discrimination (e.g. Collins et
al., 2012). In this instance, exploration of the fingerprint data consisted of two statistical approaches. The first approach followed Collins et al. (1998) for tracer selection using a two stage statistical selection procedure. This involved statistical discrimination using non- parametric statistical tests of Kruskal-Wallis and Mann-Whitney tests to verify the ability of an individual tracer to provide statistical discrimination between sources. The second stage used a stepwise minimization of Wilk’s lambda multivariate discriminant analysis to deduce the final selection of variables that could discriminate between sources. The second approach utilized Principal Component Analysis (PCA) to explore the relationship and significance of the variables accounting for the variance of sediment sources. PCA is a data reduction technique that allows multiple correlated variables to be reduced to a set of uncorrelated variables called principal components. This is done through the use of a transformation that establishes the axis in the direction of the greatest variation, the first component. Due to the close proximity and lack of sample locations C1 and C2 were combined into one 3-junction confluence.
3.5.3.1 Statistical Discrimination
Mann-Whitney tests were used for C3, C4 & C5 and Kruskal-Wallis H test were used for C1-2. These tests were carried out for each analysed element concentration to develop a rank order that would indicate the statistically significant discriminants for subsequent analysis. The null hypothesis was that the upstream sediment sources group at each confluence were the same while the alternative hypothesis accepted upstream sediment sources to be different. A 95.0 % confidence interval or an α level of 0.05 was chosen for the critical p-value. Discriminant function analysis (DFA) allows for prediction of group membership based on linear combinations of predictor variables (Eq. 3)
ܦ ൌ ݒଵܺଵ ݒଶܺଶ ݒଷܺଷǤ Ǥ Ǥ ݒܺ ܽ (Eq. 3)
Where D = discriminant function v = the discriminant coefficient a = a constant
i = the number of predictor variables
Discriminant analysis was conducted using SPSS using stepwise discriminant analysis applying a minimization of Wilks’ Lambda. Wilk’s Lambda is a measure of the between group variability to within group variability whereby minimizing the value reduces the sample grouping. It is guided by ‘F’ values which determine the entry and removal of variables as a measure of the
extent to which an individual variable contributes to group prediction. Default values of 3.84 (probability = 0.5) and 2.71 (probability = 0.1) are used for F to enter and F to remove respectively. These values provided and adequate number of selected elements the purposes of this study without the need to alter the F values. However for C4 the F values were lowered to allow additional tracers to be employed in the final solution.
3.5.3.2 Principal Component Analysis
Principal component analysis was carried out using SPSS version 21.0 (IBM Corp., 2012). A correlation matrix was used to standardize the variable measurements. Component extraction was limited to eigenvalues greater than 1 unless the percentage of variation accounted for was low, in which case additional components were added. An oblique rotation of direct oblimin was used for the rotation method as a standard method for allowing factors to be correlated. Coefficients for the PCA components were applied to the original samples for visual display as well as well as rotated loading plots for the variables
3.5.4 Geochemical indicators
In order to evaluate sediment fingerprint data a collection of geochemical analyses and displays was performed. Multi element normalised diagrams for Major, Trace and REEs were constructed to show trends in average sediment geochemistry. Major and trace elements were normalized to Upper Continental Crust (UCC) using McLennan (2001) values revised from Taylor and McLennan (1985). Post-Archean Australian Shale (PAAS) values after McLennan (2001) were also used for comparison. Rare Earth Element (REE) values were normalized using average CI chondrite values from McDonough and Sun (1995). Plots were constructed using
PetroGraph from Petrelli et al. (2005).
Major element geochemistry was explored through Chemical Index of Alteration (CIA) (Nesbitt and Young, 1982), Index of Compositional Variability (ICV) (Cox et al., 1995) and K2O/Al2O3 values to give an indication of the major element character. CIA measures the conversion of feldspars to clays by measuring the depletion of CaO, Na2O and K2O relative to Al2O3 in the silicate portion. CIA values typically range from <50 for fresh material to 100 for fully weathered material.
ܥܫܣ ൌ ሺͳͲͲሻ ቂ మைయ
మைయାைכାேమைାమைሻቃ (Eq. 4)
CaO* only refers to the CaO incorporated into the silicate portion so a correction was applied according to McLennan (1993) to approximate the CaO* value devoid of the carbonate fraction. ICV is similarly a measure of the major cation composition relative to Al2O3 using wt%. Values <1 are typical of clay minerals and > 1 is more indicative of plagioclase and feldspars.
ܫܥܸ ൌ ቂሺிమைయାమைାேమைାைାெைା்ைమሻ
మைయ ቃ (Eq. 5)
K2O/Al2O3 ratios from 0.0 – 0.3 suggest clay minerals, and 0.3 – 0.9 suggest plagioclase (Cox et
al., 1995). Sorting and sediment recycling effects can be reflected in plots of Th/Sc and Zr/Sc, whereby, high Zr/Sc values indicate zircon enrichment associated with sediment recycling as Sc preserves provenance signatures whereas Zr is strongly associated with mineral zircon which is subject to sorting due to the high mineral density. (La/Yb)N is used as a measure of LREE/HREE enrichment and low Eu/Eu* ratio a measure of depletion of plagioclase, which is enriched in Eu2+. TotREE is the sum of all REE concentrations (in ppm) to show variation in total REE concentrations across sediment source groups. Particle size analysis was undertaken on the upstream and downstream sediment samples using a Horiba Partica LA-950v2 laser scattering particle size distribution analyser.