Background

  • The chemical composition of river sediments integrates a variety of biogeochemical processes taking place in the basin, enabling their quantification on a large spatial scale.
  • In particular, studies of river chemistry have provided crucial insights into the complex feedbacks between climate, physical and chemical weathering, and the global carbon cycle (e.g., Gaillardet et al., 1999, Hilton & West, 2020). Further progress requires the explicit partitioning of competing processes, like CO2-driven vs H2SO4-driven weathering (Torres et al., 2016), or biospheric OC burial vs petrogenic OC oxidation (Hilton et al., 2014) and more accurate quantification of riverine fluxes. See this year's Ingerson Lecture by Christian France-Lanord for a nice introduction into this topic.
  • Both of these are especially difficult (but rewarding) in large river basins that spanning complex lithologies and climatic zones. In addition, sediments in large river channels are strongly affected by hydrodynamic sorting, complicating the measurement of average composition and total flux.

Methods

  • We collected water and sediment depth samples, as well as ADCP discharge data across the length of the basin, over several expeditions in 2017-2019.
  • The map on the left shows the sampling locations of the mainstem (large symbols) and the tributaries (small symbols; gray - upper basin, pink - lower basin). Chindwin tributary (green) comprises ~1/3 of the basin and is a major sediment source.
  • We used suspended sediment and particulate organic carbon concentrations (SSC and POC) to quantify the total sediment and POC fluxes of the Irrawaddy, using a hydrodynamic Rouse-based model. This work is summarized in our recent EGU 2020 presentation and a detailed account is given in a paper recently accepted in JGR:ES.
  • A number of the sediment samples were treated with sequential leaching, revealing negligible presence of carbonates and limited reactive (oxy)hydroxides. Both leached residues and unleached bulk samples are thus presented together below.
  • Sediment chemistry was determined using a newly developed lithium borate "microbead" fusion method, requiring less than 50 mg of sample, followed by ICP-AES.

Overview of data

  • Since the dataset consists of many samples collected from different sites, statistical techniques like Principal Component Analysis (PCA) can be very useful to assess which parameters (in this case oxides) contribute to most variation.
  • Here, we use PCA to reduce a multi-dimensional dataset (Si-Al-Fe-K-Na-Ca-Mg oxide concentrations) to two dimensions, while still retaining most variance (in this case 79%).
  • When applied to sediment samples across the Irrawaddy basin, most variance appears to derive from the major oxides of Si-Al-Fe (spread along horizontal axis of PC1).
  • However, sediment chemistry are compositional data (e.g. wt% oxide concentrations) that must sum to 100%, which introduces auto-correlations between elements (e.g., as SiO2 goes up, all other concentrations must go down). Another issue is that the majority of the observed variation can be dominated by the species with highest concentrations (eg. Si-Al-Fe as above), while we are more interested in the less abundant elements (like Na-K-Ca-Mg variations due to weathering or mineralogy).

Compositional data approach

  • Appropriately transforming compositional data can more clearly distinguish differences between sample sets and help disentangle different factors controlling their chemistry. Different techniques have been developed to do this (e.g., Aitchison, 1983; Tolosana-Delgado, 2012).
  • Here we use the centered-log-ratio (clr) approach, which normalizes log-transformed oxide wt% concentrations of all elements within a sample to their geometric mean.
  • Repeating PCA on clr-transformed oxide concentrations (denoted as c(oxide)) reveals which elements vary the most in relative, rather than absolute, terms (see figure on the right). The scale-driven dominance of Si-Al-Fe is removed and even more of the total variance is now explained by two PCs (87%).
  • In the Irrawaddy, upper basin sediments are more Ca-rich (blue triangles), lower basin tributaries are Al-Fe-rich (pink diamonds). Sediments at the mouth (red circles) represent a mixture, with a major contribution by the Chindwin tributary (green triangles).

Hydrodynamic sorting

  • Lipp et al. (2020) have recently proposed that the above approach can describe most global river sediment and sedimentary rock data, distinguishing between lithological provenance and chemical weathering.
  • On a single river basin scale, however, most of the variation is observed within each sampling site (see figure above). This spread is systematically correlated with grain size (see figure on the left), indicating the overwhelming influence of hydrodynamic sorting on sediment chemistry.
  • In the Irrawaddy, fine sediments are dominated by Fe-Al-Mg, while coarse sands are dominated by Si-K. These findings are consistent with quartz dominance in river bottom sands demonstrated previously (Bouchez et al., 2011) but also reveal that K-feldspar might be concentrated in the sand fraction in the Irrawaddy.

Mineralogy

  • The inferred correlation between chemistry and mineralogy can be partly confirmed using mineral grain count data of sands across the Irrawaddy basin (Garzanti et al., 2016).
  • Although data are scarce, they directly support the conclusion that coarse samples are dominated by quartz (see animated figure on the left; lower right quadrant).
  • In contrast, the more Ca-Na-rich sampled (upper right quadrant) are dominated by feldspar (likely Ca-Na-plagioclase), suggesting distinct provenance of sediments eroded in the northern headwaters.

Lithology

Chemical weathering


Thank you for reading!

Ask your questions here!



Funding source: