6.2 Unmixing Sources
In order to further characterize the evolving source of sediments to the Indus Fan we employ the unmixing software of Sundell and Saylor [2017], which analyzes the U-Pb age spectra from each of the samples and compares them with the defined end-member compositions of the different source ranges compiled from the published literature. This approach works particularly well in the western Himalaya where the sources are well defined and often unique. Data from the Tethyan, Greater and Lesser Himalaya were compiled from DeCelles et al.[2004]. Data from the Karakoram are from Le Fort et al.[1983], Parrish and Tirrul [1989], Schärer et al.[1990], Fraser et al. [2001] and Ravikant et al.[2009]. Data from Nanga Parbat are from Zeitler and Chamberlain [1991] and Zeitler et al. [1993]. Data from the Transhimalayan are from Honegger et al. [1982], Schäreret al. [1984], Krol et al. [1996], Weinberg and Dunlap [2000], Zeilinger et al. [2001], Dunlap and Wysoczanski [2002], Singh et al. [2007], and Ravikantet al. [2009].
This unmixing method uses a Monte Carlo approach to estimate the contributions from the different sources that would be required to generate the modes and modal abundances of U-Pb ages seen in the sediment samples. Because this is relatively objective the method is considered robust for analyzing potential source contributions, assuming that the sources themselves have been well characterized. The bedrock sources of the Indus catchment have significant differences between many of them and are some of the best characterized worldwide. Results of the Monte Carlo simulation are provided in Table 4, showing the output using all three statistical comparison methods, cross-correlation, the best V value in the Kuiper test. as well as the best D value in the K-S test. The method involves creating 10,000 model mixed sediments using the defined bedrock source end members. The DZMix software then compares the model with the measured spectra and retains the best 1% of these models in order to estimate which sources were contributing the sampled material. We favor the unmixing models derived from the cross-correlation approach as being geological reasonable and favored by Sundell and Saylor [2017].
The results of our unmixing calculations show a progressive provenance evolution that is consistent with that seen in the MDS diagram (Figs. 11 and 12). The very oldest sample deposited at 15.5 Ma shows a dominance of sediment eroded from the Karakoram and from the Tethyan and Greater Himalaya. Most of the Miocene samples dated between 8.2 and 7.0 Ma are more dominated by material from the Karakoram but also usually show significant Tethyan and Greater Himalayan contributions. This Himalayan component is particularly noteworthy at 7.99, 7.84, 7.78, 7.66, and 7.0 Ma during this interval. The proportion of Karakoram zircons shows a significant decrease starting no later than 5.72 Ma and again at 3.02 Ma. The sediment deposited at 3.17 Ma shows the greatest amount of modeled erosion from Karakoram sources of any sample.
From 3.02 Ma onwards the Himalaya dominate as sources to the submarine fan, with significant amounts of material from the Lesser Himalaya first appearing at 1.56 Ma. The sample dated as being deposited at 0.93 Ma is anomalous for being very similar in source signature to Tethyan and Greater Himalayan bed rocks sources. However, we note that fission track data indicate that this sample was derived from peninsular India [Zhou et al. , 2019]. The unmixing analysis largely mirrors the pattern shown by the MDS diagram, in showing a progressive long-term increase in erosion from the Himalaya relative to the Karakoram, although with significant steps making the evolution nonlinear. All of the samples contain a small amount of very young <25 Ma zircons. None of the samples analyzed show a close similarity with post-LGM river compositions. Delta samples deposited at 6.6 and 15 ka are strongly enriched in Karakoram-derived grains compared to fan sediments deposited at and after 3.02 Ma. This short term variability is interpreted to reflect the short-term changes in erosion patterns linked to monsoon strength, modulated by glacial cycles since the onset of the NHG.