<p>This paper explores the potential for using multiple particle size fractions in a hierarchical geochemical sediment fingerprinting approach to the assessment of changes in sediment sources through time within a small Mediterranean peri-urban catchment. Conventional &#160;sediment fingerprinting has focussed on the <63&#181;m fraction of fine bed-sediment on the basis that this fraction represents suspended sediment, which in turn is considered dominant over bedload in catchment sediment budgets. In reality, however, coarser sediment than 63&#181;m may form part of suspended sediment and/or occurs as relatively fast-moving fine bedload.&#160; Furthermore, sediment sources vary in their particle size distribution and, as geochemical composition can vary with particle size, it is arguable that sediment fingerprinting studies should consider use of multiple size fractions.</p><p>This study explores this approach using <63&#181;m, 63-125&#181;m, 125-250 &#181;m and 250-2000&#181;m size fractions.&#160; It focuses on the north-south flowing Ribeira dos Cov&#245;es catchment (6.2 km<sup>2</sup>), on the outskirts of Coimbra in central Portugal. The climate is humid Mediterranean. Catchment geology is 56% sandstone (in the east), 41 % marly limestone (in the west) and 3 % alluvium. Current land-use is 56% woodland, 4 % agricultural and 40% urban (mainly residential, but also including a recently constructed enterprise park (5%) and major highway (1%)). Recent urbanization has largely occupied former agricultural land.&#160;</p><p>The study adopts a multi-proxy sediment fingerprinting approach to assessment of changes in sediment sources, based on geochemical (elemental) characterization of the four different size fractions of fluvial bed-sediment and soil samples, using a Niton x-ray fluorescence (XRF) elemental analyser. Sampling of fluvial sediment was carried out at 33 sites within the stream network (including all significant tributaries, downstream sites and the catchment outlet). Samples were collected in July 2018 and November 2018 following contrasting &#8216;late-wet-season&#8217; and &#8216;end-of-dry-season&#8217; events. In July 2018, samples of potential sediment sources were collected including: (i) soil surface (0-2cm) samples at 64 locations, (ii) 17 samples from eroding channel margin sites, and (iii) 15 samples of road sediment. All fluvial and soil samples were sieved to obtain the four target size fractions. The elemental geochemistry of each sample fraction at all fluvial and source sites was derived using the XRF analyser.&#160; (These results were added to similar datasets previously obtained on three occasions in 2012-15 in a period of enhanced urban constructional disturbance). Differences (and similarities) in geochemical signatures between the different size fractions at each survey date at and between each tributary and potential source site were assessed using a range of statistical techniques.&#160; Messages arising are discussed. For each size fraction and survey date, Bayesian unmixing models were used in a hierarchical (confluence-based) fashion to assess the contributions of sub-catchments to downstream sites and the catchment outlet. Modelling results for the two 2018 events were validated by comparing them with suspended sediment records collected at five tributary locations and at the catchment outlet.&#160; Although overall, the modelling was successful in indicating and quantifying significant changes in sediment sources through time within the catchment, uncertainties in interpretation of the multiple fractions are identified and discussed.&#160;</p>
Documentation of particle-size analyzer time series, and discrete suspended-sediment and bed-sediment sample data collection, Niobrara River near Spencer, Nebraska, October 2014
