Particle size effect on geochemical composition of experimental soil mixtures relevant for unmixing modelling

Abstract

Sediment fingerprinting experiments have been used to demonstrate the sensitivity of numerical mixing model outputs to different particle size distributions in source materials and experimental sediment mixtures. The study aims to examine further grain size effects in the distribution of geochemical elements by soils through a laboratory experiment simulating mixing and sorting processes, to investigate if different size fractions are influencing fingerprinting analyses and unmixing model results. Multiple particle size fractions are analysed to understand the relationship between particle size and source signal through elemental signatures. FingerPro model was applied to unmix six experimental mixtures with known percentages contribution from three experimental sources. The experimental design comprised four different setups with a specific size fraction for sources (S) and mixtures (M). Setups A (S <63 and M <63 μm) and B (S <20 and M <20 μm) relies upon a comparable particle size fraction for sources and mixtures, while C (S <63 and M <20 with PSC) and D (S <63 and M <20) address particle size impacts simulating fine enrichment, with and without a single particle size correction factor, respectively. Tracers were extracted after applying two statistical tests, the range test (RT) and a combination of RT, Kruskal-Wallis (KW) and DFA tests thus obtaining the set of optimum tracers for each mixture. Our findings indicate that source apportionment results are sensitive to tracer selection and particle size. The most accurate source apportionment results were achieved when comparing sources and mixtures with the <63 μm grain-size fraction (setup A) by using the set of tracers extracted after RT, KW and DFA tests, (mean RMSE: 2%, AE: 2%). Larger errors were obtained progressively for setups B, C and D with better results when using more number of tracers from RT (mean RMSE: 7, 10, 13%, AE: 8, 11, 15%, respectively). The main strength of using experimental mixtures with a known contribution of the sources relies on reducing the uncertainty of the unmixing model outputs, one of the main limitations in fingerprint studies. The impact of SSA on the elemental concentration is difficult to predict because the positive linearity between them does not apply equally to all elements and this assumption needs to be constantly examined and considered for fingerprinting studies. Otherwise, the use of a single particle size correction factor could negatively affect unmixing results. The outcomes of this research will help to develop appropriate strategies for sediment fingerprinting, contributing to our knowledge of processes affecting sediment geochemistry and sediment transport across different particle sizes.