WHAT DO MAGNETIC COERCIVITY COMPONENTS REALLY TELL US ABOUT THE ORIGIN OF MAGNETIC GRAINS IN LOESS-PALEOSOL SEQUENCES?

Unmixing of isothermal remanent magnetization (IRM) curves is widely used in rock magnetic and environmental magnetic studies to discriminate between magnetic coercivity components of different origins. Acquiring IRM curves in the lab is done relatively fast (minutes), but the interpretation of the resulting coercivity components in terms of their origin remains ambiguous, due in part to the fact that IRM curves are low-dimensional (1D) magnetic characterization tools. In order to reduce this ambiguity, we combined non-parametric IRM unmixing with unmixing of non-linear Preisach (NLP) maps, which are a category of higher-dimensional (2D) magnetic characterization tools. The drawback of acquiring NLP datasets is that they are time consuming (hours), but the advantage is that they provide a more straightforward interpretation of coercivity components in terms of magnetic domain state, interactions, and constituent magnetic phases. We applied our method to loess-paleosol samples from two typical mid-latitude northern hemisphere sequences from Romania. The loess-paleosol sequences, one located along the Danube River in S Romania and the other on the Black Sea shore, exhibit different pedogenesis intensities, have distinct source areas for the dust, and experienced different climatic conditions. Both our unmixing methods provide two nearly identical end-member (EM) coercivity components in both loess-paleosol sequences. From the NLP maps we can glean that the first end member (EM1) is of pedogenic origin and is dominated by low coercivity (20-30 mT) non-interacting single domain magnetite. The second end member (EM2) is of eolian origin and exhibits a moderate coercivity (40-60 mT) given by a mixture of fine and coarse magnetite/maghemite, as well as hematite. The fact that EM1 is the same at both locations suggests that this component is independent of climate and degree of pedogenesis. The same EM2 indicates that a similar magnetic population is transported and deposited via eolian processes, irrespective of the dust source area. By combining the two unmixing methods we provide strong evidence that the widely-used non-parametric IRM unmixing method isolates physically-realistic EMs. Complementary unmixing of NLP maps is critical in elucidating the domain states of the magnetic phases present in each EM.