Using GLUE to pull apart the provenance of atmospheric dust

Abstract

Abstract Identifying the sources of aeolian dust is a crucial step in mitigating the associated hazards. We apply a Generalized Likelihood Uncertainty Estimation (GLUE) model to constrain the uncertainties associated with sediment fingerprinting of atmospheric dust in the Sistan region on the Iran-Afghanistan border, one of the world’s dustiest places. 57 dust samples were collected from the rooftop of the Zabol Department of Environmental Protection during a summer dusty period from June to October 2014, in addition to 31 surface soil samples collected from potential sources nearby, including cultivated land (n\u202f=\u202f8), uncultivated rangeland (n\u202f=\u202f7), and two dry lakes: Hamoun Puzak (n\u202f=\u202f10) and Hamoun Saberi (n\u202f=\u202f6). Dust and soil samples were analyzed for 24 tracers including 16 geochemical elements and 8 water-soluble ions. Five optimum composite fingerprints (Fe, Sr, Mn, Cr and Pb) were selected for discriminating sources by a two-stage statistical process involving a Kruskal-Wallis test and stepwise discriminant function analysis (DFA). Uncertainty ranges for source contributions of dust determined by the GLUE model showed that the dry lake Hamoun Puzak is the dominant source for all dust samples from Zabol and cultivated land is a secondary source. We found marked spatial variance in the importance of regional dry lake beds as dust sources, and temporal persistence in dust emissions from Hamoun Puzak, despite very large areas of adjacent lake beds drying during the study period. Aeolian sediment fingerprinting studies can benefit considerably from the constraints provided by modelling frameworks, such as GLUE, for quantifying the uncertainty in dust provenance data.