Evaluation of the performance of receptor statistical methods for quantifying desert dust contributions to ambient particulate for health studies and policy

There is a body of evidence on the risk to human health posed by the exposure to desert dust. However, results from epidemiological studies from different regions are inconsistent. Among possible causes of inconsistency is the scarcity of daily desert dust contributions to PM 10 and PM 2.5 levels recorded in populated areas in order to conduct rigorous epidemiological studies, but also the lack of robust and harmonized methodologies to deliver airborne desert-dust concentrations. This study evaluated the performance of the main statistical methods currently used by European countries to estimate the net load of desert dust on PM 10 and PM 2.5 levels during the occurrence of desert dust outbreaks. To this end, long-term data series (2010-2023) on PM 10 and PM 2.5 levels and composition obtained in a regional background (Montseny) and an urban background (Barcelona) monitoring site, in Northeast Spain, were used. The results identify the most appropriate method for determining the regional daily PM 10 background concentration, excluding dust-days. This involves applying a moving 50 th percentile with a 30-day time window to data from nearby regional background environments. Such PM 10 background is essential for calculating the daily net dust contribution (PM 10 net-dust ). However, applying this procedure to data from urban or industrial environments causes an overestimation of PM 10 net-dust values, which intensifies with higher local PM 10 levels. On the other hand, in the case of low PM 10 net-dust values(<3 μg/m 3 ) the relative errors are so high that it is not advisable to use these estimates of natural PM contributions to assess compliance with air quality standards. For PM 2.5 , however, the mineral dust content is much lower than for PM 10 . Consequently, applying the same methodology results in significantly greater relative errors in the PM 2.5 net-dust estimates. In this case, it is also advisable to use PM 2.5 data series obtained in regional background environments. However, if nearby regional background data is not available, the method can be applied directly to the evaluated urban or industrial datasets, but excluding traffic and industrial hotspots for the calculation of the PM net-dust . In conclusion, accurately quantifying PM net-dust is a complex issue, and it is necessary to continue improving current methods and developing new methods that allow for the most accurate estimation possible of daily desert dust contributions, especially for the low concentration ranges and for finer PM size fractions. The methodologies reported here are applicable to all regions affected by desert dust. • Methods to evaluate desert dust contributions to PM concentrations are evaluated. • Statistical receptor modelling is used and validated with PM speciation datasets. • Long term datasets (2010-2023) are used to this end. • Differences in errors and accuracy depend on the methods and the PM size fraction. • The results can be used for epidemiological studies and for policy implementation.