Stratospheric impacts on dust transport and air pollution in West Africa and the Eastern Mediterranean

North Africa is the world’s largest dust source, and dust transported toward the Atlantic and Mediterranean has major health, environmental, and economic impacts across densely populated coastal regions. Despite the importance of accurate dust forecasts for mitigating exposure and informing policy actions, current operational models generally provide reliable guidance only 2–5 days ahead. Extending predictability to subseasonal-to-seasonal (weeks to months) timescales would substantially benefit public health and economic planning. Stratospheric variability—particularly sudden stratospheric warmings (SSWs), which disrupt the polar vortex and induce a negative North Atlantic Oscillation (NAO)-like surface pattern—offers a recognized source of such predictability. While the NAO’s influence on North African dust export has been documented, the direct connection between SSWs and desert dust emission/transport has not been systematically explored. This study tests the hypothesis that SSWs modulate North African dust emission and transport, leading to regionally coherent and predictable changes in surface dust burden and air quality across West Africa and the Eastern Mediterranean on subseasonal timescales.

Prior work has established the strong societal and health impacts of Saharan dust intrusions in Europe and the Mediterranean, including associations with mortality and significant economic costs. Operational dust forecasts in Europe typically have limited lead times of several days, though specific upper-level precursors (e.g., Rossby wave breaking) can offer 5–10 day early warnings for some events. On subseasonal-to-seasonal timescales, the stratosphere contributes to surface predictability, with SSWs known to project onto a negative NAO-like pattern that drives precipitation and temperature extremes. The NAO has been shown to modulate North African dust export, but a direct, quantitative assessment of SSW impacts on dust emission, transport, and surface air quality has been lacking. This gap motivates investigating whether SSWs can systematically alter dust burdens and provide actionable predictability for air quality in affected regions.

The study combines chemical transport model simulations, reanalyses, and observations. Meteorological fields (daily sea level pressure, 10 m winds) are from MERRA-2 (1.25°×1.25°). Dust concentrations are taken from two model products driven by MERRA-2 meteorology: CESM2 (DEAD dust module; 1980–2014; ~0.9°×1.25°) and the MERRA-2 aerosol reanalysis (GOCART; 1980–2020; 1.25°×1.25°). Note that MERRA-2 dust fields are largely model-based (optical depth is assimilated, dust mass is not directly constrained). SSWs are identified from MERRA-2 by reversal of zonal-mean winds at 60°N, 10 hPa; 24 SSWs occur during 1980/81–2019/20 (24 for MERRA-2 period; 22 for 1980/81–2013/14 relevant to CESM2). SSW episodes are defined as days [1,30] after onset. For comparison, 1000 sets of non-SSW episodes are constructed by randomly sampling NDJFM dates without SSWs matched to SSW calendar dates. Composite differences and Monte Carlo tests (two-sided, p<0.10) assess significance. Model-derived dust pollution metrics include: (a) mean surface dust concentrations and (b) number of high dust days per 30-day period exceeding thresholds of 250, 500, and 1000 µg m⁻³. Regional analysis focuses on Eastern Mediterranean (30°–33.75°N, 10°–30°E) and West Africa (16.25°–23.75°N, 12.5°–23.75°E). Observational verification uses daily PM₁₀ from background stations: Finokalia, Greece (Finokalia Atmospheric Observatory) and three INDAAF sites in West Africa (M’Bour, Senegal; Cinzana, Mali; Banizoumbou, Niger), plus AERONET AOD at these sites. For observations, air quality indicators are: (i) mean PM₁₀ levels and (ii) expected number of poor air quality days per 30 days using thresholds 50 µg m⁻³ (EU) and 260 µg m⁻³ (Senegal/Tunisia). Rates are normalized by available data days to account for missing observations. Aerosol optical depth from MERRA-2 is evaluated at grid points nearest to AERONET stations. Statistical significance for regressions uses two-tailed Student’s t-tests with effective degrees of freedom; composite significance uses Monte Carlo resampling. Mortality impacts from dust-source PM₂.₅ (from models) are estimated using a short-term concentration-response function: a 10 µg m⁻³ increase in 2-day averaged PM₂.₅ associates with a 0.98% increase in total mortality. For each SSW, 2-day averaged PM₂.₅ anomalies are computed for 15 consecutive 2-day periods over days [1,30]; grid-cell mortality changes are aggregated using baseline mortality (GBD) and population (CIESIN), then summed regionally to yield excess deaths per SSW event. Uncertainties arise from concentration-response parameters and baseline mortality choices.

- SSWs produce a cyclonic sea level pressure anomaly over southern Europe/northern Africa, stronger and more eastward-reaching than the NAO regression alone, consistent with a negative NAO-like signal. This shifts surface winds to favor southwesterlies into the Eastern Mediterranean and weakens northeasterly trade winds over North Africa. - Chemical transport models (CESM2 and MERRA-2) show a robust dipolar dust response during days [1,30] after SSW onset: enhanced surface dust over the Eastern Mediterranean and reduced dust over West Africa. The negative response aligns with climatologically high winter dust regions in West Africa. - Quantitatively, in the Eastern Mediterranean, SSWs increase mean surface dust concentrations by ~20–30% and high dust days by ~20–40% (thresholds 250, 500, 1000 µg m⁻³). In West Africa, mean dust concentrations decrease by ~10–20%, and high dust days fall by ~10–30%, relative to the non-SSW mean. - Mechanisms: Over the Eastern Mediterranean, SSWs increase the fraction of southwesterly wind occurrences, enhancing northward Saharan dust transport; easterly winds (which could bring Arabian dust) decrease. Over West Africa, SSWs reduce the probability of high surface wind speeds (>5 m s⁻¹) and increase low speeds (<4 m s⁻¹), suppressing dust emission and westward transport. - Observations corroborate model results: Regressions of SLP onto station PM₁₀ show higher PM₁₀ in Greece under cyclonic anomalies and higher PM₁₀ in Senegal under anticyclonic anomalies. Station PM₁₀ and AERONET AOD indicate worsened air quality in Greece and improved air quality in Senegal during SSW episodes. - Health impacts: Per SSW event, dust-source PM₂.₅ changes lead to an estimated 680–2460 additional premature deaths in the Eastern Mediterranean and prevent 1180–2040 premature deaths in West Africa (ranges across models; CESM2 larger, MERRA-2 smaller), with impacts tracking population centers. - Predictability: SSWs are predictable 1–2 weeks in advance and their surface impacts can persist for weeks to months, offering subseasonal predictability for dust and air quality in the affected regions.

The study demonstrates that SSWs systematically modulate North African dust emission and transport, addressing the open question of whether stratospheric dynamics can inform subseasonal dust and air quality forecasts. The negative NAO-like surface response to SSWs enhances southwesterly flow into the Eastern Mediterranean, boosting dust intrusions and degrading air quality, while simultaneously weakening northeasterly trade winds over North Africa, reducing dust emission and westward transport and improving air quality in West Africa. Agreement across two independent model systems, consistency with observed PM₁₀ and AOD responses, and physically grounded wind and pressure changes support the robustness of the dipolar dust signal. The quantified mortality impacts underscore the public health relevance. Importantly, because SSWs are detectable and forecastable with 1–2 week lead times and impart multi-week surface effects, they provide a practical source of subseasonal predictability for air quality management. The findings suggest incorporating stratospheric diagnostics into dust forecasting frameworks could improve early warnings. The discussion also notes that other stratospheric extremes (strong vortex events) may induce opposite-signed responses, and spring final warmings may extend similar impacts beyond winter, highlighting a broader stratosphere–dust linkage.

This work establishes a clear, dipolar impact of sudden stratospheric warmings on North African dust and surface air quality: increased dust burden and degraded air quality in the Eastern Mediterranean, and decreased dust burden with improved air quality in West Africa. These changes are substantial (typically 10–30% relative to non-SSW conditions) and yield consequential shifts in premature mortality attributable to dust-source PM₂.₅. Given the 1–2 week predictability of SSWs and their multi-week surface impacts, the stratosphere offers a valuable, underused source of subseasonal predictability for air quality. The study advocates integrating stratospheric variability into operational dust and air quality forecasting systems. Future research should (i) disentangle and quantify differences between the SSW-induced surface response and canonical NAO impacts, (ii) assess strong vortex and final warming events for opposite or extended dust responses, (iii) reduce model spread in dust emissions and transport through improved emission parameterizations and assimilation, and (iv) expand observational networks and event-based verification to refine regional health impact estimates.

- Model uncertainty: CESM2 and MERRA-2 exhibit different dust emission magnitudes (CESM2 higher, MERRA-2 lower), reflecting broad inter-model spread. MERRA-2 dust mass is not directly constrained by assimilation, increasing uncertainty in dust fields. - Observational constraints: Station records have gaps and limited spatial coverage; potential sampling errors may affect observed responses. AOD and PM₁₀ provide partial proxies for PM₂.₅ health impacts. - Health impact estimation: Mortality calculations rely on a single short-term concentration-response function and baseline mortality/population datasets; uncertainties in β and demographic heterogeneity are not fully captured. - Temporal and regional scope: Analysis focuses on extended boreal winter (NDJFM) and specific regions (Eastern Mediterranean, West Africa); results may differ in other seasons or areas. - Dynamical attribution: While linked to a negative NAO-like response, the SSW surface signal extends beyond simple NAO projection; the precise mechanisms and remote influences remain to be fully resolved. Operational predictability of SSWs is limited to ~1–2 weeks, constraining lead time.