Wind-blown dust (WBD) emitted by the Earth's surface can have a significant effect on both climate and human health by reducing air quality. It affects the climate directly and indirectly by scattering solar radiation, modifying the cloud properties and inducing precipitation as it can also serve as cloud-condensation nuclei . Additionally, exposure to high levels of dust particles can have severe effects on human health in the respiratory and the cardiac system .

One of the major WBD emitters of concern in Europe (but also globally) is the Sahara, which contributes up to 50 % of dust emissions globally. Sahara dust is a major contributor to European atmospheric pollution as well, and its levels are critically high in southern Europe, while light dust episodes are often detected above central and western Europe . Other natural sources can be wildfires, which due to intense turbulence can generate dust emissions . WBD can be emitted also by non-vegetated areas containing fine and loose sediments when strong winds occur. Human activities contribute significantly to increasing dust generation too. Destruction of soil crust and vegetation removal in semi-arid regions, changing cultivation patterns, and new transport pathways are some of the most impactful anthropogenic activities .

With climate change, dust emissions are anticipated to increase in the future . Modified climate conditions (with the associated weather patterns) and changes in land use are the main factors affecting the dust emissions. If dry periods between the precipitation events are prolonged, then the soil of the surface is going to be susceptible to strong winds, resulting in an increase in dust emissions. assessed how the anthropogenic contribution to the emissions has affected the probability of droughts in Europe. Their results stress that the drought risk for southern Europe has already increased, although the results for central Europe are inconclusive. used two precipitation indices and showed significant increases in drought likelihood for southern Europe and decreases in likelihood in the total area of the north, resulting in values that are dependent on the geographical domain and can shift the spatially averaged values for all of Europe. On the other hand, many studies have shown that fine particles can be transported over long distances through the atmosphere and can elevate particulate matter (PM) levels in different areas of the continent, far from the source area . Hence, mineral dust emissions must be examined in connection with both the main dust centres (like the Sahara) but also in relation to emissions over non-arid areas like Europe, where the temporal distribution of precipitation and denser vegetation normally prevents the necessary soil drying for such emissions. Indeed, as said above, in a changing climate, such conditions can be more frequent.

Over Europe, very few studies accounted for the local (i.e. not advected from other continents) dust emissions. Recently, identified potential dust sources over Europe (among other areas). gave a detailed model-based estimate for the spatial and temporal variation in such emission using a mesoscale weather model (MM5) as the meteorological driver. They, however, did not compute their contribution to the total PM concentrations. calculated the emission from natural erosion and resuspension over Europe and found significant model (CHIMERE) improvement when these emissions accounted for PM. However, only considered two seasons in a selected year without taking long-term effects into account. Similarly, considered a strong European dust event originating in Ukraine, but this cannot be considered representative of the long term. Recently, looked at urban dust over Europe and calculated its impact on PM levels. Moreover, they also accounted for the impact on the formation of secondary aerosols. They showed that the urban dust source can be significant and can potentially reduce model biases. However, they were not interested in other dust sources, e.g. those originating from soils in rural/natural areas, and they only looked at 1 month and did not provide a long-term estimate.

Motivated by this, here we present a novel study to quantify the long-term dust emissions for present-day conditions over central Europe using a regional climate model coupled with a chemical transport model along with a WBD model for dust fluxes. For the correct modelling of the potential future evolution of WBD, it is crucial to first evaluate the models' ability to resolve their present-day magnitude and the associated impact on the total PM concentrations. Our study focuses on the long-term impact during a 10-year period, which allows us to obtain a representative pattern of the temporal and spatial distribution of the WBD emissions and their overall impact on PM levels. Moreover, this study will also look at the secondary impact of WBD particles on secondary aerosol components focusing on the inorganic aerosol. Indeed, there is an indication that the composition of dust particles can have an indirect impact on nitrates, sulfates and ammonia either by acting as a surface for heterogeneous reactions (e.g. ) or through the dust particles' ion composition and modulation of aqueous reactions that form nitrates and sulfates , representing an indirect pathway of impacting the overall PM levels. In this study, the main interest will be the quantification of WBD contribution to urban PM levels as urban areas already experience adverse air-pollution episodes, and it is of interest to calculate how natural emissions like WBD can potentially contribute to urban PM concentrations.

To achieve the goals of the study, we applied the chemical transport model CAMx (Comprehensive Air-quality model with Extensions) offline, driven by the regional climate model Weather Research and Forecasting (WRF). The emissions of wind-blown dust were calculated by the wind-blown dust (WBDUST) emissions model. All these models are described in detail below.

This study aimed at the potential long-term regional impact of dust emissions on PM2.5 and PM10 concentrations for the period 2007–2016. The analysis focused on central Europe and on big urban areas such as Berlin, Prague, Vienna, Munich, Budapest and Warsaw. The impact was also estimated for the secondary inorganic aerosol concentrations as constituents of PM2.5.

In our simulations, the annual average coarse and fine PM emitted and averaged over the whole domain is about 1.5 and 0.5 Mgyr-1km-2, so about 2 Mgyr-1km-2 for the total PM10. This value is 1 order of magnitude higher than that calculated by for Europe. The dust emissions show significant temporal and spatial patterns. Our dust model computed 2 times stronger emissions during the winter period than in summer (for both the fine- and coarse-dust particles). Over high-latitude areas, reported strong winter dust emissions over areas where, under dry conditions, the sublimation of snow (and eventually permafrost) occurs and the soil is more prone to saltation, while during summer, the soil is generally more moist reducing the saltation potential of soil particles. In the dust model used in this study, the three most important parameters affecting the dust emissions are the near-surface wind speed, snow-equivalent water and soil moisture. The reason for much higher winter WBD emissions can be (1) much higher modelled winter wind speeds compared to summer ones, lower soil moisture during the winter months and an underestimation of snow cover, which prevents dust events. We saw that our driving model (WRF) produced much higher winds than the ones measured, and this positive bias is largest during winter. This strong overestimation is a known feature of the BouLac planetary boundary layer (PBL) scheme used in this study, and others have reported a similar overestimation of wind speed (e.g. ). As dust emissions scale non-linearly with wind speeds that are above a certain threshold , this raises the potential for overestimating dust emissions if winds are overestimated. Our sensitivity estimates confirmed this and, even at a 5 % reduction in model wind speeds, greatly reduced the calculated WBD fluxes. We must note too that by driving WRF with the newer ERA-5 reanalysis data instead of ERA-Interim, some of the wind biases would probably be reduced as it has been shown by many (e.g. ) that ERA-5 provides somewhat lower near-surface wind speeds over Europe compared to ERA-Interim.

Regarding the modelled soil moisture, it is comparable to observed values in winter and somewhat higher during other months than that measured. This means that the strong winter emissions are probably due to high wind speeds in WRF. The last factor potentially playing a role is the snow cover, which was not evaluated in this study, but the modelled precipitation exhibited some underestimation in winter, which might result in reduced snow in our simulations (even though temperature is underestimated in winter).

Apart from the clear annual cycle, the calculated emissions show a diurnal cycle too. Daytime emissions are usually 50 %–100 % higher than nighttime ones. The reason for this is most probably the well-known cycle of wind speed with maxima occurring at noon , and similar diurnal behaviour of dust emissions were seen in other studies too (e.g. ).

The daily time series of dust emissions provide some hints regarding their distribution: while on most of the days, the coarse-mode emissions remain low (lower than 0.1–0.2 gs-1km-2), on selected days the emissions peak at much higher values (1–2 gs-1km-2). The same is true for the fine-mode dust. This points to the fact already mentioned that the dust emissions respond non-linearly to wind speeds (more specifically to friction velocity; see Eq. 2) above a certain threshold. As wind speeds are overestimated in the driving meteorological model (WRF), the emissions are probably also overestimated, or at least these strong peaks are not realistic. Indeed, the sensitivity analysis to wind speed reduction showed a very high sensitivity of WBD emissions to this parameter: making the wind speeds half of the original values removed almost all WBD emissions. This means that very accurate meteorological driving data are needed to constrain the wind-blown dust emissions. We can further expect that due to the positive wind bias and resulting overestimated WBD emissions, the actual dust emissions are closer to what calculated.

One interesting feature is evident from the modelled geographical distribution of seasonal WBD emissions. Besides large rural areas emitting dust, the largest dust sources are concentrated near large urban areas: the Ruhr area in Germany, the highly populated Benelux states and also large cities like Berlin, Prague, Budapest, etc. One has to be very cautious in interpreting this result: dust is potentially emitted only over land use categories representing potentially bare soil (if other circumstances are met), i.e. crops, shrubs, grass land, tundra and desert. “Urban” land use is, however, not treated as having dust emission potential. On the other hand, urban areas are characterized by low vegetation and thus low leaf area index (LAI). Indeed, in the LAI data used (MODIS), cities have near-zero LAI for most of the year. As land use in the WBDUST module (and also for CAMx dry deposition) is represented as fractional land use (based on CORINE data), many of the 9 km grid boxes covering urban areas are partly covered with bare soil and partly urban land use. If the LAI value is too low for such areas, dust emission can occur in the model. This was the case for densely populated areas, e.g. over the above-mentioned Ruhr area. In the case of large cities, such as Berlin, dust emissions are concentrated near the edges of the city, where grid boxes share both urban land use and bare soil. Our sensitivity analysis of the way MODIS LAI is averaged over CAMx grid cells showed that if the lowest LAI values (assuming these constitute the urban fraction of the grid cell which cannot emit WBD) are omitted from the averaging, the resulting average LAI over the grid cell is much higher, making the average WBD emissions smaller. Thus, this has to be treated as a caution to provide consistent input data for land use and other land-related parameters like LAI, preferably with a similar horizontal resolution.

Regarding the PM concentrations, the background noWBD case showed a reasonable model performance with typical correlations for PM2.5 and PM10 as achieved in other modelling studies for Europe (e.g. ). Lowest correlations are computed for the summer period, while winter ones are usually the highest. This can be explained by the more stable weather conditions in DJF, which are better resolved than summer weather, which is often marked with highly variable convective environment . The PM values are underestimated in summer and overestimated in winter, which is probably due to vertical transport that is too strong in summer and transport that is too low in winter but may be connected also to deficiencies in the monthly profiles used for annual emissions . An important goal of the model validation was to evaluate whether the inclusion of WBD emission improves model performance. This turned out to be true for summer biases, which were reduced by adding the dust load. However, the winter, which was already marked by a negative bias, is modelled with an even higher bias if dust is considered. Also, the correlations decreased significantly if wind-blown dust is included in our simulations. This can be explained by the strong peaks in the impact on PM values which are a result of strong emission peaks seen in the daily time series of FCRS and CCRS emissions. The modelled urban PM peaks are often much higher (often by a factor of 5 or even more) than measurements and thus can strongly reduce the correlation with the observed values. Also, the RMSE values increased, which can again be explained by the many outliers in the modelled PM data.

Similar to dust emissions, the WBD impact on overall PM concentrations is high near big urban centres (over Germany and the Benelux states; reaching 15–20 µgm-3), but large rural contributions are also modelled, exceeding 2 and 5 µgm-3 for PM2.5 and PM10, respectively. The contributions are largest when the largest emissions occur, and we showed an evident correlation of high concentrations of PM10 with high emissions of WBD in the coarse model. These seasonally averaged impacts are, however, strongly exceeded by the daily average values, which can be higher by 1 order of magnitude, reaching 100 µgm-3 for some cities. However, as already said, these extreme peaks are probably overestimated (due to too strong winds in WRF).

calculated the summer and autumn wind-blown dust contribution to PM due to European local sources and found a contribution of around 1–2 µgm-3 to PM10 over central Europe, which is about 2 times less than in our simulations. We can, however, expect that our result would get closer to their numbers without the positive wind bias encountered in our driving model.

Apart from the impact on the overall PM2.5/10 concentrations, our study quantified the long-term impact on the secondary aerosol components, namely the secondary inorganic aerosol (SIA) components. On a seasonal average, the impacts are rather small: an up to around 0.1 µgm-3 increase for PSO4 (mainly during winter) and PNO3 (all year round). For PNH4, we modelled decreases of a similar absolute magnitude. A much higher impact is, however, calculated for specific days as daily means. These can reach an increase of up to 0.5–1 µgm-3 for sulfates (maximum increase over Berlin in 2009 exceeding 1 µgm-3) and similar increases for nitrates. During winter 2008–2009, nitrates occasionally even decreased by up to 0.2–0.3 µgm-3. Ammonium decreased due to dust by up to -1 µgm-3 on selected days. These decreases occurred mainly on winter days.

The explanation of the above-presented SIA modifications can be explained by two types of processes: one is the heterogeneous oxidation from SO2 and N2O5 on the surface of dust particles , and the other is the catalytic oxidation enhancement by dust elements in cloud water through influencing cloud pH and thus the aqueous chemistry of SO2, HNO3 and NH3. Indeed, if we consider the impact on total SIA, we see that SIAs increased (the increases in sulfates and nitrates outweigh the decrease in ammonium). This is in line with the expectation and with previous studies dealing with the impact of dust on secondary aerosol formation .

Concretely, the increases in sulfates due to the presence of dust particles were modelled by (increases by about 1 µgm-3 during a strong dust event in China), who attributed them to dust surface heterogeneous chemistry. Also, modelled increases in sulfates over Europe due to dust. Our findings are also in line, at least qualitatively, with the recent findings of , who argued that on the “dust surface, heterogeneous drivers (e.g. transition metal constituents, water-soluble ions) are more efficient than surface-adsorbed oxidants (e.g. H2O2, NO2, O3) in the conversion of SO2, particularly during nighttime”.

Regarding the impact on nitrates, the increases are consistent with an earlier study of , who found that nitrates associate with dust and result in volatilization. The increase in nitrates can be explained also by the formation of deliquescent salts (e.g. through the reaction of crustal cations in dust with NO3- ions), as argued also by . This can potentially even lead to some overprediction of nitrates, which requires revisiting the chemical composition of dust .

Finally, the ammonium response to dust is tightly connected to the response of sulfates and nitrates. As we saw that nitrates easily associate with dust (via reaction with dust crustal anions like Ca2+), this means that less nitrate is available to react with ammonium, leading to more ammonia remaining in the gas phase . In other words, NH4+ is replaced by dust containing cations . This last author and others (e.g. ) note that ammonium could also increase due to the present of dust as a result of more sulfates forming on dust surfaces. However, in our simulations this is evidently offset by the above-mentioned replacement of ammonium by crustal cations.

In our experiments, the impact on SIA is clearly stronger using EQSAM, although the differences are not large and the overall impact on PM is not affected too much. The reason for stronger sulfate and nitrate formation in EQSAM is probably the fact that in EQSAM, the cloud pH is influenced by three cations (Mg++, Ca++ and K+), while in ISORROPIA, it is only influenced by calcium. This also explains the stronger decrease in ammonium in EQSAM being replaced by more cations.

An exception to the above-mentioned behaviour for nitrates is the winter 2009 decrease (by about 0.2 µgm-3) seen for all analysed cities. This period is not characterized by exceptional dust emissions or extreme PM values (based on our results). On the other hand, during this period, the dust impact on PSO4 is relatively large, while the impact on ammonium is very small. Thus ammonium was probably preferably neutralizing sulfates instead of nitrates, causing their reduction.

Summing up the results, we showed that the long-term impact of local wind-blown dust emissions in Europe can significantly enhance urban PM levels, especially during extreme events rather than in seasonal averages. However, our calculations probably overestimate dust emissions due to very strong winds in the driving model. We also showed that apart from the total aerosol load, dust also impacts the secondary inorganic fraction of PM, which can significantly increase on selected days.

We also have to note that the uncertainties related to different inputs used for the study cannot be judged well here. We already mentioned that the land use and the LAI input can coact (bare soil vs. low or high LAI) differently depending on the choice of these data. For example, this led to one of the highest dust emissions in our simulations being located around urban areas. We also used some default values for dust composition based on a study which measured this composition over a different geographic area. Lastly, we used only one driving model and one model for wind-blown dust emissions, so the model uncertainty also cannot be addressed. The future goal should thus be to focus on the sensitivities of wind-blown dust loads to different input data and methods to obtain a more robust long-term estimate of dust emissions and the impact on PM and its secondary components.