African Dust Particles over the Western Caribbean Part I: Impact on air quality over the Yucatan Peninsula

. On a global scale, African dust is known as one of the major sources of mineral dust particles as they can be efficiently transported to different parts of the planet. Several studies have suggested that the Yucatan Peninsula could be influenced by such particles, especially in July, associated with the strengthening of the Caribbean low level jet. Although these particles have the potential to impact the local air quality significantly, as shown elsewhere (especially particulate matter, PM), the arrival and the impact of African dust into Mexican 20 territory has not been quantitatively reported to date. Two short-term field campaigns were conducted to confirm the arrival of African dust onto the Yucatan Peninsula in July 2017 and July 2018 at the city of Merida atmospheric observatory (20.98N 89.64W). Aerosol particles were monitored at the ground level by different on-line and off-line sensors. Several PM 2.5 and PM 10 25 peaks were observed during both sampling periods, with a relative increase in the PM levels ranging between 200% and 500% with respect to the normal background. Given that these peaks were found to highly correlate with super micron particles and chemical elements typically found in mineral dust particles, such as Al, Fe, Si, and K, they are linked with African dust. This conclusion is supported by combining back trajectories with vertical profiles from radiosondes, reanalysis, and satellite images to show that the origin of the air masses 30 arriving at Merida was the Saharan Air Layer (SAL). The good agreement found between the measured PM 10 concentrations and the estimated dust mixing ratio content from MERRA-2 (Version 2 of the Modern-Era Retrospective analysis for Research and Applications) corroborates the conclusion that the degradation of the local (and likely regional) air quality in Merida is a result of the arrival of African dust. and remote reanalysis, back trajectory analysis, and complementary meteorological

(LIDAR) and satellite sensors (e.g., the moderate resolution imaging spectroradiometer (MODIS) and the visible infrared imaging radiometer suite (VIIRS)) provide the aerosol spatial distribution with altitude in terms of the aerosol optical depth (AOD), mass concentration, and particle size distribution (Zhang and Reid, 2006; 80 Jackson et al., 2013). Additionally, the cloud-aerosol lidar and infrared pathfinder satellite observations (CALIPSO) can quantify (in three dimensions) the trans-Atlantic transport of African dust (Liu et al., 2008;Adams et al., 2012;Chouza et al., 2016;Korte et al., 2018).
Another useful tool is the reanalysis from global climate models that assimilates, in a statistically optimal way, 85 satellite and ground observations. The reanalysis produces continuous, four-dimensional fields of different atmospheric variables of interest, contrasting with the observations that may be spatially and temporally sparse (Cohn, 1997;Kalnay, 2003;Rienecker et al., 2011;Schutgens et al., 2010). The use of reanalysis, considering its inherent uncertainties, has become an essential tool in the atmospheric research community (Gelaro et al., 2017). For example, the hybrid single-particle lagrangian integrated trajectory (HYSPLIT) model has been 90 successfully used to track the transport of African dust particles (e.g., Ashrafi et al., 2014;Prospero et al., 2005).
HYSPLIT uses meteorological data from different modeling sources, including the NCEP-NCAR National Centers for Environmental Prediction (NCEP)-National Center for Atmospheric Research (NCAR) reanalysis model (Stein et al., 2015).

95
The transport of African dust can also be evaluated with NASA's Global Modeling and Assimilation Office (GMAO) MERRA-2 reanalysis. MERRA-2 (Version 2 of the Modern-Era Retrospective analysis for Research and Applications) is the first multidecadal reanalysis that assimilates both meteorological and aerosol data from various ground-and space-based remote sensing sources (Gelaro et al., 2017;Randles et al., 2017). Despite some deficiencies, previous studies have demonstrated that the MERRA-2's aerosol assimilation system does 100 indeed show considerable skill in simulating numerous observable aerosol properties (e.g., Buchard et al., 2015Buchard et al., , 2016Buchard et al., , 2017Randles et al., 2017). MERRA-2 has been previously used to study the effects of aerosol particles in the earth system, in several studies focused on dust-related phenomena. For example, Buchard et al. (2017) showed the benefit of the MERRA-2 assimilation for the retrieval of the seasonality, vertical distribution, and magnitude of the dust surface concentrations during an episode of dust transport from Africa to the Caribbean.

105
Later on, Veselovskii et al. (2018) showed the consistency of the MERRA-2 aerosol products with MIE-Raman lidar observations performed in West Africa during a smoke and dust mixing event. Similarly, Grogan and Thorncrof (2019)  The in situ monitoring of aerosol properties, such as aerosol size and mass distribution, is very useful to determine their influence on local air quality and human health (Querol et al., 2019). Hence, different studies have been carried out in the Caribbean islands and Florida to quantify the impact of African dust on the local 115 https://doi.org/10.5194/acp-2020-378 Preprint. Discussion started: 16 June 2020 c Author(s) 2020. CC BY 4.0 License. air quality (Prospero, 1999;Prospero and Mayol-Bracero, 2013;Prospero et al., 2014). In Barbados, the monitoring of the atmospheric aerosol mass began in 1965, while in Miami, Florida, it began in 1974 and continues to the present (Prospero and Mayol-Bracero, 2013). In Barbados, it is estimated that 50% of the PM2.5 (i.e., with an aerodynamic diameter d<2.5 μm) and ca. 90% of the PM10 (i.e., d<10 μm) consist of African dust (Li-Jones and Prospero, 1998;Prospero et al., 2001;Reid et al., 2003a). In Miami, the mean daily mass 120 concentration of mineral dust during the summer typically ranges between 10 and 100 μg m -3 , with a large interannual variability (Prospero et al., 2001). During the Puerto Rico Dust Experiment (PRIDE) campaign carried out between 28 June and 24 July, 2000, the mineral dust concentration at the ground level was found to exceed 70 μg m -3 (Reid et al., 2003b). In the aforementioned studies, the African dust particles transported over the Atlantic affected the local air quality, exceeding the World Health Organization (WHO) guidelines for PM2.5 125 and PM10 (i.e., above 25 μg m -3 for PM2.5 and 50 μg m -3 for PM10). According to the WHO, air pollution and its effects are considered a global health priority (WHO, 2002). Several studies have linked high concentrations of mineral dust (in terms of PM2.5 and PM10) to brain, cardiovascular, and respiratory diseases (Wilker et al., 2015;Brook et al., 2010;Dominici et al., 2006). According to Goudie (2014) and Zhang et al. (2016), inhaled dust particles can cause damage to the lungs, in addition to other parts of the body such as the heart, skin, and brain.

130
For example, in Rome, Italy, the effects of PM10 on cerebrovascular diseases were found to increase by 5% due to the presence of African dust particles (Alessandrini et al., 2013). Similarly, Rodríguez-Cotto et al. (2013) found that PM2.5 and PM10 can cause asthma and allergic reactions mainly to children during the epochs of African dust in Puerto Rico. Additionally, African dust particles have been found to serve as carriers for biological material. Griffin et al. (2001)   African dust in the receptor regions (Nenes et al., 2014). The most abundant minerals present in these particles are silicates (quartz), clay minerals (kaolinite, illite, chlorite, palygorskite), feldspars (albite, anorthite) and carbonates (calcite) (Goudie and Middleton, 2006;Querol et al., 2019;Broadley et al., 2012). The major oxides in Saharan dust are SiO2, Al2O3, Fe2O3, CaO, MgO and K2O and to a lesser extent P2O5 and TiO2 (Goudie and Middleton, 2006;Linke et al., 2006). Several studies in the Caribbean have identified high levels of Fe and Al 145 in dust events (Prospero et al., 2001;Rosinski et al., 1988). Additionally, Rosinski et al. (1988) reported high percentages of Si and Mg in particles collected in the Gulf of Mexico (GoM) during July.
Although the arrival of African dust in Mexico has been suggested for decades (e.g., Bravo et al., 1982;Prospero, 1999;Lenes et al., 2012), to our knowledge, there has not been a comprehensive study,

Sampling site and field campaigns
The Yucatan Peninsula is located in the southeast of Mexico. It borders with the GoM to the north, the Atlantic Ocean to the east, and the Caribbean Sea, Guatemala, and Belize to the south. The Yucatan has characteristics that are unique to this region (Plasencia, 1998). For example, its uniform terrain, the absence of rivers, and the 160 type of soil, formed by Cretaceous sediments that do not present mineralization and are rich in calcium, commonly called "Laja de Yucatán" (Plasencia, 1998) sets the Yucatan aside from other regions of Mexico. The average temperature of the Yucatan Peninsula ranges from 25°C to 35°C (World Resource Institute, 2018) with an average annual relative humidity of 79% (INEGI, 2009). The Peninsula has a warm, semi-dry climate on the coast and a warm, sub-humid climate throughout the rest of the region, with a rainy 165 season between summer and autumn (June-October) (Orellana et al., 2009;ProAire, 2018). Precipitation in this region is mainly due to convective activity and it is influenced by the moisture advection by the trade winds (Orellana et al., 2009;ProAire, 2018).

Aerosol concentration and particle size distribution
The particulate mass concentration was monitored continuously with PM2.5 and PM10 analyzers providing realtime measurements (FH 62 C14 Thermo Scientific Inc) with a temporal resolution of one minute at a sampling 185 flow rate of 16.7 L min -1 (Thermo Fisher Scientific Inc, 2007).
The total number concentration of particles with sizes approximately larger than 50 nm was measured by a condensation particle counter (CPC 3010, TSI) at a sampling rate of 1 hz with a flow rate of 1.0 L min −1 and https://doi.org/10.5194/acp-2020-378 Preprint. Discussion started: 16 June 2020 c Author(s) 2020. CC BY 4.0 License. the aerosol number concentration as a function of particle size was monitored by an optical particle counter 190 (LasAir II 310A, MSP). The LasAir has six different size bins (0.3, 0.5, 1.0, 5.0, 10.0, and 25 μm), a flow rate of 28.3 L min -1 and a time resolution of 11 s.

Aerosol collection and chemical composition analysis
PM2.5 and PM10 aerosol particles were collected for 24 h with a Partisol model 2525 (Thermo Fisher Scientific

195
Inc.) and for 48 h with a Minivol (3380, Air metrics) on 47 mm teflon filters (Pall Science). The MiniVol and Partisol flow rates were 5.0 L min -1 and 16.7 L min -1 , respectively. After the sampling periods, the filters were placed in 60 mm Petri dishes and stored at 4°C prior to the chemical analysis.
Elemental analysis was performed on each filter using X-Ray Fluorescence (XRF) with the X-ray spectrometer 200 at Laboratorio de Aerosoles, Instituto de Fisica, UNAM (Espinosa et al., 2012). The X-ray tube was made by Oxford Instruments (Scotts Valley, CA, USA), an Rh anode and an Amptek X-123SDD spectrometer (Bedford, MA, USA) were used. The samples were irradiated for 900 s working with a current of 500 μA and resulting in a spectrum that was analyzed using the WinQXAS computer code (IAEA, 1997). The product of this analysis derived mass concentrations of Fe, Al, Si, Ca, Na, P, Mg, Mn, Ti, Cl, P, Zn, K, S, Cu, and Ni along with their 205 associated uncertainties, as described by Espinosa et al. (2010).

Meteorological and satellite data
The local and regional meteorological conditions were monitored using different approaches. The RUOA meteorological sensors were placed at the rooftop of the FC-UADY (Table 1)  Hourly total precipitable water vapor and three-dimensional 3-hourly aerosol mixing ratio data were obtained from the MERRA-2 reanalysis (GMAO, 2015a(GMAO, , 2015b. The aerosol properties in MERRA-2 were simulated 220 with the Goddard Chemistry Aerosol Radiation and Transport model (GOCART), which takes into account the sources, sinks, and chemistry of 15 externally-mixed aerosol mass mixing ratio tracers: dust (five noninteracting size bins), sea salt (five non-interacting size bins), hydrophobic and hydrophilic black and organic carbon (BC and OC, respectively; four tracers), and sulfate (SO4) (Randles et al., 2017;Buchard et al., 2017).

225
The air mass back trajectories were calculated using the HYSPLIT model from the National Oceanic and Atmospheric Administration (NOAA). In conjunction with the in situ measurements, the back trajectories were https://doi.org/10.5194/acp-2020-378 Preprint. Discussion started: 16 June 2020 c Author(s) 2020. CC BY 4.0 License. calculated considering the maximum concentration of PM reported by the PM2.5 and PM10 analyzers. The trajectories were initiated at 50, 250 and 500 m above ground level going backward in time for 13 days.
Although Kramer et al. (2020) reported that mineral dust particles arrive in Miami ca. 10 days after they are

Local evidence
Several studies have shown that air quality (PM2.5 and PM10) significantly deteriorates upon the arrival of African dust plumes (Prospero and Lamb, 2003;Prospero et al., 2001;Prospero and Mayol-Bracero, 2013;Prospero et al., 2014). Figure 2 shows the time series of the PM2.5 and PM10 concentrations for the July-August 240 period of 2017 and 2018. Some high concentration PM peaks are clearly identified, with PM2.5 and PM10 values as high as 54 μg m -3 and 135 μg m -3 , respectively. Henceforth those peaks will be referred to in our study as African dust peaks (ADPs). Note that the background concentration of PM2.5 is ~ 4 μg m -3 and of PM10 is ~ 10 μg m -3 . The ADPs found in 2017 (i.e., July 22-24, 27-28, and August 4, 6-7) resulted in an increase of 300% in PM2.5 and 500% in PM10 with respect to the background. In 2018, the ADPs (i.e., July 10-11, 13-15, 16-17, 23 245 -26, and August 9-10) exceeded 200% and 300% of the background levels of PM2.5 and PM10, respectively. The aforementioned ADPs not only exceeded the PM2.5 and PM10 thresholds suggested by the WHO (i.e., PM2.5=25 μg m -3 and PM10=50 μg m -3 , 24-h mean) but more than double them, as was the case for the August 9-12, 2018 event. Similar behavior has been previously observed in Puerto Rico, Miami, and Barbados during the arrival of African dust particles (Reid et al., 2003b;Prospero et al., 2005Prospero et al., , 2014. The mass concentrations of PM2.5 and 250 PM10 were found to be 49% and 54% higher in 2018 than in 2017, respectively, suggesting a higher frequency or intensity of African dust plumes arriving over Merida in 2018. High levels of sodium (Na, pink), chlorine (Cl, turquoise blue), sulfur (S, dark orange), and calcium (Ca, light 260 green) were found in the background samples, corresponding to >70% of the total mass. The presence of Na and Cl are expected in airborne particles at this site given the city's proximity to the GoM (i.e., 23 km away). Cerón et al. (2002) reported large concentrations of Na, Cl, and Mg that originated from sea salt, when analyzing the composition of rainwater from the Yucatan Peninsula. The high levels of S can be associated with local https://doi.org/10.5194/acp-2020-378 Preprint. Discussion started: 16 June 2020 c Author(s) 2020. CC BY 4.0 License. anthropogenic activities such as vehicular, ship, and industrial emissions (e.g., Corbett and Fischbeck, 1997; 265 Cerón-Bretón et al., 2018). Additionally, given the short distance between Merida and the GoM, it is possible that dimethylsulfide (DMS) production from plankton in the GoM can be a natural source of S, as has been shown in other studies (e.g., Rosinski et al., 1988;Kloster et al., 2006;Vallina and Simó, 2007). Finally, the presence of Ca could be related to the limestone soil prevalent in the Yucatan Peninsula and the resuspension of road dust (Plasencia, 1998;Querol et al., 2019). concentrations. Out of the 16 elements Al, Si, K, and Fe were the only ones with correlation coefficients r>0.6 (p < 0.05) for both years, as shown in Figure S1. The present results are in agreement with previous studies that showed high correlation coefficients between the aforementioned elements (e.g., Caquineau et al., 1998;Guieu et al., 2002;Trapp et al., 2010). Also, the concentration of aerosol particles with diameters between 0.5 μm and 280 25 μm, as measured by the LasAir, were found to highly correlate with the PM2.5 and PM10 concentrations, r=0.79 and r=0.87, respectively ( Figure S2). Finally, the typical background particle size distribution showed significant changes during the arrival of ADPs for particles ranging between 0.5 μm and 5.0 μm ( Figure S3). It is widely known that the typical size of African dust particles transported over long distances ranges from 0.1 μm to 20 μm (e.g., Bégue et al., 2012;Denjean et al. 2016). Overall, the high concentration of coarse particles 285 and the increase of Al, Si, and Fe during the ADPs, together with the good correlations found between the PM2.5 and PM10 concentrations with Al, Si, K, Fe, and particles larger than 0.5 μm strongly suggests that the ADPs are mineral particles associated with dust transported from Africa to Mexico.
Additionally, it is important to note that Al, Si, K, and Fe are common oxides found in African dust composed KAlSi3O8, among others (A. Goudie and Middleton, 2006;Linke et al., 2006;Broadley et al., 2012;Querol et al., 2019). Rosinski et al. (1988) reports that up to 90% of the collected airborne particles in the presence of 295 dust events in the GoM contained Al, Fe, Si. In Puerto Rico, Reid et al., (2003b) found that during dust events that reached the island, the concentrations of Si and Al on aerosol particles (> 0.74 μm) were above 10 μg m -3 and 5 μg m -3 , respectively. Similarly, Prospero et al., (2001)   Note that those gases and particles can be considered as proxies of anthropogenic pollutants generated by the 305 incomplete combustion of fossil fuels and biomass burning, as previously demonstrated for Merida (Muñoz-Salazar et al., 2020;Alvarez-Ospina et al., 2020). Also, correlation coefficients below 0.29 were found between O3 and solar radiation with the PM2.5 and PM10 concentrations. Muñoz-Salazar et al., (2020) found in Merida that ultrafine aerosol particles of secondary origin are correlated with O3, a proxy of photochemical activity and hence, of secondary particle production. Therefore, it is very unlikely that secondary organic particles could be 310 the source of the ADPs observed in Merida.
Finally, although none of the different meteorological variables monitored at the surface level were found to correlate with the PM2.5 and PM10 concentrations, as shown by the wind roses in Figure S5, easterly winds were prevalent when ADPs were observed. This is relevant since African dust can only be transported by easterly 315 winds.

Larger scale observations
To evaluate the source of the ADPs observed in Merida from a large-scale perspective, we focus on the classification of tropical air masses in the North Atlantic and the Caribbean region during the boreal summer 320 months proposed by Dunion (2011). We used HYSPLIT to estimate the trajectories of different air masses that reached Merida in the periods of July-August 2017 and 2018. HYSPLIT trajectories for the 2017 and 2018 ADPs point to an African origin and, therefore, suggest that these air masses are either MT or SAL (See Figure   S6). To differentiate the MT from SAL, we focused on their distinctly unique moisture characteristics. Dunion (2011) proposes that a threshold of 45 mm of total precipitable water vapor (PWV), which corresponds to the 325 total amount of water vapor contained in the atmospheric column from the surface to the top of the troposphere (AMS, 2000), can be used to differentiate dry from moist air masses. This value is consistent with other studies that use PWV to identify dry-air days (e.g., Hankes and Marinaro, 2016), and since deep tropical convection begins to increase above a critical PWV value of 50 mm (Holloway and Neelin, 2009).
330 Figure 3 shows the time series of PWV for the July-August 2017 and 2018 periods at each WMO radiosonde site. The black solid line shows PWV from MERRA-2 (GMAO, 2015a), together with PWV estimated from the available radiosonde profiles shown as the dashed blue lines. One caveat is that in the periods of interest, there is a striking lack of radiosonde data. Nevertheless, we can see a good agreement between the available observed PWV and that of PWV from MERRA-2. Therefore, the latter can be used as a good approximation 335 for PWV in the region to differentiate moist from dry air masses. In Figure 3, the periods where PWV is less than 45 mm are highlighted in red. These periods show dry air masses that coincide with air mass trajectories with an African origin (i.e., in 2017: July 22-24, July 27-28, August 04, and August 6-7; and in 2018: July 10-https://doi.org/10.5194/acp-2020-378 Preprint. Discussion started: 16 June 2020 c Author(s) 2020. CC BY 4.0 License. 12, July 13-15, July 16-17, July 23-26 and August 9-12), allowing us to conclude that these dry air masses have mainly SAL characteristics.

340
The arrival of African dust in Merida was also explored from the MERRA-2 dataset. Figure 4 shows Similarly, Kalashnikova and Kahn (2008) demonstrated that with the MODIS it is possible to observe the 360 evolution of African dust plumes over the Atlantic Ocean. Additionally, Kaufman et al., (2005) identified and quantified the transport and deposition of mineral dust over the Atlantic Ocean using MODIS data.

Comparison of in situ observations and reanalysis
The daily mean PM10 from MERRA-2 was estimated using the method proposed by Provençal et al. (2017).

365
The black line in Fig. 5 shows the estimated PM10, which is compared to the PM10 measured by the RUOA station in Merida (blue line). The estimated surface dust mixing ratio from MERRA-2 is also shown in red. Figure 5 shows that MERRA-2 overestimates compared to the ground-based measurements of PM10.
Nevertheless, it should be clarified that the reanalysis information corresponds to a region 0.5° x 0.625°, implying that the MERRA-2 is estimating the regional average, while the station corresponds to a local 370 measurement.
Despite these differences, Figure 5 shows that the observations at the RUOA station have variations similar to those of MERRA-2. Figure 6 shows the dispersion diagram of the daily mean surface dust mixing ratio from MERRA-2 vs. PM10 measured from RUOA station for the periods indicated in Figure 5. It shows a high 375 correlation between the estimated dust and the measured PM10 in particular for the 2018 period, which was https://doi.org/10.5194/acp-2020-378 Preprint. Discussion started: 16 June 2020 c Author(s) 2020. CC BY 4.0 License.
particularly active with constant arrivals of African dust to the region, as shown in Figure 4. A similar analysis was performed for the 3-h estimated and measured PM10, as shown in Figures S10 and S11, with identical conclusions as for 24-h averages.

Conclusions
For the first time, the arrival of African dust into Mexican territory is quantitatively verified. The arrival of African dust particles in Merida significantly degraded the local air quality as PM2.5 and PM10 concentrations increased up to 500% with respect to the background. Therefore, the presence of African dust in Merida and the Yucatan Peninsula has the potential to trigger or exacerbate several diseases, as has been reported elsewhere 385 (Wilker et al., 2015;Brook et al., 2010;Dominici et al., 2006). The arrival of African dust in other regions of the world has led to even higher PM concentrations than those found in the present study. Moreover, African dust particles can also be a serious health threat as they serve as the carrier of biological 390 material originating in Africa. If the foreign biological particles are opportunistic pathogens, they can cause a variety of diseases in the receptor regions, such as the Yucatan Peninsula. Finally, those particles can impact the development of precipitation affecting the regional hydrological cycle when they serve as efficient ice nucleating particles (Hoose and Möhler, 2012;Murray et al., 2012;Kanji et al., 2017).

395
As shown in the present study, combining ground-based off-line and on-line sensors provides robust evidence of the arrival of African dust; however, we also show that the combination of back trajectories with radiosondes, and the estimated surface dust mixing ratio from MERRA-2 are powerful tools that can be exploited when in situ information is missing, especially in developing countries where the necessary instrumentation is scarce.