Dust Characterization and Its Potential Impact during the 2014–2015 Fogo Volcano Eruption (Cape Verde)

Characterization the Fogo Abstract: Fogo (Fogo Island) is the youngest and most active volcano of Cape Verde. The last eruption occurred in 2014–2015. Aiming to assess the dust sources that impacted the air quality during the present study period, fresh lava samples were collected, while Saharan dust intrusions and transport were modeled. Rooftop dust was also collected on the island dwellings and a mineralogical and chemical characterization was undertaken. Air quality monitors were used to obtain concentrations of atmospheric particulate matter (PM) and gaseous pollutants. The mineralogical constitution was assessed by XRD and Electron Microprobe. The pseudototal chemical concentration was performed by XRF, ICP-MS and SEM; the latter includes particles morphology. During the study, WRF-CHIMERE results showed the intrusion of desert dust which affected the air quality. Lava was classiﬁed as tephritic to basanitic, with high potassium content. The Pollution Load Index for rooftop dust was >1 in all samples, suggesting an enrichment. Higher values were found in dust size fraction <63 µ m, with contamination factor pointing to high enrichment of As, Ni and Pb, and very high enrichment of Cd. The non-carcinogenic hazard estimated for children suggested that health problems may arise. The carcinogenic risk was above the target risk, mostly due to As > Pb > Co. Ingestion was the main exposure route. PM 10 concentrations exceeded the 24-h mean of 50 µ g/m 3 recommended by WHO. Nevertheless, TVOCs displayed levels lower than guidelines. The highest levels of CO 2 were recorded in more populated villages and farthest from the volcano. 2 O 6 ], which are related to the local geogenic context and eruption outputs (S4). A cluster analysis showed two main groups, in the samples of this particle size range, with K, Na, P, Al, Cu, Si and Cd forming a group with variables more inﬂuenced by the local geogenic environment, and the remaining variables associated with both a combination of natural and external contributions. These results are also reﬂected in the signiﬁcant and positive Spearman correlations ( p < 0.01) between Al/K (r = 0.958), Al/Na (r = 0.976), Al/P (r = 0.976), K/Na (r = 0.994), K/P (r = 0.992), K/Si (r = 0.861), Na/P (r = 0.982), Na/Si (0.838) and P/Si (r = 0.838), and between Co/Cr (r = 0.964), Co/Fe (r = 0.904), Co/Mg (r = 0.940), Co/Ni (0.970), Co/Zn (r = 0.855), Cr/Fe (r= 0.857), Cr/Mg (0.952), Cr/Ni (0.994), Fe/Mg (r = 0.881), Fe/Ni (r = 0.886) and Mg/Ni (r = 0.970), suggesting a common source between the pairs of variables. These ﬁndings indicate that coarser fractions are more linked to the volcanic eruption and local geogenic environment, while the ﬁner fractions are related to the


Introduction
Extreme events, such as volcanic eruptions and transport of desert dust masses, represent an important hazard to the ecosystems worldwide. Deserts are considered the main source of mineral dust to the atmosphere and influencing the air quality, with the Sahara Desert contributing to nearly half of it [1]. Studies by [2] showed that Saharan dust impacted the soils in the Canary Islands, an archipelago belonging to Macaronesia, along with Cape Verde. Sahara Desert particles transported by wind, mostly composed of silt (4 to 63 µm) and clay (<4 µm) sizes, can be deposited to distant locations, such as Cape Verde or even the American continent [3]. These particles may be associated with salts, pathogens and anthropogenic pollutants. Within the framework of CV-DUST (Atmospheric aerosol in Cape Verde region: seasonal evaluation of composition, sources and transport) research project [4], in Praia city (Santiago Island, Cape Verde), it was observed that, in 2011, the daily concentrations exceeded the air quality guidelines proposed by the World Health Fogo is an active stratovolcano with a maximum altitude at Pico do Fogo (~2829 m above sea level (asl)). According to [24], Fogo Island is covered by young volcanic rocks of Quaternary age. A collapsed caldera is located in the center of the island [25], with~8 km in diameter. This collapse gave rise to a scar, locally called Bordeira, with a vertical wall with maximum~1000 m high and a flat area of~35 km 2 (~1700 m asl), locally called Chã das Caldeiras, where two villages were destroyed during the 2014-2015 eruption (Bangaeira and Portela villages, previously evacuated) (Figure 1c). The eruptive style of the strombolian type began with an explosive phase, forming an eruptive column with~6 km high. Three eruptive craters with strombolian emissions were formed [26].
Sampling and air quality monitoring: After the volcanic eruption started, travelling and sampling planning was very short in time. Apparatus's to be taken were selected according to the possibility of being easily and quickly transported to the study area. Between 10 and 21 December 2014, data acquisition was performed in 10 human settlements, including locations to where inhabitants directly affected by the eruption were relocated ( Figure S1): Mosteiros, Achada Furna, Cova Figueira, Monte Grande, Ponta Verde, S. Filipe, Corvo, Covatina, Campanas de Baixo and Cisterno (Figure 1c). Additionally, Monte Beco, a former eruptive vent near the 2014-2015 eruption, was also included in the study. Samples and data collected were: (a) 10 samples of outdoor dust, collected from the rooftops of private houses using new plastic brooms and shovels ( Figure S2); in Monte Beco samples were not collected, since there are no dwellings; in the laboratory, samples were dried (<40 • C), homogenized, sieved (<2 mm) and partially pulverized to <170 µm in a agate mill for chemical analysis; (b) 11 samples of the thoracic fraction of particulate matter (PM 10 ), collected by low volume sampling (2 L/min) using a Deluxe pump (model 224-PCMTX8) from SKC Inc. (Dorset, UK) connected to a size-selective head, also from SKC Inc., equipped with 37 mm quartz filters. The pump was powered by a battery and included a run time programmable unit, a constant pressure controller and a low flow adaptor; (c) continuous measurements of temperature (T), relative humidity (RH), carbon dioxide (CO 2 ), carbon monoxide (CO) and total volatile organic compounds (TVOCs) in 11 locations with a 5-min resolution were performed with an IQ-610 Air Quality Probe (Gray Wolf ® monitor); and (d) a DustTrak DRX 8533 aerosol real time monitor from TSI (also with 5-min resolution) was used in the same 11 locations for simultaneous monitoring of PM 1 , PM 2.5 , PM 4 , PM 10 and total suspended particulate matter (TSP). Additionally, two fresh lava samples were collected in the two directions of the lava flows, direction N and S of the vent (Figure 1c). Each media sample was georeferenced and preserved until laboratorial processing.
During the sampling campaigns, informal questionnaires were performed to the local population, asking about habits and health outcomes before and during the eruption. As expected, respiratory problems, nose and throat irritation with severe coughing and breath difficulty, eye irritation and skin allergies were reported. In addition, nervousness and extreme anxiety episodes were communicated, as a consequence of the eruption.
Granulometric analysis: Rooftop dust, <2 mm fraction, was subject to wet sieving to quantify and analyze separately different size fractions: <63, 63-125, 125-250, 250-500, 500-1000 and 1000-2000 µm. Particles associated salts derived from the eruption magmatic acidic gases, were dissolved by the sieving process, not being considered for this study. The granulometric distribution of the dust fraction <63 µm was determined at the University of Aveiro, Portugal, using an SediGraph III Plus X-ray grain size analyzer (Micromeritics ® Instrument Corp., Norcross, GA, USA). This technique is based on the sedimentation theory (Stokes' law) and the absorption of X-radiation (Beer-Lambert law).
Electron microprobe analysis: The mineralogical characteristics of lava samples were determined using backscattered electron imaging (BSE) in an electron microprobe JEOL 8500-F (National Laboratory of Energy and Geology-LNEG, S. Mamede de Infesta, Portugal). The operating conditions obtained for the backscattered electron imaging were 15 kV excitation voltage and 30 nA beam current.
Particle morphology and semi-quantitative chemical analysis: PM 10 (collected on quartz filters) and rooftop dust were analyzed at University of Aveiro (UA, Portugal) with a Hitachi S-4100 scanning electron microscope (SEM) coupled to a Bruker Quantax 400 energy dispersive spectrometer (EDS) to assess the morphological, granulometric and semi-quantitative chemical properties of the particles. The identification of inorganic insoluble particles was performed using a mix of protocols for each particle [27].
Precision and accuracy of analyses and procedures: These parameters were monitored using accredited ACME Analytical Laboratories, LNEG and UA internal standards, certified reference material and quality control blanks. Results were within the 95% confidence limits. The relative standard deviation was between 5% and 10%.
Modeling external mineral dust emissions and transport: The CHIMERE chemistrytransport model [28], forced by 3-D meteorological fields simulated by the WRF model [29], was used in this study to model the external dust mineral contribution. This model was previously used over this geographical region, where it showed its ability to correctly reproduce the atmospheric dust load and surface concentrations from the daily to the seasonal timescales [30]. The model applied considered the processes that drive the main phases of the dust cycle: saltation and sandblasting for emissions calculations, horizontal advection and vertical transport, including advection and mixing. Two nested domains were considered (CV d1 and CV d2 ; Figure 1a) covering the main dust source areas in North Africa, with horizontal resolutions of 0.5 • and 0.175 • , respectively. The GOCART global model climatology [31] was used to set the boundary conditions for the larger domain simulation. Eight vertical layers were used extending from the surface up to 500 hPa. The simulations were performed in time slices of 24 consecutive hours, each new period being initialized by the previous one, so that concentrations are continuous in time. Spin-up runs, relative to the 10 days prior to the period of interest, were used to initialize the model. Enrichment index (Ei): Ei was calculated by averaging the ratios of elemental concentrations to the lava elemental mean concentration and divided by the number of elements set to assess enrichments in dust samples. An Ei > 1.0 indicates that, on average, the elemental concentration is above the standard level, and that anthropogenic input may have contributed to elemental enrichment [32].
Pollution Factor (CF) and Contamination Degree (CD) [33]: CF was calculated for each element (As, Cd, Co, Cr, Cu, Mn, Ni, Pb, Zn) by dividing the mean individual concentration (C i ) by its corresponding baseline value (C b , estimated from lava samples) to assess external sources. The overall CD for each sample was calculated for each of the nine elements considered: (1) CF and CD classifications are: 0 ≤ CF < 1 Low, 1 ≤ CF < 3 Moderate, 3 ≤ CF < 6 High and 6 ≤ CF Very high Contamination Factor; and 0 ≤ CD < 1.5 None to very low, 1.5 ≤ CD < 2 Low, 2 ≤ CD < 4 Moderate, 4 ≤ CD < 8 High, 8 ≤ CD < 16 Very high, 16 ≤ CD < 32 Extremely high and 32 ≤ CD Ultra high Contamination Degree.
Pollution Load Index (PLI): PLI provides a simple comparative means to assess the level of enrichment. It was determined as the nth root of the product of the nEF: ELI = (CF 1 × CF 2 × CF 3 × · · · × CFn) 1/n . PLI > 1 suggests environmental deterioration by elemental pollution [34].
Potential Ecological Risk Index (PERI): PERI is defined as the sum of risk factors, defining a potential ecological risk of an element in each sample: PERI i is the Potential Ecological Risk Index for each sample (i), EF i the monomial potential ecological risk factor, CF i the contamination factor and TF the heavy metal toxicresponse factor for each element [33]. It was initially proposed for sediments but has been extensively applied to dust (e.g., [36]). Reference values of igneous rock types, soil, freshwater, land plants and land animals were as proposed by [37]. The PERI classification is as follows: PERI < 150 Low, 150 ≤ PERI < 300 Moderate, 300 ≤ PERI < 600 Considerable and 600 ≤ PERI Very High Ecological Risk.
Non-carcinogenic and carcinogenic risk assessment: Human health risks were calculated taking into consideration that residents, both children and adults, are directly exposed to dust through three main pathways (a) ingestion, (b) inhalation and (c) dermal absorption. PM 10 (particulate matter < 10 µm) are more relevant in inhalation processes, nevertheless, coarser sizes are decomposed in the gastrointestinal track. The carcinogenic and non-carcinogenic side effects for each element were computed individually, consid-ering reference toxicity levels for each variable, as extensively described in [38]. For each selected potentially toxic element (PTE; As, Cd, Cr, Co, Cu, Pb, Mn and Zn) and pathway, the non-cancer toxic hazard was estimated by computing the Hazard Quotient (HQ) for systemic toxicity (i.e., non-carcinogenic risk). If HQ > 1, non-carcinogenic effects might occur as the exposure concentration exceeds the reference dose (RfD). The cumulative noncarcinogenic hazard index (HI) corresponds to the sum of HQ for each pathway and/or variable. Similarly, HI > 1 indicates that non-carcinogenic effects might occur. Carcinogenic risk, or the probability of an individual to develop any type of cancer over a lifetime as a result of exposure to a potential carcinogen, was estimated by the sum of total cancer risk for the three exposure routes. A risk > 1.00E−06 is classified as the carcinogenic target risk, while values >1.00E−04 are considered unacceptable [38].

Outdoor Dust-Main Potential Contributions
To estimate Fogo Island main potential sources of outdoor air contamination, and since there were not enough samples to perform an advanced multivariate statistical analysis, the geochemical and mineralogical characterization of the two 2014-2015 lava samples were considered, and the external mineral dust emissions and their transport were modeled.
The XRF chemical analysis of the lava samples revealed 44.28 and 46.68 wt% SiO 2 content, a basaltic type with high melt temperature (1000-1200 • C). Results are similar to those of other studies in the same region (e.g., 43.03-47.99% SiO 2) ) [39] and also to the present study rooftop dust (<2 mm) composition (43.01 to 44.99 wt% SiO 2 ). According to the total alkali-silica (TAS) diagram [40], this volcanic rock is tephritic to basanitic, with high potassium content. When compared to the 1995 Fogo volcano eruption lavas [41], the current samples displayed relatively higher mean SiO 2 (present study vs. 41 Figure S3, Table S1), reflected in the chemical composition. These results are in line with other studies of the same eruption event (e.g., [39]).
The modeled mean PM 10 surface concentrations within the inner domain (CV d2 ), between mid-November 2014 and 31 January 2015, is presented in Figure 2 and Figure S4. It is possible to identify two PM 10 surface concentration hotspot, one in Mauritania, where the highest dust concentrations (>400 µg/m 3 ) were simulated. According to the model results, Cape Verde PM 10 mean concentrations due to desert dust intrusions were within the range 50-150 µg/m 3 .
Desert dust storms impacted the PM levels over the Cape Verdean archipelago, starting in December 2014, revealing a remarkable increase of the PM surface level concentrations. One of the main dust storm events occurred during this study period, between 12 and 20 December. The emission of dust in the Mauritania hotspots was intensified between 9 and 11 December. In these days, dust was transported westwards with maximum surface PM 10 concentrations (>350 µg/m 3 daily mean values) being reached around 12 December in Fogo Island ( Figure S5). The model suggests that the source regions that mostly affect the surface Cape Verdean air quality are located over Mauritania (Sahara Desert). Aerosol optical thickness (AOD) values ( Figure S5), registered during the same periods in the AERONET site of Sal Island in Cape Verde, highlight the extent of long-range transport of dust from the African continent. The same tendency that is shown both by the modeled PM 10 and the AOD observations, can be seen in the partial data for this period acquired in S. Filipe by an aerosol monitor, by the National Institute of Meteorology and Geophysics (Praiacity, Cape Verde), on the same period ( Figures S6 and S7). Despite a few days without information (local electrical power problems), readings reveal a similar tendency to modeled PM 10 and AERONET results, suggesting an impact in Fogo Island air quality.

Rooftop and Ambient PM 10 Dust Characterization
The XRD analysis (Table S1), indicates that quartz was ubiquitous in all rooftop dust samples, with higher representation in coarser fractions, revealing the influence of the volcanic ash and desert transported dust. Augite and diopside were also dominant minerals in all fractions, being characteristic of Fogo volcano emissions. Secondary minerals, such as calcite, may result from volcanic weathered particles or from desert dust, present especially in the <63 µm fraction, characteristic of Sahara particles.
The chemical analysis of the rooftop dust (<2 mm fraction) and fresh lava samples revealed similar mean elemental concentration profiles (Figure 3), showing that the eruption outputs directly influenced the outdoor air quality. Minimum and maximum pseudototal concentrations of rooftop dust samples, fractions 125-250, 63-125 and <63 µm, are presented in Figure 4. A 1-way ANOVA analysis showed significant differences between the As, Cd and Cu concentrations in the three fractions (p < 0.05), with the highest minimum elemental concentrations found in fraction <63 µm, ranked Si 106,000 > Al 78,000 > Na 24,000 > K 21,000 > P 4024 > Zn 141 > Cu 78 > Cr 72 > Ni 39 > Co 35 > Pb 19 > As 4.9 > Cd 2.2 mg/kg. A cluster analysis of this fraction samples revealed two groups, a set with K, Na, Al and P variables, related to the local geogenic influence, and a second cluster divided into variables with a mixed origin of local geogenic, external and anthropogenic contributions, with Cr, Ni, Co, Mg, Fe and Mn, and another sub-cluster with As, Cd, Pb, Cu, Zn, Ca, Si and Ti, also of mixed origin and with important influence of longrange transported dust. [42] proposed Ti as a marker for Saharan dust due to rutile [TiO 2 ] and anatase from weathered soils in northern African source areas and with no specific area trend. The results of the finest fraction cluster analysis are confirmed by the Spearman correlation, with statistically significant coefficients (p < 0.01) between pairs K/Na (r = 0.941), Co/Cr (r = 0.826), Co/Ni (r = 0.826), Cr/Ni (r = 0.978) and As/Cd (r = 0.787) and Cu/Zn (r = 0.768). The authors of [43,44] found that potentially toxic elements (PTEs) are more abundant in their stable form in the finest fractions, which is in line with the outcomes of the present study. Single dust samples collected in locations in the NE region were possibly more impacted by external dust intrusions, as they presented higher contents of Fe, Al, Si and Ca than those observed lava samples and in samples from other locations, in the thinner fractions. These elements are representative of quartz and other oxides (e.g., FeO 2 ), carbonates (e.g., CaCO 3 ) and aluminosilicates (e.g., feldspar, clay minerals), characteristic of the Sahara Desert [45]. The same authors, and references therein, also proposed that the morphology of the Saharan dust is almost elliptical, with different rounded shapes. The SEM-EDS of Fogo samples found diverse individual rounded dust particles and agglomerates ( Figure S8), with (Ca + Mg)/Fe ratio > 2, suggesting an origin of the particles' origin is in northwestern Africa [46], in agreement with the external dust inputs modeled in this study. In addition, particles with a morphology typical of volcanic ash, e.g., angular, glassy, conchoidal fractures, vesicular marks, were identified in the rooftop dust ( Figure S9) (e.g., [47]). No asbestiform particles were found, thus decreasing the health impacts of these volcanic dust [48]. In locations more impacted by the emissions, especially in Covatina, several fluorite (CaF 2 ) particles were identified in the rooftop dust sample closest to the eruption. The association between F − geochemistry and health is well documented (e.g., [49,50]). Fluoride metabolism varies with solubility, structure, reactivity and release of F − ions. In humans, the main F − absorption path is the gastrointestinal tract, by inhalation and ingestion (drinking and diet), with 80-90% of consumed F − absorbed within 30 min [51]. Nevertheless, the F − content varies with individual uptake and excretion, with 45-60% of the ingested F − excreted in urine, while the remaining is recirculated into the plasma or deposited into the bones [52]. The volcanic expelled products contribute to inhalation and ingestion of dust and gases, but also through consumption of food and water enriched in potentially toxic elements. Additional studies are being performed to understand the potential impact of F − in agricultural products from Fogo Island.
The SEM-EDS analysis of the PM 10 samples collected on quartz filters ( Figure S9) also revealed the presence of sea salt [NaCl], suggesting the Atlantic Ocean as another particle source, in addition to the volcanic emissions that produce soluble salt deposited on the surface of the ash [53]. In S. Filipe, the capital of the island, the location with more traffic related emissions, soot particles were also identified.
The cluster analysis of the fraction 63-125 µm revealed three main groups of variables, with a set of K, Na, Al, P and Si linked to the local geogenic context, a group with Cr, Ni, Co, Mg, Fe, Mn and Ti linked to a mix of sources and a third set with As, Cd, Pb, Ca, Cu and Zn, possibly associated with external and anthropogenic sources. These findings are confirmed by the significant Spearman correlations (p < 0.01) between K/Na (r = 0.939), Co/Cr (0.976), Co/Ni (r = 0.976), Cr/Ni (r = 0.988), Ca/Zn (r = 0.857) and Cu/Zn (0.838) and (p < 0.05) for the pairs Mg/Ti (0.762), Ca/Cu (r = 0.0826), Ca/Pb (r = 0.786) and Pb/Zn (0.810). [54] suggested that the fine fraction <150 µm is the one that most adhere to human skin, especially in children with hand-to-mouth behaviors. According to these authors, particles < 150 µm are more likely to be ingested and dissolved, accessing the gastric system, and posing a potential risk to humans, depending on each PTE bioaccessible fraction.
The 250-125 µm fraction revealed that the higher minimum values in elements Fe 125,000 > Si 107,000 > Ti 19,000 mg/kg, are possibly linked to the abundant presence of minerals such as augite [(Ca,Na)(Mg,Fe,Al,Ti) 2 (Si,Al) 2 O 6 ], which are related to the local geogenic context and eruption outputs (S4). A cluster analysis showed two main groups, in the samples of this particle size range, with K, Na, P, Al, Cu, Si and Cd forming a group with variables more influenced by the local geogenic environment, and the remaining variables associated with both a combination of natural and external contributions. These results are also reflected in the significant and positive Spearman correlations (p < 0.01) between Al/K (r = 0.958), Al/Na (r = 0.976), Al/P (r = 0.976), K/Na (r = 0.994), K/P (r = 0.992), K/Si (r = 0.861), Na/P (r = 0.982), Na/Si (0.838) and P/Si (r = 0.838), and between Co/Cr (r = 0.964), Co/Fe (r = 0.904), Co/Mg (r = 0.940), Co/Ni (0.970), Co/Zn (r = 0.855), Cr/Fe (r= 0.857), Cr/Mg (0.952), Cr/Ni (0.994), Fe/Mg (r = 0.881), Fe/Ni (r = 0.886) and Mg/Ni (r = 0.970), suggesting a common source between the pairs of variables. These findings indicate that coarser fractions are more linked to the volcanic eruption and local geogenic environment, while the finer fractions are related to the two main events that took place during this study period, the eruption and desert dust intrusion. Other studies have also established a link between As, Cd, Cr, Co, Fe, Mn, Ni and Zn concentrations and anthropogenic activities, including the concrete nature of the rooftops and traffic related materials, presenting similar concentrations when compared to this study [55][56][57].  The <63, 63-125, 125-250, 250-500, 500-1000 and 1000-2000 µm granulometric distribution of the rooftop dust samples suggested a predominance of particles with diameter <250 µm (~56%), followed by the size fraction 125-250 µm (~30%). In two locations, other fractions prevailed: (a) Corvo, with a dominant 250-500 µm fraction (~55%); and (b) Cisterno, with a prevailing 63-125 µm fraction (~55%). Among all samples, Mosteiros was the village with higher <63 µm content (11%). This sampling site is located in the NW part the island (Figure 1), directly impacted by external dust masses that can be transported over long trajectories, namely from the African continent. [58] suggested that particles <20 µm are mainly transported by long-term suspension, and that this can occur for long periods and distances.
The quantification of different granulometric inhalable fractions <10 µm (PM 10 ; Figure 3b) was performed by Sedigraph. PM 10 represented 33 to 93% of the 63 µm fraction. The highest percentage of the 10-5 µm fraction (~35%) was found in Monte Beco, while in Mosteiros, Corvo and Cisterno the finest fraction <0.1 µm was the most represen-tative, accounting for 36.1, 46,1 and 48.5%, respectively. The granulometric results suggest an external source, since particles <10 µm represent a significant mass fraction of dust transported over long distances, originating in the Sahara Desert [59].
A DustTrak DRX 8533 aerosol real time monitor from TSI (5-min resolution) was used for simultaneous monitoring of PM 1 , PM 2.5 , PM 10 and total suspended particulate matter (TSP). Regardless of the sampling location, concentrations of the thoracic fraction of particulate matter (PM 10 ) largely exceeded the 24-h daily mean stipulated by the European Air Quality Directive 2008/50/EC and recommended by the World Health Organization of 50 µg/m 3 ( Figure 5). Fine particles, known as the alveolar size fraction (PM 2.5 ), also far exceeded the daily guideline of 25 µg/m 3 recommended by the WHO. Highest concentrations were registered in Covatina, located in Chã das Caldeiras, close to the Bordeirã 1 km high wall, which acts as a barrier and retain dust particles. These were the lowest concentrations, but still above 100 µg/m 3 and 75 µg/m 3 for PM 10 and PM 2.5 , respectively, were measured outside the eruption area, on the north-east to the south-west flanks of the volcano (Ponta Verde, Monte Grande and Cova Figueira). In the two monitoring sites closest to the eruption (Monte Beco and Covatina), PM 10 accounted for 65 to 69% of TSP. This means that almost 70% of the inhaled particles can pass beyond the larynx and ciliated airways. The percentage of thoracic particles increased to 92 to 96% at other locations. Small villages and towns within the caldera were strongly impacted by coarse ashes, which tend to settle out first. Finer particles remain in the plume and stay airborne while being transported over long distance. Thus, the TSP/PM 10 ratio tends to be higher in locations farther from the eruptive source. The alveolar size fraction encompassed between 51 and 79% of the thoracic particulate mass. The lowest PM 2.5 /PM 10 mean ratio was registered in Covatina, near the Bordeira wall, whereas the highest value was monitored in Ponta Verde, a settlement in the northwestern part of the island outside the crater rim. PM 2.5 comprised higher proportions of ultrafine particles (PM 1 ), around 97-98%, in the vicinities of the volcano, while all locations outside the Bordeira scarp presented PM 1 /PM 2.5 ratios in the 0.92-0.94 range. In São Filipe, the island capital, a Cape Verdean aerosol monitoring station (15 min. resolution) was installed between 28 November 2014 and 15 February 2015, the eruption time-lapse. Although some problems with equipment and energy (days 30 November for PM 10 , 28 November to 6 December for PM 2.5 and 15 to 19 December for both PM 2.5 and PM 10 ), some data was collected. The mean PM 10 24-h concentration guideline [5] was largely exceeded ( Figure S6 Concentrations of PM 10 and PM 2.5 above the guidelines may represent a risk to the human health [60,61] and are consistent with the volcanic eruption time-lapse and the Sahara desert dust intrusions. PM 10 -bound constituents and morphology can cause breathing problems and irritation to the respiratory system [5]. In turn, PM 2.5 can reach alveoli and induce asthma and even heart attacks, being especially disturbing in children and people with pre-existing health conditions [62,63]. To humans, fresh volcanic ashes might represent a higher hazard than mineral dust, due to morphology and chemical coatings.

CO 2 , CO and TVOCs
TVOCs include non-methane hydrocarbons (NMHC) and oxygenated NMHC. Considerable discomfort and headaches are likely if levels fall in the range of 1.2-10 ppm [64]. Only Monte Beco reached levels close to the lower limit of the worrying range ( Figure 6). There are no enforceable limits for TVOC concentrations, i.e., there is no legal threshold. Much higher VOC concentrations, varying between 2.89 and 35.3 ppm, have been reported by [65] for 13 sites in Delhi, India. However, much lower levels, in the ranges 7-173 ppb and 10-285 ppb in the summer and fall, respectively, have been registered in Los Angeles, California [66]. VOC concentrations may be highly variable, depending on the emission sources, dispersion conditions and atmospheric stability. A follow-up air quality monitoring campaign by passive sampling techniques was carried out in the Fogo Island between November 2016 and January 2017 [67]. In this later campaign, the highest concentrations of various individual VOCs were registered in Monte Grande, probably due to additional anthropogenic activities resulting from population resettlement at this village after the eruption. It has been shown that VOCs from fumaroles mainly consist of C 2 -C 5 alkanes and relatively high concentrations of aromatics (benzene and toluene) and C 2 -C 3 alkenes (mainly propene and isobutene) [68]. The main mechanism for the production of alkanes is likely related to pyrolysis of organic matter bearing sediments that interact with the ascending magmatic fluids. Alkanes are then converted to alkene and aromatic compounds via catalytic reactions (dehydrogenation and dehydroaromatisation, respectively). However, a biogenic origin for the light hydrocarbons cannot be ruled out [68]. It is expected that sulfur and halogen-bearing gaseous compounds, not measured as TVOCs, are dominant in the volcanic plume. Moreover, it should be taken into account that the volcanic plume is ejected at high altitude. Emissions from lava and the fraction of the volcanic plume that reaches the ground suffer dilution by strong winds and changes of their direction. Thus, VOCs may originate, at great extent, from other local sources, other than the volcano. These include natural sources, such as vegetation, and anthropogenic sources, which include emissions from biomass burning, solvent usage and road transport.
The average CO 2 concentrations ( Figure 6) ranged from 445 ppm at Monte Beco (1908 m asl) to 570 ppm at Campanas de Baixo (453 m asl). The highest levels were not registered near the volcano, but rather in more populated villages. CO 2 concentrations around 415 ppm have recently been recorded at the remote observatory near the summit of Mauna Loa, at an altitude of 3400 m asl, which is well situated to measure air masses that are representative of very large areas. CO 2 mixing ratios have been found to be higher in many urban centers worldwide compared to adjacent rural locations, phenomenon known as an "urban CO 2 dome" [69]. These higher mixing ratios are due in part to local traffic emissions. For example, [70] measured CO 2 concentrations in air samples collected from various locations in Paris, its suburbs and the surrounding open countryside. These data sets revealed that near-surface atmospheric CO 2 concentrations throughout the country outside Paris averaged 415 ppm, while values in the city sometimes reached as high as 950 ppm.
The CO levels were always close to the detection limit (0.5 ppm). Due to ejection at high altitude and atmospheric dilution, concentrations of gaseous pollutants seem less affected by the eruption than particles. Ash and gases can be transported either aggregated or separated in the atmosphere. Vertical separation occurs due to the eruption style or by different sedimentation velocity of ash and gases, and horizontal separation due to wind shear [71].

Environmental Impact and Human Health Risk Assessment
For the calculation of different environmental indexes, mean concentrations of lava samples were used as background values, to understand the external sources of dust enrichment. Two sets of variables were studied: (a) major elements, Al, Ca, Fe, K, Mg, Si and Ti; and (b) trace elements potentially linked to anthropogenic sources, As, Cd, Co, Cr, Cu, Mn, Ni, Pb and Zn. Three size fractions were selected (<63, 63-125 and 125-250 µm). This selection was especially done for the health risk assessment, since particles with diameter <250 µm are more likely to adhere to hands and may be ingested by hand-to-mouth behaviors [72].
The group of major elements, linked to the Fogo volcano geochemical composition, suggest a small contribution from external sources. The Enrichment Index (Ei ; Table S2) was slightly over 1, with a maximum of 1.5 for the dust fraction sized 125-250 µm of Ponta Verde. These results are confirmed by the Contamination Degree (CD), Pollution Load Index (PLI) and Potential Ecological Risk Index (PERI), which enabled to classify all locations with a low-grade enrichment, although Ponta Verde presented higher results. The Geoccumulation Index (Igeo) and Contamination Factor (CF) also showed a low external contribution in all samples and fractions, with slightly higher Mg concentrations that in lava samples.
The Enrichment index for a set of elements with a potential link to anthropogenic sources (Table S3) suggest the contribution of an external source of these elements to the dust samples, irrespective of the size fractions. The minimum values were found to be higher in the size fraction <63 µm, especially in samples from the NE sector of the island. Nevertheless, the maximum Ei values (4.9 in 125-250 µm and 4.2 in 63-125 µm fractions) were observed in coarser fractions from Ponta Verde, a village not directly impacted by volcanic emissions. The Pollution Load Index presented values above 1 in all samples, except for the 125-250 µm size fraction of dust from Cisterno, suggesting an external contribution for the selected elements. Higher PLI values were found in the size fraction <63 µm, especially in Ponta Verde and Campanas de Baixo villages. The same results were obtained with the Potential Ecological Risk Index. All the samples sized <63 µm were classified of extreme risk. The 125-250 µm size fraction of dust from Monte Grande presented the lowest PERI value (144), while the highest PERIs were obtained in Ponta Verde, Mosteiros and Campanas de Baixo (551, 491 and 334, respectively). For the size fraction 63-125 µm, the highest PERI values were found in Mosteiros (502) and Ponta Verde (492), whose samples were categorized of severe risk, whereas other samples were classified with moderate risk.
The Geoaccumulation index (S12) indicated a small external contribution for all size fractions and elements. The highest Igeo values (2.4 Cr and 2.5 Cu in dust sized 125-250 µm, and 2.0 Cr and 2.4 Cu in the 63-125 µm size fraction) were obtained in dust from Ponta Verde, classified as moderately to heavily contaminated. In the same location, As presented a maximum Igeo value of 1.5, in the size fraction <63 µm, classified as moderately contaminated. The Contamination Factor (CF) displayed a high contamination of As, Ni and Pb, and very high contamination of Cd in the size fraction <63 µm of dust from Ponta Verde and Cisterno villages. Nevertheless, the Contamination Degree presented its maximum in the 125-250 µm size fraction of dust from Ponta Verde (CD = 4.6, high contamination). Other size fractions of this location were classified with moderate contamination. Coarser dust fractions from Cova Figueira, Campanas de Baixo and Achada Furna were classified as having a moderate CD.
Being children more sensitive and prone to hand-to-mouth behavioral manners, the non-carcinogenic Hazard Quotient through Ingestion (HQ ChildIng ; Table 1) and the Hazard Index (HI ChildTotal ) were estimated for this age group. Both HQ ChildIng and HI ChildTotal results were above the threshold of 1 (mean HQ = HI = 1.4 in all samples and in the three size fractions studied), suggesting that non-carcinogenic health outcomes might occur, since exposure to elements is higher than the reference dose. The highest HQ (1.6) and HI (1.7) were found in the fraction 63 The carcinogenic risk (Table 1), for both children and adults, of all size fractions was above the target value of 1.00E−6. However, regardless of the samples, the risk was always <1.00E−4. Arsenic was the element that most contributed to the risk, accounting for 86.2 to 95.8% of the total (Figure 7), followed by Pb (1.7 to 9.7%) and Co (1.1 to 10.6%). Ingestion was found to be the main exposure route for all dust size fractions, representing 78.5 to 87.3% of the total risk posed by the three exposure routes. The finest size fraction (<63 µm) was the one with the highest risk, mostly due to As, whose concentrations accounted for 91.4 to 95.5% of the total. The lowest total risk value was found in Monte Grande, located in the interior of the island, for the 125-250 µm size fraction, while the highest was observed in the <63 µm fraction of Ponta Verde. Nevertheless, it is known that the pseudototal concentration of PTEs can overestimate the hazard and/or risk to humans and a bioaccessibility study is needed.

Conclusions
Fogo island inhabitants, particularly those living in Chã das Caldeiras, are exposed to air deterioration, especially during local volcanic eruption outputs and Sahara dust intrusion events.
Two main sources of atmospheric contamination were assessed: (i) the long-range transport of mineral dust, showing a contribution from the Mauritania region (Sahara Desert), and (ii) the volcanic eruption. The granulometric distribution of outdoor dust samples was dominated by particles with diameter <250 µm, followed by the size fractions >125 and <250 µm. The PM 10 inhalable fraction represented the highest percentage of the dust samples sized <63 µm, which can potentiate breathing problems and irritation to the respiratory tract. PM 10 and PM 2.5 concentrations largely exceeded the World Health Organization guidelines, in line with the volcanic eruption time-lapse and the Sahara Desert dust outbreaks. The highest levels of CO 2 were recorded in more populated villages.
Cumulative environmental indices calculated based on the content of Al, Ca, Fe, K, Mg, Si and Ti, allowed the categorization of dust as polluted in all locations with a moderate enrichment in the size fraction <63 µm. Individual indices suggested a high contamination of As, Ni and Pb, and very high contamination of Cd. The Non-carcinogenic Hazard Quotient and the Hazard Index from the exposure of children to elements through ingestion were above 1, suggesting that non-carcinogenic health outcomes might occur, since concentrations were higher than the reference doses. The carcinogenic risk associated with all dust size fractions was above the target risk of 1.00E-06. Arsenic was the element that most contributed to the global risk, followed by Pb and Co. Ingestion was found to be the main exposure route of dust-bound metals in all size-segregated fractions. The size fraction <63 µm was the one representing a higher risk, mostly due to the concentration of As.
The results highlight the need for continuous monitoring of air quality in this area. It is known that the total concentration of potentially toxic elements can overestimate the risk for humans. To understand the bioaccessible fraction (BAF%) in lungs, gastric and gastrointestinal phases of the outdoor dust, the Artificial Lysosomal Fluid (ALF) and Unified Bioaccessibility Method (UBM) assays are being done. The island population is strongly dependent on agriculture, livestock and groundwater. The volcanic ex-pelled products not only contributed to human health by inhalation and ingestion of dust, but also through consumption of food and water enriched in potentially toxic elements. In addition, the edible plants most consumed in the island are being analyzed and BAF% being estimated.
Author Contributions: C.C.: sampling, conceptualization, funding, methodology, formal analysis, supervision, writing-original draft. P.F.Á.: methodology, formal analysis, writing-review and editing. C.A.: methodology, formal analysis, writing-original draft. C.G.: methodology, formal analysis, writing-original draft. C.S.: formal analysis, writing-review and editing. E.F.d.S.: writing-review and editing. F.R.: writing-review and editing. All authors have read and agreed to the published version of the manuscript.

Funding:
The sampling campaign was supported by the Portuguese Foundation for Science and Technology (FCT), with an emergency financial support provided to C4G (Collaboratory for Geosciences) for the 2014-2015 Fogo volcano eruption monitoring mission. Carla Candeias is grateful to FCT by grant SFRH/BPD/99636/2014. All authors are thankful for the support to GeoBioTec (UID/GEO/04035/2019 + UIDB/04035/2020) and CESAM (UIDB/50017/2020 + UIDP/50017/2020) to FCT/MCTES through national funds, and co-funding by FEDER, within the PT2020 Partnership Agreement and Compete 2020.