Impact of Desert Dust on Air Quality: What is the Meaningfulness of Daily PM Standards in Regions Close to the Sources? The Example of Southern Tunisia

Desert dust is now recognized as a major health hazard. However, there still exists a lack of measurements of desert dust atmospheric loads in regions located in the vicinity of the major desert areas, where a growing part of the world population is living. Dust emission is a sporadic and intense phenomenon so that the classical air quality standards should not be appropriate to reflect the real population exposure to desert dust. In order to give some insight to answer this question, PM10 concentrations were continuously measured at a five-minute time step in southern Tunisia from February 2014 to February 2019. The daily and annual PM10 concentrations were first discussed according to the Tunisian air quality standards: In this relatively remote area, close to dust source, these standards were respected at the annual, but no longer at the daily scale. Measurements performed at a high temporal resolution (five minutes) allowed to discriminate the different situations that led to exceed Tunisian daily standards in southern Tunisia. In particular, intense (five-minute PM10 concentrations up to more than 1500 μg m−3) and short-lived (a few hours) dust outbreaks were regularly observed. This result raises the question of the necessity of defining specific air quality standards at the sub-daily scale in countries affected by intense desert dust outbreaks.


Introduction
During recent decades, more and more attention has been paid to the atmospheric level of particulate matter (PM), especially for their potential impact on human health. Indeed, the fine particles can be inhaled and, according to their diameter, can enter the respiratory system up to the alveoli (e.g., [1][2][3]) so that they are now considered as a major health hazard (e.g., [4][5][6][7]).
Among them, mineral dust, also commonly referred to as desert dust because of their geographical origins, is of crucial interest as it represents about 30% to 35% of the total mass of aerosols emitted into the atmosphere every year [8]. Most of the dust emissions come from the Sahara Desert, followed by Chinese and Central Asian deserts, the Arabian Peninsula, and the Australian deserts [9]. The population living in the vicinity of these regions can be exposed to very high levels of mineral dust concentrations. Considering their physical, chemical, and biological properties, the impacts of desert In this framework, a national network of air quality monitoring stations (Réseau National de Surveillance de la Qualité de l'Air (RNSQA); www.anpe.nat.tn) has been deployed. In 2011, 30 stations were already installed in the main urban and industrial sites of northern Tunisia, and seven allowed the measurement of PM 10 concentrations by means of automatic beta radiation attenuation monitors [35,36]. However, none of these stations are located in the south of the country; therefore, population exposure to high desert dust atmospheric loads is not characterized. To do so, as a first step, PM 10 concentration has to be monitored in a remote place, closer to dust sources and away from other sources of pollutants, especially from anthropogenic activities. This is why, in February 2014, a ground-based station dedicated to the monitoring of PM 10 concentration was installed in the campus of the Institut des Régions Arides (IRA) of Médenine. At the monitoring station of El Fjé, measurements are performed with a short time step (five minutes) in order to assess, as well as possible, the high concentration periods.
In this paper, PM 10 concentrations measured during the period February 2014-February 2019 at the station of El Fjé in southern Tunisia are discussed first according to the Tunisian air quality standards, but also according to the way these standards were exceeded. This allows one to conclude on the relevance of actual air quality standards used in regions downwind of active desert dust sources.

Location of the Station
The dust station is installed inside the campus of the IRA Médenine in El Fjé (33.49963N; 10.64255E), as identified by the red dot in the satellite image in Figure 1. This place is located away from any other pollutant sources, especially from anthropogenic origin (biomass burning, road traffic, etc.) and is representative of the atmospheric background of the region. Indeed, the surroundings of the station are an agricultural area, mainly dominated by olive tree plantations, the most important town, Médenine (about 70,000 inhabitants [24]), being at more than 20 km south west. The Mediterranean Sea is about 5 km to the east, and 20 km to the north, while the Matmata Mounts are about 50 km to the west. From a climatological point of view, the monitoring station is located in the north east of the Jeffara plain, in the arid part of Tunisia where mean annual rainfalls ranging between 135 and 215 mm yr −1 were recorded during the 1949-2001 period in 8 meteorological stations [38].

Instruments and Method
The monitoring station of El Fjé was set up following the model of the stations of the Sahelian Dust Transect deployed in the Sahel since 2006 ( [39]; http://www.lisa.u-pec.fr/SDT/index.php?p=1), now part of the International Network to Study Deposition and Atmospheric chemistry in AFrica (INDAAF; https://indaaf.obs-mip.fr/?rubrique38). The same kind of instruments were chosen for their capability to resist severe dusty and meteorological conditions. All these instruments were installed at 2 m height from the flat roof of the higher building of the campus (i.e., about 14.5 m above the ground level).

PM10 Concentration
PM10 concentrations were measured using a Tapered Element Oscillating Microbalance (TEOM™ 1405 from Thermo Scientific TM , Franklin (MA), USA) equipped with a PM10 inlet (Thermo Scientific TM ). This instrument, widely used in air quality networks, allows one to measure atmospheric particulate concentrations ranging from a few micrograms up to one gram per cubic meter.
The microbalance was installed in an air-conditioned room to maintain ideal conditions for the TEOM™ sensor and control units. It was positioned about 5 m vertically below the PM10 inlet. The flow rate is 3 L min −1 and the averaging time is 5 min in these operational conditions, the nominal precision of the instrument is ± 2.0 µg m −3 for 1 h averaging, and ± 1.0 µg m −3 for 24 h averaging.
The lower part of the sampling tube (60 cm length) and the measurement chamber of the TEOM™ were heated at a temperature of 50 °C in order to allow a thermal stability of the particles and the filter and to minimize water condensation interferences. This heating may induce a loss of a part of the most volatile particulate compounds, such as ammonium nitrates or condensate organic compounds [40,41]. Thus, the contribution of carbonaceous and/or anthropogenic aerosols to the

Instruments and Method
The monitoring station of El Fjé was set up following the model of the stations of the Sahelian Dust Transect deployed in the Sahel since 2006 ( [39]; http://www.lisa.u-pec.fr/SDT/index.php?p=1), now part of the International Network to Study Deposition and Atmospheric chemistry in AFrica (INDAAF; https://indaaf.obs-mip.fr/?rubrique38). The same kind of instruments were chosen for their capability to resist severe dusty and meteorological conditions. All these instruments were installed at 2 m height from the flat roof of the higher building of the campus (i.e., about 14.5 m above the ground level).

PM 10 Concentration
PM 10 concentrations were measured using a Tapered Element Oscillating Microbalance (TEOM™ 1405 from Thermo Scientific TM , Franklin (MA), USA) equipped with a PM 10 inlet (Thermo Scientific TM ). This instrument, widely used in air quality networks, allows one to measure atmospheric particulate concentrations ranging from a few micrograms up to one gram per cubic meter.
The microbalance was installed in an air-conditioned room to maintain ideal conditions for the TEOM™ sensor and control units. It was positioned about 5 m vertically below the PM 10 inlet. The flow rate is 3 L min −1 and the averaging time is 5 min in these operational conditions, the nominal precision of the instrument is ± 2.0 µg m −3 for 1 h averaging, and ± 1.0 µg m −3 for 24 h averaging.
The lower part of the sampling tube (60 cm length) and the measurement chamber of the TEOM™ were heated at a temperature of 50 • C in order to allow a thermal stability of the particles and the filter and to minimize water condensation interferences. This heating may induce a loss of a part of the most volatile particulate compounds, such as ammonium nitrates or condensate organic compounds [40,41]. Thus, the contribution of carbonaceous and/or anthropogenic aerosols to the PM 10 concentration is probably slightly underestimated when using this type of measurement. Moreover, the heating of the sampled air can lead to negative values of PM 10 concentrations when relative humidity is high, and PM 10 concentrations especially low. This can be observed during intrusions of air masses coming from the sea and lightly loaded in PM.

Sunphotometric Measurements
A Cimel sunphotometer from the AERONET (AErosol RObotic NETwork)/PHOTONS (Photométrie pour le Traitement Opérationnel de Normalisation Satellitaire) network ( [42]; http: //aeronet.gsfc.nasa.gov/) was installed in July 2014 at the monitoring station. This instrument allows one to retrieve the Aerosol Optical Depth (AOD) and the corresponding Angström exponent (α) from the measurements of the direct solar radiation at 4 wavelengths (1020, 870, 675, and 440 nm) during daytime [43,44]. In this study, data of level 1.5 (i.e., quality checked data) were used because data of level 2.0 (i.e., quality checked and cloud screened data) was only available until 27 September 2018.

Data Selection
The monthly recovery rate ranged between 65 and nearly 100% for PM 10 concentrations (see Table A1 in Appendix A), with the exception of October 2015 when the main electronic card of the TEOM™ broke down. Most of the time, missing data corresponded to particular events such as operations of maintenance on the instruments (filters changing, cleaning, control, etc.), electricity failures, or break down of the computer connections.
In order to compare to the Tunisian air quality standards, daily mean PM 10 concentrations were computed. Firstly, the hourly mean PM 10 concentration was computed from the 5 min measurements. Secondly, the daily mean PM 10 concentration was computed from the hourly mean PM 10 concentration (without accounting for the hourly negative values when there were). Considering a one-month period, 9% at a maximum (in June/July) of the hourly mean PM 10 concentration were removed for the computation of the daily mean PM 10 concentration because it was negative. For the rest of the year, negative values of the hourly mean PM 10 concentration never accounted for more than 5% of the total values for one month.
As it was mentioned in Section 2.2.1., the cases corresponding to negative values of the hourly mean PM 10 concentration were mainly cases during which PM 10 concentration was low. Thus, eliminating these negative values for the computation of the daily mean PM 10 concentration could generate a bias (i.e., could lead to overestimation of daily means). In a context of air quality, this could produce "false" excesses of air quality standards, and the impact of PM 10 could be overestimated. In order to test the impact of the method used to compute the daily mean PM 10 concentration, we replaced the negative hourly values (instead of removing them) by the nominal precision of the instrument for 1 h averaging (i.e., 2.0 µg m −3 ). In these conditions, the coefficient of determination R 2 between the daily mean PM 10 concentrations computed using these two methods was 0.9995 with a slope of 0.999 for the regression line (N = 1790). In terms of air quality, a difference of 1 day over the 5-year period was observed between the two methods when estimating the number of times the limit values relative to well-being and relative to health were exceeded. As a conclusion, we can consider that removing the negative values of the hourly mean PM 10 concentration for the computation of the daily mean PM 10 concentration does not significantly affect the results that are presented and discussed in the present paper.

Annual Mean PM 10 Concentrations
The annual mean concentrations of PM 10 in El Fjé range between 32 and 56 µg m −3 from 2014 to 2018 (Table 2). These values are lower than 60 µg m −3 for all years respecting both the limit values relative to health and to well-being defined by Tunisian air quality standards for airborne particles (Table 1). On the other hand, for the five years, the WHO air quality guidelines for PM 10 were exceeded by, at least, a factor 1.5 ( Table 2).  Figure 2 presents an inventory of the most recent available measurements of annual mean PM 10 concentrations in the countries downwind the Sahara and frequently impacted by Saharan dust outbreaks (i.e., North Africa, South Europe, and West Africa). Firstly, it can be noted that most of the available measurements are located in urban and suburban areas (63%): Only 20 stations (upon 116 in total) are classified as rural or regional background stations (see Table A2 in Appendix B). El Fjé appears as the unique rural station in Tunisia and, to a regional extent, in North Africa. The annual mean PM 10 concentrations measured in El Fjé are low (about three times lower for 2017) compared to those measured in the northern Tunisian urban centers (except Sousse-see measurements reported for Bizerte, Sfax, and Tunis in Figure 2). On the other hand, the annual mean PM 10 concentrations measured in El Fjé are 1.6 to 5 times higher than the values reported by Pey et al. [45] for south European urban centers located farther from the dust sources ( Figure 2). Finally, the annual mean PM 10 concentrations measured in El Fjé are 2.5 to 3.5 times lower than what can be measured at the rural stations in the Sahel (see measurements reported for Cinzana and Banizoumbou by Kaly et al. [46] in Figure 2), which suggests that southern Tunisia is not a dust source or a pathway for Saharan dust transport as important as the Sahel is.  [46]. Details on the location, type, and data coverage is given for each station in Table A2 (Appendix B). Rural and regional background stations are identified using bold italics font.

Daily Mean PM10 Concentrations
The Tunisian air quality standards were no longer respected at the daily time scale in El Fjé: during the period 12 February 2014-28 February 2019, the daily limit value relative to health was exceeded 40 times (i.e., 2% of all the measured PM10 concentrations), while the daily limit value relative to well-being was exceeded 107 times (i.e., 6% of all the measured PM10 concentrations) (see Table A3 in Appendix C). Figure 3 presents the monthly frequency of occurrence of days with daily mean PM10 concentrations greater than 120 and 260 µg m −3 . Daily air quality standards were more frequently exceeded during winter (DJF) and spring (MAM), than in autumn (SON) and summer (JJA), during which it is frequently null. This seasonal cycle follows the seasonal cycle of desert dust outbreaks observed in this region, strongly suggesting that desert dust aerosols were responsible for daily air quality standards excesses. Indeed, Moulin et al. [31] reported that April was a period of intense dust uplift over North Africa because of the Sharav cyclones [47], and that the major source regions for the Mediterranean were: (i) South Tunisia around 30° N for the western central basin [26], and Libyan desert just below 30° N for the eastern central basin [48]. More recently, Fiedler et al. [49] showed that above 20° N, up to 80% of desert dust emissions was due to atmospheric depressions whose frequency over the north of Africa is the highest between March and May, and then between December and February.  [45]; black = this study; blue = the 2016 WHO GUAAP Database; orange = Kaly et al. [46]. Details on the location, type, and data coverage is given for each station in Table A2 (Appendix B). Rural and regional background stations are identified using bold italics font.

Daily Mean PM 10 Concentrations
The Tunisian air quality standards were no longer respected at the daily time scale in El Fjé: during the period 12 February 2014-28 February 2019, the daily limit value relative to health was exceeded 40 times (i.e., 2% of all the measured PM 10 concentrations), while the daily limit value relative to well-being was exceeded 107 times (i.e., 6% of all the measured PM 10 concentrations) (see Table A3 in Appendix C). Figure 3 presents the monthly frequency of occurrence of days with daily mean PM 10 concentrations greater than 120 and 260 µg m −3 . Daily air quality standards were more frequently exceeded during winter (DJF) and spring (MAM), than in autumn (SON) and summer (JJA), during which it is frequently null. This seasonal cycle follows the seasonal cycle of desert dust outbreaks observed in this region, strongly suggesting that desert dust aerosols were responsible for daily air quality standards excesses. Indeed, Moulin et al. [31] reported that April was a period of intense dust uplift over North Africa because of the Sharav cyclones [47], and that the major source regions for the Mediterranean were: (i) South Tunisia around 30 • N for the western central basin [26], and Libyan desert just below 30 • N for the eastern central basin [48]. More recently, Fiedler et al. [49] showed that above 20 • N, up to 80% of desert dust emissions was due to atmospheric depressions whose frequency over the north of Africa is the highest between March and May, and then between December and February.
Despite this obvious mean annual cycle, daily mean PM 10         In order to identify the type(s) of particles responsible for the 107 excesses of the Tunisian daily limit value relative to well-being, we used as an indicator the Angström exponent (440-870 nm), α 440-870 . Indeed, Holben et al. [50] considered that α 440-870 values less than 0.4 are typical of situations dominated by coarse aerosols (sea salt, mineral dust) whereas larger α 440-870 values denote the presence of submicron aerosols such as those produced by biomass burning or other combustion processes. Figure 5 reports the daily minimum α 440-870 as a function of the daily mean PM 10 concentration for the whole period. It can be seen that, for all days (except one) with a daily mean PM 10 concentration greater than 120 µg m −3 , the daily minimum α 440-870 is systematically lower than 0.4, which means that AOD was strongly controlled by coarse particles. Kchih et al. [51] analyzed the composition of PM 10 collected in the southern Mediterranean coast of Tunis. These authors found that in all sites where PM 10 were collected, these particles were mainly composed, in mass, of crustal matter (45%-50%) and poorly of sea salt (3%-4%), which showed that aerosol chemical composition was slightly impacted by the nearby Mediterranean Sea. These results confirm that the recorded PM 10 maximums in El Fjé were due to mineral dust.

The Situations of the Excess of the Tunisian Daily Air Quality Standards
In order to identify the type(s) of particles responsible for the 107 excesses of the Tunisian daily limit value relative to well-being, we used as an indicator the Angström exponent (440-870 nm), α440-870. Indeed, Holben et al. [50] considered that α440-870 values less than 0.4 are typical of situations dominated by coarse aerosols (sea salt, mineral dust) whereas larger α440-870 values denote the presence of submicron aerosols such as those produced by biomass burning or other combustion processes. Figure 5 reports the daily minimum α440-870 as a function of the daily mean PM10 concentration for the whole period. It can be seen that, for all days (except one) with a daily mean PM10 concentration greater than 120 µg m −3 , the daily minimum α440-870 is systematically lower than 0.4, which means that AOD was strongly controlled by coarse particles. Kchih et al. [51] analyzed the composition of PM10 collected in the southern Mediterranean coast of Tunis. These authors found that in all sites where PM10 were collected, these particles were mainly composed, in mass, of crustal matter (45%-50%) and poorly of sea salt (3%-4%), which showed that aerosol chemical composition was slightly impacted by the nearby Mediterranean Sea. These results confirm that the recorded PM10 maximums in El Fjé were due to mineral dust.

Sub-Daily Variability
Considering the impact of desert dust on human health, we can wonder what differences exist between a long-term exposure to high levels of PM10 concentration and a short-term exposure to very high PM10 concentration. Indeed, in extreme cases, people and animals caught in a severe dust storm may die due to suffocation [9]. For instance, a major dust storm in April 1895 in the US state of Colorado was reported to have caused the loss of 20% of the cattle in the east of the state [52]. When examining all the days with a daily mean PM10 concentration greater than 120 µg m −3 , we noticed that the daily limit value can be exceeded in different ways. In order to investigate it, we searched for two days for which the daily mean PM10 concentration was greater than 120 µg m −3 and in the same order of magnitude, but during which the five-minute PM10 concentration dynamics was completely different. Because the number of days exhibiting a daily mean PM10 concentration below 260 µg m −3 represents more than a half of the days for which the daily limit value relative to well-being was

Sub-Daily Variability
Considering the impact of desert dust on human health, we can wonder what differences exist between a long-term exposure to high levels of PM 10 concentration and a short-term exposure to very high PM 10 concentration. Indeed, in extreme cases, people and animals caught in a severe dust storm may die due to suffocation [9]. For instance, a major dust storm in April 1895 in the US state of Colorado was reported to have caused the loss of 20% of the cattle in the east of the state [52]. When examining all the days with a daily mean PM 10 concentration greater than 120 µg m −3 , we noticed that the daily limit value can be exceeded in different ways. In order to investigate it, we searched for two days for which the daily mean PM 10 concentration was greater than 120 µg m −3 and in the same order of magnitude, but during which the five-minute PM 10 concentration dynamics was completely different. Because the number of days exhibiting a daily mean PM 10 concentration below 260 µg m −3 represents more than a half of the days for which the daily limit value relative to well-being was exceeded, we selected, as an example, two days for which a daily mean PM 10 concentration of~125 µg m −3 was recorded-12 February 2014 and 13 May 2014. Figure 6 presents the temporal evolution of the five-minute PM 10 concentrations during these two days. In the first case, on 12 February 2014, five-minute PM 10 concentrations were higher than 100 µg m −3 during more than half of the day, but never exceeding 400 µg m −3 , while, in the second case, on 13 May 2014, the daily limit value relative to well-being was exceeded because of the occurrence of a peak of very high PM 10 concentration (up to 3398 µg m −3 on five minutes) over a short time period (less than 1 h). Such differences in the five-minute PM 10 concentration dynamics were also observed during days with higher daily mean PM 10 concentrations (not shown).
case, on 13 May 2014, the daily limit value relative to well-being was exceeded because of the occurrence of a peak of very high PM10 concentration (up to 3398 µg m −3 on five minutes) over a short time period (less than 1 h). Such differences in the five-minute PM10 concentration dynamics were also observed during days with higher daily mean PM10 concentrations (not shown).
The differences observed in the temporal evolution of PM10 concentrations between the two presented days can be explained by the meteorological situation (see the detailed analysis presented in Appendix D): On 12 February 2014, high PM10 concentrations measured at the station of El Fjé resulted from dust transport from the region of the Chotts in the western part of Tunisia, while, on 13 May 2014, the same PM10 level was the consequence of local wind erosion in the Jeffara plain. These observations raise the question of the possible need to define sub-daily limits to warn populations in case of very intense dust events. Middleton [9] reported on the public information system established by the Korea Meteorological Administration (Asian dust warning system; http://web.kma.go.kr/eng/weather/asiandust/intro.jsp) to warn the population of possible health effects in case of intrusion of desert dust coming from long-range transport from China and Mongolia across the Korean peninsula. This warning system includes two levels based on the value of the hourly mean PM10 concentration: Advisory (respectively warning) when the hourly mean PM10 concentration is expected to exceed 400 µg m −3 (respectively 800 µg m −3 ) for over 2 h.
During the period February 2014-February 2019, in El Fjé, the hourly mean PM10 concentration was greater than 400 µg m −3 472 times, and greater than 800 µg m −3 170 times, so that the advisory level would have been triggered 360 times while the warning level would have been triggered 115 times. At a maximum, the hourly mean PM10 concentration was greater than 400 µg m −3 during 27 consecutive hours from 30/01/2015 22:00UTC to 01/02/2015 00:00UTC; during this period, the hourly mean PM10 concentration was greater than 800 µg m −3 during 20 h. Even if such situations are exceptional (at the hourly timescale, PM10 concentrations were, most of the time (77%), lower than 40 µg m −3 , Figure 7), they clearly suggest that people living in this region (and more generally, those These observations raise the question of the possible need to define sub-daily limits to warn populations in case of very intense dust events. Middleton [9] reported on the public information system established by the Korea Meteorological Administration (Asian dust warning system; http: //web.kma.go.kr/eng/weather/asiandust/intro.jsp) to warn the population of possible health effects in case of intrusion of desert dust coming from long-range transport from China and Mongolia across the Korean peninsula. This warning system includes two levels based on the value of the hourly mean PM 10 concentration: Advisory (respectively warning) when the hourly mean PM 10 concentration is expected to exceed 400 µg m −3 (respectively 800 µg m −3 ) for over 2 h.
During the period February 2014-February 2019, in El Fjé, the hourly mean PM 10 concentration was greater than 400 µg m −3 472 times, and greater than 800 µg m −3 170 times, so that the advisory level would have been triggered 360 times while the warning level would have been triggered 115 times. At a maximum, the hourly mean PM 10 concentration was greater than 400 µg m −3 during 27 consecutive hours from 30/01/2015 22:00UTC to 01/02/2015 00:00UTC; during this period, the hourly mean PM 10 concentration was greater than 800 µg m −3 during 20 h. Even if such situations are exceptional (at the hourly timescale, PM 10 concentrations were, most of the time (77%), lower than 40 µg m −3 , Figure 7), they clearly suggest that people living in this region (and more generally, those living close to dust sources) are regularly exposed during several hours to very high PM 10 concentrations, exceeding by far all the air quality standards. The health consequences of such intense but relatively short exposure of the population to dust events remain to be documented and the definition of air quality standards accounting for the intermittency of dust emission should be encouraged in regions where this phenomenon strongly controls air quality. living close to dust sources) are regularly exposed during several hours to very high PM10 concentrations, exceeding by far all the air quality standards. The health consequences of such intense but relatively short exposure of the population to dust events remain to be documented and the definition of air quality standards accounting for the intermittency of dust emission should be encouraged in regions where this phenomenon strongly controls air quality.

Conclusions
In order to investigate the role of mineral dust on air quality in a region where this aerosol species is dominant, PM10 concentrations were measured at a five-minute time step from February 2014 to February 2019 in a remote place of southern Tunisia (El Fjé).
Simultaneous measurements of the AOD and the Angström exponent with a sunphotometer were also performed in the station of El Fjé. These complementary measurements allowed to identify the nature of the aerosols responsible for the excess of the daily Tunisian air quality standardsmineral dust.
The results showed that, on an annual basis, the Tunisian air quality standards were respected in this part of the country contrary to what was measured for the northernmost urban and industrial stations of Tunisia: Annual means of PM10 concentration recorded in El Fjé were significantly lower than those recorded in the main coastal urban centers of northern Tunisia. On the other hand, they appeared high (up to twice higher) for a rural area when compared to stations located far from dust sources (southern Europe), but low when compared to Sahelian stations. In the latter case, this suggests that southern Tunisia cannot be considered as a continuous active source of mineral dust, nor as a regular pathway for mineral dust emitted in the Sahara. On a daily basis, Tunisian air quality standards were no longer respected.
The five-minute time step of PM10 concentration measurements allowed to identify the manner these standards were exceeded. Two contrasted cases were identified: The high daily mean PM10 concentration that was measured during these days was either due to high PM10 concentrations (some thousands of µg m −3 ) lasting several hours or due to a brief, but very high peak of PM10 concentration (> 1000 µg m −3 ) over a short time period (less than 2 h). Using concomitant measurements of mean and maximum wind speed at the same time step as PM10 concentrations should give some insight on

Conclusions
In order to investigate the role of mineral dust on air quality in a region where this aerosol species is dominant, PM 10 concentrations were measured at a five-minute time step from February 2014 to February 2019 in a remote place of southern Tunisia (El Fjé).
Simultaneous measurements of the AOD and the Angström exponent with a sunphotometer were also performed in the station of El Fjé. These complementary measurements allowed to identify the nature of the aerosols responsible for the excess of the daily Tunisian air quality standards-mineral dust.
The results showed that, on an annual basis, the Tunisian air quality standards were respected in this part of the country contrary to what was measured for the northernmost urban and industrial stations of Tunisia: Annual means of PM 10 concentration recorded in El Fjé were significantly lower than those recorded in the main coastal urban centers of northern Tunisia. On the other hand, they appeared high (up to twice higher) for a rural area when compared to stations located far from dust sources (southern Europe), but low when compared to Sahelian stations. In the latter case, this suggests that southern Tunisia cannot be considered as a continuous active source of mineral dust, nor as a regular pathway for mineral dust emitted in the Sahara. On a daily basis, Tunisian air quality standards were no longer respected.
The five-minute time step of PM 10 concentration measurements allowed to identify the manner these standards were exceeded. Two contrasted cases were identified: The high daily mean PM 10 concentration that was measured during these days was either due to high PM 10 concentrations (some thousands of µg m −3 ) lasting several hours or due to a brief, but very high peak of PM 10 concentration (> 1000 µg m −3 ) over a short time period (less than 2 h). Using concomitant measurements of mean and maximum wind speed at the same time step as PM 10 concentrations should give some insight on the local or regional origin of the dust responsible for the excess of air quality standards. Nevertheless, the use of back-trajectory analysis and space-borne measurements in complement would be necessary to confirm the obtained results.
The different ways daily standards were exceeded at the sub-daily scale question the fact that daily mean PM 10 concentration is not the most relevant indicator to quantify population exposure to desert dust. Indeed, during the period February 2014-February 2019, in El Fjé, the hourly mean PM 10 concentration was greater than 400 µg m −3 472 times, and greater than 800 µg m −3 170 times, while the daily limit value relative to health (daily mean PM 10 concentration greater than 260 µg m −3 ) was only exceeded 40 times. As a conclusion, the definition of sub-daily indicators should be then considered at the same time as the impact of the high peaks of PM 10 concentration on human health should be investigated. This would require epidemiological studies to be conducted on the association of high dust concentration at the sub-daily scale with health effect (such as irritation symptoms, hospital admission respiratory disease, hospital admission cardiovascular disease, respiratory mortality, cardiovascular mortality, or total mortality) in regions located downwind desert areas, such as the Sahara. As a first step, it should be useful to compare the respiratory and/or cardiovascular hospitalization rates in these areas with those in other urban regions in north Tunisia. Nevertheless, in the inhabited areas located downwind of deserts, atmospheric pollution due to mineral dust cannot be reduced, such as pollution due to road traffic or industries. Before precisely assessing the role of short bursts of mineral dust on public health, a simple preventive step to protect the local populations should be to set up warning systems based on existing weather and atmospheric dust load forecasts, such as the one already running in Korea. This would allow the local population to be protected during intense dust events following simple rules such as avoiding outdoor activities, wearing anti-pollution masks, etc.  (2) the Chemistry-Aerosol Mediterranean Experiment (ChArMEx) reaserch program (http://charmex.lsce.ipsl.fr) that allowed the installation of the sunphotometer. The authors gratefully acknowledge the NOAA Air Resources Laboratory (ARL) for the provision of the HYSPLIT transport and dispersion model and READY website (http://www.ready.noaa.gov) used in this publication. MODIS data used in this study were downloaded from the Giovanni online data system, developed and maintained by the NASA GES DISC. We also acknowledge the MODIS mission scientists and associated NASA personnel for the production of the data used in this research effort.

Conflicts of Interest:
The authors declare no conflicts of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results. Table A1. Percentage of available data on a five-minute time step for the TEOM™, and number of days in the month where Aerosol Optical Depth (AOD) measurement is available. ND stands for "no data".  Appendix B Table A2. Location, type, data coverage, and source of the annual mean PM 10 concentrations (in µg m −3 ) for each station presented in Figure 2. NA means that the type of station was not given in the original database.

Appendix D Detailed Analysis of the 12 February and 13 May 2014 Dust Events
At the station of El Fjé, a Gill 2D-type Windsonic anemometer (Campbell ® Scientific Instruments) is also installed to measure wind speed and direction at a five-minute time step. As measurement frequency is 10 s, maximum and mean wind speed values measured during the five-minute time period are stored on a CR200 data logger (Campbell ® Scientific Instruments, Shepshed, Loughborough, UK).
In order to determine the differences observed in the sub-daily variations of PM 10 concentrations observed on 12 February and 13 May 2014, as a first step, the temporal evolution of the five-minute PM 10 concentration and five-minute mean and maximum wind speed were analyzed ( Figure A1). It can be seen that, on 12 February 2014, the increase in PM 10 concentration occurred when wind speed was relatively low (below 4 m s −1 on average), which shows that dust was not produced locally. On the opposite, on 13 May 2014, the sudden increase in PM 10 concentration occurred almost at the same time as wind speed increases (up to~12 m s −1 for five-minute mean wind speed and up to~17 m s −1 for five-minute maximum wind speed), which suggests that dust was at least partially due to local wind erosion.
In order to confirm the regional or local origin of dust during these two days, we combined back-trajectory analysis and space-borne measurements. The origin of the air-masses was determined using the online version of the HYbrid Single Particle Lagrangian Integrated Trajectory model (HYSPLIT; [53,54]). This method is indeed widely used to determine the origin of African dust contributions to the total aerosol load (e.g., [55][56][57]) or the origin of mineral dust deposits (e.g., [33]). 72 h backward trajectories starting at 12:00UTC on both days were computed, ending at the ground level (0 m) at El Fjé, using the ensemble model (i.e., 27 backward trajectories were computed for all identified days) forced by the GDAS (Global Data Assimilation System) meteorological database at 0.5 • .
In complement to the air mass trajectories, we used the AOD at 550 nm derived from the measurements of the Moderate Resolution Imaging Spectroradiometer (MODIS) on-board the Aqua satellite: 1 • × 1 • daily maps of the merged Dark Target/Deep Blue product available in the collection 6 MODIS aerosol products over land and ocean [58] were used to determine the most probable origin of mineral dust observed during both days. To do so, AOD maps were produced for each of the days composing the 72 h backward trajectories computed for 12 February and 13 May 2014, and the most probable origin of mineral dust corresponds to the location where the air-mass trajectory crosses near the surface a region with a high AOD.
For 12 February 2014 ( Figure A2), back-trajectories indicate that the air-mass travelled close to the surface from the regions of the Chotts, located at the border between Algeria and Tunisia in the West, where AOD was high (between 0.6 and 0.8) on 11 February 2014. This analysis confirms that dust arriving at El Fjé on 12 February 2014 was not produced locally. In order to confirm the regional or local origin of dust during these two days, we combined backtrajectory analysis and space-borne measurements. The origin of the air-masses was determined using the online version of the HYbrid Single Particle Lagrangian Integrated Trajectory model (HYSPLIT; [53,54]). This method is indeed widely used to determine the origin of African dust contributions to the total aerosol load (e.g., [55][56][57]) or the origin of mineral dust deposits (e.g., [33]). 72 h backward trajectories starting at 12:00UTC on both days were computed, ending at the ground level (0 m) at El Fjé, using the ensemble model (i.e., 27 backward trajectories were computed for all identified days) forced by the GDAS (Global Data Assimilation System) meteorological database at 0.5°. West, where AOD was high (between 0.6 and 0.8) on 11 February 2014. This analysis confirms that dust arriving at El Fjé on 12 February 2014 was not produced locally. For 13 May 2014 ( Figure A3), back-trajectories arrived from the South, where AOD was low (between 0.1 and 0.2). On the other hand, AOD on 13 May 2014 was between 0.3 and 0.4 in the Jeffara plain, where the station of El Fjé is located, which confirms the local origin of the dust responsible for the high PM10 concentrations recorded during this day.