Study of mineral dust entrainment in the planetary boundary layer by lidar depolarisation technique

Abstract

Measurements on 27 June 2011 were performed over the Southern Iberian Peninsula at Granada EARLINET station, using active and passive remote sensing and airborne and surface in-situ data in order to study the entrainment processes between aerosols in the free troposphere and those in the planetary boundary layer (PBL). To this aim the temporal evolution of the lidar depolarisation, backscatter-related Angström exponent and potential temperature profiles were used in combination with the PBL contribution to the aerosol optical depth (AOD). Our results show that the mineral dust entrainment in the PBL was caused by the convective processes which ‘trapped’ the lofted mineral dust layer, distributing the mineral dust particles within the PBL. The temporal evolution of ground-based in-situ data evidenced the impact of this process at surface level. Finally, the amount of mineral dust in the atmospheric column available to be dispersed into the PBL was estimated by means of POLIPHON (Polarizing Lidar Photometer Networking). The dust mass concentration derived from POLIPHON was compared with the coarse-mode mass concentration retrieved with airborne in-situ measurements. Comparison shows differences below 50 µg/m³ (30% relative difference) indicating a relative good agreement between both techniques.

• Saharan mineral dust
• atmospheric aerosol
• lidar
• lidar depolarisation technique
• planetary boundary layer

1. Introduction

The planetary boundary layer (PBL) is the part of the troposphere directly affected by the surface. A deeper knowledge on the entrainment processes occurring at the top of the PBL is fundamental in aerosol studies, since these processes can transfer aerosol from the free troposphere to surface level. The presence of aerosol particles near the surface is highly relevant as they affect visibility (Prospero, 1999) and may have a negative impact on human health (e.g., Kwon et al., 2002; Pérez et al., 2008). A better knowledge of these entrainment processes is very important for compliance with the European legislation. For instance, EU Directive 2008/50/EC about air quality specifies the possibility to subtract the contribution of natural sources when reporting the exceedances of air quality standards such as PM₁₀ levels (i.e. particles with aerodynamic diameters less than 10 µm). This is of particular relevance in Southern European countries, which are often affected by North African mineral dust intrusions. This transport of mineral dust typically occurs as advected layers in the free troposphere (Navas-Guzmán et al., 2013) and due to the entrainment and convective processes within the PBL can significantly affect PM₁₀ mass concentrations at surface level.

In this study, we used a multiwavelength Raman lidar with depolarisation capabilities in synergetic combination with a sun-photometer for monitoring the entrainment processes of a lofted Saharan dust layer within the PBL. In-situ instrumentation was used to verify the presence of mineral dust at the surface. Furthermore, in order to estimate the amount of mineral dust at different height levels that was available to be dispersed into the PBL, the dust mass concentration profiles were retrieved by means of the Polarizing Lidar Photometer Networking (POLIPHON) based on lidar depolarisation information (Ansmann et al., 2011; Ansmann et al., 2012). In addition, this method was compared to the coarse mass concentration profiles retrieved with airborne instrumentation.

2. Experimental site

Data used in this study were measured at the Andalusian Institute for Earth System Research (IISTA-CEAMA), located in Granada (37.16°N, 3.61°W, 680 m asl). Granada is a non-industrialised medium-size city surrounded by mountains, with a population of 240 000 inhabitants. The population of the metropolitan area is around 350 000 inhabitants (source: www.ine.es). It is placed in the Southern Iberian Peninsula where continental climate prevails with high temperatures and scarce rainfall during summer and low temperature and moderate rain during winter. Main local aerosol sources are road traffic, soil re-suspension (especially during the warm-dry season) and domestic heating based on fuel-oil combustion during winter (Titos et al., 2012, Titos et al., 2014a). Main external aerosol sources are the Sahara Desert located at 500 km southward as well as industrial and urban areas in Europe (Lyamani et al., 2010).

3. Instrumentation and methodology

3.1. Passive remote sensing instrumentation

Measurements of columnar aerosol properties at daytime were obtained using a CIMEL CE-318 sun-photometer. This instrument is included in the AERONET network (Holben et al., 1998) (www.aeronet.gsfc.nasa.gov), which imposes standardisation of instruments, calibration, processing and data distribution. The sun-photometer provides solar extinction measurements at seven channels ranging from 340 to 1020 nm and sky radiances measured at 440, 675, 870 and 1020 nm. Solar extinction measurements are used to compute the aerosol optical depth (AOD) at 340, 380, 440, 500, 675, 870 and 1020 nm. The AOD_CIMEL uncertainty ranges from ±0.01 in the infrared-visible to ±0.02 in the ultraviolet channels (Eck et al., 1999). The sky radiance measurements in conjunction with AOD_CIMEL data at four wavelengths are used to retrieve aerosol micro physical properties like columnar aerosol size distribution, refractive index and single scattering albedo, ω(λ), using the algorithm by Dubovik and King (2000) with improvements by Dubovik et al. (2006). The reported uncertainties for aerosol properties are around 10–35%, for the size distribution retrievals in the 0.1 µm < r<7 µm size range, while for sizes retrieval outside of this range, uncertainties rise up to 80–100% [Dubovik and King (2000); Dubovik et al. (2006)].

In addition, continuous monitoring of tropospheric temperature profiles during the studied period was performed using a ground-based passive microwave radiometer (RPG-HATPRO G2, Radiometer Physics GmbH, Birkenmaarstraße 10, 53340 Meckenheim, Germany). The passive microwave radiometer performs measurements of the sky brightness temperature with a radiometric resolution between 0.3 and 0.4 K root mean square error at 1 s integration time (Navas-Guzman et al., 2014). The radiometer uses direct detection receivers within two bands: 22–31 and 51–58 GHz. The first-band radiation provides information about the tropospheric water-vapour profile, while the second band is related to the temperature profile. Using an inversion algorithm developed by Rose et al. (2005), temperature profiles were retrieved from brightness temperature and surface meteorological data. The accuracy of the temperature is 0.8 K within the first 2 km and 1.2 K at higher altitudes. Tropospheric profiles were obtained from the surface up to 10 km using 39 heights with vertical resolution ranging from 10 m near the surface to 1 km for altitudes higher than 7 km agl. 25 independent points with vertical resolution between 10 and 200 m are provided below 3 km agl, where the PBL is usually located over Granada (Granados-Muñoz et al., 2012). Temperature profiles were used to determine the PBL height using the parcel method (Holzworth, 1964; Granados-Muñoz et al., 2012; Wang et al., 2012; Cimini et al., 2013; Corredor-Ardoy et al., 2014) and to calculate the potential temperature in combination with the vertical profiles of pressure of the 1976 Standard Atmosphere (Krueger and Minzner, 1976) scaled at surface pressure. This method is applied only during daytime since results are not reliable under non-convective conditions. The PBL height uncertainty is around 200 and 400 m above and below 2 km, respectively, due to the aforementioned different resolution of the temperature profile derived by the microwave radiometer (Granados-Muñoz et al., 2012).

The microwave radiometer is complemented with the air temperature data measured at 2 m agl by the meteorological station. Due to the high sensitivity of the parcel method to the temperature near to the surface, the use of the data from the meteorological station provides more accurate results.

3.2. Active remote sensing instrumentation

A multiwavelength Raman lidar model LR331-D400 (Raymetrics, S.A., 32 Spartis, Metamorfosis 14452, Athens, Greece) was operated pointing to the zenith with emissions at 355, 532 and 1064 nm and detection at elastic and Raman-shifted wavelengths at 355, 532 (parallel and perpendicular components respect to the polarising beam-splitter plane) and 1064 nm, and 387, 408 and 607 nm, respectively. The telescope diameter of the LR331-D400 is 40 cm. In order to correct the overlap deviation in the multiwavelength lidar, we apply the procedure proposed by Wandinger and Ansmann (2002). In most cases, this correction allows the retrieval of the profile at 355 and 532 nm down to 320 m above the station (Navas Guzmán et al., 2011). This lidar system is part of EARLINET (European Aerosol Research Lidar Network) (Pappalardo et al., 2014) and fulfils the quality assurance protocols established by the network. A detailed description of this lidar system is included in Guerrero-Rascado et al. (2009).

Lidar data allows the retrieval of the aerosol-particle backscatter (β) and extinction coefficients (α) as well as the volume linear depolarisation ratio (δ′). β and α were retrieved following the Klett-Fernald (Fernald et al., 1972; Klett, 1981; Fernald, 1984; Klett, 1985) and Raman (Ansmann et al., 1990) methods for daytime and night-time measurements, respectively. During daytime, β derived from lidar measurements requires the assumption of the height-constant extinction-to-backscatter ratio, called lidar ratio (LR). The determination of the LR was performed by matching the integrated extinction coefficient derived by lidar (AOD_lidar) and the aerosol optical depth measured by a co-located sun-photometer (AOD_CIMEL) accepting a range of variation less than 0.03 (Landulfo et al., 2003). As AOD_lidar and AOD_CIMEL are obtained at different wavelengths, the AOD_CIMEL at 532 nm was calculated by linear interpolation. β and α statistical uncertainties were determined by means of Monte Carlo technique (Pappalardo et al., 2004; Guerrero-Rascado et al., 2009). In this work, the maximum errors of the retrieved β profiles were around 30% at the lowermost part of the profile where the maximum differences between the real and the assumed LR were expected. Typically, relative errors of the aerosol-particle backscatter coefficient derived with Raman and Klett-Fernald method are less than 15 and 20%, respectively, and relative errors of the aerosol-particle extinction coefficient derived with Raman method are in the range 10–30% (Russel et al., 1979; Franke et al., 2001; Alados-Arboledas et al., 2011; Preissler et al., 2011).

The volume linear depolarisation ratio, δ′, (Cairo et al., 1999) was retrieved from the parallel and perpendicular polarised components at 532 nm from the multiwavelength Raman lidar. The depolarisation calibration was obtained using the Δ90° calibration method presented by Freudenthaler et al. (2009) and detailed by Bravo-Aranda et al. (2013). δ′ uncertainties have been estimated to range between 0.025 and 0.045 for δ′ values of 0.004 and 0.45, respectively (Bravo-Aranda, 2014).

3.3. Ground-based in-situ instrumentation

In-situ aerosol-particle scattering and absorption properties were obtained with an integrating nephelometer (TSI INCORPORATED, 500 Cardigan Road, Shoreview, MN, 55126, USA) and a multi-angle absorption photometer (MAAP, Thermo ESM Andersen Instruments, 81 Wyman Street, Waltham, MA, 02451, USA), respectively. The nephelometer was used to measure the aerosol-particle scattering coefficient, σ_sp, at 450, 550 and 700 nm. This instrument is periodically calibrated using CO₂ and filtered air. Additionally, the raw σ_sp data are corrected for truncation and non-Lambertian illumination errors using the method described by Anderson and Ogren (1998). The uncertainty in σ_sp is of 7% (Heintzenberg et al., 2006). The aerosol-particle absorption coefficient was measured at 637 nm, σ_ap(637 nm), using a MAAP with an estimated uncertainty of 12% (Petzold and Schonlinner, 2004; Petzold et al., 2005). Air sampling for these instruments is obtained from the top of a stainless steel tube of 20 cm diameter and 5 m length (Lyamani et al., 2008Lyamani et al., 2010), located at about 15 m above the ground. More details on the instruments and the experimental setup are given by Lyamani et al. (2010).

3.4. Airborne in-situ instrumentation

Airborne measurements were carried out by the INTA (National Institute for Aerospace Technology) CASA C-212-200 atmospheric aircraft over Granada on 27 June 2011. The airborne platform obtained vertical measurements following a pseudo spiral centred at (37.19°N, 3.57°W), 8 km away from the experimental site (37.16°N, 3.61°W), with a diameter of about 4500 m. The aircraft overpassed the lidar system at heights between 1200 and 5200 m asl. Figure 1 shows the ascending track of the INTA-C212–200 aircraft during this flight, which took place around 10:45 UTC (local time: UTC + 2). The aircraft was equipped with two optical particle counters: the PCASP-100X manufactured by Particle Metrics System and the Cloud, Aerosols and Precipitation Sonde (CAPS) from Droplet Measurement Technologies (DMT) (Baumgardner et al., 2001). The latter is composed of multiple sensors but only the cloud aerosol spectrometer with depolarisation (CAS) was used in this work. Both instruments were fixed at two hard points located under the aircraft wings.

Fig. 1 Ascending track of the INTA C212-200 aircraft during the flight on 27 June 2011 around 10:45 UTC. Green line represents the vertical of CEAMA station.

CAS probe classifies the particles sizes in the range 0.51–50 µm into 30 channels logarithmically distributed. Its measuring principle is based on light-scattering at 680 nm, i.e. particles scatter light from an incident laser, and collecting optics guide the light scattered in the 4–12 range into a forward-sizing photodetector. This light is measured and used to infer particle size. The uncertainty of this probe varies between 15 and 20% for the size distributions (Baumgardner et al., 2001; Feingold et al., 2006).

The PCASP-100X (passive cavity aerosol spectrometer probe) provides size distributions of aerosol particles in the range 0.1–3 µm of diameter in 15 bins. The measuring principle is the same to the CAS. The scattered radiation between 35 and 135 degrees is collected to determine the size. The wavelength of the laser is 632.8 nm. Size distributions are provided with an uncertainty of 20% (Baumgardner et al., 2005).

Volume concentration profiles for the fine mode (radius between 0.05 and 0.5 µm) and coarse mode (radius from 0.5 to 25 µm) were obtained from the PCASP-100X and CAS sondes, respectively. A refractive index correction is applied to retrieve aerosol volume concentrations since polystyrene latex spheres and glass beads with refractive indices (from 1.51 to 1.56) different to that of the atmospheric aerosols were used for calibration. In order to retrieve the mass concentration profile from the volume concentration profile, aerosol particles were considered as spheres of density 2.6 g·cm⁻³ for Saharan mineral dust (Perez et al., 2006).

3.5. Derived aerosol-particle properties

From the aerosol properties measured with the aforementioned instrumentation, some other optical and physical properties were calculated as indicated in Table 1. The AOD-related Angström exponent at wavelengths λ ₁ and λ ₂, å _AOD (λ ₁, λ ₂), provides information about the wavelength dependence of AOD [eq. (1)]. Similarly, different Angström exponents å (λ ₁, λ ₂) related to β, α, and σ_sp with different combinations of λ ₁ and λ ₂ are defined [eqs. (2) and (3)]. The Angström exponent increases with decreasing contribution of coarse particles in the aerosol mixture and takes values around 2 when the scattering/absorption process is dominated by fine particles, while it is close to 0 when the scattering/absorption process is dominated by coarse particles (Seinfeld and Pandis, 1998; Schuster et al., 2006). Single scattering albedo was retrieved using eq. (4) at 637 nm. The PBL and FT contribution to the AOD, AOD_PBL [eq. (5)] and AOD_FT [eq. (6)], respectively, were retrieved by integrating the lidar extinction coefficient between the boundaries indicated in Table 1, using a similar procedure to the one applied to calculate the AOD_lidar.

Table 1. Aerosol-particle optical and physical properties calculated from optical properties

Property	Symbol/Equation
AOD-related Angström exponent		(1)
σ sp -related Angström exponent		(2)
α, β-related Angström exponent		(3)
Single scattering albedo		(4)
AODPBL		(5)
AODFT		(6)
Lidar ratio		(7)
Particle linear depolarisation ratio		(8)
β dust		(9)
m dust		(10)

The night-time LR profiles at 355 and 532 nm were calculated as the ratio between the Raman-retrieved extinction and backscatter coefficients [eq. (7) from Table 1]. This property depends on the size distribution, refractive index, wavelength, relative humidity and particle shape (Mishchenko et al., 1997; Ackermann, 1998). Large LR values are associated with highly light absorbing particles while low LR values are indicative of scattering particles (Muller et al., 2003; Navas-Guzmán et al., 2013).

Particle linear depolarisation ratio (δ ^p ) was retrieved from the δ′ and the particle-to-molecular backscatter coefficients ratio, R, called backscattering ratio [eq. (8)]. δ′ shows the depolarisation caused by both molecules and aerosol particles, whereas δ ^p only describes the depolarisation caused by aerosol particles providing information about the aerosol-particle shape. Large δ ^p values are related to non-spherical particles, whereas low δ ^p values are related to spherical particles (Freudenthaler et al., 2009). δ ^p varies between 0.01 for sulphates and 0.35–0.4 for ash (Ansmann et al., 2011). The uncertainty of δ ^p was estimated between 10 and 25% according to Freudenthaler et al. (2009).

The anthropogenic and mineral dust aerosol contribution to β was separated based on the work of Shimizu et al. (2004). This method can be used to separate aerosol-particle contribution with different depolarising capabilities. The dust backscatter coefficient, β _dust , is calculated using eq. (9) where the height and spectral dependence was omitted for simplicity. and represent δ ^p values for pure dust and for the anthropogenic aerosol, respectively. A value of 0.31±0.10 was chosen according to the results of SAMUM campaign (Freudenthaler et al., 2009). The value was calculated for the study area obtaining a value of 0.05±0.02 for situations dominated by anthropogenic conditions, similar to that used by Nemuc et al. (2013) of 0.03. Using these values, the dust mass concentrations, m _dust , is retrieved using POLIPHON method (Ansmann et al., 2011, 2012). POLIPHON is based on the combination of lidar depolarisation and sun-photometer measurements as indicated in eq. (10), where ρ _dust and LR _dust are the particle mass density and the dust LR, respectively. For pure dust particles, ρ _dust =2.6 g·cm⁻³ (Perez et al., 2006) and a value of LR _dust =42±6 sr was used, obtained from the Raman retrieval in the period 00:00–01:00 on 27 June (Fig. 2). The ratio C _dust /AOD _dust , called mean extinction-to-mass conversion factor, represents the ratio of column-integrated coarse volume concentration (C _dust ) to AOD for the coarse (dust) mode (AOD _dust ). C _dust was calculated from AERONET-derived level 1.5 inversion products on 27 June 2011 at 11:19 UTC (local time: UTC+2). AOD _dust was derived using the total AOD at 532 nm and the fine fraction η, provided by AERONET (O'Neill et al., 2001a, b). By applying the law of error propagation to eq. (10), the uncertainty in the mass concentration was estimated to be about 45%, considering the following individual uncertainties: 25% for the mass densities, 20% for the dust LR, backscatter coefficient and the linear particle depolarisation ratios, and 15% for the mean extinction-to-mass conversion factors (Ansmann et al., 2011).

Fig. 2 From left to right, aerosol extinction (α) and backscattering (β) coefficients and lidar ratio (LR) at 355 and 532 nm, backscatter- and extinction-related Angström exponent [å_β(355, 532 nm) and å_α(355, 532 nm), respectively] and particle linear depolarisation ratio, δ_p(532 nm), at 00:00–01:00 UTC on 27 June 2011.

4. Results and discussions

4.1. Night-time conditions

Raman lidar measurements during the night on 27 June 2011 were analysed in order to characterise the aerosol type and its vertical structure. Figure 2 shows α, β and LR at 355 and 532 nm retrieved by Raman inversion, å_α(355, 532 nm), å_β(355, 532 nm) and δ^p(532 nm) profiles at 00:00–01:00 UTC 27 June 2011 (local time: UTC + 2). å_β and δ^p profiles exhibit two well-differentiated regions: the upper region corresponds to a layer from 3 to 5 km asl, with mean and standard deviations of −0.02±0.25 for å_β and 0.26±0.05 for δ^p, whereas the lower region (1–3 km asl) is characterised by a negative correlation between å_β and δ^p, with decreasing values of å_β (1.00–0.23) and increasing values of δ^p (0.12–0.26) with height. This behaviour indicates a gradual increase of the predominance of coarse and non-spherical particles with height in this region (Fig. 2), suggesting an increase in the presence of mineral dust.

The values of å_β and δ^p in the upper region are in agreement with the values of å_β (355, 532 nm) and δ^p(532 nm) measured by Tesche et al. (2009) and Freudenthaler et al. (2009) for pure Saharan mineral dust (0.19±0.20 and 0.31±0.03, respectively). The LR mean values and standard deviations at 355 and 532 nm of the upper layer obtained here are 43±5 and 41±3 sr, respectively. These values are in the range of values provided in previous studies for medium-transport dust events (Guerrero-Rascado et al., 2009; Müller et al., 2009; Preissler et al., 2011; Bravo-Aranda et al., 2013; Preissler et al., 2013). Considering the error bars, the measured values of LR at 355 and 532 nm are in the range 53±7 and 55±7 sr for pure Saharan mineral dust presented by Tesche et al. (2009).

In order to confirm the presence of Saharan dust over Granada on 27 June, as indicated by the experimental results, 5-day backward trajectories at 00:00 and 13:00 UTC on 27 June ending at Granada are shown in Fig. 3. The backward trajectories were computed with the HYSPLIT model (Draxler and Rolph, 2014) and using the GDAS database. The air masses came from the North of Africa at 2500, 3500, 4500 and 6000 m agl. Particularly, the 4000 m agl backward trajectory comes across the Saharan Desert at heights lower than 1000 m agl between 19 and 22 June and therefore it is very likely to be loaded with mineral dust.

Fig. 3 HYSPLIT backward trajectories ending at 00:00 (left) and at 13:00 UTC (right) on 27 June 2011 at 500, 1500, 2500, 3500, 4500 and 6000 m agl (Granada, 37.16°N, 3.61°W).

4.2. Mixing process: downward entrainment of the mineral dust into the PBL

Figure 4 shows the temporal evolution of the range corrected signal at 532 nm, RCS(532 nm), the PBL height and the total AOD_CIMEL(532 nm), AOD_PBL and AOD_FT obtained on 27 June 2011. The temporal evolution of the RCS(532 nm) shows a similar aerosol-load structure at night- and daytime within the first 4.5 km. AOD_CIMEL(532 nm) is rather stable (~0.28) along daytime, whereas AOD_PBL and AOD_FT change during the morning: AOD_PBL increases from 15% at 07:30 UTC up to 58% at 12:00 UTC, whereas AOD_FT decreases from 85% down to 42% during the same period. The aerosol-load vertical structure changes along the morning in coincidence with a great enhancement of the PBL height as it can be seen in Fig. 4.

Fig. 4 Top: temporal evolution of the total AOD at 532 nm (green). Bottom: temporal evolution of PBL height (grey) and the PBL and FT contribution to AOD(532 nm) (black and red, respectively) on 27 June 2011. Colour maps represent the lidar range corrected signal at 532 nm on 27 June 2011 from 00:00 to 01:00 and 06:30 to 12:15 UTC.

Figure 5 shows the hourly mean potential temperature (θ) profiles between 07:30 and 11:30 UTC. The large increase of θ at surface level and the evolution of θ profiles are coherent with the PBL heights obtained with the parcel method. The superadiabatic behaviour close to the surface (680–690 m asl), with , together with the increase in the atmospheric instability in the layer 690–1490 m asl during the morning, leads to an increase of the convective activity. These convective processes result in variations of AOD_FT and AOD_PBL due to mineral dust entrainment from the FT to the PBL. Similar phenomena were detected in Cape Verde Island by Engelmann et al. (2011) where a vertical mixing between air masses in the PBL and the FT was detected.

Fig. 5 Hourly mean potential temperature evolution at different hours showed by the label (UTC) on 27 June 2011.

The mineral dust entrainment from the FT to the PBL is also clearly evidenced by the temporal evolution of å_β, δ^p and β_dust/β profiles. Figure 6 shows the β, å_β, δ^p and β_dust/β profiles from 07:30 to 11:30 UTC (local time: UTC + 2) up to 2.5 km asl. As it can be seen, β presents slight variations in the lowermost part of the profile during the morning, whereas å_β and δ^p show higher variability with time. This temporal evolution suggests that the dust particles arriving at higher levels (above 2 km) are getting incorporated into the lower layers of the atmosphere. Thus, at 11:30 local time å_β and δ^p values at the lowermost part of the profile are similar to those found above 2 km. Table 2 shows the mean values of δ^p(532 nm), å_β(355, 532 nm) and β_dust/β in the range 1–1.1 km asl and the values of the å_sp(450, 700 nm) at surface and the PBL height at 07:30 and 11:30 UTC. As can be seen, å_β and δ^p present an opposite trend with decreasing å_β and increasing δ^p values and β_dust/β between 07:30 and 11:30 UTC. This behaviour linked to the increase of the PBL height points to a downward entrainment of the dust into the PBL due to the convective processes. To the best of our knowledge, the combined information provided by å_β and β_dust/β profiles is used here for the first time to study the entrainment and mixing processes within the PBL.

Fig. 6 β(532 nm), å_β (355 532 nm), δ_p (532 nm) and β_d/β(%) profiles retrieved using 30-min averaged lidar signals at different hours following the label (UTC) on 27 June 2011. Only error bars corresponding to 11:30 UTC are shown for clarity.

Table 2. Thirty-minute mean and standard deviation of in-situ and lidar optical properties at surface and between 1 and 1.1 km asl, respectively, and the PBL height at two different hours during the morning of 27 June 2011

Property	07:30 UTC	11:30 UTC
δ^p (532 nm)[1–1.1 km]	0.08±0.02	0.19±0.03
Åβ(355, 532 nm)[1–1.1 km]	2.1±0.1	0.8±0.1
βdust/β [1–1.1 km] (%)	12±10	45±10
PBL height (km asl)	1.2±0.05	2.8±0.1
åsp(450, 700 nm)	1.23±0.05	0.86±0.05

In order to discard that the increase of mineral dust could be related to local sources or to advected mineral dust at low levels additional data are used. δ^p value measured on 27 June 2011 is compared to a reference measurement on 14 June in absence of transported mineral dust. δ^p is almost twice on 27 June 2011 (~0.19) compared to 14 June 2011 (~0.10) in the lowermost layer. Thus, the large difference of δ^p near the surface between 27 and 14 June shows that the presence of the mineral dust in the PBL on 27 June 2011 cannot be solely due to local mineral sources. On the other hand, Fig. 3 shows that the air masses at 500 m agl came from Europe across the Mediterranean basin indicating that the increase of mineral dust at low altitudes is not related to Saharan dust intrusion at this level.

Figure 7a and 7b show the temporal evolution of σ_sp(550 nm), σ_ap(637 nm), å_sp(450, 700 nm) and ω(637 nm) at surface level from in-situ measurements on 27 June 2011 as well as the typical conditions at the station for working days not affected by Saharan mineral dust on June. The temporal resolution on 27 June ranges between 1 and 15 minutes depending on the instrument, whereas June average values correspond to 15 minutes intervals. Shaded regions in Fig. 7 indicate the standard deviation. In Fig. 7a, it can be seen that å_sp(450, 700 nm) values range between 0.8 and 1.5 on 27 June while the June monthly mean å_sp(450, 700 nm) values range between 1.5 and 1.8. It indicates a larger predominance of coarse particles on 27 June compared to typical conditions, related to the presence of mineral dust at surface. During night-time, å_sp(450, 700 nm) and ω(637 nm) values are larger than their mean values in June (Fig. 7a), indicating the presence of mineral dust also during night-time. Even though mineral dust is detected at the surface from night-time, a significant change of ω(637 nm) and å_sp(450, 700 nm) is detected around 11:30 UTC. At this time, an increase of ω(637 nm) and a decrease of å_sp(450, 700 nm) at surface level occur in coincidence with the decrease of å_β(355, 532 nm) and the increase of β_dust/β values in the lowest part of the atmospheric column. As it can be seen in Table 2, å_β(355, 532 nm) in the range 1–1.1 km asl and å_sp(450, 700 nm) decrease from 07:30 to 11:30 UTC indicating a larger predominance of coarse particles. The agreement between the temporal evolution of å_β(355, 532 nm) at 1 km asl and å_sp(450, 700 nm) at surface is related to the strong convective processes which facilitate the mixing of anthropogenic particles originated close to the surface and the mineral dust located at higher altitudes. Additionally, as Fig. 7b shows, the trend of σ_sp(550 nm) values between morning peak and the afternoon plateau points to this hypothesis. σ_sp(550 nm) mean values for June on working days not affected by Saharan dust events decrease from 63 Mm⁻¹ at 08:00 UTC to 25 Mm⁻¹ at 13:00 UTC, due to the increasing PBL height and decreasing road traffic emissions. However, on 27 June, σ_sp(550 nm) values show a less intense decrease from 57 Mm⁻¹ (08:00 UTC) to 37 Mm⁻¹ (13:00 UTC) likely connected to the entrainment of mineral dust within the PBL.

Fig. 7 Temporal evolution of in-situ properties: (a) å_sp(450, 700 nm), ω (637 nm) and (b) σ_sp (550 nm), σ_ap (637 nm). Thick lines correspond to data on 27 June and thin lines correspond to the mean of June 2011 for working days not affected by Saharan mineral dust at surface level. Shaded regions indicate the standard deviation.

The diurnal evolution of in-situ aerosol properties at surface is affected by the presence of mineral dust. Maxima of σ_sp(550 nm) and σ_ap(637 nm) in June occur at 07:30 and 07:45 UTC, respectively, whereas the minimum of ω(637 nm) occurs at 06:30 UTC (Fig. 7). Thus, a temporal displacement of σ_sp(550 nm) and σ_ap(637 nm) maxima is observed under typical conditions (Lyamani et al., 2010). This effect was previously detected by Lyamani et al. (2010) and explained in terms of the time required for secondary particle formation by gas-to-particle conversion. However, this temporal displacement is not observed on 27 June since maxima of σ_sp(550 nm) and σ_ap(637 nm) and minimum of ω(637 nm) are coincident at 08:00 UTC. A possible explanation for this behaviour could be, among other factors, the presence of mineral dust particles at surface level that favour gaseous deposition and can supress or reduce gas-to-particle conversion processes. Thus, the entrainment of mineral dust into the PBL can significantly influence the internal mixing and particle formation processes occurring within the PBL.

4.3. Estimation of mass concentration profiles

According to the air quality report of June 2011 (Andalusian Regional Government, www.juntadeandalucia.es/temas/medio-ambiente.html), on 27 June the PM₁₀ daily limit of 50 µg·m⁻³ is exceeded and listed as harmful for human health. On this date, it is observed that convective and entrainment processes incorporate aerosol particles within the PBL from the advected mineral dust layer at higher altitudes increasing the dust mass concentration, m _dust , at surface level. Therefore, an estimation of the vertically resolved m _dust would help to evaluate the amount of mineral dust available to be incorporated into the PBL. To this aim, POLIPHON (Ansmann et al., 2011; 2012) is used to separate the mineral dust contribution from the urban-dust mixture. The m _dust profile at 10:30 UTC derived from POLIPHON is presented together with the coarse-mode mass concentration profile, m _c , derived from airborne data at 10:45 UTC (Fig. 8). As shown in Fig. 8, the dust mass concentration is in the range 30–60 µg·m⁻³ between 1 and 2 km asl, whereas the mass concentration of particles with diameters in the range 1–10 µm at surface is 25.5±1.5 µg·m⁻³ as measured with a DustTrak DRX (TSI INCORPORATED, 500 Cardigan Road, Shoreview, MN, 55126, USA) (Kim et al., 2004). These results show a larger aerosol-particle load at high altitude than at surface level. Thus, due to convection processes, the coarse mass concentration at surface level could increase leading to the exceedance of the PM₁₀ daily limit established by the European Directive observed on this day.

Fig. 8 Dust mass concentration (m _dust ) retrieved by POLIPHONat 10:30 UTC, coarse-mode mass concentration (m _c ) derived from airborne measurements at 10:45 UTC and mean mass concentration of particles with diameters between 1 and 10 µm (m _s ) (µg/m³) in the period 10:00–11:00 UTC at surface level (0.68 km asl) on 27 June 2011.

The vertical distribution of the aerosol load determined by lidar and airborne in-situ measurements generally agrees with differences between m _dust and m _c lower than 30% in the height ranges 1–2.4 and 3–5 km asl (Fig. 8). In the range 2.4–2.7 km asl, the lidar and the airborne in-situ instrumentation detect maximum peaks at different heights (3.2 and 3.35 km asl, respectively). The difference between the maximum values at these peaks is also around 30%. The differences found between m _dust and m _c can be due to the fact that the distance between the lidar station and the aircraft path was around 8 km. In addition, POLIPHON includes assumptions such as the external mixing, the height-constant LR used to retrieve the β profile and the correspondence of the mineral dust with the coarse-mode particles. Furthermore, measurement techniques are different and m _c is retrieved assuming spherical particles which, according to Ginoux (2003), may introduce large uncertainty for particles between 3 and 8 µm of radius. Discrepancies between in-situ and remote sensing data due to hygroscopic growth are not expected in this study since this effect can be neglected for relative humidity values below 40% (region below 3 km asl). In the dust layer, where the relative humidity reaches values around 55%, the hygroscopic enhancement is also negligible since dust particles have low hygroscopicity (Fierz-Schmidhauser et al., 2010; Titos et al., 2014b). The comparison of POLIPHON retrievals with the aircraft measurements shows that this method is appropriate to retrieve the mass concentration of an urban-dust mixture despite of the required assumptions. It is worthy to note that, to our knowledge, this is the first time that POLIPHON method is validated using aircraft measurements.

5. Conclusions

In this work, we use data collected during a Saharan dust outbreak on 27 June 2011. For this day, all the available instrumentation at IISTA-CEAMA station was operating and there was additional information from in-situ instrumentation on-board an aircraft. Data collected during this campaign are used to study mineral dust entrainment processes in the PBL. The synergetic use of active and passive remote sensing instrumentation allowed us to study the interaction between the dust layer located in the FT and the PBL due to the entrainment and convective processes. For this purpose, an approach based on the combined analysis of the temporal evolution of å_β and δ^p and β _dust /β profiles was used. Up to our knowledge, this kind of approach has not been presented in previous studies. The results obtained evidenced the entrainment of mineral dust within the PBL and an increase of the mineral dust presence at lower levels throughout the morning on 27 June. Additionally, the temporal evolution of the FT and the PBL contribution to the total AOD showed an increase of the PBL contribution to the total AOD, whereas the FT contribution decreased. This indicates that the convective processes enhanced the PBL which ‘trapped’ the dust layer, accelerating the downward dust entrainment. The increase of β _dust /β at 1 km asl from 12±10% to 45±10% throughout the morning indicates the relevance of this process for the downward entrainment. This work shows the importance of high spatio-temporal resolution measurements for monitoring and analysing aerosol mixing processes.

The temporal evolution of properties derived from in-situ data and lidar depolarisation and backscatter-related Angström exponent profiles demonstrated that the downward dust entrainment caused an increase of the mineral dust presence at surface level. Furthermore, the delay between aerosol scattering and absorption coefficient at the surface level disappeared, suggesting that the gas-to-particle conversion may be less important due to the gas deposition on particles surface under mineral dust presence at surface level.

In addition, the dust mass concentration was obtained by POLIPHON and, for the first time, compared to the coarse-mode mass concentration retrieved with airborne in-situ measurements. Considering the shortcomings of both methodologies, a relative good agreement (differences below 30%) was obtained. According to the results, POLIPHON is a promising tool to retrieve dust mass concentration profiles in anthropogenic-dust aerosol mixtures and can be used to estimate the amount of mineral dust available to be dispersed into the PBL due to convective activity.

Acknowledgements

This work was supported by the Andalusia Regional Government through projects P12-RNM-2409 and P10-RNM-6299, by the Spanish Ministry of Science and Technology through projects CGL2010-18782 and CGL2013-45410-R, and by the EU through ACTRIS project (EU INFRA-2010-1.1.16-262254). CIMEL Calibration was performed at the AERONET-EUROPE calibration centre, supported by ACTRIS [European Union Seventh Framework Program (FP7/2007–2013)] under grant agreement no. 262254. The authors gratefully acknowledge the NOAA Air Resources Laboratory (ARL) for the provision of the HYSPLIT transport and dispersion model and/or READY website (www.ready.noaa.gov). Also, the authors acknowledge images from the BSC-DREAM8b (Dust REgional Atmospheric Model) model, operated by the Barcelona Supercomputing Center (www.bsc.es/projects/earthscience/BSC-DREAM/). Air quality report of June was provided as a result of the Administration Assignment Agreement between the Ministry of Agriculture, Food and Environment and the Superior Council for Scientific Research National Agency for the performance of work related to the study and evaluation of air pollution by particulate matter and metals in Spain. Thanks are also due to INTA Aerial Platforms, a branch of the Spanish ICTS program, and the Spanish Air Force for their efforts in maintaining and operating the aircraft. This work was supported by the University of Granada through the contract ‘Plan Propio. Programa 9. Convocatoria 2013’. Dr. Bravo-Aranda was funded under the grant AP2009-0559. Data supporting this article are available upon request through the corresponding author.