Aerosol radiative effect during the summer 2019 heatwave produced partly by an inter-continental Saharan dust outbreak. 1. Shortwave dust-induced direct impact

. The shortwave (SW) direct radiative effect during the summer 2019 heatwave produced partly by a moderate, long-lasting Saharan dust outbreak over Europe is analysed in this study. Two European sites (periods) are considered: Barcelona, Spain, (23-30 June) and Leipzig, Germany (29-30 June). Major data are obtained from AERONET and MPLNET observations. Modelling is used to describe the different dust pathways. The dust coarse (Dc) and fine (Df) components (total dust, 20 DD=Dc+Df) are separated in the profiles of the total particle backscatter coefficient using the POLIPHON method in synergy with MPLNET measurements. This information is used to calculate the relative mass loading and the centre-of-mass height, as well as the contribution of each dust mode to the DD radiative effect (DRE). The mean dust optical depth and its Df/DD ratios are, respectively, 0.153 and 24 % in Barcelona and 0.039 and 38 % in Leipzig. The dust produced a cooling effect on the surface with a daily mean DRE (Df/DD DRE ratio) of -9.1 W m -2 (37 %) in Barcelona and -2.5 W m -2 (52 %) in Leipzig. 25 Although less intense than on surface, a cooling is also observed at the top-of-the-atmosphere (TOA), where the Df/DD DRE ratio is even though higher (45 % and 60 %, respectively, in Barcelona and Leipzig). Despite the predominance of Dc particles under dusty conditions, the SW radiative impact of Df particles can be comparable to, even higher than, that induced by the Dc ones. In particular, the Df/DD DRE ratio in Barcelona increases by +2.4 % (surface) and +2.9 % (TOA) day -1 along the dusty period. These results are especially relevant for the next ESA EarthCARE mission (planned in 2022), as devoted to 30 aerosol-cloud-radiation interaction research. for the continuous monitoring 70 of the change of dust properties during transportation, and hence, of the DRE evolution. The dust plume was firstly observed in Barcelona (BCN, Spain; 41.4ºN, 2.1ºE, 125 m a.s.l.) on 23 June, and arrived later at Leipzig (LPZ, Germany; 51.4ºN, 12.4ºE, 125 m a.s.l.) on 29 June; P-MPL measurements were performed in both stations. length). Comparable results are found in BCN under dusty conditions, when similar extinction-to-mass conversion factors are provided (Hess et al., 1998; Ansmann et al., 2019), and just depending on the strength (intense and extreme) of dust intrusions (Córdoba-Jabonero et al., 2018, 2019). However, the situation at LPZ is slightly different. The relative Dc (Df) contribution is lower (similar) than that at BCN ( 𝑀 (cid:3005) (cid:3278) = 68 % and 𝑀 (cid:3005) (cid:3281) = 9 %, in average) for the first dust episode, meanwhile 𝑀 (cid:3005) (cid:3278) and 𝑀 (cid:3005) (cid:3281) are reduced (in average, 48 % and 7 %, respectively), for the second one, despite the mean 𝑀 (cid:3013) (cid:3364)(cid:3364)(cid:3364)(cid:3364) is 1.50 times higher with 285 respect to that for the first one. Regarding the Df mass contribution with respect to the total dust mass loading, a 11 % with respect to to 𝑀 (cid:3013) is derived in BCN, that is lower than that found in LPZ (13.5 %) (see Table 3 ). The mean daily-averaged mass loading in BCN is 0.282, 0.032 and 0.314 g m -2 , respectively, for Dc, Df and DD components in average for the whole 23J-30J period; for instance, when an extreme dust situation was observed in BCN, mass loadings reached up to 2.8 g m -2 (Córdoba-Jabonero et al., 2019). In LPZ, those values are lower in comparison, representing a 35 %, 44 % and 36 %, 290 respectively, of that found in BCN for Dc, Df and DD particles. As stated before, this is consequence of the particular dust transport of dust intrusions as observed in BCN and LPZ. These results reflect the fact, as stated before for the optical properties of dust particles ( Sect. 3.2.1 ), that during the second dust episode at LPZ, dust particles (Dc and the surface is very singular. The diurnal cycle of the Dc, Df and DD DRE the surface, at TOA the atmospheric is represented in Figure 10 . Cooling occurs at both the surface and TOA for all modes (Df, Dc and DD) and at all hours of the day. The dust (all modes) produces a heating of the atmosphere during the most of hours of the day and a slight cooling (i.e., |𝐷𝑅𝐸(𝑇𝑂𝐴)| > |𝐷𝑅𝐸(𝑆𝑅𝐹)| ) close to dawn/dusk. At both the surface and TOA, the of the diurnal cycle 575 of 𝐷𝑅𝐸 (cid:3005)(cid:3030) and 𝐷𝑅𝐸 (cid:3005)(cid:3005) is similar to a “W”, showing two minima, one in the morning (06UT) and one in the afternoon (17-18UT), and a maximum at central hours of the day. These results are explained by the sensitivity analysis of SSA, asyF and SA upon the shape of the diurnal DRE cycle as performed by Osipov et al. (2015), and also by a former study of Osborne et al. (2011). The “W” shape, called MMM (min-max-min) structure by Osipov et al. (2015), is basically due to a combination of solar geometry and dust anisotropic scattering: although the radiative produced by forward-scattering particles 580 increases with increasing solar zenith angle, the decreasing solar irradiance at long slant paths (at dawn and dusk) causes that actual peaks as achieved at intermediate solar zenith angles (Osborne et al., 2011). This is valid at both the surface and TOA. The greater 𝑔 , the more pronounced this MMM structure (Osipov et al., 2015). Independently of from the Iberian Peninsula describing a left-side arch for coming from the North, slightly crossing 625 central Europe, meanwhile, for the second one, they arrived directly from the Iberian Peninsula, crossing Europe to LPZ. Indeed, differences found in the vertical optical and mass impact, and consequently in the DRE, of the dust particles are based on the singular transport of dust particles to both distant BCN and LPZ stations. Both AERONET data and MPLNET observations were used for continuous monitoring of the dust outbreak and the retrieval of the dust properties in order to calculate the DRE. By using the synergy between the POLIPHON method and polarized MPL 630 measurements, the vertical profiles of the dust coarse (Dc) and dust fine (Df) extinction coefficient (and also the mass concentration) profiles are separately obtained, and hence the Dc and Df contribution to the total dust (DD=Dc+Df) DRE is estimated. station. In BCN, a total mean daily DRE on the surface, 𝐷𝑅𝐸(𝑆𝑅𝐹) , of -9.1 W·m -2 is found with an instantaneous maximum of 665 W·m -2 , being the total mean daily 𝐷𝑅𝐸𝑓𝑓(𝑆𝑅𝐹) of -88.9 W·m -2  -1 (and an instantaneous peak of -133.7 W·m -2  -1 ). The daily Df/DD DRE(SRF) ratio is 37 %, being > 24 % for Df/DD DOD ratio; that is, in relative terms, Df particles contribute more to the total dust DRE than they do to the DOD. This is also observed in the DREff on the surface: 𝐷𝑅𝐸𝑓𝑓 is higher in absolute value for Df particles (-129.6 W·m -2  -1 ) than for Dc ones (-75.2 W·m -2  -1 ). The driving factor of that is the asymmetry factor: a lower 𝑔 value is found for the fine mode (0.63) than for the coarse one (0.86), implying that, relative to a pure forward- 670 scattering particle, the more solar irradiance is scattered in the atmosphere and thus less irradiance is reaching the surface. In these conditions, it must be highlighted that, at constant AOD, the DRE on surface for Df particles would be higher, in absolute value, than Dc ones. Along the dust 8-day event in BCN the Df/DD DRE(SRF) ratio increases at of +2.4 %·day -1 ., i.e., +17 % between the first and day of the event. That is, end of the dust period, the Df contribution to the total dust DRE on surface is 45 %, i.e., almost the


Introduction
Climate change is a concerning issue nowadays (IPCC, 2013), being extreme events (as heatwaves, droughts, intense aerosol outbreaks, etc.) linked to high perturbations in the radiative balance of the atmosphere. In particular, the degree of 35 understanding of the aerosol-induced climatological implications is still low, and the uncertainties associated to the determination of the aerosol direct and indirect radiative effects are difficult to be unambiguously estimated. This mainly rely on the change of the aerosol properties during their transport, the characterization of aerosol complex mixtures and the lack of information on aerosol-cloud interaction mechanisms (i.e., Haywood and Boucher, 2000).
Dust particles play an important role in the frame of climate forcing due to their direct effect on scattering and absorption of 40 solar radiation as well as their indirect effect by acting as both cloud condensation nuclei and ice nucleating particles (deMott et al., 2003). Radiative forcing (RF) is a proxy for climate research and policy, also linked to the change in global mean surface temperature as derived from climate models at continental, regional or local scales. In particular, significant uncertainties in the estimation of the dust-induced direct radiative effects are still present, hence the necessity to study the radiative properties of dust particles, and to adequately quantify their direct effects on the Earth-Atmosphere radiative budget (IPCC, 2013). The 45 direct radiative effect of the total dust has been widely investigated at both shortwave (SW) and longwave (LW) spectral ranges (Sokolik and Toon, 1996;Pérez et al., 2006;Balkanski et al., 2007); however, despite dust intrusions are usually dominated by large particles, the dust fine mode cannot be disregarded, and hence its relative contribution to the total radiative effect. Therefore, the individual radiative estimate for both dust coarse and fine modes must be separately evaluated, although only a few work addressed this issue. Indeed, dust coarse particles seem to mainly affect the LW radiation, being their fine mode 50 mostly responsible of the SW radiative modulation (Sicard et al., 2014b).
Mineral dust is the most abundant aerosol in the atmosphere with emissions up to 3000 Mt yr -1 (i.e., Zender et al., 2004); in particular, the Saharan desert dust represents half of that airborne abundance (Prospero et al., 2002). In addition, Saharan dust is frequently transported far from its sources to Europe and the American continent, being able to reach rather high altitudes (up to 8 km height). Hence, changes in the dust properties are expected, influencing their vertical radiative field. Indeed, there 55 is a special concern to characterize the impact of the dust intrusions in the climate of Europe. The arrival of Saharan dust intrusions over Europe is frequently observed in springtime and summertime, and mostly in southern Europe; only in very few cases those intrusions are able to reach central and northern Europe. Several studies have focused on the dust vertical distribution using ground-based lidar systems (e.g., Ansmann et al., 2003;Papayannis et al., 2005Papayannis et al., , 2008Mona et al., 2006;Córdoba-Jabonero et al., 2011), belonging to different aerosol lidar networks (EARLINET, MPLNET). Recent studies have Total PBC profiles are derived by using the Klett-Fernald (KF) retrieval (Fernald, 1984;Klett, 1985) with the P-MPL RCS measurements in constraint with the NASA/AERONET (AErosol RObotic NETwork, aeronet.gsfc.nasa.gov) Aerosol Optical Depth (AOD). Columnar AOD data are provided by the two AERONET Cimel sun-photometers, co-located with the P-MPL systems at BCN and LPZ stations. AERONET V3 L2.0 and V3 L1.5 data are available, respectively, at BCN and LPZ (see Sect. 3.3.1). PLDR profiles are obtained from the PBC and VLDR. More details of the P-MPL signal processing and optical 115 retrieval is described in Córdoba- Jabonero et al. (2018).
Additionally, once , and are determined, the mass concentration (MC) profiles of each component, ( ) (g m -3 ; = Dc, Df and ND), can be calculated following the relationship Córdoba-Jabonero et al., 2016) 120 where (g m -2 ) are the specified AERONET-based extinction-to-mass conversion factors (Ansmann et al., 2019) for each component ( = Dc, Df and ND). Note that where , and are, respectively, the particle density (g cm -3 ), the volume conversion factor (10 -12 Mm) and the mass extinction efficiency (MEE, m 2 g -1 ). The values of all those parameters, which are assumed in this work, for each component 125 are shown in Table 1 (including the corresponding references). The vertical profile of the total mass concentration, (g m -3 ) is obtained from the sum of each mass component, i.e., and their vertically-integrated mass values (i.e., mass loadings, g m -2 ) are denoted as and , that is, where ∆ is the vertical resolution of the lidar measurements (75 and 15 m, respectively, for BCN and LPZ). Table 1: Assumed values of PLDR ( ) and LR ( , sr) together with the mass conversion factor ( , g m -2 ), particle density, (g cm -3 ), volume conversion factor ( , 10 -12 Mm) and MEE ( , m 2 g -1 ) for the Dc, Df and ND components.
Regarding the vertical impact of each component, their relative contribution in terms of a mass-weighted altitude, the so-called center-of mass (CoM) height, is calculated, and its evolution is examined along each particular dusty period in BCN and LPZ.
The CoM height, , is defined similarly as in Córdoba- Jabonero et al. (2019), that is,

GAME radiative transfer model
Solar fluxes were calculated in 18 layers of the atmosphere distributed between the surface and 20 km with the radiative transfer (RT) model GAME (Dubuisson et al., 1996;. The solar spectral range was set to 0.2 to 4 μm. GAME calculates solar 150 fluxes at the boundary of plane and homogenous atmospheric layers by using the discrete ordinates method (Stamnes et al., 1988).
More details about the computation of the gas transmission functions can be found in Dubuisson et al. (2004) and Sicard et al. (2014a). The gas absorption is parametrized by profiles of pressure, temperature and relative humidity. In BCN radiosoundings launched twice a day (at 00 and 12 UT) by the University of Barcelona in collaboration with the Servei Meteorològic de 155 Catalunya, the Catalonia meteorological agency, were used. In LPZ no radiosoundings are available, thus the 6-hour profiles from the Global Data Assimilation System (GDAS) provided by the National Oceanic and Atmospheric Administration (NOAA) were used instead. In GAME aerosols are fully parametrized by the user in terms of spectrally-and vertically-resolved aerosol optical depth (AOD), single scattering albedo (SSA) and asymmetry factor (asyF). The spectrally-resolved surface albedo is another input of the model. All those latter parameters are described in Section 3.3.1.

160
GAME has been used to calculate solar fluxes for scientific purposes in several works (see e.g. Roger et al., 2006;Mallet et al., 2008;Sicard et al., 2012). It also participated in an intercomparison exercise of radiative transfer models (Halthore et al., 2005) which concluded that GAME is accurate to a few units of watt (1-3) for a flux reaching of 1000 W m -2 . Since this work is focused on the dust radiative impact, the expression of the aerosol radiative effect (ARE) is particularly defined for dust as the dust radiative effect ( ), at a given height level, , that is, where and are the radiative flux with and without dust, and the ↓ and ↑ arrows indicate whether the fluxes are downward or upward, respectively. By that definition, negative (positive) DRE values represent a cooling (warning) effect. The DRE was calculated at two climate-relevant altitude levels: at the top-of-atmosphere (TOA) and on surface (SRF). The dust contribution in the atmospheric column is quantified by the atmospheric radiative effect, ( ), which is defined as:

Air masses trajectory modelling
In order to determine the origin and pathway of the dusty air masses affecting the two stations involved in this study, a trajectory analysis is performed using two different models.  Draxler and Hess, 1998;Stein et al., 2015;Rolph et al., 2017) is used together with the Global Data Analysis System (GDAS) meteorological files (spatial resolution of 1º x 1º every 3 hours) in order to identify the source regions of the dust particles.
Hence, the dust intrusions observed over each station can be associated to Saharan desert sources by examining the HYSPLIT  (Janjic et al., 2011).
It is used to provide short-to medium-range dust forecasts for both regional and global domains. In particular, dust forecasts over both BCN and LPZ stations for the period from 23 to 30 June 2019 are examined.

185
This section is divided in three subsections: dust plume origin and transport, discussion of the results from the application of the POLIPHON algorithm to the P-MPL observations in terms of dust coarse and fine mode contributions to the optical and mass products, and the estimation of the dust direct radiative forcing.

Dust plume origin and transport
The summer 2019 heatwave as observed across Europe (Sousa et al., 2019)  A similar pattern is found by using HYSPLIT 5-day backtrajectory frequencies at 4000 m a.g.l. and back-trajectory analysis 205 depending on height, as shown, respectively, in Figures 2 and 3. At the end of the dust outbreak period, between 20 % and 90 % of air masses are coming from the Sahara desert area to the BCN station, as well as to LPZ, where the percentage is between 10 % and 40 % (of those 20 calculated, see Fig. 2). Looking at particular back-trajectories over each station in dependence of height (see Fig. 3; note that dates and times are the same as those shown in Fig. 1  The evolution of the aerosol optical properties during the dusty events as observed at BCN (23J-30J) and LPZ (29J-30J) is analysed regarding their vertical structure. Figure 4 shows the total PBC,  , and those separated into dusty,  and  , and non-dusty,  , components, together with the PLDR, , and VLDR, , for representative cases of that evolution (date and times are shown in each panel, corresponding to those HYSPLIT images as shown in Fig. 3).

235
https://doi.org/10.5194/acp-2020-1013 Preprint. Discussion started: 24 November 2020 c Author(s) 2020. CC BY 4.0 License.  In general, by looking at the successive selected panels in Figure 4-a, a gradual dust occurrence can be observed at BCN during the whole dusty event. The most incidence of dust intrusion is observed on 24J, showing a pronounced dust (DD=Dc+Df) layer, which extends from 1.5 to 5.5 km height and with a total PBC peak of 5.6 Mm -1 sr -1 , and where the Dc component is predominating. In this dust layer, values are around 0.30 from 1.5 to 3.5 km, reducing up to 0.22 at higher 245 altitudes up to 5.5 km. Also, a narrow dust layer with of 0.20 is found above at around 6 km. Next days, the DD layer with a small ND contribution also extensively ranges from 1 to 5 km (25J) and 6 (26J) km height, but with less dust incidence: values of 2.2 and 1.8 Mm -1 sr -1 are found, respectively, on 25J and 26J (approx., 30-40 % of that on 24J), with in the 0.2-0.3 range. On 27J, the DD layer descends down to around 4 km height, and reaching the surface, with of 1.5 Mm -1 sr -1 and similar values. From 28J on, the DD signature is also observed from around 4 km down to the surface; however, the most 250 DD occurrence is found from 2.0-2.5 km down, with a rather weak dust incidence: is around or less than 1 Mm -1 sr -1 (around 18 % of that found on 24J), and is lower than 0.2 from 2 km height up and from 0.2 to 0.3 from 2 km down to the surface. As stated before, this points out that the DD signature is more intense at heights lower than 2 km. On 30J-afternoon on, the DD incidence is rather low, showing values less than 0.2.
In the case of the dust intrusion as observed at LPZ, the dust pattern is different from that at BCN site, as also confirmed by 255 BSC/NMMB-Dust and HYSPLIT modelling (see Figs. 1-3). At LPZ, during the 29J-30J dust event, two dusty periods can be observed: the first one lasting from 29J at 12UT to 30J at 05UT, and the second one from 30J at 13UT to the end of that day (unfortunately, no P-MPL data are available after 18UT). Figure 4-b illustrates the evolution of the dust intrusion once arrived at LPZ on 29J. In the first dusty period, a well-differentiated two-layered structure is observed: an evident DD layer with predominance of Dc particles is clearly confined from 3.5 to 5.5 km height, and no DD signature at lower altitudes. In this dust 260 layer, values are less than 1 Mm -1 sr -1 , showing a lower dust incidence with respect to BCN, and only comparable with that present during the last days (12-18 %); however, values are higher, ranging between 0.30 and 0.39, indicating the presence of mostly Dc particles (Mamouri and Ansmann, 2017). On the second dust period, a mixing of Dc, Df and ND particles is observed: the DD layer is extended from the ground to 5 km height, approximately, but with a weak incidence ( = 0.3-0.8 Mm -1 sr -1 ). The DD signature as observed from 2 km height down only corresponds to Df particles, dominating ND aerosols 265 (  0.06), and the Dc and Df particles mostly present from 4 to 5 km (  0.27) and from 2 to 4 km height (  0.16), respectively.

Mass features: relative mass loadings and centre-of-mass height
The aerosol mass features during the dusty events as observed at BCN (23J-29J; continuous dust incidence) and LPZ (29J-30J; two separated dust episodes) are analysed in terms of the relative mass contribution of the Dc , Df and ND components,

270
and their centre-of-mass (CoM) height, as a measure of the vertical mass impact of each component. Figure 5 shows the evolution of the relative mass loading for each component, (%) ( : Dc. Df, and ND; see Eq. (7)) together with the total mass loading, (g m -2 ) (top panels), and the CoM height, (bottom panels), for each component along each particular dust event. Daily-averaged (denoted by a bar over the variable) and values found at BCN and mean (also denoted by a bar over the variable, for simplicity) those values for the two episodes as observed at LPZ are also included.

275
Regarding daily-averaged total mass loading values, a maximal of 0.66  0.42 g m -2 is found on 24J at BCN, representing 3-5 times higher than those maxima observed in LPZ (0.14  0.03 and 0.20  0.04 g m -2 , respectively, for the first and second episodes). In general, as shown in Figure 5 (top panels), Dc particles over BCN are mostly dominating ( > 80 % with respect to the total mass loading) during the 59 % of the overall dust event (23J-30J), though, prevailing for the 90 % of that period in a rather high percentage ( around 60 %); the Df presence is rather lower ( < 10 % for the 72 % of the dust length). Comparable results are found in BCN under dusty conditions, when similar extinction-to-mass conversion factors are provided Ansmann et al., 2019), and just depending on the strength (intense and extreme) of dust intrusions (Córdoba-Jabonero et al., 2018. However, the situation at LPZ is slightly different. The relative Dc (Df) contribution is lower (similar) than that at BCN ( = 68 % and = 9 %, in average) for the first dust episode, meanwhile and are reduced (in average, 48 % and 7 %, respectively), for the second one, despite the mean is 1.50 times higher with 285 respect to that for the first one. Regarding the Df mass contribution with respect to the total dust mass loading, a 11 % with respect to to is derived in BCN, that is lower than that found in LPZ (13.5 %) (see Table 3). The mean daily-averaged mass loading in BCN is 0.282, 0.032 and 0.314 g m -2 , respectively, for Dc, Df and DD components in average for the whole 23J-30J period; for instance, when an extreme dust situation was observed in BCN, mass loadings reached up to 2.8 g m -2 (Córdoba- Jabonero et al., 2019). In LPZ, those values are lower in comparison, representing a 35 %, 44 % and 36 %, 290 respectively, of that found in BCN for Dc, Df and DD particles. As stated before, this is consequence of the particular dust transport of dust intrusions as observed in BCN and LPZ. These results reflect the fact, as stated before for the optical properties of dust particles (Sect. 3.2.1), that during the second dust episode at LPZ, dust particles (Dc and Df components) are highly mixed with ND aerosols (Dc proportion is mostly reduced) in comparison with the first one, which presented a welldifferentiated dust layer between 3.5 and 5.5 km height with predominance of Dc particles ( of 78-84 % with respect to 295 the total mass loading is found for the most intense dust incidence period as occurred from 29J 18UT to 30J 02UT). In addition, these results are in accordance with Sect. 3.1, since the second dust event at LPZ corresponded to Saharan air masses coming directly from the Iberian Peninsula, when dusty conditions were present, and crossing Europe, thus allowing a higher dust mixing than that observed in BCN. However, for the first dust event (when a defined high dust layer was observed over LPZ), only air masses at higher altitudes experienced a pathway slightly crossing the Iberian Peninsula, but arriving at LPZ mainly 300 without crossing Europe (see Fig. 3), avoiding thus a high degree of dust mixing.
The dust intrusion arrived at BCN on 23J. As shown in Figure 5 (bottom panels), in the night from 23J to 24J the CoM height of the dust intrusion reaches its highest value, i.e., is around 4 km. Regarding the daily-averaged for Dc and Df particles, the evolution of their CoM heights follows a similar descending pattern from around 3 km height on 24J down to 2 km on 30J. Besides, the for Df particles is slightly higher than that for the Dc component (200-250 m difference) on 27J 305 until the end of the dust event. These two results can indicate the removal of larger particles along with the progression of the dust intrusion over BCN. In the case of LPZ, two consecutive, but different, dust episodes are observed. The first dust episode (a high well-defined dust layer) arrived at LPZ on 29J at 11UT, mostly composed of Dc particles and with a CoM height of 4.6 km (slightly higher with respect to BCN); then, it showed a constant descending evolution down to 3.7 km on 30J at 05UT.
Concerning the Df particles, their progression along this first episode is from 1.1 to 1.8 km height, peaking at 3.7 km on 310 29J at 20UT. A mean value of 4.1  0.3 and 3.0  0.9 km is obtained, respectively, for the Dc and Df components during this dust episode. After that, a complete removal of the DD particles is observed. Later on 30J, the dust signature is detected again (second DD episode, with a high aerosol mixing, as stated before) at 14UT, lasting until the end of that day; unfortunately, no data were recorded later from 18UT. The CoM height also shows a descending behaviour, and the mean values during this episode for the Dc and Df particles are, respectively, of 3.4  0.1 and 3.1  0.5 km.

315
Therefore, as also regarded before, differences in the vertical mass impact of the dust particles (their relative mass loading and CoM height) found in both distant BCN and LPZ locations are associated to the particular pathway of transported dust particles between stations (see Sect. 3.1).

Dust radiative properties
AERONET V3L2.0 data were available in BCN; but only V3L1.5 data were available in LPZ at the time of writing this article.

335
In order to illustrate the situation at both sites, Figure 6   The dust radiative properties needed to be introduced in the GAME RT model are the AOD, the single scattering albedo (SSA), , and the asymmetry factor (asyF), . These three parameters should be spectrally defined and per layer height. The vertical 360 profiles of dust coarse/fine mode extinction coefficient at 532 nm are obtained from the application of POLIPHON method to the continuous hourly-averaged P-MPL measurements (see Sect.2.2). These profiles are integrated into 18 layer-mean DODs for both the coarse and fine mode. The spectral layer-mean DOD is calculated from the layer-mean DOD at 532 nm using . The dust SSA and asyF as well as the surface albedo (SA), , are taken from AERONET. These three properties are interpolated at the model wavelengths up to the highest AERONET wavelength (1020 nm) and assumed constant 365 to the AERONET 1020-nm value beyond that limit. The asyF is given separately for the coarse and the fine mode.  Fig. 6) are taken as representative for each day. The same periods are considered for the SSA, asyF and SA. Their spectral dependence is represented in Figure 7 and their corresponding values at 440 nm are reported in Table 2. The SSA values found (around 0.94 at 440 nm, see Fig. 7a) are representative of moderately absorbing dust particles. The typical spectral behaviour of SSA for dust is expected to be increasing with increasing wavelength (Dubovik et al., 2002; 380 Sicard et al., 2016). On the one hand, in BCN, the spectrally increased between 440 and 675 nm and variations lesser than The spectral behaviour of the asymmetry factor is shown in Figure 7b, separately for the coarse and the fine modes. The forward scattering is much more pronounced for large particles ( = 0.86) than for small particles ( = 0.61-0.63) independently of the wavelength. This result implies that at constant AOD and low solar zenith angle (SZA) and independently of the wavelength, the solar radiation scattered to the surface will be greater for the coarse mode than for the fine mode. The 395 spectral decreases with increasing wavelength for both size modes. The asyF for the coarse mode is similar in BCN and LPZ, which indicates that the scattering properties of this mode will have a similar effect on the radiative effect retrievals at both sites. The forward scattering of the fine mode at wavelengths greater than 675 nm is slightly higher at BCN ( = 0.59) than at LPZ ( = 0.52). This result implies that at near-infrared wavelengths (> 675 nm), for constant AOD and low SZA, the solar radiation scattered to the surface by fine particles will be greater at BCN than at LPZ.

400
The surface albedo (see Figure 7c) Table 3. The fine-to-total ratio (Df/DD) of the daily DRE varies between 28 and 46 %, being 37 % in average over the whole dust event, that is, the Df/DD ratio of DRE produce a little more than one third of the total DRE. This result can be interestingly related to the Df/DD ratio of the daily DOD (24 %), meaning, in relative terms, that the dust fine particles contribute more to the total DRE than they do to the DOD. Figure 10 nicely illustrates the dust radiative effect on the surface, 435 ( ), vs. DOD (see Fig. 9a), and the values of DREff are included in Table 3. Figure 9a shows instantaneous ( ) for both coarse (red) and fine (blue) modes as a function of their respective DOD.
By using linear regression analysis (regarding DRE=0 with DOD=0), the dust DREff corresponds to the slope of the linear fittings. In BCN, the ( ) over the whole event is -75.2 and -129.6 W m -2  -1 for the coarse and fine mode, respectively, producing a total dust DREff of -88.9 W m -2  -1 . It can be clearly seen that at constant DOD the dust fine mode 440 produces a higher enhancement of DRE than the dust coarse mode. The main difference between the parametrizations for the radiative properties of both modes is the asymmetry factor: values of 0.865 and 0.629 are reported for the coarse and fine modes, respectively. That lower value found for the fine mode with respect to the coarse one implies that, in relation to a pure forward-scattering particle, more solar irradiance is scattered in the atmosphere by Df particles and thus less irradiance is reaching the surface. Another difference is the vertical distribution of each of those coarse-and fine-mode dust layers.

445
However, the height of the dust layer is not expected to have a relevant impact on the ( ) (Liao et al., 1998).
During the most intensive days of the event in BCN, 24J-27J, the total dust DREff on the surface,

465
In LPZ, the event is much weaker than in BCN. Under dusty conditions, for the first (from 29J-pm to 30J-am) and second (30J-pm) dust episodes in LPZ, the DOD is, respectively, 0.083 and 0.067 (see Table 2). On both days, the fine-to-total (Df/DD) ratio of the DOD is approximately of one-third (one-fourth in BCN). The dust instantaneous DRE (see Fig. 8) is on the order of magnitude of the DRE in BCN on the first day of the outbreak (23J), with peaks of -7.4 and -7.8 W m -2 for the coarse and fine mode, respectively. The daily DRE, as averaged over the two days, is -1.2 and -1.3 W m -2 for the coarse and 470 fine mode, respectively, yielding to a total dust DRE of -2.5 W m -2 , and in particular, the radiative contribution of the Df particles was of 52 % (37 % in BCN) with respect to . Despite the radiative impact of the mineral dust in LPZ is small (because of the low dust loading), this result is remarkable. It shows that, in some given circumstances, dust fine mode contribution to the DRE is comparable to that of the coarse mode, whereas mineral dust is usually regarded as a coarsedominating aerosol. The increase of the Df/DD ratio of daily DRE in LPZ (52 %) with respect to BCN (37 %) is due to the 475 gravitational settling of the largest dust particles during a longer transport. Indeed, according to Section 3.1, dust particles arriving at LPZ on 29J and 30J at 4500 m were travelling over the Iberian Peninsula for 3-4 days before (see Fig. 2c). In terms of radiative efficiency, the dust Dc and Df DREff, and , in LPZ, as averaged over the two days 29J-30J, is, respectively, -89.5 and -157.9 W m -2  -1 (-75.2 and -129.6 W m -2  -1 in BCN; see Fig. 9a and Table 3). There are two main differences between BCN and LPZ parametrizations: the spectral is slightly larger in BCN than in LPZ (see Fig. 7b) and 480 the spectrally-integrated surface albedo is lower in BCN than in LPZ (see Fig. 7c). At constant DOD, both differences have an opposite impact on the dust DRE on the surface: the first one (higher in BCN) will yield a weaker cooling effect (i.e. a larger radiative efficiency, as indeed observed), while the second one (smaller surface albedo in BCN) will yield a stronger 485 (2015) showed that a SA decrease from 0.35 to 0.25 (which is approximately the difference in SA between LPZ and BCN at the near-infrared wavelengths; see Fig. 7c) yields to a difference of the SW DRE on the surface less than 3 W m -2 . In addition, the time evolution of the instantaneous Df/DD DRE ratio is shown in Figure 8 for BCN. On the surface, this ratio (blue colour) shows a diurnal cycle, whose shape changes from day-to-day. The mean DRE value over the whole dust event of these instantaneous Df/DD ratios on the surface is 39 %. By discarding the first day (23J) when the dust arrived at BCN, an 495 increase of the Df/DD DRE ratio with time is observed. The best linear fit as calculated between 24J and 30J presents a positive slope of +0.10 %·hr -1 , that is, +2.4 %·day -1 . In other terms, the contribution of the dust fine mode to the total dust DRE on the surface increases steadily along the dust event, being the increase of +17 % between the beginning of the event (on average, 28 % on 24J) and its end (45 % on 30J). ), as defined in Eq. (10), is also analysed. The instantaneous dust DRE at TOA vs. time is shown in Figure 8 and vs. DOD in Figure 9b, and daily and maximal values are reported in Table 4. In BCN, the dust DRE at TOA is negative along all the dusty period. The overall mean daily DRE at TOA was -3.2 and -2.6 W m -2 and the instantaneous maxima are -24.8

Dust direct radiative effect at TOA and in the atmosphere
(24J) and -19.3 W m -2 (25J) for the coarse and fine mode, respectively. For the total dust, the overall mean daily DRE at TOA, 505 ( ), is -5.8 W m -2 , and an instantaneous maximum of -42.7 W m -2 is reached on 25J. For the coarse mode, the instantaneous dust DRE at TOA ( ( ), green bars in Fig. 8) was smaller than that on the surface (in terms of daily values, ( ) represented 56 % of that found on the surface). For the fine mode, this difference is less pronounced: This difference in daily DRE of 20 % between the dust coarse-and fine-mode is the same than that observed in BCN, and the reasons for it are those already mentioned.
The fine-to-total (Df/DD) ratio of the daily ( ) in BCN varied between 36 and 56 % (and between 56 and 67 % in LPZ) and is 45 % (60 % in LPZ) in average over the whole dust event. These ratios are higher than those found on the surface 525 (37 and 52 %, respectively, for BCN and LPZ over the whole dust event). These results indicate that, likewise on the surface, Df particles contribute more to the total ( ) in LPZ than in BCN because of the gravitational settling of the largest dust particles during a longer transport, as stated before (see Sect. 3.1), and that, in relative terms, their contribution is stronger at the TOA than on the surface. This is especially relevant for satellite remote sensing instrumentation, which is mostly sensible to SW wavelengths, since its measurements can be likely affected by dust contamination (Marquis et al., 2020).

530
Since the is lower at the TOA than on the surface, the at the TOA logically decreases as compared to that on the surface. This is observed in Figure 9, where the slope of the best linear fitting is less steep at the TOA than on the surface, i.e. ( ) is negatively higher than ( ). In BCN (LPZ) the daily ( ) and ( ), as averaged over the whole dust event, is -43.9 and -98.6 W m -2  -1 (-49.9 and -116.9 W m -2  -1 ), respectively. The total dust  (Table 4) and those found on the https://doi.org/10.5194/acp-2020-1013 Preprint. Discussion started: 24 November 2020 c Author(s) 2020. CC BY 4.0 License. surface (Table 3) (see Eq. (10)). It can be seen, by looking at Figure 8, that the dust produces generally a heating of the atmosphere (most of the red bars are positive in Fig. 8).  being stronger than that on the surface (in blue colour). The best linear fitting as calculated between 24J and 30J presents a positive slope of +0.12 %·hr -1 , i.e. +2.9 %·day -1 . On average, the contribution of the Df particles to the total dust DRE at the 565 TOA increases +20 % from 39 % on 24J to 59 % on 30J. Likewise, a slightly smaller positive slope of +0.10 %·hr -1 (i.e.,+2.4 %·day -1 ) is found for the Df/DD ( ) ratio, with a mean value of 39 % in the same period; the Df contribution to the total dust DRE on the surface (SRF) increases +34 % along the same dust period in BCN.

Diurnal cycle of the dust direct radiative effect
In order to analyse the diurnal cycle of , the day of 26J is selected, since the dust plume vertical distribution is relatively 570 stable and the AOD almost constant along that day (see Fig. 6). In consequence, the shape of the diurnal cycle of the radiative effect at the surface is very singular. The diurnal cycle of the Dc, Df and DD DRE on the surface, at TOA and in the atmospheric column is represented in Figure 10. Cooling occurs at both the surface and TOA for all modes (Df, Dc and DD) and at all hours of the day. The dust (all modes) produces a heating of the atmosphere during the most of hours of the day and a slight cooling (i.e., | ( )| > | ( )|) close to dawn/dusk. At both the surface and TOA, the shape of the diurnal cycle 575 of and is similar to a "W", showing two minima, one in the morning (06UT) and one in the afternoon (17-18UT), and a maximum at central hours of the day. These results are explained by the sensitivity analysis of SSA, asyF and SA upon the shape of the diurnal DRE cycle as performed by Osipov et al. (2015), and also by a former study of Osborne et al. (2011). The "W" shape, called MMM (min-max-min) structure by Osipov et al. (2015), is basically due to a combination of solar geometry and dust anisotropic scattering: although the radiative effect produced by forward-scattering particles 580 increases with increasing solar zenith angle, the decreasing solar irradiance at long slant paths (at dawn and dusk) causes that actual peaks as achieved at intermediate solar zenith angles (Osborne et al., 2011). This is valid at both the surface and TOA.
https://doi.org/10.5194/acp-2020-1013 Preprint. Discussion started: 24 November 2020 c Author(s) 2020. CC BY 4.0 License. the overall dusty period in both stations, although a higher Df mass contribution with respect to the total dust mass loading is found in LPZ (13.5 %) than in BCN (11 %). The mean daily-averaged total mass loading is higher in BCN (0.314 g m -2 ) than in LPZ (36 % of that found in BCN). As stated before, this is a consequence of the particular dust transport of the dust intrusions to BCN and LPZ, which is also reflected in the vertical impact of the dust intrusions over each station according to the dailyaveraged CoM height. In BCN, the evolution of the CoM height follows a similar descending pattern from around 3 km on 655 24J down to 2 km height on 30J, but the mean CoM height for Df particles is slightly higher than that for the Dc component with the progression of the dust intrusion over BCN. In the case of LPZ, the mean CoM height of the dust particles is located higher than in BCN (i.e., at 3-4 km). Since the dust intrusion in LPZ lasted only for two days and observed for two differentiated episodes each day, that descending behaviour as in BCN is unobserved.

660
In the context of the particular dust scenario as observed in BCN (continuous and progressive dust particles coming from the Sahara region) and LPZ (two close but separated dust episodes: the first introducing a well-defined high decoupled dust layer with mostly Dc presence, and the second presenting a high degree of dust mixing), the DRE (and DREff) are calculated on the surface (SRF) and at the TOA, and also the atmospheric DRE (and its efficiency), in each station.
In BCN, a total mean daily DRE on the surface, ( ), of -9.1 W·m -2 is found with an instantaneous maximum of -54.5 665 W·m -2 , being the total mean daily ( ) of -88.9 W·m -2  -1 (and an instantaneous peak of -133.7 W·m -2  -1 ). The daily Df/DD DRE(SRF) ratio is 37 %, being > 24 % for Df/DD DOD ratio; that is, in relative terms, Df particles contribute more to the total dust DRE than they do to the DOD. This is also observed in the DREff on the surface: is higher in absolute value for Df particles (-129.6 W·m -2  -1 ) than for Dc ones (-75.2 W·m -2  -1 ). The driving factor of that is the asymmetry factor: a lower value is found for the fine mode (0.63) than for the coarse one (0.86), implying that, relative to a pure forward-670 scattering particle, the more solar irradiance is scattered in the atmosphere and thus less irradiance is reaching the surface. In these conditions, it must be highlighted that, at constant AOD, the DRE on surface for Df particles would be higher, in absolute value, than for Dc ones. Along the dust 8-day event in BCN the Df/DD DRE(SRF) ratio increases at a rate of +2.4 %·day -1 ., i.e., +17 % between the first and the last day of the event. That is, at the end of the dust period, the Df contribution to the total dust DRE on surface is 45 %, i.e., almost the same as for the Dc particles.
On the other hand, a total mean daily DRE at TOA (the atmospheric DRE) of -5.8 W·m -2 (+3.3 W·m -2 ) is found (instantaneous DRE peak of -42.7 W·m -2 ). Regarding the DREff, a total mean daily value at TOA and the atmospheric one of -58.0 W·m -2  -1 and +30.9 W·m -2  -1 , respectively, are estimated, with an instantaneous DREff peak at TOA of -122.5 W·m -2  -1 . The daily Df/DD DRE ratio at TOA is 45 %, which is higher than that found on the surface (37 %). Hence, the contribution of the Df particles is stronger at the TOA than on the surface. Along the 8-day dust event the Df/DD DRE(TOA) ratio increases at a rate 680 of +2.9 %·day -1 ., i.e. +20 % between the first and the last day of the event. Then, at the end of the event the Df/DD DRE(TOA) ratio is 59 %, that is, the Df contribution to the total dust DRE is higher than that for the Dc particles. Regarding the atmospheric DRE, the Df/DD ratio is very similar to that estimated for DOD; additionally, the atmospheric DRE is found to be independent of the dust mode considered.
The results at LPZ are a kind of extension of what is observed in BCN, because the main origin of the dust intrusion is from 685 the Iberian Peninsula, which was already under Saharan dusty conditions, arriving later at LPZ on 29J. Hence, dust particles were travelling for a longer period to LPZ, potentially experiencing a more pronounced gravitational settling of the largest particles. Although the total dust radiative cooling impact is much lower in LPZ (i.e., the total dust DRE, in average, is the 27.5 % and 26 % of that in BCN, respectively, on the surface and at the TOA), the relevance of the DRE in LPZ relies on a two-fold aspect. Firstly, the two close in time dust episodes were caused by different dust air masses pathways coming from 690 the Iberian Peninsula, as described before, leading to a completely decoupled high dust layer for the first one, and a more mixed dusty environment for the second one, despite their similar columnar SSA pattern, which suggested a certain dustpollution mixing in both episodes. Indeed, this is an example of the advantage of using lidar measurements in characterizing aerosol complex scenarios: both DD and ND components were present in both episodes, but only 'mixed' in the second one.
Second, in average, the Df/DD DRE ratio on the surface is 52 %, which is higher than that 37 % found in BCN, likewise that 695 observed at the TOA, where the Df/DD DRE ratio is 60 % in LPZ and 45 % in BCN. This can confirm the gravitational settling of the largest dust particles during a longer transport, as mentioned before, and then leading to a higher contribution of Df particles to the total dust DRE in comparison with BCN. Moreover, the total dust DREff on the surface is -113.4 W m -2  -1 , in LPZ, a value higher with respect to that found in BCN (-88.9 W m -2  -1 ). That apparent increase, in absolute value, is because of the spectral behaviour of , which is slightly smaller in LPZ than in BCN, and then, at constant AOD, a larger cooling 700 effect (i.e., a larger radiative efficiency) is produced.
 A dust-induced cooling effect is observed, being the DRE efficiency higher, in absolute value, on the surface than at the top-of-atmosphere (TOA).
 Despite Dc particles usually dominate under intense dusty conditions, the contribution of Df particles to the total dust DRE, both on surface and at the TOA, can be relevant, even higher than that for Dc particles along the overpassing of the 720 dust outbreak over the station; additionally, that Df contribution is higher at the TOA than on surface.
 Consequently, although the dust cooling effect is lower in LPZ with respect to BCN, the Df contribution to the total dust DRE is higher in LPZ than in BCN because of the progressive loss of large particles by gravitational settling during their longer transport to the LPZ station.
In general, results obtained in this work are especially relevant for the next ESA EarthCARE mission (launch planned in 2022), 725 which is focused on radiation-aerosol-cloud interactions, but also for satellite remote sensing instrumentation, which is mostly sensitive to SW wavelengths, since its measurements can be likely affected by dust contamination. In addition, the determination of the dust ice nucleating particle (INP) concentration, once separated dust and non-dust components, is ongoing; this is in relation with the indirect dust radiative forcing, representing an added-value in aerosol-cloud-radiation research. At present, an additional study is on-going on the longwave (LW) direct radiative effect of both Dc and Df particles 730 on the surface and at TOA in order to assess the net radiative impact of dust during the Saharan dust intrusion described in this work.
Data availability. Part of the data used in this publication were obtained as part of the AERONET and MPLNET networks and are publicly available. For additional data or information please contact the authors.