Aerosol radiative impact during the summer 2019 heatwave produced partly by an inter-continental Saharan dust outbreak – Part 2: Longwave and net dust direct radiative effect

. This paper is the companion paper of Córdoba-Jabonero et al. (2021). It deals with the estimation of the longwave (LW) and net dust direct radiative effect (DRE) during the dust episode that occurred between 23 and 30 June, 2019, and coincided with a mega-heatwave. The analysis is performed at two European sites where polarized-Micro-Pulse Lidars ran continuously to retrieve the vertical distribution of the dust optical properties: Barcelona, Spain, 23-30 June, and Leipzig, 20 Germany, 29-30 June. The radiative effect is computed with the GAME radiative transfer model separately for the fine- and coarse-mode dust. The instantaneous and daily radiative effect and radiative efficiency (DREff) are provided for the fine-mode, coarse-mode and total dust at the surface, top of the atmosphere (TOA) and in the atmosphere. The fine-mode daily LW DRE is small (< 6 % of the shortwave (SW) component) which makes the coarse-mode LW DRE the main modulator of the total dust net DRE. The coarse-mode LW DRE starts exceeding (in absolute values) the SW component in the middle of the 25 episode which produces positive coarse-mode net DRE at both the surface and TOA. Such an unusual tendency is attributed to increasing coarse-mode size and surface temperature along the episode. This has the effect of reducing the SW cooling in Barcelona up to the point of reaching total dust net DRE positive (+0.9 W m -2 ) on one occasion at the surface and quasi-neutral (-0.6 W m -2 ) at TOA. When adding the LW component, the total dust SW radiative efficiency is reduced by a factor 1.6 at both surface (on average over the episode, the total dust net DREff is -54.1 W m -2 τ -1 ) and TOA (-37.3 W m -2 τ -1 ). A sensitivity 30 study performed on the surface temperature and the air temperature in the dust layer, both linked to the heatwave and upon which the LW DRE strongly depends, shows that the heatwave contributed to reduce the dust net cooling effect at the surface DD is concerned, most of the DD (cid:1830)(cid:1844)(cid:1831) (cid:3015)(cid:3006)(cid:3021) instantaneous values at SRF and TOA are negative and remain lower than 20 W m -2 in absolute value. The daily means are -3.0 (BCN) and -1.40 (LPZ) W m -2 at SRF and -2.3 (BCN) and -0.35 (LPZ) W m -2 at TOA. At SRF the highest daily DD (cid:1830)(cid:1844)(cid:1831) (cid:3015)(cid:3006)(cid:3021) are -7.4 and -1.5 W m -2 in BCN and LPZ, respectively. In one occasion in BCN (on 28 June) the daily DD (cid:1830)(cid:1844)(cid:1831) (cid:3015)(cid:3006)(cid:3021) at SRF takes a positive value +0.9 W m -2 , meaning that on a daily basis the dust radiative effect produces a net warming at SRF. This is quite an unusual results. Our values are

presents a sensitivity study on the relationship between heatwave (by means of surface temperature and air temperature in the dust layer) and coarse-mode longwave and total dust radiative effects. Conclusions are given in Section 4.
2 Dust radiative properties in the longwave spectral range and GAME parametrization
Gaseous absorption (H2O, CO2, O3, N2O, CO, CH4 and N2) is treated from the correlated k distribution (Lacis and Oinas, 1991). More details of the longwave module of GAME can be found in Sicard et al. (2014a). GAME presents the advantage of the complete representation of the longwave aerosol scattering, in addition to their absorption. The moderate spectral resolution of GAME makes it possible to account for the spectral variations of aerosol properties, especially in the infrared 85 window. The spectral optical properties of aerosols are defined for each atmospheric layer where dust is present: the single scattering albedo (SSA) and the asymmetry factor (asyF) are assumed constant vertically; the extinction coefficient ( ) varies with altitude. GAME outgoing (i.e. leaving the terrestrial atmosphere) longwave radiation (OLR) was validated through comparison with CERES (Clouds and the Earth's Radiant Energy System) OLR measurements in 11 cases of dust intrusion in Barcelona (Sicard et al., 2014a). Their results indicate a bias between simulated and measured OLR of -0.8 % and a root mean 90 square error of 2.52 W m -2 .
The convention followed for the definition of the dust direct radiative effect (DRE) at either the surface (SRF) or the Top-Of-the-Atmosphere (TOA) is the one of Eq. (9) of Córdoba-Jabonero et al. (2021). The atmospheric (ATM) DRE is the difference between the TOA DRE and the SRF DRE. To simplify abbreviations the shortwave, longwave and net (shortwave + longwave) DRE are noted , and , respectively. 95 GAME is used to calculate the instantaneous longwave radiative effect in a hourly basis between 5 and 19 UTC over the 8 days of the event in Barcelona (23-30 June) and over 2 days in Leipzig (29 June, Episode 1, and 30 June, Episode 2; see Córdoba-Jabonero et al., 2021). In the rest of this section the dust microphysical and radiative properties are sometimes averaged over the whole event. For Barcelona this means from 23 to 30 June or from 24 to 29 June in order to avoid the beginning and the end of the dust intrusion when abrupt changes in the microphysics and radiative properties are expected. In 100 Leipzig the dust microphysical and radiative properties are averaged over the afternoons of 29 and 30 June (when the dust was present, see Figures 5 and 6 of Córdoba-Jabonero et al., 2021), respectively noted 29J-pm and 30J-pm, corresponding to the two dust episodes described in the former reference. In all cases the averaging period is always indicated.

Dust microphysics
The dust radiative properties in the longwave spectral range were calculated using a Mie code in the range 4 -50 μm. The 105 input of our Mie code is the geometric median radius, , and its standard deviation, , of the lognormal distribution, the particle number, and the spectrally-resolved refractive index. The code calculates the extinction coefficient normalized to that at 532 nm (the wavelength of the P-MPL systems) spectrally-resolved in the range 4 -50 μm, ⁄ , the single scattering albedo and the asymmetry factor. Although the particle number is provided in input, it has no effect on our calculations since SSA and asyF are intensive parameters and the extinction is normalized to that at 532 nm, thus the dependence on the particle 110 https://doi.org/10.5194/acp-2021-419 Preprint. Discussion started: 14 June 2021 c Author(s) 2021. CC BY 4.0 License.
number is removed. We distinguish between the dust coarse-mode (Dc) and the dust fine-mode (Df). The total dust (DD) is the sum of both Dc and Df components.
The spectral refractive index (real and imaginary part) is the same than in Sicard et al. (2014a) and comes from measurements of long range transport mineral dust taken in Meppen in western Germany (Volz, 1983). The table giving the refractive index as a function of the wavelength was found in Krekov (1993). The spectral variation of both real and imaginary 115 parts of the refractive index can be seen in Figure 1 of Sicard et al. (2014a).
For each of the coarse and fine modes, the geometric median radius and the standard deviation were estimated from column-integrated AERONET retrievals. AERONET provides volume median radius, , and its standard deviation, . The geometric median radius and standard deviation were calculated as follows:  (Figure 1b). It is known that dust aging encompasses many processes that can alter the dust chemical composition, shape and size. In particular, when the transport of mineral dust occurs over polluted regions anthropogenic inorganic acids can be absorbed by the dust surface forming hygroscopic salt compounds that coat the dust particles (Abdelkader et al., 2015;Athanasopoulou et al., 2016). Dust also favors the formation of secondary pollutants (Querol et al., 2019;Xu et al., 2020b).
Both processes (acid absorption and secondary aerosol formation) can lead to dust particle growth in different size ranges.

135
Secondary aerosol formation is enhanced in stagnant (low winds) and high humidity conditions (Xu et al., 2020b). NCEP (National Centers for Environmental Prediction) reanalysis horizontal wind in BCN (not shown) reveals strong winds (~18 m   Figure 1) has been applied to the BCN retrievals of and over the period 145 24-29 June in order to avoid the beginning and the end of the dust intrusion when abrupt changes in the microphysics are expected. Coarse and fine modes have opposite tendencies: increases at a rate of almost +10 % day -1 and its scatter interval ( ) also increases (~ +2 % day -1 ); slightly decreases (at a rate of ~ -1 % day -1 ) and its scatter interval ( ) also decreases (~ -2 % day -1 ). The coarse mode radius increase and widening reflects the possible size growth mentioned earlier that might have occurred during the dust transport. This fact is also nicely illustrated by Figure 2 in which both daily Dc and 150 Df normalized size distributions are represented on a day-by-day basis. One sees clearly how the Dc radii increase between 24 and 29 June, while the Df radii decrease during the same period. Although the decreasing rate of the Df radius along the event is small (< 1 % day -1 ), its representation in a logarithmic scale (X axis) in Figure 2 shows clearly that it is noteworthy.

Dust radiative properties in the longwave spectral range
With the microphysics defined in the previous section, spectral dust radiative properties ( , SSA and asyF) in the longwave 155 spectral range are calculated with the Mie code on a hourly basis at both sites. Results are presented in Figure 3 separately at BCN and LPZ and for Dc and Df. In BCN a color code is used for each curve corresponding to the number of hours past since the beginning of the period considered: 23 June at 00 UTC. In LPZ the average of 29J-pm (Episode 1) and 30J-pm (Episode 2) is shown. The effect of the increasing along the dust event in BCN is visible on the Dc extinction plot (Figure 3a, top panel): ⁄ is smaller at the beginning of the event (blue-green curves below the mean) than at the end (red-brown 160 curves above the mean). The main differences between Dc and Df are: the Df extinction is at least three orders of magnitude smaller than the Dc one (in LPZ it is even smaller); Df SSA converges rapidly towards 0 with increasing wavelengths and so does Df asyF. Differences between BCN and LPZ are essentially due to the differences in the dust radii at each site: the mean Dc extinction in BCN is slightly lower than in LPZ ( is smaller in BCN than in LPZ) and the mean Df extinction in BCN is larger than in LPZ ( is twice larger in BCN than in LPZ). SSA and asyF are similar at both sites. The profiles of extinction coefficients at 532 nm for both coarse and fine modes were taken from the POLIPHON inversion performed by Córdoba-Jabonero et al. (2021). and are averaged in 1-km layer-mean values 170 from 0 to 10 km and then input in GAME. In each layer GAME calculates the LW spectral extinction by multiplying the measured by the calculated normalized extinction, ⁄ . Figure 4 represents the Dc and Df dust optical depth (DOD) at 532 nm at both sites. In BCN Dc and Df DOD peak at 0.420 and 0.090, respectively. In LPZ both episodes are much more

175
To get an idea of the dust stratification we plot in Figure    estimated by Papayannis et al. (2008) from lidar measurements in BCN.

195
Finally the rest of input in GAME are the atmospheric profiles and the Earth surface properties. The gas absorption is parametrized from profiles of pressure, temperature and relative humidity. In BCN, radiosoundings launched twice a day (at 00 and 12 UTC) by the University of Barcelona in collaboration with the Servei Meteorològic de Catalunya, the Catalonia meteorological agency, were used. No radiosoundings are available in LPZ, thus the 6-hour profiles from the Global Data Assimilation System (GDAS) provided by the National Oceanic and Atmospheric Administration (NOAA) were used instead.

200
The Earth surface is assumed lambertian in GAME and its albedo in the longwave spectral range was set to a constant value of 0.017. This value corresponds to the climatological value found by Sicard et al. (2014a) in Barcelona from CERES measurements in the spectral window 8.1 -11.8 μm and averaged over spring and summer seasons in the period June 2007 -May 2012. The same value was used at both sites.
We used the hourly land surface temperature (LST) V2 product provided globally at 5-km resolution by the Copernicus

205
Global Land Service (https://land.copernicus.eu/global/products/lst). The LST V2 datasets are estimated from TOA brightness temperatures from the infrared spectral channels of a constellation of geostationary satellites, Meteosat Second Generation being the one covering Europe. Its estimation further depends on the albedo, the vegetation cover and the soil moisture. Further details about the temperature retrieval can be found in Freitas et al., (2013). The LST V2 product user manual and validation reports can be found at https://land.copernicus.eu/global/products/lst?qt-lst_characteristics=5#qt-lst_characteristics. Figure

Results
The results are first discussed for the longwave spectral range and for the whole spectral range (net effect = shortwave + longwave) in terms of dust direct radiative effect (Section 3.1.1), then in terms of radiative efficiency (Section 3.1.2) and a 220 deeper analysis is performed at BCN site (Section 3.2). All shortwave magnitudes were taken from the Part 1 paper by  Table   1 and at TOA in Table 2. Both size modes produce a positive radiative effect at both SRF and TOA. Except on a couple of occasions on 25 June, at SRF is always larger than at TOA resulting in a negative atmospheric , indicator of a cooling effect of the atmosphere.

235
Independently of the atmospheric level where it is estimated, the Df is small: it is in general more than one order of magnitude smaller than Dc , and it is smaller at TOA than at SRF. In BCN LW Df instantaneous values do not exceed +0.6 W m -2 (at SRF, Figure 7) and the daily values are smaller than +0.42 W m -2 ( contribution of Df to the total dust is smaller than 5-6 % ( Figure 7c). In average over the period 24-30 June (23 June is discarded because it is the day the dust arrived in BCN) it is 3.6 and 2.5 % at SRF and TOA, respectively, and it seems to slightly decrease with time along the event (negative slope of the fittings in Figure 7c). Hence Df is expected to have a very little effect on with respect to Dc . This is also true with respect to Df as nicely illustrated in Figure 8 which shows for Dc, Df and DD the daily LW/SW DRE ratio in absolute value at SRF, TOA and ATM.

245
In the middle plot (Df) the LW/SW DRE ratio is not higher than 10.0 and 5.2 % at SRF and TOA, respectively. The average over the whole event is 6.7 and 3.7 % at SRF and TOA, respectively. Note that the decrease of the Df-to-DD ratio with respect to time (opposite of what was observed for , see Córdoba-Jabonero et al. (2021)  June) showing an increase in the morning, a maximum reach at 12 or 13 UTC and a decrease afterward. This pattern of the total dust is mostly due to the coarse mode since the fine mode is much smaller than the coarse one. The diurnal cycle of DD (or Dc) is more pronounced at TOA than at SRF. It is well known that the OLR has a marked diurnal cycle (Slingo et al., 1987) and that it is highly correlated to the surface heating and cooling due to the diurnal cycle of 280 insolation (Chung et al., 2009). For these statements are less straightforward. When holding all parameters constant except the surface temperature, Osborne et al. (2011) showed with Saharan dust measurements performed in Mauritania and Niger in June 2007 that the shape of the diurnal cycle of mimicked that of the surface temperature. In our analysis the surface temperature is also the only variable with a diurnal cycle with the shape of a cosine curve, so we believe that the cosine or inverted-V shape of DD observed on the last four days of the dust event is related to the diurnal cycle of the surface 285 temperature.
The instantaneous dust ( ) is presented as a function of time in both BCN and LPZ in Figure   9 separately for Dc, Df and DD and at SRF, TOA and in the atmosphere. Daily values of Dc, Df and DD are reported at SRF in Table 1 and at TOA in  The objective of this last paragraph is to complete for the net radiative effect the discussion about the diurnal cycle of made in the companion paper by Córdoba-Jabonero et al. (2021). Figure 10 shows the diurnal cycle of for Dc, Df and DD at SRF, TOA and in the atmosphere for 26 June. Dc and DD are also reproduced. Df is not reproduced since it is nearly equal to Df (Df is small). We find again that the diurnal cycle of the Dc and 320 DD is singular at SRF and TOA with the shape of a "W", showing two minima, one in the morning (around 6 UTC) and one in the afternoon (17-18 UTC), and a maximum at central hours of the day. This shape is the one of the diurnal cycle of (the diurnal cycle of is rather flat on 26 June) which is due to a combination of solar geometry and dust anisotropic scattering (Osborne et al., 2011;Osipov et al., 2015). We refer to Córdoba-Jabonero et al. (2021)

LPZ.
is thus partly counterbalanced by . At TOA during the central hours (11-13 UTC), when Dc reaches its 340 minimum, DD even produces a quasi-neutral effect (-0.7 < DD < -0.2 W m -2 ). At SRF, the instantaneous DD is negative and it is larger (in absolute value) than 12 W m -2 ; the daily DD (Table 1)

Dust direct radiative efficiency
The dust direct radiative efficiency (DREff) is assessed on a daily basis (and at the event scale) by calculating the best linear fit forced to 0 of the scatterplot of instantaneous values of one full day (and of the whole dust episode) of vs. .

355
First the longwave dust direct radiative efficiency is discussed at SRF and TOA: see Figure 11 showing all instantaneous vs. at SRF and TOA separately for Dc/Df and BCN/LPZ, and Table 1 and Table 2 reporting the daily LW DREff values; and then the net direct radiative efficiency is discussed compared to the SW and LW components separately for Dc, Df and DD and at SRF, TOA and in the atmosphere: see Figure 12 showing only the linear fits (for the sake of clarity) of SW, LW and net DRE components vs. , and Table 1 and Table 2 reporting the daily net DREff values. As expected 360 Df LW DREff is small at SRF and TOA and at both sites (Df LW DREff < +5.3 W m -2 τ -1 ). At the event scale at SRF Dc LW DREff in BCN (+44.3 W m -2 τ -1 ) and in LPZ (+41.3 W m -2 τ -1 ) are similar, while at TOA they differ significantly (+26.5 and +48.5 W m -2 τ -1 , respectively). There are two reasons for that difference: 1) in average the dust coarse particles are larger in LPZ than in BCN (see Section 2.1 and Figure 1) inducing larger extinction coefficients in the longwave spectral range (Figure   3a and c) and 2) the center of mass of the dust plume is much higher in LPZ than in BCN ( Figure 5) which produces higher 365 LW DRE at TOA (Dufresne et al., 2002;Sicard et al., 2014a). According to Sicard et al. (2014a), when maintaining the aerosol optical depth constant, Dc does not vary much for > 1 μm. This result lets us think that the main reason why Dc LW DREff at TOA in LPZ is higher than in BCN is related with the height of the dust plume. When Df and Dc are considered together, DD LW DREff drops down (with respect to the Dc value) to +34.9 (BCN) +28.2 W m -2 τ -1 (LPZ) at SRF and to +20.8 (BCN) and +32.6 W m -2 τ -1 (LPZ) at TOA.

370
To study the effects of each particle size mode (Dc and Df) and spectral components (SW and LW) we now look at Figure   12 which has been produced with the data from BCN for the whole event ( 1.6 with respect to the SW one. When both size modes are added together (DD, right plots of Figure 12), a reduction of the SW radiative efficiency is also observed when the LW component is taken into account. At SRF and TOA, we observe a 380 decrease of the net DD radiative efficiency with respect to the SW one of a factor 1.6: -54.1 (net) vs. -88.9 W m -2 τ -1 (SW) at SRF, and -37.3 (net) vs. -58.0 W m -2 τ -1 (SW) at TOA. Interestingly, the ratio of the total dust SW-to-NET radiative efficiency is the same (1.6) at both SRF and TOA. It is a value significantly larger than 1 which highlights again the importance of considering the longwave component in studies focused on dust radiative effects. Finally, in the atmosphere (Figure 12i

Heatwave and dust cooling/warming effect of the Earth-Atmosphere system
How the net direct radiative effect of mineral dust is modified when the dust intrusion occurs simultaneously with a heatwave?
This question is the motivation of this section which focuses on the results in BCN. We will start with some features of the June 2019 heatwave. Data provided by the Copernicus Climate Change Service show that the European-average temperature 395 for June 2019 was higher than for any other month of June on record (Copernicus, 2021). Average temperatures were more than 2°C above normal and if we consider the 5-day period 25-29 June the temperatures were 6 to 10ºC above normal, with Europe and it is during the last one, from 25 to 29 June, that the highest temperatures were reached. Towards the end of June, this anticyclone merged with a high-pressure system located over northeastern Europe that had previously produced extreme 410 temperatures, and extended towards the northeast as a strong subtropical ridge intensified by a low-pressure system located over the eastern Atlantic. This subtropical ridge injected very warm air and mineral dust from the Sahara region towards western Europe, and this air overheated during its transport while traveling over previously warmed land (Sousa et al., 2019;Xu et al., 2020a). Sousa et al. (2019) classified the heatwave of 25-29 June 2019 as a mega-heatwave (Barriopedro et al., 2011) because of its outstanding duration, intensity and spatial extent. Other climate studies involving multi-model methodologies 415 have come to the conclusion that the June 2019 heatwave had been most probably triggered by anthropogenic climate change, and that such heatwaves could become more widespread, long-lasting, and severe over Europe in the future (Ma et al., 2020;Vautard et al., 2020).
As seen in the previous section, the total dust net radiative effect is modulated by the dust coarse mode LW radiative effect. Figure 13a shows on the same plot the daily Dc LW/SW DRE ratio and the daily Df and DD at SRF for the 420 BCN site and for the period 23-30 June; Figure 13b shows the same magnitudes at TOA. In Figure 13c Figure 6 and Figure 13c). As a rule of thumb the mean anomaly over the episode is estimated to be +6 ºC.

445
For comparison the difference between the maximum daily LST (32.7ºC reached on 28 June, Figure 6) and the mean over the episode (28.8ºC) is 3.9ºC. In HW2 and HW3 6ºC have been subtracted to the temperature profiles in the dust layer. The different scenarios allow to estimate how the dust radiative effect would have been modified: HW1, if the surface temperature had not been so high; HW2, if the air had not been so warm; and HW3, if the dust episode had not been accompanied by a heatwave ("normal" surface and air temperature). Figure 14a and b show the daily DD and the daily Dc LW/SW 450 DRE ratio at SRF and TOA, respectively, for the three scenarios as well as in the heatwave conditions (this study). Figure 14c shows the daily LST (2019) Figure 14a and b) the use of the climatological LST has no 455 impact on the Dc LW/SW DRE ratio at SRF (and thus on the DD either). Contrarily the impact at the TOA is quite noticeable: towards the end of the episode, the reduction of LST in HW1 yields logically a reduction in Dc , thus a reduction in the Dc LW/SW DRE ratio and thus an amplification of the dust cooling effect at TOA. Say the other around, the effect of a high LST during the heatwave is to increase the amount of radiation which escapes to space. Since the radiation budget at SRF is nearly unchanged, the effect in the atmosphere is to reduce the heating of the atmosphere. In HW2 (blue lines 460 in Figure 14a and b) the use of a decreased air temperature in the dust layer has an impact at both SRF and TOA and of opposite sign. The decrease of the air temperature has a direct impact on the gaseous transmittance. According to Dubuisson et al. (2004) and following the correlated k-distribution (Lacis and Oinas, 1991) used in GAME, a decrease in air temperature yields a decrease in the absorption coefficient. This will result in more LW radiation propagating upward in the atmosphere and thus  coincided with a mega-heatwave. It also investigates the effect of the heatwave on the dust radiative effect. The radiative effect was calculated with the GAME radiative transfer model separately for the fine-and coarse-mode dust. The dust radiative properties in the longwave spectral range were calculated with a Mie code and particle microphysics from AERONET. The dust fine-and coarse-mode vertical distribution as well as the shortwave DRE were taken from the companion paper. Two Independently of the atmospheric level where it is estimated, the instantaneous Df are low and do not exceed +0.6 W m -2 in BCN and +0.2 W m -2 in LPZ. The respective daily values do not exceed +0.42 and +0.08 W m -2 . In average over the whole event the contribution of Df daily to the total dust is not higher than 4 %. Most of the dust is thus produced by the coarse-mode particles. Instantaneous Dc at SRF in BCN reaches a maximum of +11.4 505 W m -2 and the daily values a maximum of +9.8 W m -2 . In LPZ the instantaneous and daily values reach maxima at +3.3 and +1.3 W m -2 , respectively. Dc at TOA is smaller than at SRF which produces a negative (cooling) effect in the atmosphere. Df in BCN represents 6.7 (3.7) % at SRF (TOA) of its SW counterpart. The Dc LW/SW DRE ratio in BCN increases from ~80 % at the beginning of the episode up to 170 % towards the end of it. Such an unusual tendency is attributed to increasing coarse-mode size and surface temperature along the episode. In the last four days of the episode Dc 510 is larger than Dc (Dc LW/SW DRE ratio > 100 %) at both SRF and TOA. This is a singular result of this study which has the effect of reducing considerably the SW cooling.
The results of the dust net DRE are discussed in terms of daily values which are the magnitude that matters for assessing the effect of aerosols on the Earth-Atmospheric radiative budget. The fine mode is similar to since ≪ . In BCN Dc at SRF is negative at the beginning of the event and positive afterwards (peak at 515 +3.1 W m -2 ). In LPZ Dc is slightly negative (mean of -0.25 W m -2 ). At TOA the same tendency is observed in BCN (peak at +3.1 W m -2 ). But in LPZ Dc is also positive at TOA (mean of +0.50 W m -2 ). This stronger effect of the dust at TOA (vs. SRF) in LPZ is related with the quite high dust plume in LPZ (> 3.3 km). Overall, the total dust net DRE is even positive (+0.9 W m -2 on 28 June) indicating a total dust net warming at the surface, which contrasts with the 520 "traditional" dust cooling effect usually observed in clear sky conditions.
As far as dust direct radiative efficiency is concerned, Df LW DREff is small so that Df net DREff is nearly equal to Df SW DREff. Dc LW DREff is positive which contributes to decrease significantly the net DREff with respect to the SW one.
Overall, the total dust net DREff is: -54.1 (-37.3) W m -2  -1 at SRF (TOA) in .2) W m -2  -1 in LPZ. The higher dust direct radiative efficiency in LPZ (vs. BCN) is due to the SW component, already discussed in the companion 525 paper. At TOA the difference between of DD net DREff between both sites is caused by the larger compensation of the LW component in BCN (vs. LPZ). Finally it should be noted that, in BCN at the event scale and for the total dust, the inclusion of the LW DRE in the calculation of the total net DREff reduces a factor 1.6 the SW DREff at SRF and BOA and a factor 1.8 in the atmosphere. These reduction factors significantly larger than 1 highlight the importance of considering the longwave component in studies focused on mineral dust radiative effects.

530
In order to evaluate the impact of the heatwave that accompanied the dust intrusion on the dust radiative effect, we perform a sensitivity study on the surface temperature and the air temperature in the dust layer, both linked to the heatwave and upon which the LW DRE strongly depends. Three scenarios are considered: 1) the LST is set to climatological values, 2) the air temperature in the dust layer is reduced of 6 ºC, and 3) a combination of the first two scenarios. Our findings show that the increase of LST and air temperature in the dust layer caused by the heatwave 1) provoked a reduction of the dust net 535 cooling effect at the surface, 2) left unchanged the dust net cooling effect at TOA, and 3) consequently reduced the dust net heating of the atmosphere. The situation at the surface is a vicious circle: the heatwave reduces the dust cooling effect which, in turn, may increase some critical variables associated to the heatwave (e.g., LST and air temperature). The effect of the heatwave on the dust radiative effect is reverse as it contributes to cool down the atmosphere. Since recent studies have warned that mega-heatwaves such as the one studied in this work might become more frequent in the future, the novel results presented 540 in this paper call for more research on the subject.
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.