Dust Aerosol Radiative Effects During a Dust Event and Heatwave in Summer 2019 Simulated with a Regional Climate Atmospheric Model over the Iberian Peninsula

Abstract

Mineral dust particles significantly influence the Earth’s climate through direct and semi-direct radiative effects. This study investigates these effects and their meteorological impacts during a dust intrusion and heatwave over the Iberian Peninsula in summer 2019 using a regional climate model. Three simulations with different spectral nudging configurations are evaluated. During the central period, the mean direct and semi-direct radiative effects in the shortwave spectrum at the top of the atmosphere (bottom of the atmosphere) are −0.4 ± 0.4 (−3.9 ± 2.3) Wm⁻² and +0.1 ± 1.7 (−0.1 ± 1.9) Wm⁻², respectively. In the longwave spectrum, these effects are +0.1 ± 0.1 (+0.3 ± 0.1) WmWm⁻² and 0.0 ± 0.6 (+0.9 ± 1.1) Wm⁻², respectively. The semi-direct effect mitigates 18.8% of the dust-induced warming in the full atmosphere and alters meteorological variables. The liquid water path decreases by −0.2 ± 4.5 mg m⁻², the cloud fraction in the upper (lower) troposphere reduces (increases) by −0.2 ± 1.2 (+0.1 ± 1.3) %, and the near-surface air temperature drops slightly by −0.2 ± 0.2 °C. The results highlight substantial spatial variability and underscore the importance of considering semi-direct radiative effects in radiative analysis.

1. Introduction

The Mediterranean basin is one of the European regions with the highest aerosol loads due to air masses carrying many different aerosol types, especially mineral dust, because of its proximity to the Saharan Desert [1,2]. Atmospheric dust particles are generally mobilized in the Saharan Desert and the Middle East during the local dry season from October to April [3], and they pass over the Eastern Mediterranean and Southern Europe at heights generally not greater than 6 km above sea level [4]. Dust aerosols influence a wide range of physical, chemical and biochemical atmospheric processes, from the marine and terrestrial biosphere to human life. Dust storms transport nutrients such as phosphorus and trace elements such as iron, manganese, titanium and aluminium, which supply micronutrients to ecosystems [5,6] and can enhance their bioproductivity. Mineral dust is thus one of the main drivers of oceanic primary productivity, forming the basis of the marine food web. In contrast, the effect of dust on human health is highly damaging, as dust particles can pass easily into the lungs and bloodstream, where they can increase the risk of dying from heart disease, stroke, lung cancer, chronic obstructive pulmonary disease and lower respiratory infections [7,8]. Moreover, if a dust event occurs during a heatwave, its harmful effects on human health may be intensified. The coupling between dust intrusions and heatwaves is becoming more frequent. Since 2000, the percentage of dust events accompanied by a heatwave has reached 39% and 14% over the Western and Eastern Iberian Peninsula, respectively [9].

Atmospheric dust particles cause direct, indirect and semi-direct radiative effects on the climate. The direct radiative effects consist of the absorption and scattering of radiation [10,11]. Indirect radiative effects are related to changes in cloud properties by affecting the cloud albedo, lifetime and precipitation processes [12,13]. The semi-direct radiative effect is associated with the rapid adjustments induced by dust aerosol radiative effects on the surface energy budget, the atmospheric profiles and cloudiness. Dust absorption can cause the atmosphere to warm and thus affect cloud dynamics, leading to cloud dissipation [14,15]. This effect depends on the height of the dust aerosol relative to the cloud and the type of cloud [14,16,17]. Atmospheric dust particles can also modify the atmospheric stability in the boundary layer and the free troposphere by either suppressing or enhancing convection [18,19]. To date, there is limited confidence in accurately determining the sign and magnitude of rapid global-scale adjustments caused by absorbing aerosols such as Saharan dust, as the current models differ in their responses and are known to under-represent some of the most important relevant cloud processes [20]. Additionally, the significant spatial and temporal variability in these adjustments, together with the need to use regional climate atmospheric models to quantify them, further complicates the estimation of semi-direct effects.

An outstanding heatwave event coupled with a Saharan dust intrusion occurred in June and July 2019. This 14-day event set new temperature records in several places in Europe. In particular, the month of June 2019 was noted as one of the hottest months ever recorded at the European level [21]. According to a detailed article issued by the Spanish State Meteorological Agency [22], the average temperatures were more than 2 °C above normal, and, if we consider the 5-day period of 25–29th June, the temperatures were 6 to 10 °C above normal, with local differences being even higher in Northeastern Spain, France and the United Kingdom. In addition, the heatwave of summer 2019 was classified as a mega-heatwave in [9,23], being the subject of several studies. Refs. [23,24] quantified the role of dynamical and thermodynamical processes in triggering this extreme event and investigated how changes in these processes observed over the last few decades may have affected the occurrence probability of such extreme events. Refs. [25,26] used climate models to quantify the role of the anthropogenic contribution. Ref. [9] studied the relationship between heatwaves and Saharan warm air intrusions in the Iberian Peninsula (IP) in the long-term context. Ref. [27] calculated the shortwave dust direct radiative effect during 23–30th June 2019 at two sites in Barcelona, Spain and Leipzig, Germany. Their companion paper [28] described the longwave and net dust direct radiative effects.

The objective of this paper is to extend previous [27,28] works by analyzing this dust and heatwave event with simulations of the regional climate model called CNRM-ALADIN64 (ALADIN, Aire Limitée Adaptation dynamique Développement InterNational) over the Iberian Peninsula. Taking advantage of working with a regional climate atmospheric model, special attention is paid to the semi-direct effects of dust aerosols and their responses regarding the local meteorology. The instrumentation used is explained in Section 2. The CNRM-ALADIN64 model is presented in Section 3. An evaluation of the CNRM-ALADIN64 simulations with dust aerosols is carried out against different observations and previous studies in Section 4. The results are shown in Section 5, and the conclusions are presented in Section 6.

2. Observational Data

2.1. Aerosol Robotic Network

The NASA Aerosol Robotic Network (AERONET https://aeronet.gsfc.nasa.gov/, accessed on 8 September 2024) is a collaborative network of ground-based sun/lunar photometers established by NASA and LOA-PHOTONS (CNRS). For over two decades, AERONET has provided a long-term, continuous and publicly accessible database of aerosol optical and microphysical properties. This database supports aerosol characterization research, satellite retrieval validation and synergism with other observational datasets. The network ensures the standardization of the instruments, calibration procedures, data processing and distribution.

In this work, Version 3.0 and Level 2.0 (cloud-screened and quality-assured, with pre-field and post-field calibration applied) solar Aerosol Optical Depth (AOD) products are used for the time period from 20th June to 5th July 2019. To validate the hourly AOD at 550 nm obtained from the CNRM-ALADIN64 model, a two-step process was implemented. First, the AERONET Ängstrom exponent, calculated at 440–675 nm wavelengths, was applied to the AERONET AOD measurements at 675 nm [29] to calculate an AERONET-equivalent AOD at 550 nm. Second, as the sun photometer provides instantaneous measurements, hourly averages were calculated from the instantaneous data to ensure temporal agreement. The spatial distribution of the AERONET stations over an orographic map of the Iberian Peninsula is shown in Figure A1, included in Appendix A.

2.2. Visible Infrared Imaging Radiometer Suite

The VIIRS instrument (https://ladsweb.modaps.eosdis.nasa.gov/missions-and-measurements/viirs/, accessed on 8 September 2024) is one of five instruments aboard the Suomi National Polar-Orbiting Partnership (Suomi NPP) satellite, launched in 2017. VIIRS is a visible and infrared radiometer equipped with 22 spectral bands ranging from 0.412 to $12 \mathsf{\mu }$m. These include 9 visible/near-infrared bands plus a day/night pan band and 8 mid-infrared and 4 longwave infrared bands. The VIIRS products, managed by the Level-1 and Atmosphere Archive & Distribution System Distributed Active Archive Center (LAADS DAAC), contribute to advancing our understanding of global climate change by enabling measurements of cloud and aerosol properties, ocean and land surface temperatures, surface albedo, wildfires and more [30].

In this study, we use the NOAA20 VIIRS Deep Blue Level 3 daily aerosol dataset, which has a horizontal grid resolution of 1° × 1°. This product provides satellite-derived measurements of the AOD at 550 nm over both land and oceans as gridded aggregates with daily global coverage [31].

2.3. Micro-Pulse Lidar Network

The NASA Micro-Pulse Lidar Network (MPLNET, Madison, WI, USA) is a global network of micro-pulse lidar (MPLNET lidar, http://mplnet.gsfc.nasa.gov (accessed on 14 May 2025)) systems designed to measure the vertical structure of clouds, aerosols and boundary layer heights [32,33]. All MPLNET sites currently use the MPLNET lidar, which was developed at NASA’s Goddard Space Flight Center (GSFC, Greenbelt, MD, USA) in the early 1990s. The MPL instrument was patented and later licensed for commercial production in the mid-1990s. These instruments collect data continuously, both day and night, over long time periods, from locations worldwide. Many MPLNET sites are co-located with AERONET stations. MPLNET data have supported numerous studies and applications, including domestic and international aerosol and cloud research [34,35] and climate change and air quality investigations [36], as well as providing support for NASA satellite missions and aerosol modelling and forecasting [37].

The lidar system used in this study is a polarized micro-pulse lidar (P-MPLNET lidar) integrated within NASA’s MPLNET. The Barcelona MPLNET lidar is located on the roof of the Remote Sensing, Antennas, Microwaves and Superconductivity Group (CommSensLab, https://ors.upc.edu/, accessed on 8 July 2024) building in Campus Nord of the Universitat Politècnica de Catalunya (41.38°N, 2.11°E; 115 m a.s.l.), approximately at 1 km from Serra de Collserola and 7 km from the sea.

In this study, we employed the Aerosol (AER) product, Version 3 (V3, released in 2021) and Level 1.5 (L15, near-real-time, quality-assured). This product provides data with a 1 min temporal resolution and a 75 m vertical resolution, enabling the characterization of the aerosol layer closest to the surface. The MPLNET AER product includes sun/lunar AOD measurements derived from photometer observations, which are used to invert lidar signals and retrieve aerosol properties during both daytime and nighttime. Key properties included in the MPLNET AER product are aerosol extinction, backscatter, the column lidar ratio and other relevant properties [34,35].

2.4. Solar Radiation Network

The Solar Radiation Network (SolRad-Net, https://solrad-net.gsfc.nasa.gov/, accessed on 8 September 2024) is a network of ground-based sensors that provides continuous high-frequency solar flux measurements in near-real time for the scientific community. Designed as a companion program to AERONET, the SolRad-Net instrument is co-located with AERONET locations. Each SolRad-Net site is initially equipped with two flux sensors: a Kipp and Zonen CM-21 pyranometer (0.305–$2.8 \mathsf{\mu }$m) for measuring the total solar spectrum and a Skye Instruments SKE-510 PAR (Photosynthetically Active Radiation) energy sensor (spectral range: 0.4–$0.7 \mathsf{\mu }$m).

In this study, measurements from the Kipp and Zonen instrument were employed at Level 1.0, which corresponds to unscreened data without final calibration, for the time period from 20th June to 5th July 2019. The Kipp and Zonen CM-21 units are ISO 9060 [38] Secondary Standard thermopile pyranometers, featuring a receiving element protected by two concentric Schott K5 glass domes. Additional information about this instrument is available at (https://www.kippzonen.com/Product/14/CMP21-Pyranometer).

2.5. Clouds and the Earth’s Radiant Energy System Project

The Clouds and the Earth’s Radiant Energy System (CERES, https://ceres.larc.nasa.gov/, accessed on 8 September 2024) project provides observations of the Earth’s radiation budget through measurements obtained from CERES instruments aboard the Terra, Aqua and Suomi National Polar-Orbiting Partnership (S-NPP) and NOAA-20 (formerly JPSS-1, named after the Joint Polar Satellite System-1 mission) satellites [39]. The primary goals of the CERES project are (1) to produce a long-term and integrated global climate data record for detecting decadal changes in the Earth’s radiation budget from the surface to the top of the atmosphere; (2) to enhance the understanding of the temporal and spatial variability in the Earth’s radiation budget and the role played by clouds and other atmospheric properties; (3) to support climate model evaluation and improvement through model–observation intercomparisons.

The processed dataset in this study is the CERES instantaneous Single Scanner Footprint (SSF) product at Level 2. Specifically, we use the observed top-of-the-atmosphere upward fluxes in the longwave spectral range (5–$35 \mathsf{\mu }$m), for the time period covering 20th June to 5th July 2019. The CERES measurements were either from the AQUA satellite (overpass over Barcelona between 12:00 and 13:00 UTC) or from the TERRA satellite (overpass over Barcelona between 10:00 and 10:30 UTC). This variable is provided at a spatial resolution of 20 km at nadir [40].

2.6. Meteorological Service of Catalonia

Radiosondes are launched twice daily (at 00:00 and 12:00 UTC) by the Meteorological Service of Catalonia (Meteocat), with the technical support of the Faculty of Physics at the University of Barcelona, at a location less than 1 km away from the MPLNET lidar site. The radiosondes provide vertical profiles of the pressure, altitude, temperature, relative humidity, dew point, wind speed and wind direction. In this study, only the altitude, pressure, temperature and dew point profiles are used. The specific humidity has been calculated from the water vapor and specific humidity equations, using the pressure and dew point variables provided by the radiosondes.

3. The CNRM-ALADIN64 Regional Climate Model

The present study is carried out with the regional climate atmospheric model CNRM-ALADIN64 [41]. We use here version 6.4 of ALADIN, which has a similar physical package to the global climate model ARPEGE–Climate [42] used in the Coupled Model Intercomparison Project Phase 6 (CMIP6) initiative [43]. It is a bi-spectral semi-implicit semi-Lagrangian model, with a 12.5 km horizontal resolution. ALADIN includes a longwave (LW) radiation scheme based on the rapid radiation transfer model (RRTM), which divides the spectrum into 16 bands covering the range from 10 to 3250 cm⁻¹ (equivalent to approximately 3.08 to 1000 μm) [44] and a shortwave (SW) scheme initially developed by [45], which has a finer spectral resolution of six bands, covering from 0.2 to 4 μm. Both schemes incorporate the effects of greenhouse gases and the direct and semi-direct effects of seven types of aerosols (sea salt, desert dust, sulfates, black carbon, organic matter, nitrate and ammonium aerosols) [46]. ALADIN is able to use a spectral nudging method, described in [47], which enables it to keep large scales from the boundary forcing and thus impose the true natural climate variability that is essential to accurately represent dust events. In this study, three types of simulations are performed with different spectral nudging: SN-All with spectral nudging for the temperature, humidity, surface pressure and wind; SN-Wind, keeping only the spectral nudging for wind; and SN-Non without spectral nudging. Spectral nudging serves to constrain the large-scale meteorological evolution within the simulation [48]. The function used imposes a constant rate above 700 hPa and a relaxation zone between 700 and 850 hPa, while the levels below 850 hPa are free. The spatial wavelengths are similarly nudged beyond 400 km, with a relaxation zone between 200 and 400 km. Thus, this method gives the model enough freedom to generate the aerosols at the surface while keeping the large-scale conditions that are essential to simulate the real chronology. More information about the other parameterizations used is presented in [41]. The model lateral boundary conditions are provided by the ERA5 reanalysis [49]. An ensemble of 2 simulations for each configuration (SN-All, SN-Wind and SN-No) has been carried out considering the presence and absence of atmospheric dust particles in the atmosphere. The semi-direct effect is calculated as the difference between the simulations with and without dust particles and the dust direct radiative effect. The simulations cover the period from 20th June to 5th July 2019, during a dust event and heatwave that hit the Iberian Peninsula. The simulation domain covers the Mediterranean basin and surrounding regions (large parts of Europe and the Saharan Desert), as in [41]. However, the analysis presented here is restricted to a smaller area, covering the Iberian Peninsula and extending over latitudes between 33 and 47° and longitudes between −12 and 5°. The objective is to analyze the direct and semi-direct radiative effects of dust aerosols over the Iberian Peninsula and to disentangle the effects of the dust event from those of the heatwave.

4. Evaluation of ALADIN Simulations

In this section, an evaluation of the simulation with dust aerosols during the dust event and heatwave in summer 2019 is carried out against different available observations and simulations performed with a 1D radiative transfer model [50,51]. In particular, the total AOD over the Iberian Peninsula, the vertical distribution of dust extinction, the aerosol optical scattering properties, the shortwave downward (SW DW) and longwave upward (LW UP) radiative fluxes and the direct radiative effects calculated in [27,28] at the Barcelona lidar station are evaluated.

4.1. Total Aerosol Optical Depth

The temporal evaluation of the total AOD at 550 nm obtained from the ALADIN simulations that consider the Saharan dust aerosols are first evaluated against AERONET and VIIRS observations in different locations over the Iberian Peninsula, as shown in Figure 1.

Figure A1, presented in Appendix A, depicts the spatial distribution of the AERONET stations over an orographic map of the Iberian Peninsula. As shown in Figure 1, a slight discrepancy is observed between the AERONET and VIIRS measurements. VIIRS provides daily observations with a lower spatial resolution, making it more difficult to represent the more complex and variable AOD patterns that AERONET captures. Specifically, VIIRS underestimates the daily AOD measured by AERONET by −1.7%, a phenomenon also noted in previous studies like [52,53]. This comparison was performed by averaging the daily AOD data measured by AERONET and considering only the days when both datasets were available. Despite the differences between satellite-based and ground-based observations, on average, this underestimation remains relatively small. The ALADIN model best represents the dust event at locations like Badajoz, Valladolid and Murcia. The model successfully reproduces the dust event in the Eastern Peninsula, where the dust load was higher, while also performing reliably in regions with a flat orography, such as the center and the Western Peninsula, independently of the dust load. The influence of orography on the accuracy of the simulations has been highlighted in previous studies [54,55], and this effect is evident in this study for locations such as Barcelona and Zaragoza. Regarding the temporal evolution, the largest discrepancies in the AOD between the model and the observations are observed at the beginning and final stages of the dust event.

In order to continue the analysis and identify which simulation best reproduces the dust event, the scatter plots of the total AOD at 550 nm simulated with the ALADIN model and measured by AERONET and VIIRS observations over the Iberian Peninsula are shown in Figure 2.

In Figure 2, it can be observed that the ALADIN simulations fit better with the AERONET measurements than with the VIIRS observations. Generally, ALADIN slightly underestimates the AOD with the simulations performed with spectral nudging and overestimates the AOD with the simulation without it. When evaluating the AOD simulated against AERONET observations, the highest R² calculated is 0.52 for the S-Dust-SN-All simulation, which is lower than the ∼0.75 reported in [55]. In comparison, the results with VIIRS are slightly worse, showing a bias that is approximately 5% lower than that calculated with the AERONET measurements. The highest R² in the AOD evaluation using VIIRS observations is 0.24 for the S-Dust-SN-All simulation, being also significantly lower than the ∼0.75 reported in [55]. In both evaluations, the S-Dust-SN-All simulation, which incorporates spectral nudging on multiple meteorological variables, shows the best agreement with the measured AOD. However, the S-Dust-SN-Wind simulation also performs well, yielding similar results to those of S-Dust-SN-All, suggesting that the spectral nudging on the wind plays a particularly influential role in the simulations. The poorer performance in the AOD simulation by the ALADIN model with respect to [55] might be attributed to the fact that this study evaluates simulations for a specific dust event lasting only 16 days, rather than an entire summer as in [55].

In brief, ALADIN is able to accurately represent the complex AOD pattern during the dust event, and the simulations that best fit the AERONET and VIIRS observations are those with spectral nudging (S-Dust-SN-All and S-Dust-SN-Wind). The simulation without spectral nudging (S-Dust-SN-Non) differs the most from all simulations, overestimating the AOD mainly at the center and the end of the dust event.

4.2. Dust Extinction

As the dust direct and semi-direct effects also depend on the vertical distribution of the dust particles, the vertical distribution of the simulated dust extinction at 532 nm is evaluated. Specifically, the average dust extinction from the simulation that includes Saharan dust aerosols is compared with the MPLNET product at the Barcelona lidar station, as shown in Figure 3. To isolate the dust contribution, the POLIPHON algorithm was applied to the MPLNET data [27,28].

The dust outbreak that struck Europe in June and July 2019 was preceded by a high-pressure system and extreme heat over Northeastern Europe. By 24th June, the low-pressure system located over the Eastern Atlantic had stretched and intensified the subtropical ridge. The deepening of the Atlantic low occurred simultaneously with the intensification and westward shift of a Siberian low-pressure system towards Scandinavia by the end of June. The last stage of the low-pressure system featured a transient evolution towards Southeastern Europe, until it dissipated [9]. On 20th June, the AERONET station in Barcelona recorded an average AOD of 0.2, which increased to 0.35 on 24th June. The average daily AOD value for the whole period was 0.21. The dust particles reached Barcelona at altitudes up to 6 km, with an average dust extinction value of 12 Mm⁻¹ within the first 6 km. Figure 3 shows that the height of the dust particles is similar to that simulated by the ALADIN model. However, the dust extinction estimated by the ALADIN model shows overestimation in the upper layers and underestimation in the lower layers, as seen in other studies [55]. The dust extinction coefficient simulated without spectral nudging (S-Dust-SN-Non) shows the poorest agreement with the MPLNET measurements, resulting in a bias of +65.9% and an R² = 0.89. The best results are obtained with the simulation with spectral nudging only for the wind (S-Dust-SN-Wind), with a bias = +7% and an R² = 0.78. According to these results, the slight relaxation of the spectral nudging in the ALADIN simulations favours the accurate representation of the vertical distribution of the dust particles. However, the absence of spectral nudging in the simulations does not lead to a significant improvement in the results obtained. In summary, the ALADIN model has demonstrated its ability to represent accurately the vertical distribution of dust aerosols. However, it is important to note that relying on a single vertical profile is insufficient to draw definitive conclusions about the ability of the model to estimate the dust vertical distribution. This type of comparison should be performed in other locations to validate the findings.

4.3. Optical Scattering Properties

As the radiative effects also depend on the optical scattering properties of the dust particles, the daily total aerosol asymmetry factor and single scattering albedo at 550 nm obtained from the simulation that considered the Saharan dust aerosols are evaluated against AERONET observations over the Iberian Peninsula, as shown in Figure 4.

In Figure 4, it can be discerned that the values of the asymmetry factor are between 0.6 and 0.8. The mean and standard deviation of the AERONET observations are 0.70 ± 0.03, being characteristic of dust aerosols [56,57]. The asymmetry factor simulated by ALADIN generally does not follow the same trend as that measured by AERONET, presenting a scatter pattern with low R² values. In contrast, negative low bias and RMSE values are obtained for all simulations, with the simulation with spectral nudging only for wind (S-Dust-SN-Wind) being the one that best represents the asymmetry factor measured by AERONET. The values of the single scattering albedo are found between 0.88 and 1. The mean and standard deviation of the AERONET observations are 0.95 ± 0.02, being characteristic of dust aerosols [56,57,58,59,60]. The single scattering albedo also does not fit the trend observed by AERONET, reflecting an R² close to 0. Despite this, the simulations show small bias and RMSE values, also reflecting the slight underestimation of this property. The simulations with the best single scattering albedo are those with spectral nudging (S-Dust-SN-All and S-Dust-SN-Wind). In summary, the ALADIN model reproduces well the optical scattering properties of aerosols but is unable to reproduce the daily trends that AERONET observes. Daily variations in anthropogenic emissions could be responsible for this mismatch between the simulations and observations.

4.4. Radiative Fluxes

The downward and upward radiative fluxes simulated with the ALADIN model are evaluated against observations from a pyranometer belonging to the SolRad-Net network located at the Barcelona lidar station in a spectral range of SW (0.305–2.8 μm) at the bottom of the atmosphere and against CERES measurements over the Iberian Peninsula in a spectral range of LW (5–35 μm) at the top of the atmosphere, as shown in Figure 5.

The cloud mask is created with AERONET data and the low-troposphere cloud area fraction product from the ALADIN model. If there are no AERONET data or the cloud area fraction in the low troposphere is greater than +1%, the atmosphere is considered to be cloudy at that time. In the case of the evaluation over the Iberian Peninsula, only the cloud area fraction product in the low-troposphere is considered. Figure 5 shows generally good agreement between the simulations performed by the ALADIN model and the observations measured by the pyranometer at the surface in Barcelona. It can be seen that the ALADIN model often underestimates the shortwave downward radiation at the surface for all simulations. These cases occur mostly in the early hours of the day (6–10 UTC), during the time period from 20 to 25th June, revealing the difficulties of the model in representing the sunrise period during the first few days of the dust event. For the simulation with the most restrictive spectral nudging (S-Dust-SN-All), the evaluation is the best, with an ${\mathrm{R}}^2$ = 0.94 and a bias <+1%. In contrast, for the other two simulations, the ALADIN model, on average, modestly underestimates the shortwave downward radiative flux at the surface, with the bias reaching −6% for the simulation without spectral nudging (S-Dust-SN-Non). The evaluation of the longwave radiative flux at the top of the atmosphere is not as straightforward as that of the shortwave downward radiative flux at the surface, resulting in considerably worse agreement between the simulations and the observations made by the CERES satellite. While, in many cases, they cluster closely around the unit slope curve, others are more widely scattered. Nonetheless, they collectively exhibit a linear dependence. The mean and standard deviation of the observations are 266 ± $43 {\mathrm{W/m}}^2$. The linear regression has quite high values for the slope and ${\mathrm{R}}^2$, resulting in the best simulation being the one with the most restrictive spectral nudging (S-Dust-SN-All), with a = 0.99 and ${\mathrm{R}}^2$ = 0.63. However, the biases of the simulated longwave upward radiative fluxes at the top of the atmosphere are not very large, being −5.6% for the most restrictive spectral nudging simulation (S-Dust-SN-All) and −12.6% for the worst simulation (S-Dust-SN-Non). This mismatch between the simulations and observations may be caused by the consideration of a larger land area—in this case, the Iberian Peninsula—or by the fact that the CERES observations are set in a 0.1 × 0.$1°$ area around the simulation grid point and not exactly in the same location. Moreover, the CERES satellite could observe a slightly different atmospheric situation and may cover part of the Mediterranean Sea; additionally, the accurate representation of water vapor in the atmosphere remains a current challenge [61]. Nevertheless, it can be concluded that the simulation that best fits the observations is the one with the most restrictive spectral nudging (S-Dust-SN-All).

4.5. Direct Radiative Effects

Having demonstrated the ALADIN model’s capability to reproduce, with high accuracy, the temporal evolution, the vertical distribution, the optical scattering properties of dust particles and the radiative fluxes, we proceed to evaluate the simulations of the dust direct radiative effect. This assessment is performed with the results shown in [27,28] for the Barcelona lidar station during the main event period (from 23rd to 30th June), under clear-sky conditions. This calculation is conducted using a radiative transfer model called GAME [50,62].

In Figure 6, it can be observed that there is a large difference in the dust direct radiative effects simulated with the ALADIN and GAME models, particularly in the longwave spectrum. These discrepancies may arise from the different methods used by each model to solve the radiative transfer equations and the specific inputs employed in the simulations. As seen in the previous sections, the AOD and optical scattering properties of the aerosols simulated by ALADIN are similar to those measured by AERONET, although they are not identical. Moreover, the LW dust optical properties (which have not been evaluated in this study) may also differ. These discrepancies could partly explain the variations observed in the dust direct radiative effects. Additionally, differences in the atmospheric profiles, surface albedo and surface temperature further contribute to these variations. Notably, the albedo input in the ALADIN and GAME simulations differs by 0.01 in both spectral ranges. More pronounced differences in the LW direct radiative effect may also be caused by the difference in the spectral range of the simulations and by the complexity of the correct representation of water vapor in the atmosphere, as seen in the evaluation of the longwave upward radiative flux at the top of the atmosphere with CERES. For instance, the GAME model uses radiosonde observations to characterize water vapor, while the ALADIN model relies on ERA5 reanalysis data. Now, we proceed to evaluate each component.

The SW dust direct radiative effect at the top of the atmosphere (TOA) ranges from 0 to −$10 {\mathrm{Wm}}^{-2}$ and that at the bottom of the atmosphere (BOA) ranges from 0 to −$34 {\mathrm{Wm}}^{-2}$. These negative direct radiative effects are more pronounced at BOA than at TOA, producing the net warming of the full atmosphere, characteristic of Saharan dust intrusions [55,58,59,60]. The minima in the SW direct radiative effect correspond to the maxima in the AOD, since the direct radiative effect increases with the aerosol load. At TOA, ALADIN underestimates the dust direct radiative effect, whereas, at BOA, it overestimates it. The minimum bias between ALADIN and GAME at TOA occurs with the simulation with the most restrictive spectral nudging (S-Dust-SN-All), showing a bias of −64.5%. Conversely, at BOA, the simulation with spectral nudging only for the wind (S-Dust-SN-Wind) achieves the minimum bias of +50.2%. The LW direct radiative effect at TOA ranges from 0 to $6 {\mathrm{Wm}}^{-2}$, and, at BOA, it ranges from 0 to $10 {\mathrm{Wm}}^{-2}$. These calculations indicate the weak warming of the atmosphere at both levels, characteristic of non-absorbing dust particles [28,59]. This warming is slightly more pronounced at BOA than at TOA, being consistent with findings from other studies [28,59,60]. At both levels, ALADIN underestimates the dust direct radiative effect. The minimum bias between ALADIN and GAME at TOA occurs with the simulation with the most restrictive spectral nudging (S-Dust-SN-All), with a bias of −90.5%. Conversely, at BOA, the simulation without spectral nudging (S-Dust-SN-Non) achieves the minimum bias of −83.3%. In summary, the dust direct radiative effect differs considerably from that obtained with the GAME model, but both models give reasonable magnitudes of the radiative effect for a Saharan dust intrusion, in agreement with the literature [27,28,55,57,58,59,60,63,64]. In this study, the LW dust direct radiative effect modeled by ALADIN represents between 16 to 24% at the TOA and between 6 and 7% at the BOA of the SW component. In contrast, the dust radiative effects calculated with the GAME model show significantly higher values, being 62% and 69% at TOA and BOA, respectively. These results indicate that the LW direct radiative effect also has a substantial impact on the Earth’s radiative balance, comparable to that of the SW radiative effect, and should not be overlooked [64].

4.6. Atmospheric State

The atmospheric state is evaluated using meteorological variables such as the temperature, pressure and specific humidity, alongside radiosonde data at the Barcelona lidar station. As radiosonde data are only available at 00 and 12 UTC, this assessment has been restricted to these hours. The temporal evolution of the near-surface air temperature and pressure is evaluated, as well as the average of the vertical profiles of the temperature and specific humidity, as shown in Figure 7.

Figure 7 shows good agreement between the near-surface temperature simulated by the ALADIN model and that measured by the radiosondes. During the whole event, the average near-surface air temperature measured by the radiosondes is 27 ± $4 °$C. The ALADIN model estimates it at noon very well, but generally underestimates it at midnight. The simulation that best fits the near-surface air temperature measured by the radiosondes is the simulation with the most restrictive spectral nudging (S-Dust-SN-All), with a bias of −5.3%, as expected due to the fact that, in this simulation, the spectral nudging includes the temperature. The surface pressure simulated by the ALADIN model shows a similar trend to that measured by the radiosondes, with an ${\mathrm{R}}^2$ = 0.89 and a bias = −0.2% for the best simulation, which has the most restrictive spectral nudging. The average pressure measured by the radiosondes is 1018 hPa.

To complete the evaluation of the atmospheric conditions, the vertical profiles of the temperature and specific humidity are evaluated with radiosonde data at 00 and 12 UTC, and the mean error of all vertical profiles is calculated for levels between 300 and 1000 hPa. On the one hand, the mean error of the temperature is between −2 and +$0.5 °$C. The levels with the highest error are the levels closest to the surface, between 900 and 1000 hPa, and the medium levels, between 500 and 600 hPa, corresponding to the heights of Saharan dust intrusions in Barcelona [27]. In general, ALADIN tends to underestimate the temperature, being more pronounced with the simulation without spectral nudging (S-Dust-SN-Non). The simulation that has only spectral nudging for the wind (S-Dust-SN-Wind) best represents the temperature, with a mean bias of −0.6% for all levels. On the other hand, the mean error of the specific humidity varies between −0.004 and +0.001, being a wider range than for the temperature. In general, ALADIN underestimates the specific humidity for levels below 900 hPa and tends to overestimate it slightly for levels above 900 hPa. The simulation that best reproduces, on average, the specific humidity is that with the most restrictive spectral nudging (S-Dust-SN-All), with a mean bias of +19%.

4.7. Summary of Evaluations

This section discusses the set of evaluations performed on the ALADIN simulations with observations and previous studies, as shown in

Table 1. Summary of the evaluations of the ALADIN model simulations against various observations and previous studies. The best metrics obtained for each evaluation are highlighted in bold.

Evaluations	Reference Dataset	Simulations
		S-Dust-SN-All			S-Dust-SN-Wind			S-Dust-SN-Non
		a	${\mathrm{R}}^{\mathbf{2}}$	bias (%)	a	${\mathrm{R}}^{\mathbf{2}}$	bias (%)	a	${\mathrm{R}}^{\mathbf{2}}$	bias (%)
AOD	AERONET	0.67	0.52	−8.0	0.68	0.49	−11.7	0.32	0.27	+19.8
AOD	VIIRS	0.41	0.24	−13.4	0.42	0.24	−17.0	0.17	0.10	+15.7
${\sigma }_{ext}$	MPLNET	0.78	0.83	+17.8	0.83	0.78	+7.0	0.57	0.89	+65.9
AsyF	AERONET	0.38	0.07	−2.6	0.50	0.11	−2.3	0.32	0.04	−1.8
SSA	AERONET	0.09	0.01	−1.7	0.09	0.01	−1.7	0.07	0.01	−2.1
SW DW Flux	SolRad-Net	0.99	0.94	+0.4	0.98	0.90	−0.9	0.93	0.86	−6.2
LW UP Flux	CERES	0.99	0.63	−5.6	0.78	0.53	−0.8	0.41	0.19	−12.6
SW TOA DRE	GAME	0.51	0.15	−64.5	0.48	0.12	−68.2	1.38	0.13	−74.3
LW TOA DRE	GAME	3.08	0.28	−90.7	5.08	0.42	−91.7	5.39	0.29	−92.4
SW BOA DRE	GAME	0.22	0.42	+62.2	0.29	0.53	+50.2	0.09	0.03	+66.5
LW BOA DRE	GAME	0.18	0.00	−84.8	0.71	0.02	−86.7	1.88	0.15	−83.3
${\mathrm{P}}_o$	Radiosounding	0.80	0.89	−0.2	0.75	0.89	−0.1	0.63	0.78	−0.1
${\mathrm{T}}_o$	Radiosounding	0.61	0.76	−5.3	0.63	0.76	−6.6	0.57	0.71	−7.0
Recount best values		6	8	7	8	7	5	0	2	3

.

In Table 1, it can be observed that the simulations with the best evaluation metrics are S-Dust-SN-All and S-Dust-SN-Wind. Specifically, S-Dust-SN-All achieves the most accurate AOD results, with the highest ${\mathrm{R}}^2$ values and the lowest bias. However, S-Dust-SN-Wind demonstrates the best slope of the linear regression between the observations and simulated values, indicating its capability in representing trends. Optical scattering properties are better reproduced by S-Dust-SN-Wind, although S-Dust-SN-All obtains comparable metrics. Radiative fluxes are also more accurately represented by S-Dust-SN-All. Despite the previous good evaluations, the direct radiative effects calculated by the GAME model show generally poor agreement with all ALADIN simulations. Nevertheless, S-Dust-SN-All exhibits the best evaluation metrics for the TOA, while S-Dust-SN-Wind and S-Dust-SN-Non show better performance for the BOA. Finally, the surface pressure and near-surface air temperature assessments are very similar between the simulations with spectral nudging.

Overall, while all simulations with spectral nudging show reasonable agreement in key variables like the AOD, optical scattering properties, radiative fluxes and meteorological variables, challenges remain in capturing direct radiative effects according to the GAME model, where large biases persist. S-Dust-SN-All emerges as the most balanced configuration, with the highest ${\mathrm{R}}^2$ and the lowest bias in most evaluations, while S-Dust-SN-Wind presents the best slope. The poor performance of S-Dust-SN-Non underscores the importance of including spectral nudging in the simulation to enhance the model fidelity.

In addition to the evaluation metrics presented in Table 1, an interesting alternative approach for further model evaluation involves the use of advanced observational techniques such as the GRASP (Generalized Retrieval of Aerosol and Surface Properties) algorithm. GRASP offers advanced observational products that provide the detailed separation of aerosol components, including dust, as well as information on the vertical distribution and radiative properties. These features make GRASP particularly suitable for direct comparison with simulated results from the ALADIN model, especially with respect to the aerosol composition and radiative effects. Future work could benefit from a systematic evaluation between ALADIN simulations and GRASP-derived products, as demonstrated in recent studies [65,66,67], to improve the model’s representation of aerosol–radiation interactions and the vertical structure of aerosols.

5. Dust Radiative Effects and Impacts on Heatwave

After evaluating the ALADIN model simulations against several reference datasets, it can be concluded that the model represents the Saharan dust and the heatwave event with high accuracy. Among the simulations, the one with the most restrictive spectral nudging (S-Dust-SN-All) demonstrates the best agreement with reference datasets across most evaluations. Similarly, the simulation with spectral nudging applied only to wind (S-Dust-SN-Wind) yields comparable results to S-Dust-SN-All. Consequently, the analysis of the direct and semi-direct radiative effects of dust is based primarily on the S-Dust-SN-All simulation, while key results from the S-Dust-SN-Wind simulation are presented in Appendix B. The analysis of dust direct and semi-direct effects will be divided into two parts. The first part involves a spatial analysis, through temporal averaging over the central time period of the event, which covers 23rd to 30th June 2019. The second part consists of a temporal analysis, spatially averaged over the entire Iberian Peninsula and selecting three Spanish representative AERONET stations: Barcelona, Madrid and El Arenosillo. This dual analysis is necessary because the dust event did not impact the Iberian Peninsula uniformly; as seen previously, different regions experience dust arrival gradually. In order to properly analyze the dust direct and semi-direct radiative effects, spatial and temporal analyses are carried out for the total AOD, as shown in Figure 8.

Figure 8 confirms that the dust particles do not reach the Iberian Peninsula uniformly. The eastern region experiences the most intense dust event, with an average dust AOD of 0.08 ± 0.05 across the entire Iberian Peninsula during 23rd to 30th June 2019. The highest dust load values occur during this period, which corresponds to the central part of the dust event. Initially, dust particles reach Barcelona, where a maximum dust AOD of 0.31 is measured on 26th June. Subsequently, Madrid experiences a maximum dust AOD of 0.23 on 27th June, followed by El Arenosillo with a maximum dust AOD of 0.11 on 28th June. As the Saharan dust intrusion advances across the Iberian Peninsula, the load of dust particles decreases due to deposition processes.

5.1. Direct and Semi-Direct Radiative Effects

First, the averages of the direct, semi-direct and total radiative effects of dust particles over the Iberian Peninsula in the shortwave and longwave spectra, at the top of the atmosphere and at the bottom of the atmosphere, have been calculated for the central period of the event, which covers the period from 23rd to 30th June 2019, shown in Figure 9.

Figure 9 shows, in SW, a weak negative dust direct radiative effect at TOA over the whole Peninsula, resulting in an average of −0.4 ± $0.4 {\mathrm{Wm}}^{-2}$. This average is consistent with the typical SW DRE of −$0.4 {\mathrm{Wm}}^{-2}$ reported for the Earth’s climate [68]. This relatively weak cooling effect can be attributed to the low average dust particle AOD during this period (see Figure 8). In contrast, at BOA, a stronger negative dust direct radiative effect is found over the whole Peninsula, resulting in an average of −3.9 ± $2.3 {\mathrm{Wm}}^{-2}$. These values are in line with other studies, where the direct radiative effects at TOA (BOA) ranged from −47 (−66) to −1 (−3) ${\mathrm{Wm}}^{-2}$ [55,57,59,63,64,69,70,71,72]. Moreover, according to the efficiency provided by [58] for a dust event measured at El Arenosillo, the SW TOA dust DRE would be −$4 {\mathrm{Wm}}^{-2}$, a higher value than the one found during the dust event under analysis. At both levels, the dust semi-direct effect is changing, reflecting large spatial variability. On average, at TOA, the dust semi-direct radiative effect opposes the direct one, and, at BOA, they are accumulated. As a result, we obtain a mean SW TOA total negative dust radiative effect of −0.3 ± $1.8 {\mathrm{Wm}}^{-2}$ and a mean SW BOA total negative dust radiative effect of −4.0 ± $3.1 {\mathrm{Wm}}^{-2}$. According to the literature, positive SW dust semi-direct radiative effects are found at both levels, not only at TOA. Ref. [14] obtained an SW TOA (BOA) dust semi-direct effect of +0.91 (+2.0) ${\mathrm{Wm}}^{-2}$ for global climatology over land in summer; [73] obtained between +0.9 (+0.7) and +2.4 (+2.6) ${\mathrm{Wm}}^{-2}$ over Europe for absorbing aerosols; [70] obtained between +5 and +10 (approximately +4.8) ${\mathrm{Wm}}^{-2}$ over the Iberian Peninsula in summer for all aerosols; [74] obtained between +0.2 and +$0.6 {\mathrm{Wm}}^{-2}$ globally over land for dust and sea salt aerosols; and [72] obtained between +0 and +$5 {\mathrm{Wm}}^{-2}$ over the Iberian Peninsula.

In LW, at TOA, a weak positive dust direct radiative effect is found over the whole Peninsula, resulting in an average of +0.1 ± $0.1 {\mathrm{Wm}}^{-2}$. This average is consistent with the typical LW DRE of +$0.25 {\mathrm{Wm}}^{-2}$ reported for the Earth’s climate [68]. At BOA, a similar pattern in the dust radiative effect is obtained with respect to TOA, but with higher values, resulting in an average of +0.3 ± $0.1 {\mathrm{Wm}}^{-2}$. These positive direct radiative effects are in line with other studies where the dust direct radiative effects at TOA (BOA) have yielded values between +0.2 (+0.6) and +9 (+18) ${\mathrm{Wm}}^{-2}$ [51,57,59,63,64,69,70]. At both levels, the dust semi-direct radiative effect adds to the direct one in the Eastern Peninsula and it opposes it in some parts of the center and the Western Peninsula. As a result, we found a mean LW TOA (BOA) total positive dust radiative effect of +0.1 ± 1.6 (+1.2 ± 1.2) ${\mathrm{Wm}}^{-2}$. This contrasts with other studies, where a negative LW dust semi-direct radiative effect is generally found. Ref. [70] obtained, at both levels, a large range of LW semi-direct effect values that ranged between −3 and +$2 {\mathrm{Wm}}^{-2}$ over the Iberian Peninsula in summer for all aerosols; [74], at TOA, found negative values between −0.4 and −$0.6 {\mathrm{Wm}}^{-2}$ globally over land for dust and sea salt aerosols; [72], at TOA, also found negative values between −2 and +$0 {\mathrm{Wm}}^{-2}$ over the Mediterranean region. Contrary to the previous studies, ref. [14] obtained a positive LW TOA semi-direct radiative effect of +$0.1 {\mathrm{Wm}}^{-2}$ but a negative LW BOA semi-direct radiative effect of −$0.2 {\mathrm{Wm}}^{-2}$ for a global climatology over land in summer.

In the absence of global observational estimations, climate model simulations have reported, at TOA, a net positive global annual mean dust semi-direct effect [68]. These estimations vary by over an order of magnitude, ranging from +0.01 to +$0.16 {\mathrm{Wm}}^{-2}$, depending on the climate model used [68,74,75]. Specifically, the dust net TOA SDRE has been estimated at 0.07 ± $0.07 {\mathrm{Wm}}^{-2}$. This positive value is consistent with the result obtained in this study, where a mean dust net TOA SDRE of +0.1 ± $2.2 {\mathrm{Wm}}^{-2}$ was found over the Iberian Peninsula.

In general, Figure 9 exhibits substantial variability in the radiative effect across the Iberian Peninsula, despite being averaged over the central period of the event, which covers 23rd to 30th June 2019. This variability is more pronounced in the semi-direct radiative effect compared to the direct radiative effect, primarily because the dust event, as previously mentioned, did not uniformly or simultaneously affect the entire study region. Additionally, the diverse orographic and atmospheric conditions over the Iberian Peninsula contributed significantly to this variability. A detailed analysis of the total radiative effect of the dust event across the different provinces of the Iberian Peninsula is presented in Appendix C, where a notable decrease in variability across many regions is observed, attributed to the finer spatial resolution of the analysis.

To continue the analysis, Figure 10 shows the averages of the direct, semi-direct and total radiative effects of dust particles over the Iberian Peninsula in the shortwave and longwave spectra, in the whole atmosphere, which have been calculated for the central period of the event, which covers 23rd to 30th June 2019.

In Figure 10, it can be observed that, in SW, the direct radiative effect produces net warming in the full atmosphere, with an average value of +3.5 ± $2.1 {\mathrm{Wm}}^{-2}$. The semi-direct radiative effect, while generally adding to the direct effect, is comparatively weaker, resulting in an average of +0.2 ± $0.5 {\mathrm{Wm}}^{-2}$. Consequently, the total radiative effect is positive, with an average of +3.7 ± $2.3 {\mathrm{Wm}}^{-2}$. In contrast, in LW, the direct radiative effect produces net cooling in the full atmosphere, with an average value of −0.3 ± $0.1 {\mathrm{Wm}}^{-2}$. The semi-direct radiative effect also adds to this one, except in the northwest region, which has an average of −0.9 ± $0.9 {\mathrm{Wm}}^{-2}$. Taking both contributions, the total radiative effect produces net cooling in the full atmosphere of −1.1 ± $1.0 {\mathrm{Wm}}^{-2}$. All values have been averaged over the Iberian Peninsula.

When both spectral ranges are combined, the results indicate that, in the direct effect, the SW component dominates across the full atmosphere, resulting in net warming of +3.2 ± $2.0 {\mathrm{Wm}}^{-2}$ over the Iberian Peninsula. In contrast, in the semi-direct radiative effect, the LW component dominates, producing net cooling of −0.6 ± $0.9 {\mathrm{Wm}}^{-2}$. Consequently, the semi-direct radiative effect mitigates 18.8% of the warming caused by dust particles due to absorption and scattering phenomena in the full atmosphere. As a result, net warming of +2.6 ± $1.8 {\mathrm{m}}^{-2}$ is observed on average over the Iberian Peninsula.

For further discussion, the temporal distribution of the daily dust radiative effects is analyzed, over the Iberian Peninsula and at three representative AERONET stations—Barcelona, Madrid and El Arenosillo—as shown in Figure 11.

Figure 11 shows the importance of considering the semi-direct effect in atmospheric radiative analysis. In SW, at TOA, negative values for both the direct and semi-direct radiative effects are generally observed for all locations and days. Over the Iberian Peninsula, the DRE remains consistently weakly negative during this period, while the SDRE exhibits slightly lower values, reaching a minimum of −$10.7 {\mathrm{Wm}}^{-2}$ on the 27th. In Barcelona, the 25th was particularly notable, with an exceptionally high DRE of +$5.8 {\mathrm{Wm}}^{-2}$, coupled with a remarkable SDRE of −$30.2 {\mathrm{Wm}}^{-2}$. However, some exceptions with positive values are found, such as the SDRE of +$9.5 {\mathrm{Wm}}^{-2}$ on the 27th in Madrid and the significant SDRE of +$18.7 {\mathrm{Wm}}^{-2}$ on the 24th in El Arenosillo. In SW, at BOA, the dust radiative effects show higher variability and magnitudes compared to TOA, especially in Barcelona, where the dust effects seem to be more intense around the 26th, coinciding with the maximum AOD found. An example of this is on the 25th in Barcelona, where the daily DRE reaches a value of −$17.0 {\mathrm{Wm}}^{-2}$ and the SDRE is −$36.2 {\mathrm{Wm}}^{-2}$. Over the Iberian Peninsula, the DRE remains consistently weakly negative throughout the period, while the SDRE exhibits slightly higher values, reaching a minimum of −$2.2 {\mathrm{Wm}}^{-2}$ on the 26th. In Madrid, the radiative effects are less intense but still significant, with a DRE of −$17.6 {\mathrm{Wm}}^{-2}$ on the 27th. In El Arenosillo, a lower DRE of −$8.0 {\mathrm{Wm}}^{-2}$ is observed on the 28th and a negative SDRE of −$19.5 {\mathrm{Wm}}^{-2}$ on the 30th.

In LW, at TOA, the radiative effects are generally positive and smaller, but some differences between stations are identified, with Barcelona being the most remarkable for its higher SDRE values, mainly on the 25th at +$2.1 {\mathrm{Wm}}^{-2}$. Over the Iberian Peninsula, notably lower daily SDRE values are obtained, not reaching +$1 {\mathrm{Wm}}^{-2}$. In Madrid, the values are similar to those of Barcelona, but with smaller variations, except for the 27th, with a negative SDRE of −$4.0 {\mathrm{Wm}}^{-2}$. In Arenosillo, the values are mostly close to zero, except for the 29th, with a negative SDRE of −$1.4 {\mathrm{Wm}}^{-2}$. In LW, at BOA, the dust radiative effects follow a pattern similar to that of TOA, but with higher magnitudes. In Barcelona, the 25th is once again remarkable, with an SDRE of +$9.8 {\mathrm{Wm}}^{-2}$. Over the Iberian Peninsula, the SDRE reaches its maximum value of +$2.4 {\mathrm{Wm}}^{-2}$ on the 26th. In Madrid, a maximum SDRE of +$4.0 {\mathrm{Wm}}^{-2}$ is discerned on the 28th, while El Arenosillo has a peak of +$6.0 {\mathrm{Wm}}^{-2}$ on the 29th.

Overall, the Iberian Peninsula shows very small radiative effect values due to the large spatial averaging. Barcelona exhibits the highest absolute values of radiative effects, associated with the highest dust load observed at this location. Madrid presents more moderate variations, whereas El Arenosillo is characterized by more pronounced peaks on specific days. The period from the 25th to 27th stands out as the interval with the most significant anomalies in the DRE and SDRE values.

To conclude the analysis, the temporal averages and standard deviations of the daily direct and semi-direct radiative effects at the three representative AERONET stations and over the Iberian Peninsula are shown in

Table 2. Temporal averages and standard deviations of the daily dust direct (DRE) and semi-direct (SDRE) radiative effects (${\mathrm{Wm}}^{-2}$) at three representative AERONET stations—Barcelona, Madrid and El Arenosillo—and over the Iberian Peninsula, during the period 23rd to 30th June 2019.

Location	Shortwave Spectrum				Longwave Spectrum
	TOA		BOA		TOA		BOA
	DRE	SDRE	DRE	SDRE	DRE	SDRE	DRE	SDRE
Barcelona	−0.5 ± 2.7	−3.3 ± 11.0	−11.8 ± 8.1	−4.3 ± 13.0	+0.2 ± 0.1	+0.8 ± 0.8	+0.7 ± 0.4	+3.3 ± 2.9
Madrid	−0.3 ± 0.3	+1.2 ± 3.4	−7.1 ± 5.4	+1.5 ± 3.2	+0.2 ± 0.1	−0.7 ± 2.1	+0.5 ± 0.4	+0.8 ± 1.9
El Arenosillo	−0.6 ± 1.0	−0.2 ± 9.4	−2.7 ± 3.1	−0.7 ± 11.4	0.0 ± 0.0	−0.5 ± 0.6	+0.2 ± 0.2	+1.7 ± 3.3
Iberian Peninsula	−0.4 ± 0.4	+0.1 ± 0.9	−3.9 ± 3.4	−0.1 ± 1.1	+0.1 ± 0.1	0.0 ± 0.3	+0.3 ± 0.2	+0.9 ± 0.9

.

In Table 2, it can be observed that the DRE remains with the same sign at all locations, whereas the SDRE exhibits variations in sign. In SW, at both levels, the DRE is negative, with higher values at BOA. At TOA, the strongest average DRE is found in El Arenosillo (−$0.6 {\mathrm{Wm}}^{-2}$), while Barcelona shows the highest variability (±$2.7 {\mathrm{Wm}}^{-2}$). The SDRE is minimal in Barcelona, characterized by significant variability (−3.3 ± $11.0 {\mathrm{Wm}}^{-2}$), whereas Madrid exhibits the highest positive SDRE value (+$1.2 {\mathrm{Wm}}^{-2}$). At BOA, Barcelona experiences the strongest negative DRE, coupled with the highest variability (−4.3 ± $13.0 {\mathrm{Wm}}^{-2}$). The SDRE is slightly larger in absolute magnitude than at TOA, with the most negative effect in Barcelona (−$4.3 {\mathrm{Wm}}^{-2}$) and a positive effect in Madrid (+$1.5 {\mathrm{Wm}}^{-2}$).

In contrast to SW, the LW radiative effects are predominantly positive. At TOA, the DRE values remain low but positive, reaching a maximum of +$0.2 {\mathrm{Wm}}^{-2}$ in both Barcelona and Madrid. The SDRE varies significantly across the regions, showing a maximum in Barcelona (+$0.8 {\mathrm{Wm}}^{-2}$), while being negative in Madrid (−$0.7 {\mathrm{Wm}}^{-2}$) and El Arenosillo (−$0.5 {\mathrm{Wm}}^{-2}$). At BOA, the positive DRE is more pronounced, with Barcelona exhibiting the highest impact (+$0.7 {\mathrm{Wm}}^{-2}$), followed by Madrid (+$0.5 {\mathrm{Wm}}^{-2}$) and El Arenosillo (+$0.2 {\mathrm{Wm}}^{-2}$). The SDRE remains consistently positive, reaching a maximum value in Barcelona (+$3.3 {\mathrm{Wm}}^{-2}$). The variability in the LW SDRE is generally higher at BOA, particularly in El Arenosillo (±$3.3 {\mathrm{Wm}}^{-2}$).

Overall, Table 2 highlights the local differences in the dust radiative effects. Barcelona experiences the strongest negative DRE in the SW, along with the most variable SDRE, while Madrid and El Arenosillo exhibit more moderate trends. The average impact over the Iberian Peninsula suggests a dominant negative SW DRE, a positive LW DRE and a generally positive SDRE in both spectral ranges.

5.2. Dust Impact on the Heatwave

This section analyzes the average change in the state of the atmosphere by means of meteorological variables, namely the liquid water path (LWP), near-surface air temperature (${\mathrm{T}}_o$), high-troposphere cloud fraction (HCF), low-troposphere cloud fraction (LCF), eastward near-surface wind speed (UAS) and northward near-surface wind speed (VAS) over the Iberian Peninsula, as shown in Figure 11.

These changes in the meteorological variables were calculated as the differences between two simulations: one considering the presence of dust particles in the atmosphere and the other excluding them (S-Dust and S-Non). These variations are therefore attributed exclusively to the dust event, discarding any influence from the heatwave. In Figure 12, we can observe an increase in the liquid water path over the Eastern Iberian Peninsula, contrasted by a slight decrease in the central regions and more pronounced variations in the west. The near-surface air temperature shows a decrease across nearly the entire Iberian Peninsula, except in the northwest and southwest. Changes in cloud fraction show greater variability compared to the near-surface air temperature, with localized effects in the Eastern Peninsula at high altitudes and in the Western Peninsula at low altitudes. Notably, a slight increase in cloud fraction is observed in the low troposphere, particularly accentuated along the Mediterranean coast and the Strait of Gibraltar. The near-surface wind speed also evidences significant variability, with a decrease in the northward component over most of the territory and a slight increase in the eastward component over the Iberian Peninsula, except in the center.

Over the Iberian Peninsula, on average, the liquid water path decreased by −0.2 ± 4.5 mg ${\mathrm{m}}^{-2}$. This result diverges from previous findings in the literature. For example, ref. [76] reported a general increase in the liquid cloud water path between 0 and +2.5 g ${\mathrm{m}}^{-2}$ for mineral dust in South Africa, while [77] found an increase in the liquid water path between +6 and +7% for biomass-burning particles over Africa. A mean decrease in the near-surface air temperature of −0.2 ± $0.2 °$C was observed, with drops of less than $1 °$C found in several locations across Southwestern France. These findings are in line with [78], which found a 2 m level temperature change between −6 and −$1 °$C over the Iberian Peninsula; [73], which found a 2 m level temperature decrease of −$0.14 °$C for absorbing aerosols in Europe; [79], which found a surface temperature reduction of between −0.5 and −$0.2 °$C in France, Germany and Italy for dust particles; [70], which found a surface temperature decrease of −$0.4 °$C over the Iberian Peninsula in summer; [76], which found a maximum decrease in surface temperature of between −1.1 and −$0.7 °$C for mineral dust over South Africa; [80], which found a 2 m level temperature decrease of between −1 and −$0.5 °$C for different aerosols over Europe; and [77], which found cooling between −2 and −$1 °$C for biomass-burning particles over Africa.

For the high-level cloud fraction, a decrease of −0.2 ± 1.2% is observed, contrasting with [14], which found an increase of +0.55% for a global climatology over land in summer, but it is similar to [74], which reported values between −0.48 and −0.22% for dust and sea salt particles over land globally. Conversely, an increase in the low-level cloud fraction of +0.1 ± 1.3% is discerned, which aligns with the slight increase of +0.75% obtained in [14] for a global climatology over land in summer, the rise between +0.07 and +0.2% reported in [74] for dust and sea salt particles over land globally and the moderate increase of between +2 and +4% reported in [77] for biomass-burning particles over Africa. Finally, the near-surface wind speed reveals a slight decrease in the northward component of −0.1 ± 0.1 m/s and a negligible change for the eastward component of 0.0 ± 0.1 m/s. This result partially fits with [76], which obtained a reduction in wind speed of between −0.22 and −0.17 m/s for mineral dust over South Africa.

In general, notable differences between the eastern–central and western regions of the Iberian Peninsula are evident, driven by the strong variability in the changes in the atmospheric variables. A detailed analysis of the atmospheric state changes across the different provinces of the Iberian Peninsula is presented in Appendix D. This analysis reveals a marked decrease in variability across different regions, attributed to the finer spatial resolution of the analysis. However, as shown in Figure 11, it can be concluded that the eastern and central regions of the Peninsula, along with Northern Galicia, exhibit a slight increase in the liquid water path, accompanied by enhanced cloudiness at the low troposphere. This combination of effects has also been reported in previous studies [81,82] and is typically associated with a positive dust net TOA SDRE. Additionally, significant variability is observed in the high-tropospheric cloudiness changes, which could be attributed to the formation or suppression of cirrus clouds, depending on the location and local meteorological conditions. These results are consistent with the findings in [68], which reported that the overall annual mean dust SDRE tends to induce a decrease in high-tropospheric cloudiness, except during summer. Moreover, a reduction in the northward component of the near-surface wind speed is found across much of the Peninsula, potentially indicating enhanced atmospheric stability associated with the presence of dust particles [70], which could have implications for air quality degradation. Furthermore, a general decrease in the near-surface air temperature is observed, possibly associated with the negative SW BOA total radiative effect induced by dust particles (see Figure 9), although this air cooling remains minor compared to the +$6 °$C temperature rise caused by the heatwave [28].

For further discussion, the temporal distributions of the daily changes in the meteorological variables are analyzed, over the Iberian Peninsula and at three representative AERONET stations—Barcelona, Madrid and El Arenosillo—as shown in Figure 13.

Figure 13 shows the importance of the coupling between the radiative scheme and the meteorological model in order to study all of the possible effects of dust particles. In Barcelona, the LWP reaches a maximum of +42.7 mg ${\mathrm{m}}^{-2}$ on the 25th. In contrast, the changes in Madrid and over the Iberian Peninsula are more moderate and generally negative. Meanwhile, in El Arenosillo, there is notable variability, particularly from the 28th onwards, with a maximum of +17.6 mg ${\mathrm{m}}^{-2}$). In terms of the near-surface air temperature, all locations except El Arenosillo show a general decrease during the event. This reduction is more accentuated in Barcelona, with a minimum of −0.8 °C on the 26th, followed by Madrid with a minimum of −$0.7 °$C on the 27th. In contrast, El Arenosillo exhibits a general rise in the near-surface air temperature, with slight fluctuations, leading temperature increases of +$0.3 °$C on the 25th and 28th.

Turning to cloud fraction changes, Barcelona shows slight fluctuations in HCF, with a shift to positive values from the 25th. The maximum reduction of 3.2% occurred on the 24th. In the low troposphere, the changes are more moderate and show an increase on the 24th and 25th, with a maximum of +4.8%. In Madrid, there is a general decrease in HCF, except on 24th, where it increases by 1.1%. In contrast, at lower levels, the changes are negligible. In El Arenosillo, the changes in the cloud fraction are more accentuated, especially in the low troposphere, with a significant increase of +13.6% on the 29th, just one day after the peak in the dust aerosol load. Regarding the near-surface wind speed, the fluctuations are considerable at all locations, making the average negligible over the Iberian Peninsula. In Barcelona, both components of the near-surface wind speed increase on the 26th, with a maximum of +0.1 and +0.3 m/s for the northward and eastward components, respectively. In Madrid, however, larger changes are observed. The northward component achieves a maximum of +0.1 m/s on the 26th, while the eastward component reaches a minimum of −0.8 m/s. El Arenosillo shows similar trends to Madrid, with a temporal delay. The northward component reaches a maximum of +0.2 m/s on the 28th, while the eastward component reaches a minimum of −0.6 m/s on the 27th, followed by an increase of +0.4 m/s on the 29th.

Overall, the meteorological variables across the Iberian Peninsula exhibit minimal changes due to the large spatial averaging. However, Barcelona shows more pronounced variations, with a maximum increase in the liquid water path of +42.7 mg ${\mathrm{m}}^{-2}$ on the 25th, accompanied by a +4.8% rise in the low-troposphere cloud fraction. These changes may be linked to the observed minimum in the SW SDRE (Figure 11) and the simultaneous maximum in the LW BOA SDRE. Additionally, Barcelona experiences the largest decrease in the near-surface air temperature, reaching −$0.8 °$C on the 26th. This temperature drop coincides with the minimum in the SW BOA DRE and the peak AOD recorded at this location. Madrid exhibits more moderate variations but stands out with the second-lowest temperature found (−$0.6 °$C on 27th) and a decrease in the near-surface wind speed around this day, with reductions of −0.3 and −0.8 m/s in the northward and eastward components, respectively. These variations may be associated with the SW SDRE maxima observed on the 27th (Figure 11) and the simultaneous minimum in the LW TOA DRE. In contrast, El Arenosillo shows a general increase in the near-surface air temperature throughout the study period. Notable changes include a substantial rise in the liquid water path (+16.7 mg ${\mathrm{m}}^{-2}$), low-troposphere cloud fraction (+13.6%) and eastward wind speed (+0.4 m/s) on the 29th and 30th. These variations could be associated with the negative SW SDRE values observed on these days (Figure 11) and the corresponding positive LW BOA SDRE values.

To conclude the analysis, the temporal distribution of the daily changes in the meteorological variables is examined at the three representative AERONET stations—Barcelona, Madrid and El Arenosillo—and over the Iberian Peninsula, as summarized in

Table 3. Temporal averages and standard deviations of the daily changes in meteorological variables, namely the liquid water path (LWP), near-surface air temperature (${\mathrm{T}}_o$), cloud fraction at high troposphere (HCF), cloud fraction at low troposphere (LCF) and near-surface wind speed northward (VAS) and eastward (UAS) components, at three representative AERONET stations—Barcelona, Madrid and El Arenosillo—and over the Iberian Peninsula, during the period 23rd to 30th June 2019.

Location	Meteorological Variables
	$\Delta$LWP (mg ${\mathrm{m}}^{-\mathbf{2}}$)	$\Delta$${\mathrm{T}}_{\mathbf{o}}$ (°C)	$\Delta$HCF (%)	$\Delta$LCF (%)	$\Delta$VAS (m/s)	$\Delta$UAS (m/s)
Barcelona	+6.7 ± 15.1	−0.4 ± 0.2	−0.4 ± 1.2	+0.8 ± 1.7	0.0 ± 0.1	0.0 ± 0.1
Madrid	−0.5 ± 0.8	−0.3 ± 0.3	−0.5 ± 1.2	0.0 ± 0.0	−0.1 ± 0.2	−0.2 ± 0.3
El Arenosillo	+2.1 ± 6.7	+0.1 ± 0.1	−0.5 ± 0.6	+2.2 ± 5.7	0.0 ± 0.2	0.0 ± 0.3
Iberian Peninsula	−0.3 ± 0.6	−0.2 ± 0.1	−0.2 ± 0.3	+0.1 ± 0.5	−0.1 ± 0.0	0.0 ± 0.0

.

In Table 3, it can be observed how the impact of dust particles changes depending on the location. Barcelona and El Arenosillo experience an increase in the liquid water path (averaging +6.7 and +2.1 mg ${\mathrm{m}}^{-2}$, respectively), while Madrid and the whole Iberian Peninsula undergo a decrease in the LWP. This behavior is likely produced by sea breezes, because Barcelona and El Arenosillo are coastal locations. Barcelona also exhibits the largest decrease in the near-surface air temperature, with an average reduction of −$0.4 °$C, a change that is relatively minor compared to the +$6 °$C temperature increase associated with the heatwave during the Saharan dust event [28]. Madrid has the same trend, with a lower negative average temperature of −$0.3 °$C.

In terms of cloud cover, the most significant changes are observed in coastal areas, where there is a reduction in high-troposphere cloudiness (Barcelona with an average of −0.4% and El Arenosillo with −0.5%), likely due to the dissipation of cirrus clouds, and an increase in low-troposphere cloudiness (Barcelona with an average of +0.8% and El Arenosillo with +2.2%). The near-surface wind speed changes are not particularly striking during the central period of the dust event, but a slight decrease was noted in Madrid (with an average of −0.1 and −0.2 m/s for the northward and eastward components of the wind speed), which could be associated with an increase in atmospheric stability caused by the dust event [70].

Overall, Table 3 highlights the local differences in the dust impact on the meteorological variables. The most notable variations occur in Barcelona and El Arenosillo, particularly in the LWP and low-troposphere cloud fraction. In contrast, Madrid experiences milder changes in all variables, except for a noticeable reduction in the wind speed.

6. Conclusions

This study examines the dust aerosol radiative effects over the Iberian Peninsula during a dust event and heatwave in summer 2019 using a regional climate atmospheric model (CNRM-ALADIN64 called ALADIN) developed at Centre National de Recherches Météorologiques. First, an evaluation of the different ALADIN simulations characterized by different spectral nudging is presented with observations and previous studies, showing its ability to accurately reproduce the spatial and temporal distribution of this dust event and heatwave. Among the simulations, the one with the most restrictive spectral nudging (S-Dust-SN-All) demonstrates the best agreement with the reference datasets across most evaluations. Similarly, the simulation with spectral nudging applied only to wind (S-Dust-SN-Wind) yields comparable results to S-Dust-SN-All. Consequently, the analyses of the dust direct and semi-direct radiative effects are performed exclusively with the S-Dust-SN-All simulation (results for S-Dust-SN-Wind are shown in the Appendix).

In SW, the dust direct radiative effect over the Iberian Peninsula is weakly negative at TOA, with an average of −0.4 ± $0.4 {\mathrm{Wm}}^{-2}$. At BOA, this effect is stronger, with an average of −3.9 ± $2.3 {\mathrm{Wm}}^{-2}$. In the full atmosphere, general warming is observed, with an average of +3.5 ± $2.1 {\mathrm{Wm}}^{-2}$. The semi-direct radiative effect at both levels adds to the direct one across a large part of the eastern and central regions of the Iberian Peninsula, but it opposes in the Western Peninsula. On average, this results in values of +0.1 ± $1.7 {\mathrm{Wm}}^{-2}$ at TOA and −0.1 ± $1.9 {\mathrm{Wm}}^{-2}$ at BOA. In the full atmosphere, a warming effect in most parts of the Iberian Peninsula is evidenced, with an average of +0.2 ± $0.5 {\mathrm{Wm}}^{-2}$.

In LW, the direct effect is slightly positive at TOA (+0.1 ± $0.1 {\mathrm{Wm}}^{-2}$) and BOA (+0.3 ± $0.1 {\mathrm{Wm}}^{-2}$). As a consequence, a net cooling effect is observed in the atmosphere, with an average of −0.3 ± $0.1 {\mathrm{Wm}}^{-2}$. The semi-direct effect adds to the direct effect in the Eastern Peninsula, while opposing in the central and western regions. On average, this results in values of 0.0 ± $0.6 {\mathrm{Wm}}^{-2}$ and at BOA of +0.9 ± $1.1 {\mathrm{Wm}}^{-2}$. In the full atmosphere, a stronger cooling effect is observed, with an average of −0.9 ± $0.9 {\mathrm{Wm}}^{-2}$. When both spectral ranges are considered, in the full atmosphere, the direct radiative effect produces net warming of +3.2 ± $2.0 {\mathrm{Wm}}^{-2}$, while the semi-direct radiative effect induces cooling of −0.6 ± $0.9 {\mathrm{Wm}}^{-2}$, mitigating 18.8% of the warming produced by the dust particles due to absorption and scattering processes.

These general patterns over the Iberian Peninsula are complemented by localized analyses from AERONET stations. Barcelona exhibits the highest absolute values of radiative effects, associated with the highest dust load observed at this location. Madrid experiences more moderate variations, whereas El Arenosillo is characterized by more pronounced peaks on specific days. The period from the 25th to 27th stands out as the interval with the most significant anomalies in the radiative effects.

Third, the changes in the meteorological variables are analyzed during the central period of the event. Over the Iberian Peninsula, the liquid water path increases in the east, decreases slightly in the center and shows more pronounced variations in the west (average of −0.2 ± 4.5 mg ${\mathrm{m}}^{-2}$). The near-surface air temperature generally decreases, except in the northwest and southwest (average of −0.2 ± $0.2 °$C). The cloud fraction changes are more variable than temperature changes, with localized effects at the high troposphere in the east and low troposphere in the west (average of −0.2 ± 1.2 and +0.1 ± 1.3%, respectively). The near-surface wind speed shows significant variability, with a decreased northward component across most regions and a slight increase in the eastward component, except in the center (average of −0.1 ± 0.1 and 0.0 ± 0.1 m/s, respectively). For instance, Barcelona experiences the largest increase in the liquid water path (averaging +6.7 mg ${\mathrm{m}}^{-2}$). This location also shows the largest decrease in the near-surface air temperature, with an average reduction of −$0.4 °$C, a change that is relatively minor compared to the +$6 °$C temperature increase associated with the heatwave during the Saharan dust event. In terms of cloud cover, the most significant changes are observed in coastal areas, where a reduction in high-troposphere cloudiness and an increase in low-troposphere cloudiness are found. The near-surface wind speed changes are not particularly striking during the central period of the dust event, but a slight decrease was noted in Madrid. Significant variability has been observed in all results across the Iberian Peninsula, primarily due to the gradual arrival of dust particles and the non-homogeneous distribution of them. Consequently, the direct and semi-direct radiative effects, along with the changes in meteorological variables, have been calculated across the different provinces of the Iberian Peninsula.

This study highlights the critical role of the semi-direct radiative effect of dust aerosols in modifying key atmospheric variables such as the temperature, humidity, cloudiness and wind speed. An accurate representation of this effect can substantially reduce radiation biases in forecasting models during mineral dust outbreaks, leading to more reliable weather predictions. We also emphasize the importance of spectral nudging as a technique to constrain large-scale atmospheric dynamics while allowing smaller-scale processes to develop freely, thereby improving models’ consistency with observations. Nonetheless, further investigation is needed to fully understand its impact on the estimation of the semi-direct effect in regional climate modelling. Future work should focus on analyzing additional dust events over the Iberian Peninsula to quantify the average semi-direct radiative effect and assess the potential of dust to mitigate heatwave events.