We analyzed the 2000–2022 data of PM₁₀ and PM_2.5 recorded using the governmental air quality monitoring networks of Spain and Portugal. The data from Spain were recorded at 330 air quality stations distributed across the 17 autonomous communities and the autonomous city of Ceuta, whereas the data from Portugal were collected from 11 stations distributed across the Norte, Centro, Lisbon, Vale do Tejo, Alentejo, Algarve, Madeira and Azores regions.

These stations are integrated in the European air quality monitoring network, which is the largest European infrastructure for PM₁₀ and PM_2.5 monitoring and follows standardized methods for measurements, quality assurance (QA) and quality control (QC) (EN-16450:2017 and EN-12341:2015). At these stations, high-temporal-resolution (10 to 60 min) PM₁₀ and PM_2.5 data are obtained using automatic monitors based on different principles of measurement, such as beta attenuation, tapered element oscillating microbalance and optical particle sizers (Rodríguez et al., 2012), with technical specifications accomplishing the EN-16450:2017 standard. Data regarding the 24 h average PM₁₀ and PM_2.5 values are also obtained following the gravimetric reference method (EN-12341:2015) used for QA and QC assessments and for converting the PM₁₀ and PM_2.5 data obtained with the automatic devices to gravimetric equivalent data by using the European reference protocols (EN-16450:2017). The PM₁₀ and PM_2.5 data that we used were provided by the Ministry of Ecological Transition and Demographic Challenge and by the air quality departments of the autonomous communities of Spain and the Agência Portuguesa do Ambiente.

For each specific study event, we also used data from the National Center for Environmental Prediction/National Center for Atmospheric Research (NCEP/NCAR) meteorological reanalysis (Kalnay et al., 1996); the Modern-Era Retrospective analysis for Research and Applications, Version 2 (MERRA-2), model dust reanalysis (Gelaro et al., 2017); and the satellite Visible Infrared Imaging Radiometer Suite (VIIRS) image sensor onboard the Suomi polar satellite.

We analyzed the data of PM₁₀ and PM_2.5 (PM_x) of 330 air quality monitoring stations (AQMSs) in Spain and 11 AQMSs of Portugal. We found that during the duxt events the 0.5 and 1 h resolution data of PM₁₀ increased to reach a rather constant “saturation” value that in most cases was somewhat lower than 1000 µgm-3 (in many cases between 900 and 1000 µgm-3); no values above this threshold appear in the data records. In these cases, PM₁₀ concentrations remained constant during the period of extremely high dust concentrations (typically 5 to 30 h), a behavior that was not generally observed in PM_2.5, which exhibited regular variability (with values <1000 µgm-3) and even increases in the periods when PM₁₀ remained (un-consistently) constant. This behavior can be observed in the time series of PM₁₀ (Fig. 3a1–a3) and PM_2.5 (Fig. 3b1–b3) linked to the dx-01, dx-02, dx-04, dx-05 and dx-06 events (dx-03 is not included in Fig. 3 for the sake of brevity). This saturation threshold close to ≈ 1000 µgm-3 is the upper operation limit of some PM₁₀ monitors and is also the top value of the validation data flag in some commercial data-recording software used in many governmental air quality monitoring networks, which assume that PM₁₀ concentrations above this threshold may suffer underestimation due to the high load of particles (e.g., particles accumulated in the filter tape of the beta instrument or particle coincidence problems in the optical particle sizers leading to a loss of sensitivity) and consequently do not record values above this threshold. At some AQMSs the saturation threshold was found at 500 µgm-3 and even at 200 µgm-3. In fact, the EN-16450, EN-12341 and EN-14907 accreditations for PM₁₀ and/or PM_2.5 monitors available in the European market are obtained for specific ranges, whose most frequent upper limit is 1000 µgm-3 (e.g., Met One™ BAM-1020; Comde Derenda™ APM 2; and Thermo Fisher Scientific™ 5014i, 5030i SHARP, TEOM 1405-F, and 1405-DF), although it can be as low as 200 µgm-3 for some equipment (e.g., FAI™ Swam 5a) and in contrast can be as high as 10 000 µgm-3 for other devices (e.g., PALAS™ Fidas 200 and 200E) according to the certifications agencies (e.g., TÜV; see https://www.qal1.de/, last access: 19 December 2023). In all these cases of PM₁₀ data affected by saturation, we reconstructed the PM₁₀ concentrations with the method described in Fig. 4, which we have called PM_x evaluation and reconstruction method based on ratios during extreme dust events or “duxt-r”.

In this duxt-r method we performed the following steps. (1) We first identified the invalid 1 h (or 1/2 h or 10 min time resolution) PM₁₀ data affected by saturation and the associated invalid PM_2.5 / PM₁₀ ratios, highlighted with red circles in the example of the dx-04 event shown in Fig. 4a and b. (2) We second estimated the PM_2.5 / PM₁₀ ratio during the PM₁₀ saturation period by linear interpolation between the last valid PM_2.5 / PM₁₀ data point before saturation and the first valid PM_2.5 / PM₁₀ data point after saturation [R2.5/10(i)], highlighted with green points in Fig. 4b. As a result of this interpolation the PM_2.5 / PM₁₀ ratios we used are not constant, changing hour by hour, e.g., from 0.184 to 0.162 in the example shown in Fig. 4b. (3) Finally, the 1 h (or 1/2 h or 10 min) resolution PM₁₀ concentrations were determined with Eq. (1). The method was validated using a comparison with data recorded at a few PM₁₀ monitors not affected by saturation (described below). 1PM10=PM2.5R2.5/10(i) We found that the PM_2.5 / PM₁₀ ratios during the duxt events were within the range 0.16 to 0.22 at most of the AQMSs, a value lower than that observed during regular dust events (typically ≈ 0.3) and much lower than that observed in environments affected by secondary particle formation and vehicle exhaust and other combustion emissions (typically within 0.6–0.9) (Rodríguez and López-Darias, 2021).

In our database, we replaced the PM₁₀ saturated data with the new PM₁₀ reconstructed data; i.e., the red (saturated) points shown in Fig. 4c were replaced by the green reconstructed points. At each site (AQMS), the consistency of the reconstructed data was assessed by analyzing the scatter plot of the PM_2.5 vs. PM₁₀ data (Fig. 4d1–d4). As an example, the results obtained in AQMSs located in Almería province (Mediterraneo AQMS) and Murcia (Mompean AQMS) in southeastern Spain, Salamanca province (Salamanca-6 AQMS) in central northern Spain, and in Fuerteventura (Canary Islands) is shown (Fig. 4d1–d4). With this method, the red PM₁₀-saturated data shown in Fig. 4d1–d4 were replaced by the green (reconstructed) data.

Because of the technical manufacturing specifications, the automatic PM₁₀ and PM_2.5 monitors of two AQMSs were able to record valid and consistent PM₁₀ and PM_2.5 data higher than 1000 µgm-3. We used these records to validate this methodology (Fig. 3e1–e4). At these sites, we reconstructed the PM₁₀ concentrations above 1000 µgm-3 as if they had experienced saturation, and we then compared the reconstructed and measured PM₁₀ concentrations. For this comparison we also included data between 500 and 1000 µgm-3 for the purpose of having a larger dataset (i.e., all PM₁₀ data >500 µgm-3). We found that the difference between reconstructed and measured PM₁₀ concentrations ranged between 2 % and 9 % (Fig. 4e1–e4). At a few other AQMSs using beta attenuation devices able to provide PM10>1000 µgm-3, we found a low PM₁₀ variability above this threshold (indicating loss of sensitivity due to mass overload in the filter tape) that was inconsistent with the variability in PM_2.5 and that resulted in PM_2.5 / PM₁₀ ratios similar to those affected by the saturation as described above (red circles in Fig. 3a). At these sites we also reconstructed the PM₁₀ with the duxt-r method (Fig. 4). Finally, the PM₁₀ data measured with the automatic monitors (measured and reconstructed) were converted to gravimetric equivalent data by intercomparison with PM₁₀ data obtained with the gravimetric reference method (EN-12341:2015; 24 h sampling, available during 7 to 25 d per month depending on the AQMS) using the standardized procedure (EN-16450:2017).

The new dataset obtained with this duxt-r method (PM₁₀ measured and reconstructed and then converted to gravimetric equivalent) shows that 1 h average PM₁₀ data that appeared as “saturated” at 1000 µgm-3 actually reached values close to 6000 µgm-3 in the dx-02 event; close to 1400 µgm-3 in the dx-03 event; close to 2000 µgm-3 in the dx-01, dx-04, and dx-05 events; and between 3000 and 4500 µgm-3 in the dx-06 event (Fig. 3c1–c3).

By applying this methodology, we reconstructed a total of 1690 hourly PM₁₀ data points: 1537 hourly PM₁₀ data points belonged to (i) 48 AQMSs in Spain, distributed between the regions Canary Islands (39), Andalusia (5), Murcia (1), Castile y León (2), and Madrid (2), while (ii) 153 hourly PM₁₀ data belonged to 7 AQMSs in Portugal, distributed between the regions Lisbon and Vale do Tejo (6) and Alentejo (1). The data we reconstructed with this method are already available in public databases of the governmental air quality networks of Spain, the Ministry of Ecological Transition and the European Environment Agency. The PM₁₀ data of the other 44 AQMSs that also experienced PM₁₀ saturation could not be reconstructed due to the lack of simultaneous PM_2.5 measurements, which included a total of 655 hourly data points; these AQMSs were located on the Canary Islands (14), Andalusia (10), Extremadura (2), Castile y León (11), and Murcia (7). Just to illustrate the huge importance of reconstructing the data, a brief comparison (for a few AQMSs) of the 24 h average PM₁₀ concentrations calculated with saturated PM₁₀ data vs. reconstructed PM₁₀ data: (1) 948 vs. 3069 µgm-3 on 15 March 2022 in Almería province (Mediterraneo AQMS), (2) 740 vs. 1840 µgm-3 on 23 February 2020 on Gran Canaria (Playa del Inglés AQMS), (3) 1238 vs. 1684 µgm-3 on 23 February 2020 in Tenerife (Tomé Cano AQMS), (4) 577 vs. 1421 µgm-3 on 23 February 2020 in Tenerife (Piscina Municipal AQMS), and (5) 527 vs. 621 µgm-3 on 15 March 2022 in Granada province (Palacio de Congresos AQMS). The maximum 1 h average PM₁₀ and PM_2.5 recorded during dx-01 to dx-06 are in a selection of AQMSs shown in Figs. S1 and S2 in the Supplement.

The first two events occurred in February 2020 (Fig. 5): 3–5 February 2020 (dx-01; Fig. 5a1) and 22–29 February 2020 (dx-02; Fig. 5a2). Throughout a period of 6 weeks (from mid-January to ending February) a blocking anticyclone established over Iberia, i.e., the Iberian Peninsula (Fig. 5f and h), resulting in anomalous easterly wind over central Algeria (wind anomaly not shown for the sake of brevity), a scenario favorable for dust events (Alonso-Pérez et al., 2011b). On 3–5 February 2020 a cyclone reached Cabo Verde, the low-to-high (low over Cabo Verde, high over Iberia) dipole configuration (Fig. 5f) resulted in a strong pressure/geopotential gradient and winds that prompted dust emissions and a dense plume of Saharan dust that expanded over the Atlantic to the Canary Islands and toward the Azores (Fig. 5g). Across the Canary Islands the dx-01 event resulted in (i) 1 h average PM₁₀ and PM_2.5 concentrations within the range 300–2100 µgm-3 (Fig. 3c1) and 100–400 µgm-3 (Figs. 3b1; S1a1 and a2), respectively, and (ii) 24 h average PM₁₀ and PM_2.5 concentrations within the range 100–535^x µgm-3 (Figs. 5a1 and 6a1; ^x = maximum at Tenerife, San Miguel Tajao AQMS) and 50–165^x µgm-3 (Figs. 5a2 and 6a2; ^x Tenerife, Tomé Cano AQMSs), respectively. For Madeira the dust impact was smoother, with 24 h average PM₁₀ concentrations within the range 50–115 µgm-3 (Fig. 5c), due to the island remaining outside of the core of the dust plume (Fig. 5g). Dust events prompted by the summer North African Dipole Intensity (NAFDI) were originally introduced by Rodríguez et al. (2015). This concept of meteorological L-to-H dipoles has also been found to drive dust events in the Middle East (Kaskaoutis et al., 2015, 2017) and the June 2020 Godzilla duxt event (Francis et al., 2020).

On 22 February 2020, a new cyclone reached again the region of Cabo Verde, resulting in a similar L-to-H dipole meteorology (Fig. 5h), which prompted a dusty jet stream in the subtropical North Atlantic, impacting the Canary Islands (Fig. 5a2, h and i). This scenario caused the dark orange skies, record-breaking temperatures, wildfires linked to strong dry winds, closure of maritime and air navigation space, and the widespread flight cancelations described above (Fig. 1) (Cuevas et al., 2021). During this dx-02 event, extreme PM_x concentrations were recorded on the Canary Islands, with (i) 1 h average PM₁₀ and PM_2.5 concentrations within the range 2000–5254^x µgm-3 (Figs. 3c1; S1b1 and b2) (^x Gran Canaria, Arinaga AQMS) and 400–1129^x µgm-3 (^x Gran Canaria, Mercado Central AQMS) (Figs. 3b1 and S1), respectively, and (ii) 24 h average PM₁₀ and PM_2.5 concentrations within the range 600–1840^x µgm-3 (Figs. 5b1 and 6b1) (^x Gran Canaria, Playa del Inglés AQMS) and 200–404^x µgm-3 (Figs. 5b2, 6b2) (^x Gran Canaria, Playa del Inglés AQMS), respectively. Concentrations (24 h average) of PM₁₀ during this dx-02 event were up to 6 times higher than the upper limit of PM₁₀ during the regular dust events on the Canary Islands (≈ 300 µgm-3) and also much higher than the extraordinary 680 µgm-3 recorded during the dust event of 26 January 2000 (Viana et al., 2002). During this dx-02 episode, the highest 1 h PM₁₀ (3500–5254 µgm-3) and PM_2.5 (800–1129 µgm-3) and 24 h average PM₁₀ (1200–1840 µgm-3) and PM_2.5 (230–404 µgm-3) concentrations were recorded in the AQMSs located in the central part of the dust plume, on Gran Canaria and Tenerife (Fig. 6b1 and b2) (Fig. S1). After 24 February 2020, the Saharan dust plume shifted northward over the Atlantic, reaching mainland Spain (Fig. 5k) and resulting in (24 h average) PM₁₀ concentrations within the range 70–155 µgm-3 (Fig. 5d1) in central Spain (Madrid, Extremadura and Castilla–La Mancha regions), 70–75 µgm-3 (Fig. 5d1) in central Portugal and 80–200 µgm-3 in eastern Spain (Comunidad Valenciana region) (Fig. 5e). PM_2.5 concentrations within the range 20–33 µgm-3 in central Spain and central Portugal (Fig. 5d2). The reanalysis of MERRA-2 properly tracked these two dx-01 and dx-02 events, with dust and dust_2.5 (i.e., dust in the PM_2.5 fraction) within the range of the PM₁₀ and PM_2.5 concentrations recorded at the AQMSs (see orange circles in Fig. 5b–d).

The third duxt event (dx-03: 15–21 February 2021) (Fig. 7d) was also caused by the intense wind (Fig. 7c) linked to a L-to-H dipole meteorology (Fig. 7b), with the associated blocking anticyclone located over Iberia and a cyclone over the Sahel (Fig. 7b). The impact on the Canary Islands occurred during 15–19 February (Fig. 7a1 and a2) and on Madeira during 16–18 February 2021 (Fig. 7a3). On the Canaries, the highest 1 h average PM₁₀ (1000–1352 µgm-3) and PM_2.5 (200–326 µgm-3) were recorded on Gran Canaria, Tenerife, La Gomera and La Palma (Fig. S1c1 and c2). The event of 16 February 2021 resulted in 24 h average PM₁₀ and PM_2.5 concentrations within the range 400–711^x µgm-3 (Figs. 6 and 7a1) (^x Las Galletas AQMS, Tenerife) and 80–205^x µgm-3 (Figs. 6 and 7a2) (^x Las Galanas, La Gomera), respectively. Subsequently, the dusty air mass tracked the northward anticyclonic circulation, resulting in (24 h average) PM₁₀ concentrations within 80–180 µgm-3 on Madeira (Fig. 7a3), reaching central mainland Portugal and Spain (18–21 February 2021; Fig. 7e) and resulting in 24 h average PM₁₀ and PM_2.5 concentrations within the range 75–150 µgm-3 (Fig. 7a4) and 20–55 µgm-3 (Fig. 7a4 and a5), respectively. The MERRA-2 reanalysis properly tracked the dx-03 event, except during 16–17 February on Madeira and 21 February in central mainland Spain, when it clearly overestimated dust concentrations (see orange circles in Fig. 7a1–a4). A few days later, 23–28 February 2021, another dust event impacted across central to northern and eastern Europe due to the eastward shift of the low-to-high dipole and the resulting northward dust transport across the central Mediterranean (Meinander et al., 2023; Peshev et al., 2023).

Another two extreme dust events occurred in January 2022 (Fig. 8). The dx-04 event impacted the Canary Islands during 14–17 January 2022 (Fig. 8a1), resulting in (24 h average) PM₁₀ and PM_2.5 concentrations within the range 275–883^x µgm-3 (Figs. 8a1 and 6d1) (^x Tenerife, Casa Cuna AQMS) and 60–136^x µgm-3 (Figs. 8a2 and 6d2) (^x Tenerife, Caletillas AQMS) and 1 h average PM₁₀ and PM_2.5 concentrations reaching values within 1200–2170 and 240–550 µgm-3 across the Canaries (Fig. S1d1 and d2), respectively. A few days later, the dx-05 event occurred (29 January–1 February 2022; Fig. 8a2 and a3), again impacting the Canary Islands, resulting in (24 h average) PM₁₀ and PM_2.5 concentrations within the range 314–1055^x µgm-3 (Figs. 7b1 and 5e1) (^x Fuerteventura, El Charco AQMS) and 70–199^x µgm-3 (Figs. 8b2 and 6e2) (^x El Charco AQMS) and 1 h average PM₁₀ and PM_2.5 concentrations reaching values within 1000–2520 and 400–545 µgm-3 in the eastern islands (Fig. S1e1 and e2), respectively. On Madeira, PM₁₀ concentrations ranged within 80–225 µgm-3 during dx-04 and dx-05 (Fig. 8b3). In both cases a massive dust plume was transported northward, approaching Spain and Portugal (Fig. 8d1 and d2). As in the previous cases, these duxt episodes were caused by a low-to-high dipole meteorology, with an anticyclonic core over Europe expanding to North Africa and a cyclone south of the Canary Islands (Fig. 8c1 and c2). In the two events, MERRA-2 clearly overestimated dust concentrations (Fig. 8b1 and b2), suggesting that the model may be transporting dust too low in the atmosphere (O'Sullivan et al., 2020).

Finally, the sixth duxt event (dx-06: 15–20 March 2022) first impacted mainland Spain and Portugal and subsequently impacted the Canary Islands (Fig. 9). The event was also prompted by a meteorological low-to-high dipole, linked to the location of Cyclone Celia over Morocco and an anticyclonic core over the central Mediterranean (Fig. 9b). The resulting dusty jet (Fig. 9c–e) expanded from southeastern to northwestern mainland Spain and Portugal on 15 and 16 March 2022. Once in northern Spain, the dust plume split in two branches, a branch that traveled eastward across central Europe tracking the anticyclonic circulation at the north of the high circulation (Qor-El-Aine et al., 2022) and another branch that traveled southward over mainland Portugal to the Canary Islands tracking the cyclonic low circulation (Fig. 9e and f).

In Iberia, PM_x concentrations experienced a sharp increase from their regular background levels (10–30 µgm-3) to 24 h average PM₁₀ and PM_2.5 values within the range (i) 500–3070 and 100–700 µgm-3 in the southern regions of Spain and Portugal (Murcia, Andalusia, Algarve and Alentejo) (Fig. 9a7, a8, and 10), (ii) 200–1000 and 60–160 µgm-3 in the central parts of Spain and Portugal (Castilla–La Mancha, Madrid, Extremadura, and Lisbon and Vale do Tejo) (Figs. 9a5, a6 and 10), (iii) 200–1000 and 60–260 µgm-3 in central northern Spain (Centro and Castile y León) (Figs. 9a3, a4 and 10), and (iv) 150–500 and 75–130 µgm-3 in northern Portugal and Spain (Norte, Cantabria and Galicia) (Figs. 9a1, a2 and 10) during 15–16 March 2022, respectively. In Lisbon, these extremely high PM₁₀ and PM_2.5 concentrations were even registered in an indoor environment (Gomes et al., 2022). After several days traveling across thousands of kilometers, the dust plume impacted the Canary Islands during 17–20 March 2022 (Fig. 9f), resulting in (24 h average) PM₁₀ and PM_2.5 values within the range 150–430 µgm-3 (Fig. 9a9) and 30–80 µgm-3 (Fig. 9a10), respectively. A regular to intense (no extreme) dust event impacted southern Spain on 24–25 March 2022 (PM₁₀ and PM_2.5 of 50–420 and 30–120 µgm-3, respectively) (Fig. 9a7 and a8) and the Canary Islands on 24–25 March 2022 (PM₁₀ and PM_2.5 of 40–80 and 20–35 µgm-3, respectively).

This historic dx-06 event (Fig. 1d and e) started on the evening of 14 March 2022 (>21 h) with a dust inflow in southeastern Spain that led to 24 h average PM₁₀ and PM_2.5 concentrations within the range 70–260 and 25–43 µgm-3 (Fig. 10a1 and b1). On 15 March 2022, the massive dust plume moved from southeastern Spain, where it led to 24 h average PM₁₀ values within the range 580–3070 µgm-3 toward the west and northwest of Iberia, resulting in (24 h average) PM₁₀ concentrations within the ranges 825–950 µgm-3 in central Spain, 600–650 µgm-3 in central Portugal, and 440–810 µgm-3 in northern Portugal and northwestern Spain (Fig. 10a2). Concentrations of PM_2.5 (24 h average) were within the range 139–690 µgm-3 in southeastern Spain, 25–60 µgm-3​​​​​​​ in central Portugal, 100–260 µgm-3 in central Spain, and 50–130 µgm-3 in central northern Portugal and northwestern Spain (Fig. 10b2). These extremely high PM_x concentrations were also recorded indoors (Gomes et al., 2022). During 16 March 2022, high PM₁₀ and PM_2.5 values were still recorded (Fig. 10a3 and b3), with the highest PM₁₀ concentrations linked to the still ongoing dust inflow in southeastern Spain (800 µgm-3) and the southward transport of dust in southern Portugal (300–330 µgm-3). Because of the massive dust load, solar energy production in Spain dropped by 50 % (Micheli et al., 2024). Details on this event, such as the maximum hourly PM₁₀ and PM_2.5 concentrations (Fig. S2) and the names of the AQMS, are provided in the Supplement.

The analysis of the 2000–2022 time series of (24 h average) PM₁₀ (Fig. 11) and PM_2.5 (Fig. 12) data show that the duxt events we report here are record-breaking episodes in mainland Spain, continental Portugal and the Canary Islands. The massive dusty air mass that blackened the Iberian Peninsula during 15 and 16 March 2022 (dx-06; Figs. 11 and 12) resulted in the highest PM₁₀ and PM_2.5 concentrations ever recorded on a regional scale across northern Spain (Cantabria region; Fig. 11a), central northern Spain (Castile y León region; Figs. 11c and 12a), central Spain (Castilla–La Mancha, Extremadura and Madrid region; Fig. 10d and a), southern Spain (Andalusia region; Figs. 11e, 10f and 12c) and continental Portugal (Fig. 11g). In central Spain (Castilla–La Mancha, Extremadura and Madrid), regular Saharan dust events typically induce (24 h average) PM₁₀ concentrations within the range 40–140 µgm-3 (highlighted with black arrows in Fig. 11d) (Pey et al., 2013; Rodríguez et al., 2001). Anomalous intense dust events such as that which occurred on 22 February 2016 (Sorribas et al., 2017) resulted in PM₁₀ concentrations 200–380 µgm-3, whereas during the dx-06 event AQMSs of this region recorded (24 h) PM₁₀ concentrations within 300–949^x µgm-3 (^x Villa del Prado AQMS in the Madrid region) (white arrow in Fig. 11d). After analyzing the 2001–2011 time series, Pey et al. (2013) concluded that Saharan dust events inducing PM10>100 µgm-3 (24 h average) are actually rare in the western Mediterranean. The impact of the dx-06 event is not observed in Cataluña since it did not reach northeastern Spain (Fig. 11b).

On the Canary Islands, the most intense Saharan dust events ever recorded occurred in the period 2020–2022 and were linked to the duxt events described in this study. Since 2005–2020, intense Saharan dust events have regularly been associated with (24 h average) PM₁₀ and PM_2.5 values with 200–400 µgm-3 (Fig. 11h) and 100–200 µgm-3 (Fig. 12d), respectively. This PM₁₀ range is similar to (1) that of total suspended particles during Saharan dust events of the period 1998–2003, based on AQMS data not yet normalized to the European standards (Alonso-Pérez et al., 2007, 2012; Viana et al., 2002), and to (2) that of total dust at Izaña Observatory (Tenerife) during the 1987–2014 Saharan dust events (Rodríguez et al., 2015). In contrast, in the period 2020–2022, the duxt events described here led to PM₁₀ and PM_2.5 concentrations within the ranges (24 h average) 600–1840 µgm-3 (Fig. 11g) and 200–404 µgm-3 (Fig. 12d), respectively. The three most intense dust events ever recorded on the Canary Islands, exceeding the threshold of 600 µgm-3 of PM₁₀ as 24 h average, in descending order of magnitude, are as follows.

The dx-02 event on 23 and 24 February 2020, where the (24 h average) PM₁₀ values averaged at all AQMSs were between 531 and 930 µgm-3 (average of 34 AQMSs distributed over the seven Canary Islands). On 22, 23 and 24 February 2020, a total of 6, 25 and 12 AQMSs recorded a (24 h average) PM₁₀ concentration between 600 and 1840^x µgm-3 (^x Gran Canaria, Playa del Inglés AQMS). Prior to this event, the (24 h average) PM₁₀ concentrations had only exceeded 600 µgm-3 at just one AQMS (618 µgm-3 at Las Galanas during the dust event on 28 March 2018; Fig. 11g).

In mainland Spain and continental Portugal, the dx-06 event is the most intense dust event ever recorded. In Spain, a total of 20 AQMSs distributed across southeastern, central and central northern Spain registered (24 h average) PM₁₀ concentrations within the range 600 to 3070^x µgm-3 (^x Mediterraneo AQMS in Almería province). In Portugal, four AQMSs located in the central regions registered (24 h average) PM₁₀ concentrations within the range 600 to 648^x µgm-3 (^x Chamusca AQMS in Lisbon and Vale do Tejo). The 2-decade time series of PM_x also offers other interesting data. In many regions of Spain there is a clear decreasing trend in PM₁₀ concentrations from 2000 to 2020 linked to a reduction in emissions following introduction of air quality policies (Fig. 11a, c, and e) (Li et al., 2018; Querol et al., 2014), suggesting that desert dust may have an increasing relative contribution to PM₁₀ concentrations, as is also indicated by recent projections (Gomez et al., 2023).

In winter, North African dust is regularly transported southward to tropical latitudes (Merdji et al., 2023). In this season, dust transport northward, to the subtropical North Atlantic and southern Europe, occurs during rather short periods under specific meteorological scenarios described in previous studies (Flaounas et al., 2015; Fluck and Raveh-Rubin, 2023a; Rodríguez et al., 2001). Winter extreme Saharan dust events have been observed in the southern Sahara, from Mauritania to Niger along the Sahel, associated with PM₁₀ and PM_2.5 concentrations within the ranges 800–5000 and 600–1300 µgm-3 (Fluck and Raveh-Rubin, 2023b; Marticorena et al., 2010), respectively, induced by H-to-L dipoles, meaning an anticyclonic high-pressure high core over the Atlantic and western North Africa and a cutoff low over the Mediterranean, which is a meteorological configuration that results in strong southern winds (Fluck and Raveh-Rubin, 2023b).

The six 2020–2022 duxt events we report here were induced by the L-to-H meteorological dipoles formed by cyclones located at the southwest of a blocking anticyclone over western Europe. Figure 13a1–f1 show the anomaly of the 500 hPa geopotential height during the onset of the six duxt events (Fig. 13a2–f2). All duxt events occurred during Northern Hemisphere meteorological anomalies that resemble the anomalies in atmospheric circulation that have been linked in previous studies to global warming (Fig. 13): (i) subtropical anticyclones expanded and shifted to higher latitudes (Cherchi et al., 2018; Cresswell-Clay et al., 2022); (ii) anomalous low pressures expanding northward beyond the tropical belt, resembling tropical expansion (Seidel et al., 2008; Yang et al., 2020, 2023) (e.g., dx-01, dx-02, dx-03 and dx-04); and (iii) mid-latitudes amplified Rossby waves due to the concatenation of cutoff-low cyclones and anticyclones, pointing to a weakening of the polar vortex (e.g., dx-03, dx-05, dx-06 and the event on 23 March 2022) (Mann et al., 2017; Screen and Simmonds, 2013). The duxt events observed in the eastern Mediterranean in March 2018 (Solomos et al., 2018) and in March 2020 (Mifka et al., 2023), in North Africa in June 2020 (Bi et al., 2023; Francis et al., 2020), in Uzbekistan in November 2021 (Xi et al., 2023), and in China in March 2021 (Gui et al., 2022; Liu et al., 2023) occurred in the context of cyclones, blocking anticyclones and dipoles linked to amplified mid-latitude Rossby waves.

The winter blocking anticyclone that we observed over western Europe and the western Mediterranean during the duxt events (Fig. 13a1–g1) fits with the picture of the industrial-era eastward expansion and shift of the North Atlantic anticyclone starting in the 1850s and accelerating over the last few decades (Alonso-Pérez et al., 2011a; Cresswell-Clay et al., 2022), a trend that is expected to continue as the concentrations of greenhouses gases increase according to the CMIP5 multi-model simulations (Cherchi et al., 2018). The low pressures to the southwest of the blocking anticyclone that we observe during the duxt events (Fig. 13b1–e1) are expected to increase as the tropics expand northward in the forthcoming decades (e.g., Fig. 1f of Cherchi et al., 2018, for the 2075–2100 period).

The anomalies linked to the duxt events are also evident in the trajectory of the cyclones that finally form the low-to-high dipoles leading to the extreme dust events. Red circles in Fig. 13a1–g1 indicate the location of the cyclones from days before (negative number) to days after (positive number) the duxt event. Due to the anticyclonic blocking over western Europe, the mid-latitude North Atlantic cyclones did not follow their regular path across Europe or the Mediterranean. The cutoff lows forming the low-to-high dipoles reached this region (Canary Islands, Cabo Verde or inner Sahara) via the following two main paths.

They deviated southward from the regular mid-latitude westerly circulation in the North Atlantic as a result of the blocking anticyclone over western Europe. Subsequently, these cutoff lows may stay over long periods (up to 12 d) in the subtropical and tropical northeastern Atlantic (near the Canary Islands and Cabo Verde). This is the case for the dipoles of events dx-02, dx-04, dx-05 and dx-06 and also for the intense dust event on 23 March 2022. The cyclones of the events dx-02, dx-04 and dx-05 stayed (blocked by the anticyclonic situation) near the Canary Islands and Cabo Verde for 10, 12 and 8 d (Fig. 13b1, d1 and e1), respectively.

The observed sharp increase in dust transport to the western Euro-Mediterranean region during the 2020–2022 winter seasons has been also been associated with the meteorology linked to dipoles and blocking anticyclones (Cuevas-Agulló et al., 2024).