Global SST patterns and SST anomalies during El Niño events of different temporal and spatial types are constructed using the merged Hadley–NOAA/OI dataset, which has a horizontal resolution of 1∘×1∘ from 1870 to 2017 (Hurrell et al., 2008). The monthly ERA5 reanalysis data (Hersbach et al., 2020) are applied to evaluate the simulated meteorological parameters during El Niño events.

Hourly observations of PM10 (particulate matter less than 10 µm in diameter) concentrations in China from 2015 to 2021 derived from the China National Environmental Monitoring Centre (CNEMC) and the Deep Blue aerosol products (Platnick et al., 2015) from the Moderate Resolution Imaging Spectroradiometer (MODIS) on the Terra satellite, including monthly aerosol optical depth (AOD) at 550 nm and the Ångström exponent (α) from 2001–2020, are applied to evaluate the performance of dust simulation in the model. The satellite dust optical depth (DOD) is calculated following Yu et al. (2021).

We first clarify the definition of different temporal and spatial types of El Niño events here. The notation of year0 is used to denote the first year of El Niño development, and Jan0, Feb0,…, and Dec0 indicate the individual months of that year, while year1,2,… and Jan1,2,…, Feb1,2,…, …, and Dec1,2,…, respectively, denote the following years and months therein. The Niño 3.4 index is defined as area-mean anomalies of detrended SST in the Niño 3.4 region (5 ∘ S–5 ∘ N, 170–120 ∘ W). The Niño 3/4 index (INiño3/INiño4) is the same as the Niño 3.4 index, but in the Niño 3/4 region (5 ∘ S–5 ∘ N, 150–90 ∘ W; 5 ∘ S–5 ∘ N, 160 ∘ E–150 ∘ W).

For the classification of different temporal types, following Wu et al. (2019), El Niño events are firstly selected if any of the 3-month running average Niño 3.4 index during Oct0–Feb1 is greater than 0.75 ∘C. Then the LD El Niño event is identified once any of the Niño 3.4 index during Oct1–Feb2 is higher than 0.5 ∘C; otherwise, it is a SD El Niño event.

Following Yu et al. (2019), the El Niño events, selected with 3-month running average Niño 3.4 indices higher than 0.5 ∘C for five consecutive months, are classified into different spatial types based on the EP El Niño index (IEP) and the CP El Niño index (ICP). The definition of these indices is given below. 1IEP=INiño 3-α×INiño 42ICP=INiño 4-α×INiño 33α=0.4,INiño 3×INiño 4>00,INiño 3×INiño 4≤0 If the mean IEP is greater than the ICP during Oct0–Feb1 of an El Niño, then it is an EP El Niño event; otherwise, it is a CP El Niño event. Note that mixed El Niño events that are not considered separately in this study also exist.

The time series of the Niño 3.4 index derived from Hadley–NOAA/OI 1870–2017 data is shown in Fig. S1 in the Supplement. Using the definitions described above, for El Niño with different temporal types, 22 SD El Niño events and 8 LD ones are extracted during this time period; for El Niño with different spatial types, 26 EP El Niño events and 8 CP ones are extracted. The mechanisms leading to different types of El Niño are given in Sect. S1 in the Supplement.

To investigate the impacts of El Niño of different spatial and temporal types on dust aerosol in China, this study utilizes the US Department of Energy (DOE) E3SMv1 (Golaz et al., 2019). As a model developed from the well-known CESM1 (Community Earth System Model version 1), E3SMv1 provides significant improvements to the atmospheric component, including processes associated with aerosol, cloud, turbulence, and chemistry (Rasch et al., 2019). We choose the horizontal resolution of about 1∘ and 30 vertical layers. E3SMv1 predicts aerosols including mineral dust, sea salt, sulfate, primary and secondary organic aerosols, and black carbon in the four-mode Modal Aerosol Module (MAM4) (Wang et al., 2020). E3SMv1 represents dust-related processes in the atmosphere and land model components (Feng et al., 2022). Dust emissions are calculated at each model time step according to the wind erosion dust scheme proposed by Zender et al. (2003), which is related to 10 m wind speed, surface soil moisture content, soil erodibility, vegetation cover, and threshold friction velocity.

The following simulations are performed. A “CLIM” experiment applying the prescribed climatological mean of monthly SST during 1870–2017 is integrated for 30 years. Four sets of sensitivity simulations, “SD”, “LD”, “EP”, and “CP”, are driven by the monthly SST representing the composite of SD, LD, EP, and CP El Niño events, respectively, which is generated by adding the mean monthly SST anomalies from Jul0 to Jun1 of the SD, LD, EP, and CP El Niño events (Fig. S1), respectively, to the climatological SST between 60 ∘ S and 60 ∘ N. All the sensitivity experiments have three ensemble members with diverse initial conditions branched from different years of the CLIM simulation, and the results are based on the ensemble mean. All sensitivity experiments are run for 13 years, with the first 3 years as model spin-up and the last 10 years used for analysis. The differences in model fields between the sensitivity simulations and CLIM represent the influences of El Niño events with different spatial and temporal types on dust aerosols. All other external factors such as greenhouse gas concentrations, insolation, anthropogenic aerosols, and their precursor emissions are held at present-day conditions (year 2014). The SST anomalies relative to the 1870–2017 climatology during SD, LD, EP, and CP El Niño events are shown in Fig. 1.

To evaluate the model performance in dust simulation, we compare the simulated near-surface dust concentration and dust optical depth (DOD) over China with observed PM10 concentrations and satellite-retrieved DOD, respectively. The model can reproduce the spatial distribution of springtime dust in China, with high dust concentrations in northwestern China and low dust concentrations in southern and northeastern China (Fig. S2). The spatial correlation coefficient between the simulated dust concentrations in E3SMv1 and observed near-surface PM10 concentrations is +0.55. However, the model strongly overestimates dust concentrations over the source regions, which were also reported in many previous studies using the E3SMv1 and CESM (the predecessor of E3SMv1) (Wang et al., 2020; Wu et al., 2019). The high model bias near the sources is also confirmed by comparing DOD between model simulation and satellite retrieval. It suggests that the dust emissions are overestimated in northwestern China in the model. The high bias is partly related to the dust treatment in the model that dust is emitted into a shallow model bottom layer in E3SMv1 for increased model vertical resolution (Wang et al., 2020). In addition, stronger 10 m wind speed simulated by the model compared to the observation (Fig. S3) also contributes to the higher dust loading. However, we also note that the E3SMv1 underestimates the transport of dust from source regions (Wu et al., 2020; Feng et al., 2022); thus the dust over eastern China is comparable to observations.

The simulated effects of the four types of El Niño with different spatial positions (EP and CP) and durations (SD and LD) on the DJF ground-level dust concentrations are shown in Fig. 2. As for different spatial types of El Niño events, the effects on DJF dust concentrations in China are similar, with an increase in dust concentrations of 5–50 µg m-3 over central-eastern China during EP and CP El Niño compared to the climatological means. The spatial pattern of dust changes is consistent with previous modeling studies (Lee et al., 2015; Li et al., 2021). Although the influences of EP and CP El Niño on the DJF dust concentrations resemble each other in the spatial patterns over China, the magnitudes of the influences are different. During CP El Niño relative to the climatological mean, dust concentrations increase more significantly over central-eastern China, with values of 20–50 µg m-3, which is 5–20 µg m-3 higher than those during EP El Niño relative to the climatological mean. The large increase during CP El Niño relative to the climatological mean is also more widespread than that during EP El Niño relative to the climatological mean. Compared to CP El Niño, dust concentration over central-eastern China decreased slightly during the EP El Niño, but the changes are mostly insignificant.

As for different temporal types of El Niño events, their effects on DJF dust concentrations over China are quite different. SD El Niño events cause an increase in DJF near-surface dust concentrations of 20–100 µg m-3 in northern China and about 5–20 µg m-3 in southern China, whereas during LD El Niño events, winter dust concentrations have a decrease of about 5–50 µg m-3 in northern and northeastern China relative to the climatology, and no significant change is shown in southern China. In contrast to LD El Niño events, SD El Niño events have positive DJF dust concentration anomalies of 5–20 µg m-3 in southern China and a maximum over 100 µg m-3 in northern China and the Gobi Desert. Furthermore, DJF dust concentrations over the Taklamakan Desert, one of the largest dust sources in China, have an increase during LD El Niño events and an insignificant decrease during SD El Niño events.

Overall, these changes in dust concentrations indicate that CP El Niño events have stronger and more widespread impacts on DJF dust concentrations than EP El Niño relative to the climatological mean, and the SD and LD El Niño events exert opposite impacts on DJF dust in China.

Meteorological factors such as 10 m wind speed, relative humidity, and atmospheric circulation play a dominant role in altering dust concentrations by altering emissions, atmospheric transport, and wet scavenging of dust (Csavina et al., 2014). Dust changes are also controlled by the El Niño-related changes in atmospheric circulation and precipitation (Gong et al., 2006; Hara et al., 2006). The 10 m wind speed, atmospheric circulation, relative humidity, precipitation anomalies, and related processes during EP, CP, SD, and LD El Niño are investigated here to reveal the mechanisms of the influence of the four types of El Niño on dust over China.

During the CP, EP, and SD El Niño, DJF mean 10 m wind speed increases in the Gobi Desert and northwestern China compared to the climatological mean (Fig. 3), which favors the local dust emission over these regions, whereas for the LD El Niño event, the positive 10 m wind speed anomaly is greatly weakened compared to the other three types of El Niño events, and negative 10 m wind speed anomalies are triggered in the Gobi Desert and northern China (Fig. 3e), which is not conducive to dust emission during the LD El Niño events. The CP El Niño events trigger stronger positive 10 m wind speed anomalies (0.1–0.3 m s-1) than the EP El Niño events over the Gobi Desert and northern China (Fig. 3c), which could lead to a greater local dust emission. Compared to the LD El Niño, SD El Niño events produce significant positive 10 m wind speed anomalies of approximately 0.3 m s-1 in the Gobi Desert and northern China (Fig. 3f), which is consistent with the increase (decrease) in local DJF dust concentrations during the SD (LD) El Niño (Fig. 2). This suggests the importance of 10 m wind speed in the dust changes during the El Niño events in China.

Figure 4 shows the atmospheric circulation anomalies for the four El Niño events. All types of El Niño have negative anomalies of sea level pressure (SLP) in central-eastern China, except the LD El Niño, which shows a negligible SLP change in winter. Meanwhile, during the EP, CP, and SD El Niño events, anomalous Mongolian cyclones can strengthen the local ascending flow to lift more dust particles into the free atmosphere. The anomalous northwesterly during CP and SD El Niño (Fig. 3b and d) can transport these dust aerosols to central-eastern China, leading to the strong increases in dust concentrations there (Fig. 2b and d), while during the LD El Niño, the lower atmosphere in the Gobi Desert and northern China is controlled by a weak anomalous high pressure accompanied by an anomalous southeasterly that weakens the prevailing northwesterly in winter and hinders the vertical lifting and southward transport of dust.

Our previous work has confirmed the ability of E3SM in capturing the atmospheric circulation over central-eastern China in El Niño with different durations (Zeng et al., 2021). Here we further evaluate the circulations in E3SM simulations during EP and CP El Niño events by using ERA5 reanalysis data. The anomalous DJF mean 10 m wind speed and 850 hPa wind fields in the typical EP El Niño (2006/07) and CP El Niño (2014/15) relative to the climatology (1950–2017) from ERA5 are presented in Fig. 5. Although the increase in 10 m wind speed over northwestern China in the EP El Niño simulated in the model is inconsistent with the ERA5 results, E3SM does capture the large increase in wind speed over the Gobi Desert during the CP El Niño relative to the climatological mean and EP El Niño. Moreover, the anomalies in wind fields during EP and CP El Niño (i.e., anomalous southerly during EP El Niño and anomalous northwesterly during CP El Niño) can be simulated by E3SM. It suggests that the atmospheric circulation features over central-eastern China during different types of El Niño are roughly captured by the model. However, we note that there are notable differences in atmospheric circulation over many regions of East Asia. It can be partly attributed to the model bias in reproducing the atmospheric responses to El Niño. The observations can also be induced by other climate factors besides El Niño, leading to a potential inconsistency in El Niño impact between model and observation.

The effect of relative humidity (RH) on dust concentration is also essential, considering that a decrease in RH leads to a decrease in the threshold friction velocity at high RHs (>40 %), which further enhances dust emission flux and atmospheric concentration (Csavina et al., 2014). Both EP and CP El Niño events have negative anomalies in DJF RH in the Gobi Desert (Fig. 6a and b). The decrease in RH reduces the dust threshold friction velocity and favors dust emission from the Gobi Desert. The CP El Niño produces more pronounced and widespread negative RH anomalies in the Gobi Desert and northwestern China than the EP El Niño. It gives approximately 3 % stronger negative RH anomalies (Fig. 6c), resulting in stronger and more widespread increases in DJF dust concentrations during the CP El Niño event (Fig. 2c). As for El Niño with different duration, the SD El Niño leads to significant decreases in DJF RH of about 3 % near the southern part of the Gobi Desert, while increases in RH are located over the northern part of the Gobi Desert during the LD El Niño (Fig. 6g and j), likely resulting in opposite changes in dust emissions. The ERA5 reanalysis data also show the same RH variations during the different spatial and temporal types of El Niño as the E3SM simulations described above (Fig. S4). Among all four types of El Niño events, RH anomalies are consistent with the distribution of dust concentration anomalies, which indicates that RH plays an important role in affecting variations in dust emissions and concentrations in China during El Niño.

Figure 7 shows the simulated changes in DJF dust emissions during different El Niño events. During the EP and CP El Niño, DJF dust emissions are enhanced in the Gobi Desert and northwestern China relative to the climatological average. The dust emission increase is larger during the CP El Niño than the EP El Niño, which is consistent with the higher positive DJF dust concentration anomalies during the CP El Niño. Furthermore, the SD El Niño causes a significant increase in dust emissions of about 0.5 g m-2 d-1 in the Gobi Desert compared to CLIM, while the LD El Niño causes a decrease in dust emissions. These suggest that different types of El Niño events alter the DJF dust emissions in China by changing the 10 m wind speed and RH, which is an important cause of the variation in DJF dust concentrations in China.

Furthermore, a reduced DJF precipitation during both EP and CP El Niño events (Fig. S5) should weaken the wet removal of dust from the atmosphere in northern China. However, only insignificant decreases in wet deposition appear in part of northern China, and significant increases in wet deposition are located in central and southern China related to increases in dust loading during EP and CP El Niño events (Fig. S6). It suggests that El Niño impact on dust concentrations is mainly through changing the emission and transport of dust rather than the scavenging in winter.

The changes in near-surface dust concentrations over China in the following spring during the decaying phase for different spatial and temporal types of El Niño are also examined (Fig. 8). During the following spring, all El Niño events trigger large positive anomalies of March–April–May (MAM) dust concentrations in northern China. However, the increases in dust concentrations during the EP, CP, and SD El Niño relative to the climatological average fail the 90 % significance test, indicating that the effects of these types of El Niño events on the dust pollution in northern China in the following spring are uncertain, likely related to the large internal variability in the climate system. In contrast to the strong reduction in dust concentrations over the Gobi Desert and northern China during the LD El Niño in DJF, the effect in MAM reverses to a significant increase in dust concentrations over these regions by 50–100 µg m-3 (Fig. 8e). It suggests that the weaker intensity but longer duration of LD El Niño than the SD El Niño can significantly affect spring dust aerosols in China.

During LD El Niño events, MAM 10 m wind speed significantly increases over the Gobi Desert (Fig. S7), which facilitates the local dust emissions, although RH only shows an insignificant decrease over the dust source region (Fig. S8). It can be confirmed by the significant increases in MAM dust emissions by about 0.5 g m-2 d-1 over the Gobi Desert and northwestern China during LD El Niño events (Fig. 9). Then the strengthened northwesterly brings more dust to northern China during LD El Niño events (Fig. S9). Along the transport pathway, the weakened precipitation (Fig. S10) partly reduces the dust wet removal (Fig. S11), leading to the strong increase in MAM dust concentration over northern China during the LD El Niño. However, this effect is largely overwhelmed by the increased dust wet removal due to the emission-induced increase in dust concentrations.