The HYSPLIT model provided by the Australian Meteorological Agency and the National Oceanic and Atmospheric Administration Air Resources Laboratory is performed to study dust aerosol’s transport path. In this paper, the meteorological data input for the HYSPLIT model is derived from the Climate Data Center 1 (CDC1) meteorological data, and other information such as the transport altitudes of dust aerosol is obtained from the products of CALIPSO VFM.

The WRF-Chem model (Weather Research and Forecasting model coupled to Chemistry, version 3.9.1) is widely used in simulating the transport and the deposition amount of dust aerosols (Chen et al., 2017). In this paper, to cover the major source of dust aerosol (Taklimakan desert) and the deposition area for dust aerosol (the Northwestern Pacific Ocean), we set the model domain with 300 × 180 grids, the center of the domain is geographically located at 125°E and 40°N (black box in Figure 1), the vertical grid has 30 levels from 50 hPa to the ground. The meteorological fields input for the WRF-Chem model are provided by NCEP Climate Forecast System Version 2, the spatial and temporal resolution for meteorological data are 0.5 ° × 0.5 ° and 6-hourly. The chemical species’ initial and boundary conditions are derived from Community Atmosphere Model with Chemistry, and the anthropogenic emissions were set based on Emissions Database for Global Atmospheric Research inventory data. The simulation period in our study is 9–30 May 2019, and we set the first 7 days of simulation as a model spin-up for the dust case in our research.

In this study, we calculated the vertical sub-grid-scale turbulent flux caused by vortex transport using the boundary layer scheme of Yonsei University, and combined the Monin-Obukhov surface scheme with the Noah land surface scheme (Liu et al., 2016). Morrison’s two-moment microphysics scheme and Grell-Freitas integration scheme are used to simulate cloud microphysics and convection processes. We used GOCART scheme (Ginoux et al., 2001) to simulate and calculate dust emissions, and adopted four-dimensional assimilation method in the process of model simulation to reduce meteorological simulation errors. The mean bias is 1.30 m s⁻¹, which is within the acceptable range (Kumar et al., 2014). The model simulation results were adopted to determine the consistency between the increase of Chl-a concentration and dust deposition in the study area.

Aerosol optical depth (AOD) can present the spatial distribution of atmospheric aerosol, especially dust (Liu et al., 2013; Zhang et al., 2019). Figure 2 shows the distribution of daily AOD observed by Himawari-8 satellite, we can see that there is a severe aerosol event in the Taklimakan Desert beginning on 21 May 2019 according to the AOD data (shading areas in Figure 2), aerosol layers with AOD greater than 0.8 move eastward with the wind according to the distribution of wind fields at 500 hPa (black arrow in Figure 2), passing over the Sea of Japan, and reach the northwest Pacific Ocean on 24 and 25 May as shown in Figure 2. During the whole period, the maximum value of AOD can reach 1.2 observed by the Himawari-8 satellite.

Figure 3 is the vertical distributions of aerosol types obtained by CALIPSO VFM products during 21–25 May 2019, its scan trajectory during this process is shown in Figure 4. As shown in Figure 3, dust aerosols distribute from 1 km to about 7 km over the dust source region (Taklimakan Desert). In the downwind region, dust aerosols distribute from 1 km to about 6 km. When dust aerosols transport to the Pacific Ocean, they distribute from 0 km to about 10 km. Based on the CALIPSO VFM results, dust aerosols account for more than 85% of all aerosols during 21–25 May 2019. Thus, dust particles are the dominant aerosol type in the whole process, the values of AOD in Figure 2 can be related to dust concentration in the atmosphere.

A forward trajectory analysis simulated by the HYSPLIT model is conducted to find the transport paths of dust aerosols during 21–25 May 2019. The duration for the model is 93-h, and the initial height for the case is 1.5, 2.5, and 3.5 km where dust aerosols are observed on 21 May according to the results in Figure 3A. The CDC1 is used as the meteorological input at 20:00 UTC on 21 May. The model simulation shows that dust aerosols originate from the Taklimakan Desert on 21 May 2019, then spread eastward, passed over the east of China, and reach the western part of the Pacific Ocean on 24 and 25 May 2019, consistent with the results in Figure 2.

Meteorological condition plays an important role in dust aerosol horizontal transport (Uno et al., 2009). Figures 6A–E presents the potential height field and temperature field at 500 hPa during 21–25 May 2019. We can see that there is a strong cold advection at the height of 500 hPa, which is conducive for sweeping dust aerosol from the northwestern region across Northeast China, and reaching the Pacific Ocean.

To sum up, based on the results in Figures 2,3, Figures 5,6, dust aerosols rise in the Taklimakan Desert, moving eastward at high altitudes under the effect of the westerly winds, passing over the downwind regions, and deposition in the Pacific Ocean.

The spatial distribution of Chl-a concentrations from 24–30 May 2019 is studied with the observations from the Himawari-8 satellite which is in geostationary earth orbit (GEO). The satellite in GEO can provide high temporal resolution of Chl-a concentrations compared to polar orbit satellites. Figure 7 shows the spatial distribution of Chl-a concentrations from 24–30 May 2019. It is obvious that Chl-a concentrations massively increase on 26 May due to the deposition of dust, and lasted until 28 May. However, there are some missing values of Chl-a concentrations because of the existence of clouds, because Himawari-8 observes the Chl-a concentrations in the visible band. To better investigate the dust aerosol’s effect on Chl-a concentrations, we define a relatively small region (45–50°N, 145–155°E) as shown in Figure 7 (Black box) with data relatively integrity as the research region.

The seawater velocity represents the horizontal movement of the ocean. To investigate the oceanic transport effect on the position of deposition dust aerosols, we use seawater velocity derived from the GOFS3.1 product to calculate the longitudinal and latitudinal displacement for the research region. Figure 8 shows the eastward and northward seawater velocity in the surface ocean (depth <100 m) from 12 May to 4 June 2019. The positive value means seawater flows eastward or northward. The total longitudinal and latitudinal displacement from 24 May to 4 June is about 80 km based on the results in Figure 8, respectively. Thus, the change in longitude and latitude caused by seawater movement is less than 6.7 km/day. However, the area for the research region is roughly 550 km (longitude length) and 750 km (latitudinal length). Therefore, we conclude that the movement range of dust deposition carried by seawater movement is far smaller than that of the research region, so we assume that dust deposition remains in the research region and provide nutrients for phytoplankton.

Besides nutrient input present in the aerosols, PAR and SST also can impact the Chl-a concentrations (Luo et al., 2020; Wang et al., 2021). To evaluate those factors’ effect on Chl-a concentrations, the variation of the daily mean values of those factors (such as PAR, dust deposition amount, SST) are drawn over the research region using the Himawari-8 data, ERA-5 datasets, and the merged L3 global ocean color products during 24 May to 4 June 2019 (Figure 9).

In Figure 9, the dotted lines are the 5-year averaged values for Chl-a concentrations (black) and POC (blue) over the same period in the research region. The WRF-Chem model simulate that the average dust deposition is 179.79 μg m⁻² (Table 1) on 24 May, and the simulation results began to decrease on 25 May (463.84 μg m⁻²), which was consistent with the change of Himawari-8 AOD from 24 May (0.22) to 25 May (0.36). We also can see that the Chl-a concentrations increased on 26 May after the dust deposition on 24 May as shown in Figure 9. The response time is 2 days, which is similar to the iron enrichment experiment (Tsuda et al., 2003). However, dust aerosols continuously deposit into the ocean from 24 to 30 May, and the Chl-a concentrations only increase from 26 to 28 May as shown in Figure 9 and Table. 1. One reason for this phenomenon is the missing observation of Chl-a concentrations due to clouds and other factors on 29 May, another reason is that the nutrient needed by the phytoplankton is oversaturated, so Chl-a concentrations do not increase after 30 May. We also can see that Chl-a concentrations have two peaks during this period, this is due to the missing values of Chl-a concentrations. Furthermore, during the response process, Chl-a concentrations increase from 0.56 mg m⁻³ (25 May) to 1.54 mg m⁻³ (26 May), which is an increase of nearly 175%. Moreover, the Chl-a concentrations (1.54 mg m⁻³) are 256% larger than their mean values during the same periods from 2016 to 2019 (0.433 mg m⁻³). After 29 May, the values of Chl-a concentrations decrease to less than 0.45 mg m⁻³.

Suitable SST is also helpful in increasing Chl-a concentrations. In our case, the mean values of SST are 4.13, 4.01, 3.78, and 3.97 during 25–28 May. It is noted that SST shows a downward trend during Chl-a concentrations increase, the decrease in SST could easily be associated with oceanic upwelling, which can bring nutrients to the surface ocean. To investigate the possible contribution of upwelling in this case, we analyzed the upwelling velocity from SODA 3.4. Our results indicate that the mean value of upwelling velocity is about 0.053 m/d from 24 May to 28 May, which is lower than the mean value of upwelling velocity (0.085 m/d) in May 2019. Thus, the nutrient provided by the rising movement of seawater is lower than the mean values of nutrients in May 2019. Therefore, both the nutrients transported by oceanic upwelling and SST are not the main factors that lead to the increase in Chl-a concentrations.

Iwasaki (2020) indicate that typhoon also can directly or indirectly affect the Chl-a concentration in the Pacific Ocean through precipitation and strong sea surface winds. However, there was no typhoon process during May 21 to 25 May 2019 as shown in Figure 6, thus typhoon is not a factor that lead to the change in Chl-a concentrations for this case.

On the other hand, PAR presents a slight increase trend on 26 May compared with the value on 25 May, then shows a decreasing trend during 26 May to 30 May. During this period, the maximum value of Chl-a concentrations occur on 26 May, which means that dust deposition and PAR both promote the growth of phytoplankton on 26 May. However, the value of PAR on 25 May (55.76 Einstein/(m²·day)) is larger than that on 27 May (48.29 Einstein/(m²·day)), while the value of Chl-a concentrations on 25 May (0.56 mg m⁻³) is smaller than that on 27 May (0.90 mg m⁻³) as shown in Table. 1, suggesting PAR is not the main factor that leads to the increase in Chl-a concentrations for this case. To sum up, based on the results discussed above, dust aerosol deposition could be the major reason for the increase in Chl-a concentrations in this case.

When dust deposition stimulates the growth of phytoplankton, it can trigger ocean biological pumps, leading to more absorption of CO₂ from the atmosphere to the ocean (Pan et al., 2011). In this paper, we use the variations in POC to assess marine biological pumps caused by dust aerosol deposition. As shown in Figure 9, POC shows a similar trend to those of Chl-a concentrations. the POC increased on 26 May after the dust deposition on 24 May. The response time is 2 days, which is similar to the iron enrichment experiment (Tsuda et al., 2003). In addition, the POC has two peaks during this period, which is consistent with Chl-a concentrations. During the response process, POC increased from 162.36 mg m⁻³ (25 May) to 1,580.52 mg m⁻³ (26 May), increasing nearly 873%. Moreover, the value of POC (1,580.52 mg m⁻³) was 644% higher than its mean value in this period from 2016 to 2019 (212.25 mg m⁻³).