The AR samples used for incubation experiments were collected with a cellulose acetate filter membrane (Whatman 41) on the Laoshan campus of Ocean University of China (36°9’39”N, 120°29’29”E) on 30 June 2019. During the sampling period, the AQI was 57–83 μg m^–3, indicating that the air quality is moderate. The detailed air quality conditions are shown in Table 1.

In the summer of 2019, three onboard microcosm experiments were conducted during the cruise of R/V Beidou in the YS and the northern part of the ECS (Figure 1A). The initial seawater at U1, U2 and U3 was collected from surface layers (∼3–5 m below the water surface) and SCM layers (captured by the CTD profile data) using Niskin bottles with Sea Bird CTD-General Oceanic Rosette assembly (Table 2). The site name is abbreviated as Ui_Sur and Ui_SCM, respectively, and i refers to the site number, i.e., 1/2/3.

Based on the content of N in AR aerosols, three treatments were conducted in triplicate for the incubation experiments: (1) control, no AR addition; (2) low AR addition, the added amount of AR was expressed in the unit of N (i.e., 1 μmol N L^–1) at U2, indicating that the added amount of AR contains 1 μmol N L^–1; (3) high AR addition, 1.7 μmol N L^–1at U3, and 2 μmol N L^–1 at U1. AR sample was first ultrasonically extracted in deionized water at 0°C for 1 h, and the leaching solution including particles was added to the incubation bottles directly (Guo et al., 2013). Apart from inorganic nutrients, the aerosol additions could stimulate mixotrophic dinoflagellates by promoting the utilization of organic matter (Granéli et al., 1999; Heisler et al., 2008; Lin et al., 2012). With N addition of 1 μmol L^–1, 1.7 μmol L^–1, and 2 μmol L^–1 by AR particles, the P concentrations added by AR particles were only 17.98, 30.56, and 35.95 nmol L^–1. The added amount of AR was determined according to the deposition flux of N (574.0–970.0 mg m^–2 event^–1) divided by an averaged mixed layer of 30 m (Shi et al., 2012; Sin et al., 2013). Nutrient enrichment experiments were set up to illustrate nutrient limitation in original seawaters and interpret phytoplankton response to AR additions (Supplementary Table 1).

The sampled seawater was passed through 200-μm sieves, mixed well in a clean 120-L high-density polyethylene barrel, and then dispensed into clean (acid-washed) 20-L polycarbonate incubation bottles. These incubation bottles were placed into three microcosm devices filled with continually updated surface seawater to keep the incubation system temperature relatively stable (Zhang et al., 2019). To explore the impact of microsized cells on the phytoplankton community, seawater from incubation bottles of each treatment was filtered through 20-μm sieves when the contribution of microsized Chl a to total Chl a in the control exceeded 20%, and then, the filtered seawater (hereafter FS) was transferred uniformly into three 2-L bottles to continue the cultivation. The incubations unfiltered with 20-μm sieves were defined as the unfiltered seawater (hereafter UFS, Figure 1B). Bottles were shaded to have approximately 40% light attenuation, matching light levels at depths on 3–5 m as previously used (Zhang et al., 2019, 2020). All experiments ran for 5–6 days.

We used one-way ANOVA to assess whether there was a significant difference in the Chl a concentration between the control and treatments (Andersen et al., 2020) and evaluated the nutrient limitation in surface seawater at the three sites. Statistical analysis was performed using IBM SPSS Statistics 20 (SPSS 20.0). CANOCO software (version 5.0) was used to analyze the relationships between environmental factors and phytoplankton. The detrended correspondence analysis (DCA) used species-sample data showed that the first axis of gradient was less than 3. Therefore, redundancy analysis (RDA) was the better choice.

In general, trophic statuses in surface seawater at U1_Sur, U2_Sur, and U3_Sur and in SCM seawater at U3_SCM were lower than those in SCM seawater at U1_SCM and U2_SCM. Phytoplankton at U1-3_Sur were colimited by N and P based on the significant increase in Chl a after N + P addition relative to the control treatment (on days 4–6, p < 0.05, Supplementary Figure 1).

At U1_Sur, the concentrations of N + N, PO₄^3–, and Si(OH)₄ were 0.19, 0.01, and 2.85 μmol L^–1, respectively. APA was 35.3 nmol (L h)^–1 in the original seawater, consistent with the reported value of ∼35 nmol (L h)^–1 in this region (Wang et al., 2014). The total Chl a concentration was as low as 0.38 μg L^–1, of which picosized cells (73%) contributed to the most of the total Chl a, followed by nanosized cells (18%) and microsized cells (9%) (Table 2). Chloropicophyceae and Dinophyceae codominated the community, with a relative abundance of 33 and 60%, respectively.

At U2_Sur, U3_Sur and U3_SCM, low concentrations of N + N and PO₄^3– were observed, i.e., 0.08-0.11 and ∼0.01 μmol L^–1, respectively. The APA was less than 18.4 nmol (L h)^–1, and the total Chl a concentration (0.55-0.62 μg L^–1) was more than 1.4 times higher than that at U1_Sur. Picosized Chl a contributed to the most of the total Chl a (47-65%), and the contribution of microsized Chl a was less than 15%. Dinophyceae dominated the communities with a relative abundance of ≥85% (Table 2).

Seawaters at U1_SCM and U2_SCM contained abundant nutrients, with 1.70-2.59 μmol L^–1 of N + N, 0.11-0.17 μmol L^–1 of PO₄^3–, and 4.78-4.90 μmol L^–1 of Si(OH)₄. The APA at U1_SCM and U2_SCM [3.4-3.7 nmol (L h)^–1] was of the same orders of magnitude as those of U2_Sur and U3_SCM. The total Chl a concentration (≥0.95 μg L^–1) was more than 1.7 times higher than that of U1_Sur and U2_Sur. Picosized phytoplankton (52-62%) were the primary contributors to the total Chl a. Dinophyceae dominated the communities with a relative abundance of ≥89% (Table 2).

The N:P ratio after AR addition increased from 11:1-20:1 to 44:1-54:1 at all sites in the surface incubated seawater and at U3_SCM in the SCM incubated seawater. Due to the sufficient nutrient stock in the original seawater at U1_SCM and U2_SCM, the N supplied by AR addition only increased the N:P ratio from 10:1-24:1 to 16:1-28:1. Because of the low contents of PO₄^3– and Si(OH)₄ in the AR, the changes in the concentrations of PO₄^3– and Si(OH)₄ in AR-amended seawater were slight at all sites.

During the incubations at U1-3_Sur and U3_SCM, the concentrations of N + N did not change significantly in the control and AR treatments (Figure 2). In the AR treatments, the N + N concentrations remained relatively stable and were significantly higher than those in the control treatments. In contrast, the concentrations of PO₄^3– were close to the detection limit in all treatments. The maximum consumption of Si(OH)₄ in the control and AR treatments was less than 13% at U1-3_Sur and U3_SCM. At U1_SCM and U2_SCM, the concentrations of N + N and PO₄^3– decreased gradually by more than 85% in the control and AR treatments relative to the original values. The concentrations of Si(OH)₄ decreased sharply in the control (>28%) and AR treatments (>69%) on days 3-5 at both sites (Figure 2 and Supplementary Figure 2).

There were no obvious differences in the concentrations of N + N, PO₄^3–, and Si(OH)₄ between FS and UFS in the control and AR treatments at U1-3_Sur and U3_SCM. In the FS at U1_SCM and U2_SCM, the consumption of N + N and Si(OH)₄ in AR treatments was lower than that in the UFS at the end of the incubations (p < 0.05, Figure 2 and Supplementary Figure 2).

In the UFS, the APA in the AR treatments was 1.3-75.5 times higher than that in the control at the end of the incubations (p < 0.05, Figure 3), and this phenomenon could be observed in both surface and SCM incubations. In contrast, there was almost no difference in APA between control and AR treatments on the last day of FS incubations at all sites except U1_SCM. For the AR treatments, the APA in the UFS at all sites was generally higher (1.6-7.3 times) than that in the FS (Figure 3).

At U1-3_Sur, the concentration of the total Chl a in the AR treatments was generally higher than that in the control. At U1_SCM, the total Chl a concentration in the AR treatments was more than 1.2 times higher than that in the control on days 2-5. At U2_SCM and U3_SCM, there was no significant difference in Chl a between the control and AR treatments (Figure 4). The responses of phytoplankton of different sizes varied with AR addition. At U1_Sur, the dominant size of phytoplankton changed from picosize to nanosize (46% contribution to total Chl a). A similar pattern in the size shift toward larger cells also occurred at U2_Sur and U3_Sur, although picosized cells always dominated the contribution to total Chl a (Figure 5). At U1_SCM and U2_SCM, the dominant contributor of phytoplankton in AR treatments was picosized cells on days 1-2 and shifted to microsized cells (≥49% contribution to total Chl a) on day 5. At U3_SCM, the dominant contributor was always picosized phytoplankton during the incubations (Figure 5).

For AR treatments, the concentrations of nanosized and picosized Chl a in UFS were generally higher than those in FS at U1_Sur and U3_Sur. Specifically, at U1_Sur, nanosized and picosized Chl a concentrations in the UFS enriched with AR were 2.4 and 1.8 times higher than those in the FS on day 5. At U3_Sur, the picosized Chl a concentration in UFS enriched with AR was 2.7 times higher than that in FS on day 6. Similar to the incubations with surface seawater, the picosized Chl a concentration enriched with AR in FS (0.10 μg L^–1) at U3_SCM was lower than that in UFS (0.25 μg L^–1) on day 6. In contrast, at U1_SCM, nanosized and picosized Chl a concentrations in FS (4.72 and 0.36 μg L^–1) were higher than those in UFS (2.89 and 0.22 μg L^–1) on day 5. At U2_SCM, there was no significant difference in the concentrations of size-fractionated Chl a between UFS and FS (Figure 4).

The ASVs at all sites assigned to phytoplankton could be classified into 25 groups of eukaryotic microalgae at class level (level 3). The ASV richness of Dinophyceae (dinoflagellates) accounted for ≥60% of phytoplankton in the original seawater at each site. In terms of UFS, Dinophyceae dominated the phytoplankton community in the AR treatments at all sites (Figure 6). The dominant class changed from Dinophyceae to Chloropicophyceae (71%) in the control on day 5 of the incubations (corresponding to the maximum Chl a concentration) at U1_Sur. Chloropicon spp. was the main component of Chloropicophyceae (Supplementary Figure 3). With AR addition, the relative abundance of Dinophyceae increased to 46% being the dominant phytoplankton (Figure 6). The succession of phytoplankton communities at U3_SCM was similar to that at U1_Sur. Dinophyceae maintained the dominant status in the control and AR treatments at the rest of the incubation sites (Figure 6).

Aerosol additions generally stimulated phytoplankton growth and shifted the phytoplankton size structure toward larger cells in surface incubated seawaters (Figure 5). This is justified because of the established supplementary relationship between nutrients (primarily N) supplied by AR and phytoplankton requirements. Such fertilization effect has also been widely reported in the previous studies (Cottingham, 1999; Liu et al., 2013; Zhang et al., 2019). In contrast, AR additions had a limited effect on phytoplankton size structure at U3_SCM, although its trophic status was similar to those in surface seawaters (Figure 5). This is ascribed to the photo-acclimation of phytoplankton under the condition of low irradiance in SCM layer (Fujiki and Taguchi, 2002; Fu et al., 2018). With the abrupt enhancement of light intensity (from SCM to surface), phytoplankton in the incubated seawater need to readjust to the new environment and thus showed a limited response to AR additions.

Interestingly, AR addition had a significant fertilization effect on phytoplankton growth at U1_SCM (Figure 4), which was characterized by the higher trophic status among these sites. Note that there was a shift in dominant phytoplankton from picosized cells to large cells during the incubations at U1_SCM, which was different from the sustaining dominance of picosized cells at U3_SCM (Figure 5). Moreover, large phytoplankton can better acclimate to the abrupt increase in light intensity compared with picosized cells, due to their stronger self-shading capacity by the pigment (package effect) to reduce light absorption (Marañón, 2015). At U2_SCM, the stimulation effect of AR addition was not as obvious as that at U1_SCM (Figure 4). This is because there was an obvious shift of dominant algae from Coscinodiscophyceae in the control to Dinophyceae in the AR treatments at U2_SCM (Figure 6). The obvious succession in phytoplankton community while slight change in Chl a under the condition of aerosol enrichment was also observed in eutrophic seawaters of the ECS (Meng et al., 2016). The substantial input of N relative to P supplied by AR addition increased the N:P ratio from 24:1 in the original seawater to 28:1 in the AR treatments at U2_SCM, which was more favorable for the growth of dinoflagellates (Zhang et al., 2019; Table 2). At U1_SCM, in contrast, the N:P ratio ranged between 10:1 and 16:1 in the control and AR treatments, leading to the increase in relative abundance of diatoms (primarily Coscinodiscophyceae) (Zhang et al., 2019). Collectively, in contrast to the consistent phytoplankton response to AR addition in surface seawater, the impact of AR addition in SCM seawater is complicated, which is closely related to nutrient concentration and structure in seawater.

The substantial N supplied by AR had the potential to alleviate and even alter N pressure of phytoplankton in the incubated seawater (Figure 2). As a result, the relatively P-deficient environment created by AR additions made it possible for phytoplankton to induce acclimatization mechanisms to copy with P stress (Moore et al., 2013). As shown in Figure 3, the APA value in the AR treatments was higher than that of the control at the end of the incubations at all sites, indicating that AR could enhance the utilization of DOP to increase P bioavailability in the incubated seawater. Such phenomenon was supported by the good correlation between PO₄^3– and APA (Figure 7). In apart from the establishment of P-deficient environment, AR additions also provided a considerable amount of soluble Fe and Zn, which acted as cofactors of phosphohydrolytic enzymes (Mills et al., 2004; Mahaffey et al., 2014). The promotion effect of AR addition on the utilization of DOP shows the acclimation of phytoplankton to the overwhelming N input relative to P in coastal waters (Zhang et al., 1999; Zamora et al., 2010; Xing et al., 2017) and is conductive to understanding P biogeochemical cycles in the perspective of atmospheric deposition.

In terms of the phytoplankton community composition, Dinophyceae generally dominated the community in UFS enriched with AR at all sites (Figure 6). With relatively high tolerance to nutrient-deficient environments and the potential to utilize DOP by inducing the expression of the gene for the synthesis of AP (Lin et al., 2012), dinoflagellates showed an advantageous growth in the AR treatments. This is also the reason why nutrient limitation had a slight impact on the growth of Dinophyceae based on RDA (Figure 7). In addition, due to the selective feeding of micrograzers, picosized phytoplankton suffer from a higher grazing pressure (Cottingham, 1999; Strom et al., 2007). In contrast, large dinoflagellates have the ability to keep themselves away from the prey of the dominant zooplankton species Paracalanus parvus (Hexanauplia) through particle rejection behavior (reject particles as food, Supplementary Figure 4; Huntley et al., 1986; Tiselius et al., 2013).

As described in “Changes in Total and Size-Fractionated chlorophyll a” section at U1_Sur, U3_Sur, and U3_SCM characterized by lower trophic statuses, the nanosized and picosized Chl a concentrations in FS enriched with AR were lower than those in UFS. In contrast, we did not observe similar results at U1_SCM and U2_SCM characterized by higher trophic statuses (Figure 4).

Nutrients, irradiance, and temperature are considered the three major factors that affect phytoplankton growth (Laws et al., 2000; Litchman, 2007). There was no difference in light and temperature between UFS and FS, and thus, nutrients play a key role in causing the lower nanosized and picosized Chl a concentrations in FS. At U1-3_Sur and U3_SCM, there were no obvious differences in N + N and PO₄^3– between UFS and FS (Figure 2). Meanwhile, we found that the APA values in UFS were 1.6-7.3 times higher than those in FS (Figure 3), indicating that microsized cells played an important role in increasing P bioavailability in the incubated seawater (Sebastián et al., 2004). This was supported by the positive linear relationship between APA and the contribution of microsized cells to the total Chl a (Figure 8A). As an extracellular enzyme, AP enters the environment through autolyzing or organisms excreting (Štrojsová et al., 2003). Besides, in contrast to FS, nanosized and picosized Chl a concentration in UFS increased linearly with relative change of APA (Figure 8B). Therefore, under the impact of AR addition, microsized cells have the ability to favor the growth of nanosized and picosized cells by increasing P bioavailability in seawater. The result at U1_SCM characterized by higher trophic statuses could also support this argument. On days 1-2 of the incubations, there was no difference in the nanosized and picosized Chl a concentrations in the AR treatments between UFS and FS when PO₄^3– was sufficient in the seawater, but lower nanosized and picosized Chl a concentrations in FS were measured when PO₄^3– was exhausted on day 3 (Figures 2, 4). Besides, dinoflagellates and green algae with the capacity of utilizing DOP also showed advantageous growth in the P-deficient condition (Figure 6).

There are other factors that might have caused the mismatch between UFS and FS in the concentration of nanosized and picosized Chl a. For example, the biodiversity of the community decreased after removal of microsized cells, which increased the difficulty for the community to reestablish a new balance (Dyke et al., 2007). Nanosized and picosized phytoplankton may adopt a strategy to survive in unstable habitats, e.g., by producing spores that can be dormant temporarily and revive at an appropriate time (Nayaka et al., 2017). However, these inferences cannot account for the change in nanosized and picosized Chl a at U1_SCM and U2_SCM characterized by higher trophic statuses (Figure 4). Therefore, our study provides a new clue from the perspective of nutrient utilization to illustrate how microsized phytoplankton affect the growth of nanosized and picosized ones.