The DuBi (Dust–Bioaerosol) field sampling campaign was conducted over an extended period and across a broad spatial domain. From 2011 to 2021, sampling was performed at multiple sites across East Asia, including China, Japan, South Korea, and Mongolia (see Table 1 for details). These sites encompass a wide range of surface types and climatic conditions, providing strong regional representativeness and ensuring comparability among observations (Fig. 1).

This field observation campaign substantially enhanced the temporal and spatial coverage of bioaerosol measurements in East Asia. As a result, the compiled dataset enables a more comprehensive characterization of bioaerosol distributions across diverse environmental settings, thereby improving our understanding of the regional features and evolution patterns of bioaerosols in East Asia.

A custom-designed bioaerosol sampler was employed for field collection to ensure operational stability and methodological consistency across sites. Instruments of this type have been widely used in bioaerosol research (Maki et al., 2010; Qi et al., 2023). To construct the sampler, a 13 mm polycarbonate membrane with 0.2 µm pore size (Whatman 111106 and GTTP01300) was inserted inside a 13 mm Swinnex filter holder (Millipore SX0001300) fitted. Polycarbonate membranes offer advantages such as high chemical resistance, high thermal stability, and a low tendency to adsorb proteins or extractable substances (Ferguson et al., 2019). Their optical transparency also enables accurate microscopic detection. A micro air pump (AS ONE MAS-1, Japan) with a standard flow rate of 12 L min⁻¹ was connected to the sampler to draw atmospheric microorganisms onto the filter membrane (Fig. 2a, b).

Prior to sampling, both the filters and filter holders were sterilized. The assemblies were first rinsed with 75 % ethanol, followed by ultrapure water to remove residual detergent. Subsequent processing of the filters and holders was conducted in a laminar-flow hood, where they were immersed in absolute ethanol and ultrapure water to further remove adhered organic residues. The dried filters and holders were then assembled and sterilized by autoclaving at 121 °C for 20 min, followed by ultraviolet irradiation for 30 min (Xue et al., 2024; Qi et al., 2021). The sterilized components were sealed in pre-sterilized collection tubes and sterile bags until use. During sampling, operators wore masks and sterile gloves, and the samplers were mounted at a height of 1.5 m a.g.l. Real-time flowmeter readings obtained immediately before the micro-pump was turned on and immediately before it was turned off were recorded as the initial and final sampling flow rates, respectively. After sampling, filters were labeled with the sampling location and time, stored at -20 °C during transport, and subsequently transferred to -80 °C freezers for long-term preservation.

Bioaerosol particle concentrations were determined using DAPI staining (4',6-diamidino-2-phenylindole, D9542, Sigma), followed by fluorescence microscopy. 4 % paraformaldehyde (250 µL) was introduced to each filter holder containing the membrane to fix the samples for 1–2 h. The filters were then rinsed with sterile ultrapure water to remove residual fixative. Under light-protected conditions, DAPI solution (10 µg mL⁻¹, 250 µL) was introduced to stain the samples for approximately 15 min (Maki et al., 2014), followed by a final rinse with sterile water.

During slide preparation, the membrane was treated with non-fluorescent immersion oil (IMMOIL-F30CC, Olympus) for wetting. The coverslip was then pressed firmly, and excess immersion oil was removed using a dust-free wipe. Samples were examined using a fluorescence microscope (BX53 and DP72, Olympus) with an excitation wavelength range of 340–390 nm. A fluorescence microscope (BX53 and DP72, Olympus) providing a 340–390 nm excitation wavelength range was employed to examine the samples (Tang et al., 2018).

As shown in Fig. 2e–g, four categories of particles were distinguishable under the fluorescence microscope based on their emitted colors: blue, yellow, white, and black. DAPI-stained microbial cells produced strong blue or blue-green fluorescence, whereas mineral particles appeared white (Maki et al., 2013). Black particles were attributed to black carbon (Liu et al., 2023). Yellow particles were generally interpreted as non-DNA-containing particles (Kepner and Pratt, 1994). However, other studies have shown that water-soluble salts (Liu et al., 2023) and organic matter (Mostajir et al., 1995) can also exhibit yellow fluorescence. In this study, most yellow-green particles were associated with mineral dust and were therefore considered to be mineral particles enriched in water-soluble salts and organic material. In this study, bioaerosols were operationally defined as particles exhibiting cellular morphology and emitting blue-green fluorescence after DAPI staining. Accordingly, the reported bioaerosol concentrations represent the fraction of atmospheric biological particles that met both the morphological and fluorescence-based identification criteria, rather than the full range of primary biological aerosol particles (PBAPs), Viruses were outside the detection scope of this method. Pollen and plant fragments were not separately identified or quantified and would only have been included when they met both identification criteria. Hereafter, “bioaerosol concentration” refers specifically to the particle concentration determined using this operational definition. For each filter membrane, 20 randomly selected microscopic fields were imaged, and all particles belonging to the four categories were manually counted using the Cell Counter plugin in ImageJ.

Bioaerosol number concentrations, C (particles m⁻³) were calculated using Eq. (1): 1C=S1×N0S0×V, where S1 is the filtration area of the filter membrane (µm²), S0 is the field-of-view area of the microscope (µm²), V is the sampled air volume (m³), and N0 is the average number of bioaerosol particles observed per microscopic field of view (Maki et al., 2019).

The 16S rRNA gene has been widely used to characterize microbial community composition and diversity (Giles et al., 2023; Nagarajan et al., 2023). The UltraClean Soil DNA Kit (MoBio, San Diego, CA) was employed to extract bacterial DNA in this study. Sequencing was performed on Illumina NovaSeq 6000 and Illumina MiSeq platforms in PE250, PE150, or PE300 modes. The bacterial 16S rRNA gene V4–V5 hypervariable region was amplified using the 515F/907R primer pair (forward: GTGCCAGCMGCCGCGGTAA; reverse: CCGTCAATTCMTTTRAGTTT) (Fang et al., 2017). Paired-end sequencing produced reads exceeding 250 bp for each sample, which were subsequently sorted according to their unique barcode sequences. Blank controls were processed alongside the samples during both DNA extraction and PCR amplification. No PCR amplicons were detected in these controls, demonstrating that the entire experimental workflow was free from detectable contamination (Qi et al., 2023; Karstens et al., 2019).

The Quantitative Insights into Microbial Ecology pipeline (QIIME2; version 2024.2) was used to assign taxonomic identities to the raw sequences (Bolyen et al., 2019). High-quality sequences were denoised and resolved into amplicon sequence variants (ASVs) at single-nucleotide resolution. Taxonomic classification was assigned using the SILVA v138 reference database (Pin et al., 2021). ASVs classified as chloroplasts, mitochondria, archaea, eukaryotes, or unassigned were removed prior to downstream ecological analyses.

In this study, the Shannon and Chao1 indices were used to quantify the alpha diversity of bacterial communities. These indices were calculated using the vegan package in R (v4.4.4). Differences in alpha diversity among surface types were assessed using the Wilcoxon rank-sum test to evaluate statistical significance (Loos et al., 2024).

The integrated dataset compiled in this study includes the following data products (Fig. 3):

Meteorological variables, such as air temperature, dew point, wind speed, wind direction, were obtained from surface meteorological observation stations. Because relative humidity (RH) is not directly reported at these stations, RH was derived using the Magnus formula based on temperature and dew-point measurements. (China stations: https://www.ncei.noaa.gov/pub/data/noaa/isd-lite/, last access: 22 August 2026; stations outside China: https://www.ncei.noaa.gov/data/global-summary-of-the-day/access/, last access: 22 August 2026);

To characterize the general relationships between bioaerosol concentration, NDVI, and air temperature across East Asia while minimizing the influence of dust transport, this analysis was restricted to samples collected under non-dust conditions. As shown in Fig. 4a, bioaerosol concentrations were significantly negatively correlated with NDVI. In regions with low NDVI, sparse vegetation cover and dry, loose surface soils may facilitate the wind-driven entrainment of soil-associated microorganisms and other biological particles into the atmosphere (Qi et al., 2023). In contrast, denser vegetation in regions with high NDVI may enhance the interception of airborne particles and promote dry deposition, thereby contributing to lower atmospheric bioaerosol concentrations (Zhai et al., 2022). Although vegetated surfaces may also emit pollen and plant fragments, these particle categories were not separately identified in the present study. The observed correlation therefore represents the net relationship between NDVI and the operationally defined DAPI-positive particle concentration rather than the response of individual PBAP categories.

Figure 4b indicates that the relationship between temperature and bioaerosol concentration is non-linear. When temperatures are below approximately 20 °C, bioaerosol concentrations show a positive correlation with temperature. However, as temperature increases further, bioaerosol concentrations decline markedly. A possible explanation is that, within an optimal temperature range, warming enhances atmospheric mixing and air movement, thereby promoting the dispersal of microorganisms into the atmosphere and resulting in higher concentrations (Cavicchioli et al., 2019). At excessively high temperatures, however, conditions become unfavorable for microbial growth and reproduction, and the abundance of microorganisms from certain sources may decrease accordingly (Zhang et al., 2022b).

Figure 5 shows the temporal variations in bioaerosol concentrations in Yuzhong, China, on 6 and 15 November 2020. As illustrated, bioaerosol concentrations on both days exhibit pronounced diurnal patterns rather than random fluctuations. Concentrations are lowest in the early morning (07:00–09:00), followed by a gradual increase, reaching a peak in the afternoon (15:00–17:00), and then decreasing thereafter. This pattern suggests that bioaerosol concentrations are governed by well-defined diurnal processes.

The concentrations of NO₂ and CO also display clear temporal variations with similar overall trends, suggesting that these two trace gases may share common sources. However, the temporal variations of NO₂ and CO on the two sampling days are not fully consistent with those of bioaerosol concentrations. This discrepancy implies that short-term variations in bioaerosol concentrations during the sampling period were not primarily controlled by anthropogenic emissions or changes in the atmospheric boundary layer height (Li et al., 2022). Instead, they are more likely associated with surface biological emissions and atmospheric dynamical processes. These processes include daytime increases in surface temperature that enhance microbial activity, intensified release from soil and vegetation surfaces, and strengthened surface turbulence that facilitates the entrainment of particles into the atmosphere (Gashi et al., 2025; Sharma et al., 2022; Zhang et al., 2022a; Kulmala et al., 2023).

Figure 6 illustrates diurnal-scale variations in bioaerosol concentrations and related physical parameters in Yonago, Japan, from March to June 2015. As shown, bioaerosol concentrations exhibit pronounced high-frequency fluctuations on the diurnal timescale, indicating strong modulation by short-timescale processes. Within each month, concentrations consistently increase and then decrease, forming a recurrent “single-peak, within-month” pattern throughout March–June. This suggests that bioaerosols do not behave as random background noise but instead represent an aerosol component characterized by a well-defined temporal structure.

During the sampling period, a prominent concentration peak occurred in mid-April, substantially exceeding levels observed in other months and persisting for several consecutive days rather than representing a single-day anomaly. Given that no corresponding extreme peaks are observed in the other variables shown in the figure, this event is unlikely to have been driven solely by particulate matter (PM) or gaseous pollution. Instead, it more plausibly reflects a regional-scale, concentrated release from biological sources.

AOD at 550 nm reaches near-maximum values during the major bioaerosol concentration peaks and varies largely in phase with bioaerosol concentrations. The temporal variations in PM₁₀ and PM_2.5 concentrations are closely consistent with those of bioaerosol concentrations, with PM_2.5 exhibiting a slight lag. This indicates that, during this event, bioaerosols and non-biological particles were influenced by shared controlling processes rather than varying independently.

In contrast, NO₂ showed concurrent increases with bioaerosol concentrations during some peak periods, while O₃ exhibited no obvious temporal correspondence. These results suggest that gaseous pollutants were not the dominant drivers of the observed bioaerosol variability during the sampling period, although NO₂ may reflect local atmospheric conditions associated with certain bioaerosol enhancement events.

To investigate the dominant taxa at different hierarchical levels across varying underlying surface types, this study selected the five taxa with the highest relative abundances for each surface type for analysis. The underlying surface was classified into six categories based on NDVI: NDVI < 0.1 was defined as very low vegetation cover; 0.1 ≤ NDVI < 0.2 as low vegetation cover; 0.2 ≤ NDVI < 0.4 as medium–low vegetation cover; 0.4 ≤ NDVI < 0.6 as medium vegetation cover; and 0.6 ≤ NDVI ≤ 0.8 as high vegetation cover. In addition, because the sampling campaign included aircraft measurements, an additional category representing the upper atmosphere was included.

Figure 7a presents the bacterial community composition at the phylum level. Proteobacteria dominate the community, with a relative abundance of 60.1 %, followed by Firmicutes (24.8 %). Among the identified phyla, Chloroflexi are primarily associated with areas of low vegetation cover (43.0 %) and very low vegetation cover (33.2 %). Actinobacteria and Acidobacteriota exhibit similar distribution patterns, occurring mainly in low-vegetation (33.9 % and 35.3 %, respectively) and very low vegetation cover regions (31.3 % and 28.0 %, respectively). These three phyla represent dominant bacterial groups in soils under varying vegetation conditions (Zhou and Wang, 2023). As soil and vegetation are important sources of airborne microorganisms, contributions from soil sources to atmospheric microbial populations become more pronounced under conditions of low vegetation cover (Archer et al., 2023; Xie et al., 2020).

Figure 7b shows the bacterial community composition at the genus level. The dominant taxon at the genus level is Lactococcus, with a relative abundance of 23.4 %. Among the identified genera, Staphylococcus (87.8 %), Alcaligenes (67.3 %), and Enterococcus (49.6 %) are predominantly distributed in the upper atmosphere. Paraburkholderia (61.6 %) is mainly distributed in areas with high vegetation cover. The pronounced enrichment of Staphylococcus, Alcaligenes, and Enterococcus in upper-atmosphere samples may be attributed to their strong environmental tolerance and diverse source characteristics, which facilitate survival during atmospheric vertical transport and under selective filtering processes in upper-atmosphere environments (García-Solache and Rice, 2019; Onyango and Alreshidi, 2018; Machado et al., 2023; Pedrosa-Silva and Venancio, 2023). Paraburkholderia, a genus typically associated with plants, receives sustained and stable biological inputs in regions with high vegetation cover, making it more likely to accumulate in near-surface air over densely vegetated areas (Esmaeel et al., 2018; Eberl and Vandamme, 2016).

As shown in Fig. 8a and c, the mean values of both the Shannon and Chao1 indices across different underlying surface types follow the same decreasing order: low vegetation cover > very low vegetation cover > medium–low vegetation cover > high vegetation cover > medium vegetation cover. This pattern indicates that α diversity and species richness are highest in areas with low vegetation cover, whereas areas with medium vegetation cover exhibit the lowest diversity. Variations in α diversity across vegetation cover gradients reflect the combined effects of source inputs, multi-source mixing, and ecological filtering processes (Mantoani et al., 2024; Lu et al., 2024; Gashi et al., 2025). Low vegetation cover areas integrate both soil- and vegetation-derived biological sources and are typically associated with stronger surface disturbance, resulting in the highest species richness and diversity (Nie et al., 2024; Mu et al., 2020; Archer et al., 2023). In contrast, medium vegetation cover regions suppress soil resuspension while not yet developing the complex phyllosphere-associated bioaerosol sources characteristic of highly vegetated areas. Consequently, these regions experience fewer source inputs and stronger environmental filtering, leading to the lowest α diversity (Zhai et al., 2022; Robinson et al., 2021).

Figure 8b and d further show that both the Shannon index (r= -0.68, P=0.015) and the Chao1 index (r= -0.818, P=0.0011) are negatively correlated with NDVI, with the Chao1 index exhibiting a stronger correlation. Based on the differing sensitivities of these two indices, it can be inferred that, under low-NDVI conditions, atmospheric bacterial communities are more strongly influenced by external source inputs, leading to increased species richness. However, many of the newly introduced taxa occur at low relative abundances and therefore have a limited effect on community evenness, resulting in a smaller magnitude of change in the Shannon index compared to the Chao1 index.