Introduction

The frequency of severe haze episodes in East Asia has increased substantially over recent decades due to increased pollutant emissions, especially under stagnant meteorological conditions favoring high secondary aerosol formation and accumulation (Cai et al. 2017; Kim et al. 2016; Lu et al. 2016; Wang and Chen 2016; Yu et al. 2011; Zhang et al. 2015). Levels of fine particulate matter during such events cause adverse human health effects, reduce visibility, and change radiation budget (Oberdörster et al. 2005; Peel et al. 2013; Zheng et al. 2015).

Ground-level particulate matter (PM) mass concentration, either PM10 (diameter < 10 μm) or PM2.5 (diameter < 2.5 μm), is used to indicate particulate air pollution, because such particles can be inhaled by humans and cause respiratory diseases (Liu et al. 2016; Mehta et al. 2013; O’Neill et al. 2012; Yin et al. 2017). Aerosol optical depth (AOD), which is a quantitative measure of the extinction of solar radiation by the column-integrated aerosol load, is also widely used, not only to understand how the Earth’s climate is radiatively forced (Kim et al. 2008; Shindell et al. 2013; Schulz et al. 2006), but also for air quality purposes, in particular with regard to regional hotspots of particulate pollution, because of its large-scale spatial measurements by satellites (Chew et al. 2016; Engel-Cox et al. 2006; Kim et al. 2007). A number of recent papers have demonstrated the strong possibility of deriving surface PM concentrations from satellite-based AOD measurements to overcome the spatial limitations of ground-level PM10 and PM2.5 concentration measurements (Lv et al. 2016; Ma et al. 2014; Seo et al. 2015; You et al. 2015; Zheng et al. 2015).

Despite the increasing trend of severe haze events and increased pollutant emissions, several studies indicated declining trends for surface PM concentrations in China and Korea over the past decades due to continued emission control efforts (Cheng et al. 2013; Qu et al. 2010; Wang et al. 2014). In contrast, satellite observations and model simulations showed an increasing trend in AOD over East Asia (Pozzer et al. 2015; Streets et al. 2009; Yu et al. 2011; Zhang and Reid 2010).

Although many studies have reported qualitative changes in PM concentration and AOD, the estimated trends depend significantly on the data period and some discrepancies have been found for certain areas. In this study, therefore, we investigate spatial–temporal variations and trends for column aerosol AOD and surface PM10 concentration during 2004–2011 for six selected regions of interest in East Asia. Descriptions for PM10 and AOD data and the trend analysis method are given in the “Data and trend analysis” section. The “Results and discussion” section discusses the observed trends of PM10 concentrations and AOD and their causes, with spatial and temporal variations for the six selected regions. Finally, the research findings are summarized in the “Summary and conclusion” section.

Data and trend analysis

Monthly mean AOD and Ångström exponent (ÅE) from Moderate Resolution Imaging Spectroradiometer (MODIS, collection 6) onboard Aqua were analyzed, which were derived from the most recently updated land (algorithm; dark-target and deep blue) and ocean aerosol algorithms (Levy et al. 2013; Remer et al. 2005). To further ensure the data quality, MODIS AOD were cross-checked with ground-level AOD observed at Aerosol Robotic Network (AERONET) stations in East Asia.

PM10 concentrations calculated from the air pollution index (API) recorded at 27 Chinese urban stations during 2004–2011 were used in this study. Zhang et al. (2003) provide detailed descriptions for calculating PM10 concentrations from APIs and their validation. It should be noted that PM10 concentrations after 2012 are not used in this study, because the PM10 concentration data extracted from the air quality index (AQI) were inconsistent with API. PM10 concentrations at six stations in Korea and Japan during 2004–2014 were directly measured using the β-ray attenuation method, which follows the US EPA automated equivalent PM10 method (EQPM-1102-150).

Trends in AOD and PM10 concentration were calculated by Weatherhead’s least mean square (LMS) fit method (Eq. 1), which has been widely used for estimating long-term trends in aerosol characteristics (Collaud Coen et al. 2013; Streets et al. 2009; Weatherhead et al. 1998).

$$ {Y}_t=m+\rho \cdotp {X}_t+{S}_t+{A}_t,\kern3em t=1\dots n $$
(1)

Here, Y t represents monthly mean aerosol variables, t is time (months), m is a constant, ρ is the magnitude of the trend per year, and X t (=t/12) represents the linear trend function. S t (=\( \sum_{j=1}^4\Big[{\beta}_{1,j}\mathit{\sin}\left(2\pi jt/12\right)+{\beta}_{2,j}\mathit{\cos}\left(2\pi jt/12\right) \)]) is the seasonal component, where β 1 , j and β 2 , j are seasonal variation parameters. The noise term A t (=α · A t − 1 + ε t ) is modeled as autoregressive of the order of 1. Here, α is the autocorrelation coefficient of the data noise, and ε t represents independent random variables with mean zero and common variance \( {\sigma}_{\varepsilon}^2 \). The trend is statistically significant at the 5% significance level (95% confidence level) when |ρ/σ ρ | is > 2 (Tiao et al. 1990). Here, σ ρ (the standard deviation of the slope) is calculated by Eq. (2):

$$ {\sigma}_{\rho}\cong \frac{\sigma_{\varepsilon }}{\left(1-\alpha \right)}\frac{1}{n^{1.5}}=\frac{\sigma_A}{n^{1.5}}\sqrt{\frac{1+\alpha }{1-\alpha }}\kern0.5em $$
(2)

σ ε is the standard deviation of ε t , σ A is the standard deviation of A t , and n is the number of years.

Results and discussion

Spatial and temporal variations in AOD and PM10 concentration

Figure 1 shows 11-year (2004–2014) averaged MODIS AOD at 550-nm wavelength with the locations of surface PM10 concentration monitoring stations. In this study, we investigate changes in columnar aerosol properties in six regions (R1, North China; R2, South China; R3, Northeast China; R4, Yellow Sea and Korea Peninsula; R5, East Sea and Japan; and R6, Sea of Okhotsk), which are chosen based on the temporal variability of aerosols related to synoptic meteorological patterns, as well as emissions (Figs. S1 and S2).

Fig. 1
figure 1

MODIS-derived AOD at 550-nm wavelength over East Asia averaged across 2004–2014. The boxes and circled numbers represent the six regions and ground stations, respectively

Full size image

The highest annual mean AOD (0.65 ± 0.17) was observed at R1, mainly due to the heavy influence of aerosol pollutants emitted from several megacities (e.g., Beijing, Tianjin, and Hefei) and industrial complexes all year round (Chan and Yao 2008). Additionally, dust particles transported from deserts in Inland China and Mongolia contributed to elevation of AOD especially in spring (Qu et al. 2010). The annual mean AOD in R2 (0.50 ± 0.12) was lower than that observed in R1, resulting from more precipitation over R2 (Fig. S2). AOD over R3 (0.28 ± 0.13) was affected by dust particles from deserts (Nguyen et al. 2015), pollution aerosols from R1 by southwesterly winds (Fig. S1), and aerosols from biomass burning originating from Mongolia/Russia and R3 (Fig. 3). R4, which is adjacent to the downwind side of R1, showed an annual mean AOD of 0.41 ± 0.12. Similar AODs were reported in R5 (0.29 ± 0.10) and R6 (0.26 ± 0.16). While column aerosol loadings in R5 are likely due to aerosols transported mainly from R1 and R4 (Kim et al. 2007), the lowest AOD featured in R6 appeared to be influenced by sea salt and biomass burning aerosols transported from R3 and from grasslands and forests in Mongolia and Russia, especially during April–July (Fig. 3).

The main features of monthly variations in AOD, ÅE, and PM10 concentration for the six regions are as follows (Fig. 2):

  1. 1.

    In North China (R1) and its downwind region (R4), AOD maxima (R1, 0.87 ± 0.11; R4, 0.59 ± 0.08) appeared in June due to an accumulation of aerosols under stagnant regional-scale circulation (Figs. 3 and S1). Various processes, such as enhanced secondary aerosol formation, hygroscopic growth of hydrophilic aerosols, and biomass burning aerosols, may also contribute to high AOD during June, as discussed by Kim et al. (2007). ÅE during spring and winter at R1 (0.84 ± 0.07) and R4 (1.07 ± 0.08) were lower than the corresponding values during other seasons in response to the coarse dust particles transported from Inland China and Mongolia (Shao and Dong 2006; Logan et al. 2013). Contrary to AOD, elevated PM10 concentrations were apparent from October to June over R1 and from January to May over R4. This is attributable to the increase of fossil fuel usage for residential heating (Qu et al. 2010), as well as an accumulation of air pollutants under slow-moving, high-pressure systems, and intensive temperature inversion during colder months (Chan and Yao 2008; Xie et al. 2015). The invasion of wind-blown Asian dust is another important factor affecting ground-level PM10 concentration.

  2. 2.

    The peak in AOD during March–April (0.70 ± 0.07; see also Fig. S1) in South China (R2) appeared 3 months earlier than in R1. This is associated with the accumulated air pollutants under a stagnant, high-pressure system (Fig. S2). The frequent burning of agricultural wastes in R2 is thought to be another possible reason for high AOD during March–April (Fig. 3; van der Werf et al. 2010). From May, however, AOD in R2 decreased to approximately 0.37~0.52 due to frequent precipitation (Fig. S2). Monthly mean ÅE over R2 ranged from 1.0 to 1.4, with a peak on September (1.41 ± 0.06). The overall mean ÅE of 1.20 ± 0.14 at R2 was slightly higher than that of R1 (1.01 ± 0.23) because the influence of Asian dust is slightly less in R2 (Shao and Dong 2006). Meanwhile, an almost identical monthly variation of PM10 concentration was observed between R1 and R2, although PM10 concentration over R2 (80 ± 20 μg m−3) was lower than that over R1 (119 ± 33 μg m−3).

  3. 3.

    In Northeast China (R3), the maximum AOD of 0.46 ± 0.21 occurred during March with a relatively low ÅE (0.97 ± 0.10) due to the strong influence of Asian dust transported from inland China and Mongolia (Shao and Dong 2006). A monotonous increase in ÅE from April (1.24 ± 0.11) to July (1.66 ± 0.04) is attributable to the changes of wind direction from northwesterly to southwesterly, leading to significant increase of transported fine particles from R1 (Figs. S1 and S2). Distinct increase of PM10 concentration was found during December–January (mean 181 ± 49 μg m−3).

  4. 4.

    In the East Sea and Japan (R5), AOD showed consistent increase from January to April, followed by a consistent decrease. The monthly variation in ÅE was quite similar to those at R1 and R4. The overall mean PM10 concentration was estimated as 23 ± 7 μg m−3, with a distinct increase in March–May due to the influence of Asian dust (Shao and Dong 2006).

  5. 5.

    Over the Sea of Okhotsk (R6), increased AOD was apparent during April–June, peaking in June (0.52 ± 0.11). This increase in AOD can be explained by Siberian smoke drifts across R3 and Russia to the Sea of Okhotsk (Cottle et al. 2014).

Fig. 2
figure 2

Monthly variation of MODIS-derived AOD at 550 nm (blue closed circles), ÅE at wavelength 470–660 nm (red open circles), and PM10 (bar) with one standard deviation error bar for six regions

Full size image
Fig. 3
figure 3

Spatial distributions of monthly averaged MODIS-derived fire pixel counts, and wind vectors at 850-hPa pressure level over East Asia during 2004–2014. Black boxes indicate the six selected regions given in Fig. 1

Full size image

Observed trends in AOD and ÅE

Figure 4 shows time series of monthly mean AOD/ÅE and its trends for the six selected regions. Monthly mean AOD from AERONET Cimel sun/sky radiometer measurements in each region is also plotted. The monthly variation in AOD was very similar between MODIS and AERONET for all regions. Spatial distributions of annual and seasonal MODIS AOD trends for the period 2004–2014 are shown in Fig. 5. Note that the trends discussed in Figs. 4 and 5 are statistically significant at the 95% confidence level.

Fig. 4
figure 4

Time series of monthly mean MODIS AOD (blue solid line) and ÅE (red solid line) for six selected regions. The dashed lines are LMS regression fit to the data. The monthly mean values of AOD with one standard deviation from AERONET sun/sky radiometer at Beijing and Xianghe (R1), Hong Kong and Chen Kung (R2), Ussuriysk (R3), Gwangju (R4), and Shirahama and Osaka (R5) are superimposed

Full size image
Fig. 5
figure 5

Spatial distribution of a annual and b seasonal MODIS AOD trends for every 1° × 1° during 2004–2014 over East Asia

Full size image

In R1, significantly increasing trends were found for both AOD (+ 1.02% year−1) and ÅE (+ 1.06% year−1). AOD exhibited an especially strong increasing trend (+ 3.96% year−1) in megacities and heavily industrialized areas during winter. This increase is associated with elevated emissions of sulfate as well as other chemical species due to increased coal burning for residential heating during cold months (Chan and Yao 2008; Lalitaporn et al. 2013; Wang et al. 2005). We found a decreasing AOD trend in R2 (− 0.60% year−1), especially over the coastal regions, in all seasons except spring; in contrast, ÅE in R2 showed an increasing trend (+ 0.96% year−1). AOD trends remain significantly positive in R3, but no clear trends in ÅE are observed in the 11-year dataset. The increasing trend in AOD was observed only during summer, which might be related to transport of pollution particles from R1 by seasonal southeasterly winds, as discussed in the “Spatial and temporal variations in AOD and PM10 concentration” section (Figs. 3 and 5b). Although no significant trends were found for AOD and ÅE in R4, R5, or R6, the pattern of declining AOD over R4 and R5 during summer and autumn generally matched heavy precipitation regions. In the case of R6, AOD showed an increasing trend during spring, which is mainly due to pollution and biomass burning aerosols transported from R3 and Russia.

Changes in column AOD and surface PM10 concentration: a comparison

Table 1 summarizes the means and trends for MODIS AOD and PM10 concentration at 33 selected stations (R1, 15 urban stations; R2, 10 urban stations; R3, 2 urban stations; R4, 1 urban and 2 background stations; and R5, 3 background stations) over the period 2004–2011.

Table 1 Mean and trend estimates of MODIS-derived AOD and PM10 concentration over the period 2004–2011. For calculating the AOD trend of each station, monthly MODIS AOD was used for each 1° grid scale encompassing each selected station. Italicized numbers indicate trends that are statistically significant at the 95% confidence level
Full size table

In R1, AOD showed an increasing trend of + 0.64% year−1, whereas PM10 concentration decreased (− 2.60% year−1) at five urban sites in the Bohai Sea economic rim (#3 Beijing, #4 Tianjin, #5 Dalian, #6 Yantai, and #7 Qingdao), where mean AOD and PM10 concentration were observed to lie within the ranges 0.57–0.83 and 76–185 μg m−3, respectively. Five stations (#10 Nanjing, #11 Nantong, #12 Suzhou, #13 Shanghai, and #14 Hangzhou) located in the northern part of the Yangtze River Delta economic rim showed increasing AOD trend (+ 1.32% year−1), but decreasing PM10 concentration trend (− 2.83% year−1), which were about two times greater than those of five sites in the Bohai Sea economic rim. PM10 concentration showed positive trend (+ 3.51% year−1) at two urban stations (#19 Ningbo and #20 Wenzhou) in the southern part of the Yangtze River Delta economic rim, although no trend was observed in AOD for either site. The AOD trend in the mid-Yangtze River region was estimated as + 0.76% year−1 at #9 Hefei and + 1.03% year−1 at #15 Wuhan. The PM10 concentration trends were negative at Wuhan (− 4.05% year−1), #16 Changsha (− 7.31% year−1), and #17 Nanchang (− 0.87% year−1). However, we cannot explain the positive trend in PM10 concentration observed at Hefei (+ 3.70% year−1).

We found decreasing trends in both AOD (− 1.32% year−1) and PM10 concentration (− 0.77% year−1) for three stations (#21 Fuzhou, #22 Xiamen, and #23 Shantou) along the southeast coast of R2. Two stations located near the Pearl River Delta economic rim in R2 (#24 Shenzhen and #25 Guangzhou) have much larger declining trends in AOD (− 2.21% year−1) and PM10 concentration (− 4.23% year−1) than the other sites in R2.

In R3, positive AOD (+ 8.10% year−1) and negative PM10 concentration (− 2.28% year−1) trends were found at two stations (#26 Changchun and #27 Harbin). In R4, only background site (#30 Gosan) showed significant increasing trend in AOD (+ 1.98% year−1); however, considerable changes in PM10 concentration (− 5.15% year−1) were observed at urban site (#28 Seoul). No significant trends were found for background stations (#31 Yusuhara, #32 Oki, and #33 Sado-seki) in Japan (R5), with the exception of the PM10 concentration trend (− 2.74% year−1) at Sado-seki station.

Overall, 18 of the 20 stations in R1, R3, and R4 showed positive AOD trends, but significantly declining PM10 concentrations were still observed during 2004–2011. Here, we should note that observed trends in both AOD and PM10 concentrations at 12 out of 18 stations were statistically significant at the 95% confidence level. The possible reasons for this contradictory trend between AOD and PM10 concentration at 18 out of 20 stations are described as follows:

  1. 1.

    The reductions of coarse particle emissions (e.g., suspended fugitive dust) in megacities, enforced by strict policies on reducing air pollution (e.g., Action Plan for Air Pollution Prevention and Control), can directly reduce ground-level PM10 concentrations. However, several studies has reported no significant changes in PM2.5 concentrations in megacities in China due to continuous primary fine particle emissions and secondary particle formation (Lang et al. 2017; Lv et al. 2016; Xie et al. 2015; Zheng et al. 2015). This implies that AOD, which is defined as an integral of aerosol extinction coefficient with altitude, has also remained stable or has increased slightly in response to the amount of submicron particles, which have high light-extinction efficiency (Seinfeld et al. 1998).

  2. 2.

    AOD is determined by the aerosols distributed within a column of air, that is, in and above the atmospheric boundary layer. Figure 6 shows the percentage of AOD for every kilometer from the surface to 5 km altitude for the six selected regions, which derived from space-borne lidar Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP) measurements onboard Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO). The percentage of AOD from the surface to 1 km was approximately 33% (60% at 2 km) in R1. Similar ratios were also observed in other regions. As indicated in Fig. 6, large amounts of aerosols remain in the free troposphere, which does seem to affect AOD measurements.

  3. 3.

    Increased fine particle emissions from biomass burning activities in R1 may contribute to AOD more than to surface PM10 concentration, due to its favorable size for scattering light. Figure 7 shows annual mean MODIS-derived fire counts over three regions. Although accumulated fire counts in R2 and R3 showed large annual and inter-annual variations, no statistically significant trend in fire counts was found. However, R1 showed distinct increasing trend, especially during summer. Thus, increased emission of aerosols from biomass burning can be attributed to the positive AOD trend shown in R1 (Fig. 5).

Fig. 6
figure 6

AOD percentage per kilometer altitude for each region, estimated from space-borne lidar CALIOP measurements from 2007 to 2014

Full size image
Fig. 7
figure 7

Annual variation in fire pixel number from 2004 to 2014 for three regions: a R1 (North China), b R2 (South China), and c R3 (Northeast China)

Full size image

Meanwhile, as mentioned in the “Observed trends in AOD and ÅE” section, no coherent trends for PM10 concentration were found among stations in R2 although trends for AOD were declining (Fig. 5 and Table 1). Some possible explanations for these contradictory observations include a trend toward increasing precipitation over R2, which is also accompanied by increases in heavy rainfall, precipitation intensity, and frequency of precipitation events (Awan et al. 2015; Duan et al. 2013).

Summary and conclusion

Multi-year records of AOD and PM10 concentration for six selected regions of interest in East Asia were analyzed for spatio-temporal variations and trends.

Different features of monthly/seasonal variations in both AOD and PM concentration were observed among the selected regions, which were associated with distinct changes in regional-scale seasonal circulations. Enhanced AOD over North China and its downwind regions (Yellow Sea and Korea) appeared in June, compared with March–April over South China. Increased PM10 concentrations in both North and South China were observed from late fall to spring. In Northeast China, AOD peaked during March, whereas high PM10 concentrations occurred during December–January.

A considerable decreasing trend in both AOD and PM10 concentration was observed in South China, which is potentially related to increased precipitation. However, a significant increasing trend in AOD was found in North China and Northeast China (particularly in 18 out of 20 megacities), whereas surface PM10 concentrations over most megacities in these two regions declined almost linearly. The reductions in PM10 concentrations can be attributed to large reductions in the emission of near-surface coarse particles due to a series of strict control measures. However, an increase in AOD indicates that highly scattering submicron secondary particles have not been meaningfully reduced. Space-borne lidar CALIOP measurements revealed that approximately 60~70% of AOD was contributed by the aerosols present above the atmospheric boundary layer (> 1 km altitude).

Our findings suggest that strict restrictions on the emission of PM2.5 and its precursor gaseous are required in order to improve particulate air quality and to reduce adverse human health effect, so as to further reduce their radiative effects. Especially, further studies on the physicochemical and optical characteristics of fine particles (i.e., PM2.5) are needed to develop effective strategies and actions for reducing these effects, although we have not addressed recent changes in PM2.5 concentrations over East Asia due to the lack of continuous observations at multiple locations.