Spatial and Temporal Variations of Aerosol Optical Thickness over the China Seas from Himawari-8

Abstract

Six years of hourly aerosol optical thickness (AOT) data retrieved from Himawari-8 were used to investigate the spatial and temporal variations, especially diurnal variations, of aerosols over the China Seas. First, the Himawari-8 AOT data were consistent with the AERONET measurements over most of the China Seas, except for some coastal regions. The spatial feature showed that AOT over high latitude seas was generally larger than over low latitude seas, and it is distributed in strips along the coastline and decreases gradually with increasing distance from the coastline. AOT undergoes diurnal variation as it decreases from 9:00 a.m. local time, reaching a minimum at noon, and then begins to increase in the afternoon. The percentage daily departure of AOT over the East China Seas generally ranged ±20%, increasing sharply in the afternoon; however, over the northern part of the South China Sea, daily departure reached a maximum of >40% at 4:00 p.m. The monthly variation in AOT showed a pronounced annual cycle. Seasonal variations of the spatial pattern showed that the largest AOT was usually observed in spring and varies in other seasons for different seas.

1. Introduction

Atmospheric aerosols refer to some suspension of liquid, solid, or mixed particles in the air with a size distribution ranging from 0.01 to 100 μm [1]. Aerosols not only affect air quality and human health [2,3], but also impact global climate change through direct and indirect radiative forcing [4]. Additionally, aerosols can be transported over long distances and deposited into the ocean under certain weather conditions. These deposited aerosols can impact (positively or negatively) nutrients and toxic element availability, ocean biogeochemistry, primary productivity, and carbon cycling [5]. However, understanding the role of aerosols as a contributing factor in these issues is difficult due to the continuous spatial and temporal variations of aerosols [6].

Aerosol optical thickness (AOT) is defined as the integrated extinction of radiation caused by aerosol absorption and scattering of light over the total atmospheric column. This is a fundamental measurement of the optical properties of aerosols. Because AOT is closely related to aerosol loads, it is often used as an indicator of air quality. It is also an important parameter in the quantitative calculation of radiative forcing and deposition flux. Therefore, AOT is widely used to study the effects of aerosols on the environment, climate, and ecology [7].

The China Seas are among the most productive seas in the world and are integral in the climate system in the region [8]. The characteristics of aerosols over the China seas are very complex, due to that they are usually a mixture of aerosols from marine and continental and natural and anthropogenic sources. There are usually two ways to obtain AOT: ground-based measurements and satellite remote sensing retrieval. Many studies on aerosols over the China Seas used ground-based observations over the last 30 years. The distinct temporal and spatial distribution of aerosols over the East China Sea were first investigated by approximately ten cruises during 1987–1992 [9]. AOT over the Yellow Sea was approximately 0.1, without obvious diurnal variation based on automatic sun photometer CE318 observations on an island [10]. Measurements taken on a ship using a handheld solar photometer MicroTops-II revealed that AOT over the Yellow Sea and East China Sea was approximately 0.2–0.4 under clear sky conditions in spring and increased to 0.8 when the haze layer appeared, [11]. AOT over the northern coast of the South China Sea was above 0.5, and decreased to 0.2 further from the mainland over the open ocean [12]. Recently, the long-term characteristics of aerosol optical properties and volume size distribution over the East China Sea were analyzed based on four AERONET (Aerosol Robotic NETwork) site measurements [13]. These studies have provided a preliminary understanding of the source, distribution, and variation of aerosols over the China Seas. However, because of the sparse and uneven spatial and temporal distribution of ground measurements, the conclusions can be inconsistent or even contradictory and cannot reveal the distribution and variation of AOT over large regions.

Although polar-orbiting satellite remote sensing is relatively rough in terms of temporal resolution compared to ground-based measurements, it has the advantage of offering long-term continuous observations over wide areas, which can compensate for the deficiencies of ground-based measurements. It is becoming an increasingly important means to investigate the distribution and variation of aerosols over regional and global oceans on different temporal scales (monthly, seasonal, or yearly). AOTs are usually retrieved from multi-wavelength sensors carried by polar-orbiting satellites, such as the Advanced Very High Resolution Radiometer (AVHRR) [14], the Sea-Viewing Wide Field-of-View Sensor (SeaWiFS) [15], the moderate resolution imaging spectroradiometer (MODIS), and the visible/infrared imager radiometer suite (VIIRS) [16,17].

The lifespans of aerosols range from a few hours to weeks, resulting in substantial spatial and temporal variations in AOT [18]. The temporal resolution of polar-orbiting satellite sensors, which pass approximately twice per day, makes it difficult to capture the rapid variation and transposition process of aerosols. In addition, optical remote sensors are susceptible to clouds and rain, which further reduces the effective coverage of AOT from polar-orbiting satellite observations. Fortunately, geostationary satellite data have the advantage of high temporal resolution. Cloud interference can be avoided by merging multiple observations and effectively increasing the coverage of observations. Therefore, geostationary satellites are powerful tools for studying the rapid variation of AOT over oceans. Aerosol-sensitive bands were absent from the early geostationary satellite sensors, such as the Geostationary Meteorological Satellite (GMS)-5 and Multifunction Transport Satellite (MTSAT), which lowered their accuracy and limited their application [19]. However, in recent years, with the continuous development of geostationary satellite technology and retrieval algorithms, the accuracy of AOT retrieved from geostationary satellites has significantly improved and provides a great opportunity for the study of the spatial and temporal distribution of AOT, especially diurnal variation [20]. Some studies have developed improved AOT retrieval methods for Himawari-8, but they have always focused on land-based regions rather than ocean areas [21].

The aim of this study is to investigate the spatial and temporal variations of AOT over the China Seas based on hourly observations from July 2015 to July 2021 from the latest generation of geostationary meteorological satellite, Himawari-8. The remainder of this paper is structured as follows: Section 2 introduces the physical geography of the study area (the China Seas) and the dataset used in this study; Section 3 validates the Himwari-8 AOT over the China Seas and then determines and discusses the spatial and temporal characteristics of AOT; finally, Section 4 presents the conclusions.

2. Data and Methods

2.1. Study Area

The marginal seas are transition zones for the spread of atmospheric pollutants from continents to the oceans. Therefore, meteorologists and oceanographers are both interested in these regions. The atmosphere over the China seas are the main pathway to transport some aerosols from Asia to the Pacific. Aerosol deposition in these areas, which are important fishing sites, significantly impacts the water environment and ecology. Therefore, to understand the spatial and temporal variation of AOT over the China Seas is helpful in determining the source, transposition, and deposition of aerosols. The China Seas in this study include the Bohai Sea (37–41°N, 117–122°E), the Yellow Sea (33–39°N, 119–127°E), the East China Sea (23–33°N, 117–130°E), and the northern part of the South China Sea (10–23°N, 105–122°E), as shown in Figure 1.

The Bohai Sea (BS) is the northernmost semi-enclosed inland sea of China with a shallow water depth of approximately 18 m. It is composed of Bohai Bay, Laizhou Bay, Liaodong Bay, and the central sea basin. The maximum distance from north to south is 550 km, the maximum distance from east to west is 230 km, and the total area is approximately 77,000 km². The Yellow Sea (YS), located between the Chinese mainland and the Korean Peninsula, has an average water depth of 44 m, a distance from north to south of approximately 870 km, a width of approximately 556 km from east to west, and a total area of 380,000 km². The East China Sea (ECS) covers an area of approximately 770,000 km². The distance from the northeast to the southwest is approximately 1296 km, the east to west width is approximately 740 km, and the average water depth is 350 m. It borders the provinces of Shanghai, Zhejiang, and Fujian to the west; it opens in the north to the YS, the Nomzaki Cape from the east of Jeju Island in South Korea to Kyushu, Japan, and the Korean Strait to the northeast. The south is connected to the South China Sea and passes through the Taiwan Strait. Since the BS, YS, and ECS are all located on the east side of China, they are collectively referred to as the East China Seas. The South China Sea (SCS) is in the southern part of mainland China; it is the largest marginal sea in the western Pacific Ocean, with a total area of 3.5 million km² and an average water depth of 1212 m. It connects the Pacific Ocean to the east and the Indian Ocean to the southwest. Since the southern portion of the SCS is far away from the mainland and the spatiotemporal variation of AOT over it is relatively small, this research focuses only on the northern portion of the SCS (NSCS).

The BS and YS have a warm temperate monsoon climate, ECS has a subtropical monsoon climate, and the NSCS has a tropical monsoon climate. The summer monsoon and winter monsoon in these areas alternate: monsoonal winds in the summer blow predominantly from the southeast over the East China Seas and reverse to the north in the winter. For the NSCS, the monsoonal winds blow predominantly from the southwest in summer and change to the northeast in winter. In addition, the China Seas often suffer from different levels of cold air, extratropical cyclones, tropical cyclones, and typhoons, which also influence the spatial and temporal distribution of aerosols.

2.2. Himawari-8 AOT

The geostationary meteorological satellite Himawari-8 was launched in October 2014 and the official operation started in July 2015. It is located at 140.7°E, approximately 35,800 km distance from the Earth. It covers the entire East Asia and Pacific region (80°E–160°W, 60°S–60°N) at an interval of ten minutes. The primary sensor is the Advanced Himawari Imager (AHI), which is comparable to the Advanced Baseline Imager (ABI). The AHI is a 16-channel multi-spectral imager, including three visible light channels, three near-infrared channels and ten infrared channels, with a spatial resolution of 0.5–2.0 km [19].

Level 2 and level 3 aerosol products from Himawari-8 are available on the JAXA Himawari monitor website (H8-AOT/JAXA). The temporal resolutions of the level 2 and 3 products are 10 min and one hour, respectively. The spatial resolution was 0.05° of equal latitude-longitude gridding. Level 2 H8-AOT/JAXA was retrieved from a pre-calculated surface reflectance database. First, the optimum channels for aerosol retrieval were selected automatically by considering the uncertainty in the TOA reflectance resulting from surface reflectance uncertainty. To avoid the influence of water-leaving radiance over the ocean, only channels longer than 800 nm (860 and 1600 nm) were used for the algorithm. The sun glitter reflectance (direct sunlight reflected by the sea surface) was added to the TOA reflectance calculation. The sea surface reflectance was calculated based on the model developed by Cox and Munk [22] using wind speed from the Japan Meteorological Agency global analysis (GANAL) data. Geometries of the sensor and solar are also considered in this step to reduce the surface bidirectional effects. Second, an external mixture of fine and coarse aerosols with monomodal lognormal volume size distributions was assumed through a cluster analysis with AERONET measurements. Finally, a look-up table for every 1 nm in the range of 300–2500 nm was pre-calculated and weighted using the response function for Himawari-8. The level 3 product is a collection of level 2 AOT observations with minimum cloud contamination developed using the difference between aerosol and cloud spatiotemporal variability characteristics. It should be noted that the hourly AOT is an estimated value at a certain target time, rather than an average over one hour. For more details about the AOT retrieval algorithm, please refer to [23]. The retrievals include four quality flags: “very good,” “good,” “marginal,” and “no confidence.” Only level 3 H8-AOT/JAXA data labeled “very good” from the latest version 3.1, from July 2015 to July 2021, were used in this study.

2.3. AERONET AOT

AERONET is a worldwide ground-based sun photometer observation network established and maintained by NASA and France (https://aeronet.gsfc.nasa.gov/, accessed on 5 August 2021). The dataset contains more than 600 sites around the world, providing long-term and continuous monitoring of the optical, microphysical, and radiation characteristics of aerosols. The low uncertainty (0.01–0.02) and high frequency observation (once every 15 min) [24] of AERONET AOT make it suitable for satellite retrieval evaluation. AERONET products usually have three levels: level 1.0 is the raw data with the instrument calibrated and corrected, level 1.5 refers to level 1.0 data with automatic cloud screening, and level 2.0 has additional postal calibration and manual checking. In this study, AOT measurements from nine sites from the latest updated AERONET Version 3 [25] level 2.0 dataset were selected. The nine AERONET sites over the China Seas, as shown in Figure 1, were selected for H8-AOT/JAXA validation, and the information is summarized in

Table 1. Information of the nine AERONET sites used in this study.

Site Name	Longitude (°E)	Latitude (°N)	Altitude (m)	Time Period	Surface Type
Baengnyeong	124.630	37.966	136	July 2015–August 2016	Island
Socheongcho	124.738	37.423	28	October 2015–July 2021	Roof deck
Anmyon	126.330	36.539	47	July 2015–November 2019	Island
Gosan_SNU	126.162	33.292	72	July 2015–September 2016	Island
Ieodo_Station	125.182	32.123	29	July 2015–August 2019	Roof deck
Okinawa_Hedo	128.249	28.867	60	March 2019–July 2021	Cape
Cape_Fuguei	121.538	25.297	15	November 2016–July 2021	Cape
Dongsha_Island	116.729	20.699	5	July 2015–July 2021	Island
Tai_Ping	114.362	10.376	4	July 2015–July 2021	Island

.

3. Results and Discussion

3.1. Validation of Himawari-8 AOT

In this section, the hourly level 3 H8-AOT/JAXA is validated against the ground-based AERONET AOT. To statistically ensure a better balance between sample size and correlation quality, the collocation is defined as the average of AERONET measurements within ±15 min around satellite observation and at least 10 validated values of H8-AOT/JAXA observations within a matching window of 5 × 5 pixels (average of 25 pixels) centered on the AERONET site over the China Seas [26]. Figure 2 shows a two-dimensional density plot of the H8-AOT/JAXA validation against AERONET measurements over the China Seas from July 2015 to July 2021.

The statistics used include matchup numbers, Pearson correlation coefficient I, linear regression, mean absolute error (MAE), mean relative error (MRE), root mean square error (RMSE), the fraction within the expected error (within EE), above the expected error (AEE), below the expected error (BEE), and relative mean bias (RMB). These were calculated as follows:

$$R=\frac{{{\displaystyle \sum }}_{i=1}^n\left(AO{T}_{318i}-\overline{AO{T}_{318i}}\right)\left(AO{T}_{\mathrm{H}8i}-\overline{AO{T}_{\mathrm{H}8i}}\right)}{\sqrt{{{\displaystyle \sum }}_{i=1}^n{\left(AO{T}_{318i}-\overline{AO{T}_{318i}}\right)}^2}\sqrt{{{\displaystyle \sum }}_{i=1}^n{\left(AO{T}_{\mathrm{H}8i}-\overline{AO{T}_{\mathrm{H}8i}}\right)}^2}}$$

(1)

$$MAE=\frac{{{\displaystyle \sum }}_{i=1}^n\left|AO{T}_{318i}-AO{T}_{H8i}\right|}{N}$$

(2)

$$MRE=\frac{{{\displaystyle \sum }}_{i=1}^n\left|\frac{AO{T}_{318i}-AO{T}_{H8i}}{AO{T}_{318i}}\right|}{N}\times 100\%$$

(3)

$$RMSE=\sqrt{\frac{{{\displaystyle \sum }}_{i=1}^n{\left(AO{T}_{318i}-AO{T}_{H8i}\right)}^2}{N}}$$

(4)

$$R=Within\_EE=\frac{\left(0.85AO{T}_{318i}-0.05\le AO{T}_{H8i}\le 1.15AO{T}_{318i}+0.05\right)}{N}\times 100\%$$

(5)

In the above equations, $AO{T}_{318i}$ is the AERONNET observation and $AO{T}_{H8i}$ is the matched H8-AOT/JAXA. The number of matchups was approximately 3893 and they showed high levels of agreement with each other (R = 0.914). The matched data were evenly distributed on both sides of the 1:1 line, which was close to the regression line (y = 0.93x + 0.07. The MAE and RMSE values were 0.081 and 0.111, respectively, and the largest concentrations of the data points were found at 0.1 < AOT < 0.4. Although approximately 65% of the matchups fell in the EE, a slight overestimation was also observed (AEE = 32% and RMB = 1.296).

Considering that the inhomogeneity surface around the ground-based sites (their varying surface types are shown in Table 1), additional site-scale validation was performed. Figure 3 presents the validation of H8 AOT/JAXA against ground measurements for each AERONET site. These sites are mainly located in the YS (Baengnyeong, Socheongcho, and Anmoyon), ECS (Gosan_SNU, Ieodo_Station, Okinawa_Hedo, and cape_Fuguei_Station), and NSCS (Dongsha_Island and Tai_Ping). The NSCS showed the best performance in terms of high correlation (R = 0.89), percentage falling within EE (approximately 75.8–89.2%), and low errors (MAE = 0.03–0.07, RMSE = 0.05–0.1). In contrast, although the H8 AOT/JAXA over the YS agreed well with the AERONET AOT (R = 0.93), relatively poor evaluation metrics were identified (MAE = 0.07–0.12, RMSE = 0.09–0.14) and only about 52–69% of the matchups fell within EE, with others mostly falling above EE, which indicates overestimation in these areas. The performance over the ECS lies between the NSCS and the YS and mainly depends on the distance of the sites from the mainland. For example, the performance of Okinawa_Hedo, which is far away from the mainland, is similar to that of the NSCS, while the others are similar to the YS.

Uncertainties in satellite retrieved AOTs are mainly from variations in surface albedo, aerosol types, and local environmental conditions. H8 AOT/JAXA particularly overestimated coastal sites. This might be due to underestimation of the surface reflectance or errors in the aerosol schemes; the surface reflectance of coastal water is usually higher than that of open ocean because of the higher concentration of sediments. On the other hand, the overestimation of AOT may be caused by the misuse of the assumed absorbing aerosol model, whereas in reality, aerosols are mixed with non-absorbing fine mode aerosols from anthropogenic emissions [27]. There is also some overestimation at open ocean sites, which is most likely due to residual cloud contamination [28]. Additionally, temporal and spatial matching windows also influenced the comparisons. For open ocean sites, the surface and atmospheric conditions can be relatively homogeneous within a certain temporal and spatial window. In contrast, the spatial window of the coastal sites may include different surface types, and the aerosol properties may be susceptible to human activity and land emissions in the temporal window.

In general, the comparisons suggest that the H8 AOT/JAXA is consistent with the AERONET AOT at the pixel level without cloud contamination within certain spatiotemporal windows over most of the China Seas, except for some coastal regions.

3.2. Spatial Distribution

Figure 4 shows the spatial distribution of the averaged AOT retrieved from Himawari-8 over the past six years (July 2015 to July 2021) over the study area. The most significant spatial patterns are: (1) AOT over high latitude seas is generally higher than that over low latitude seas, and (2) high AOT is distributed in strips along the coast and decreases gradually with increasing distance from the shore. High AOT values (>0.4) were observed over the BS, YS, northwestern ECS, and coastal waters of the NSCS. The low AOT (<0.2) is mainly found in the open ocean, such as the regions of 10–18°N and 110–130°E in the NSCS. This indicates that the aerosols over these regions are mainly from natural sources such as ocean surface splashing and evaporation of sea salt particles. A series of AOT transition zones at 0.2–0.4 lies between the high- and low-value zones.

Extremely high AOT values were observed over the Subei Shoal, Yangtze River Estuary, and Hangzhou Bay. The average AOT in these areas was above 1.0, higher than that of the nearby continent, and decreased gradually with distance from the source. These high AOT values were not consistent with the pattern of aerosol diffusion from the land to the ocean. A possible reason for this may be that there is an underestimation of coastal sea surface reflectivity when AOT is retrieved from Himawari-8. The shallow and extremely turbid water of these areas leads to relatively high reflectance [29]; however a very small reflectivity value is used for these regions, which are used for the open ocean surface when AOT is retrieved. This results in unrealistically high AOT values.

Figure 5 shows the frequency of occurrence distribution of AOT for the Chinese seas. A bin size value of 0.1 was used to generate the AOT histograms. Descriptive statistics such as the number of observations, mean, median, skewness, kurtosis, and modal value for each sea are also presented in the figure. AOT was mainly distributed over the BS between 0.4 and 1.5, with an average of approximately 0.55, a median 0.72, and 0.4 occurred the most frequently, accounting for about 43%. The histogram distribution of AOT over the YS was a little lower than that over the BS, with 0.56 and 0.47 for the mean and median, respectively, and the highest frequency value was approximately 0.52. There were also more extreme high values (1.0–1.5), accounting for approximately 6% of the AOT values. The histogram of the ECS was mainly distributed between 0.2 and 1.0, with an average value of approximately 0.33, a median of 0.26, and 0.20 as the highest frequency, accounting for approximately 57%. The histogram of the NSCS was mainly distributed in the range of 0.2–0.7, with an average of approximately 0.29, a median of 0.23, and a highest frequency of 0.2, accounting for approximately 72% of the AOT values.

Figure 6 plots the variation of AOT with distance from the coastline. It indicates that the distribution of aerosols over the China Seas may be affected by the continent. The black square points in the figure are average AOT values for every 5 km within 50 km from the coastline. The extreme high AOT values larger than 1 were excluded from the average because of the reasons discussed earlier in Section 3.2. Within the first 15 km offshore, AOT decreased rapidly nearly linearly as the distance increased. The high values of AOT close to the coast are mainly due to more intense human activities over coastal regions. The coastal regions of China are one of the most highly industrialized and urbanized regions in the world. On the one hand, anthropogenic aerosols are generally in fine modes and hence have longer residence times. On the other hand, the relatively low elevation and flat surface of the coastal areas with less forest coverage make it easier for aerosols to be transported to the costal seas under suitable weather conditions. The AOT values then slowly drop from 20 to 50 km from the coast. As the distance from the coast increases, the particle size of aerosols changes due to coagulation, condensation, and cloud cycling processes and the aerosols are finally deposited into the ocean. These processes lead to a decrease in the aerosol concentration and therefore a decrease in the AOT values. When the offshore distance exceeds 50 km, AOT remains at a low level, which basically reflects the AOT values of the maritime background aerosol.

3.3. Temporal Variation

The temporal variation at different time scales, including diurnal, monthly, and seasonal, are investigated and discussed below.

3.3.1. Diurnal Variation

The high temporal resolution of the Himawari-8/AHI sensor provides a unique opportunity for the analysis of AOT on a diurnal scale. To minimize the impact of cloud contamination and to improve the reliability of diurnal variation analysis, only eight validated consecutive hourly observations with high quality were selected. All selected observations were captured from 9:00 a.m. to 4:00 p.m. local time throughout the day under clear sky conditions. The spatial distribution of the number of selected days in the four seasons during certain periods is shown in Figure 7. The East China Seas have full coverage in all seasons, with the maximum occurrences appearing in spring (even more than 100 days), followed by summer and autumn (approximately 50 days), and minimum in winter (no more than 20 days). For the NSCS, the number of observations is masked due to the contamination of clouds or precipitation, and there are fewer validated consecutive data, especially in summer.

The coefficient of variation (CV), which is defined as the standard deviation of the eight consecutive hourly observations each day divided by the average of these observations in this study, was computed for each pixel during different seasons to investigate the spatial pattern of the diurnal variability of AOT over the China Seas (see Figure 8). The CV values were relatively small (usually within 0.2) in all seasons over the BS and YS, where AOT is large. The CV in summer (reaching 0.4) was larger than that in other seasons over the ECS. Generally, the maximum CV values (above 0.6) can be observed during spring over the NSCS, where AOT is small. The high levels of CV in the NSCS were likely due to contamination of the cloud edges.

Figure 9 shows the average diurnal cycles of AOT and departure from the daily mean over the China Seas: (a) and (e) BS, (b) and (f) YS, (c) and (g) ECS, and (d) and (h) NSCS. The diurnal variations over the four seasons over each sea are shown in different colors. Only days with eight consecutive validated hourly observations from 9:00 a.m. to 4:00 p.m. local time were chosen for the diurnal variation calculation. During this period, diurnal variations of AOT were evident over all seas, and the variation was most pronounced in spring and summer. It shows the characteristics of the two peaks at both ends in the morning and afternoon, and the low at noon. It always decreases gradually starting at local time from 09:00 in the morning, remains stable at noon (11:00–12:00), and then begins to increase steadily in the afternoon. The increase slows down after 3:00 p.m. and reaches its maximum at 4:00 p.m. This is consistent with the previous measurements over the East China Seas, which also shows that AOT is smaller at noon and much larger in the morning and evening [10,11]. Note that the magnitude of variation varies across different seas. The percentage departure from the daily mean was calculated with reference to [30]. It allows for comparison of different seasons across different seas. The departure from the daily mean shows very similar patterns among the East China Seas. It generally ranges within ±20% of its mean, which is also consistent with the coastal sites on the western Korean Peninsula [30]. AOT increased sharply in the afternoon and reached a maximum (>40%) at 4:00 p.m. over the NSCS.

Several factors may have caused the diurnal variation in AOT over the seas. First, the daily circulation of sea and land breezes varies directly with the diurnal variation in AOT. Sea breezes usually begin several hours after sunrise and disappear several hours after sunset during the daytime. Land breezes follow the sea breezes and continue until the sea breeze forms again the next day. Sea and land breezes may appear at the same time a few hours after sunrise (sunset) at different altitudes. The land winds transport continental aerosols, which contain urban and industrial emissions, to the coast, resulting in rising concentrations of aerosols over the sea, which reached a first peak at 9:00 a.m. during the statistical period. The wind direction then shifted and sea winds appeared and AOT declined until noon. Second, solar radiation causes the intensity of photochemical reactions to continually increase as the day progresses, and the concentration of secondary particles increases owing to the increasing solar radiation. Third, convective activities usually increase in the afternoon, leading to increased water vapor content, and AOT can increase significantly as the ambient relative humidity increases [31]. We also observed that the diurnal cycle of RH was closely related to the sea surface temperature, with relative humidity the lowest at noon [32]. An increasingly high relative humidity accelerates the conversion of secondary aerosols. The impacts of these factors on the diurnal variation of AOT cannot be accurately ranked using the available information. In conclusion, AOT increased in the afternoon and reached the second peak of the statistical period at 4:00 p.m. The growth rate falls after 3:00 p.m., which may be due to the weakening of solar radiation.

3.3.2. Monthly Variation

Figure 10 shows the time series of the monthly average AOT for the four sea areas from July 2015 to July 2021. Overall, the long-term variation in AOT showed a pronounced annual cycle. For the BS, there was usually a peak above 0.6 in July in the summer; however, the spring peak was more unpredictable. There were different reasons for the peaks in spring and summer. The main reason for the summer peaks is that as the sea surface temperature and relative humidity rise in the summer, the scattering of growing aerosol particles also intensifies, leading to an increase in AOT. The peak in spring is mainly caused by dust storms, which usually break out in arid Asian land and transport to the sea surface [33]. As the surface temperature of the arid Asian land increases from winter to spring, the frozen soil melts and the sand softens, which are conducive to the rise of dust aerosols. The dust particles are blown up by strong ground winds formed by the cold front, transported to the middle troposphere by the strong updraft, and then transported to the downstream area under the action of high-altitude westerly jets. In addition, the low rainfall in the spring over arid areas is conducive to the long-distance transportation of dust aerosols. Because of the varying frequency and intensity of dust storms each year, the spring sea surface AOT is high in some years (such as 2016 and 2018), while in some other years it has not increased significantly (such as 2017).

The monthly variation of AOT over the YS was similar to that of the BS. It is also located on an important passageway of dust transport from the Asian landmass to the northwestern Pacific Ocean. In the ECS and NSCS, AOT rises significantly from January to March, peaking at more than 0.4 in March, gradually declining from April to August, and then remaining stable from September to December. The summer peaks in the ECS and NSCS are much lower than those in the BS and YS. This is because the ECS and NSCS have a wider coverage and are relatively farther away from the mainland, meaning the emissions of land-source natural aerosols are significantly lower in summer. In addition, there is much more wet deposition of aerosols over these areas in summer due to abundant summer precipitation.

The linear fits in Figure 10 also show a gradual decline in AOT for various seas from July 2015 to July 2021. A significant decrease in Himawari-8AOT of ~0.17 (28%) was observed over the BS, followed by 0.14 (24%) over the YS, 0.02 (5%) over the ECS, and 0.03 (9%) over the SCS. A 10–20% decrease in AOT over the main Chinese coastal outflow regions from 2008 to 2015 was observed based on MODIS and MISR retrievals [34]. It was also found that MODIS AOT over the ECS gradually increased from 2003 to 2009 but then decreased from 2009 to 2013 [35]. It was found that the implementation of the Clean Air Action Plan (2018–2020) and the lockdown measures for COVID-19 in 2019 led to a markedly decreasing of air pollution emissions in the central and eastern part of the continent [36]. However, the obvious trendline break points were not found over the China Seas during this period based on the Himawai-8 AOT.

3.3.3. Seasonal Variation

Figure 11 shows the seasonal distribution and variation of the AOT over the China seas from July 2015 to July 2021 based on Himawari-8 observations. The value of each pixel is the average of all validated AOT during each season of the study period: winter (December, January, and February), spring (March, April, and May), summer (June, July, and August), and autumn (September, October, and November). For the BS, the largest AOT was observed in summer, the second largest in spring, then autumn, and the minimum was observed in winter. AOT was generally above 0.5, reaching as high as 0.8 over some coastal areas of Bohai Bay and Laizhou Bay in summer and spring. However, the reasons for the high AOT in these two seasons are different. The high AOT in spring was mainly due to the impact of Asian dust storms; the BS is located on important pathways of dust storms in the Northwest Pacific [33]. The reason for the high AOT in summer is that the higher temperature and humidity during the summer are ideal conditions for gas and particle conversion, which increases the average size of the particle and leads to larger particle extinction efficiency and AOT [13]. Wet deposition processes, such as snowfall, are highly efficient in removing aerosols from the atmosphere. Snowfall usually occurs in the BS during winter because of the cold vortex and its special topography [37]. Therefore, the minimum of AOT over the BS was in the winter.

For the YS, the largest AOT was in spring, followed by summer, winter, and autumn. The YS is also located on the pathways of Asian dust storms that are transported to the Northwest Pacific, leading to a high AOT in spring. In summer, increasing temperature and humidity also lead to a high AOT. AOT in winter is usually higher than that in autumn because of the increasing urban pollution, biomass burning, and less snowfall as compared to the BS. The seasonal variation in the spatial pattern of the ECS is similar to that over the Yellow Sea. The trend of seasonal variation of AOT over the East China Seas from Himawai-8 is consistent with MODIS in general [35]. AOT values in spring and summer were significantly higher than those in winter and autumn.

For the NSCS, the largest AOT was in spring, followed by winter, autumn, and summer. The NSCS suffers from southeast monsoon winds in summer, which prevent the transport of land source aerosols from the continent to the seas. Precipitation is high in summer, and aerosols are efficiently removed by wet deposition. The northeast monsoon usually begins in autumn (mid-September to late-October) and ends in spring (May) the following year. It takes land source aerosols from the continent to the seas and leads to high AOT in coastal regions. AOT is generally above 0.6 and can be as high as 0.9 over the Beibu Gulf in the northwest of the NSCS in spring. The Beibu Gulf is a semi-enclosed bay that only opens south to the NSCS. The coastline and special topography play important roles in aggravating the concentrations of aerosols, leading to a high AOT.

The coverage of extremely high AOT over the shoal-water area along the northern coast of Jiangsu Province, Yangtze River Estuary, and Hangzhou Bay in spring and winter were higher than in summer and autumn. As mentioned in Section 3.2, higher wind speeds in winter and spring lead to vertical mixing and increase the surface reflectance, which may result in a larger positive bias of AOT retrieval for Himawari-8.

4. Conclusions

The spatial and temporal variation of AOT over the China Seas, including the BS, YS, ECS, and NSCS, were investigated based on hourly level 3 H8 AOT/JAXA observations from 2015 to 2021. The principal conclusions drawn from this study are summarized below:

(1)

The H8 AOT/JAXA is consistent with the AERONET AOT at the pixel level without cloud contamination within a certain spatiotemporal window over most of the China Seas, except for some coastal regions. The H8 AOT/JAXA retrievalsagree very well with AERONET measurements, with a high correlation coefficient of 0.914 and small MAE and RMSE values of 0.081 and 0.11, respectively. Approximately 65% of the matchups fall in the EE, and the largest concentration of the matchups appear at 0.1 < AOT < 0.4. The performance of H8 AOT/JAXA over the open ocean was better than that over coastal seas.

(2)

The most significant spatial patterns are that AOT over high latitude seas are generally larger than those over low latitude seas, and AOT is distributed in strips along the coastline and decreases gradually with increasing distance from the coast.

(3)

Based on the number of pixels that have eight consecutive validated hourly observations from 9:00 a.m. to 4:00 p.m. local time each day, the diurnal variation of AOT was proven. AOT decreases gradually starting from 9:00 a.m. in the morning local time, remains stable at noon (11:00 a.m.–12:00 p.m.), and then begins to increase steadily in the afternoon. The rate of increase slows after 3:00 p.m. and AOT peaks at 4:00 p.m. The percentage daily departure of AOT showed a very similar pattern over the East China Seas and generally ranged within $\pm$20%, while in the NSCS, AOT increased sharply in the afternoon and reached a maximum (>40%) at 4:00 p.m.

(4)

The long-term monthly variation in AOT shows a pronounced annual cycle. For the BS and YS, there is usually a peak above 0.6 in July in summer, however the emergence of spring peaks is more unpredictable. The summer peaks in the ECS and NSCS drop more sharply than those in the BS and YS. A gradual decline in AOT for various seas has been observed since 2015.

(5)

The seasonal variation of the spatial pattern shows that for the BS, the largest AOT was observed in summer, followed by spring, autumn, and winter. For the YS and ECS, the maximum AOT was observed in spring, followed by summer, winter, and autumn. For the NSCS, the largest AOT was observed in spring, followed by autumn, winter, and summer.