**3. Aerosols monitored by satellites over the ocean**

#### **3.1. Data**

Aerosol robotic network (AERONET) is a ground‐based optical aerosol monitoring network used to ensure the quality of aerosol products [29]. We obtained the AOT data from the AERONET project web page (http://aeronet.gsfc.nasa.gov). Data quality control is employed to eliminate errors caused by measurement conditions, such as clouds and biomass burning events [30]. There are a total of 554 measurement sites in the world. The site locations are distributed over islands, coastal areas, inland sites and mountains. The AOT was measured during the 2006 winter and 2007 autumn cruises over the ECS using a handheld multi‐band sun photometer (MICROTOPS manufactured by the Solar Light Company). A total of 17 stations during the winter cruise, and 18 stations during the autumn cruise were measured with much more measurements during the transit between stations. To compare values between SeaWiFS and AOT, a matched data set was assembled using a time window for the satellite overpass of ±2 h from the in situ measurement. The locations of the matched dataset are shown in **Figure 3**.

**Figure 3.** Some locations used in the paper. (1) Red dots represent the locations of matched MICROTOPS measured AOT and SeaWiFS data. (2) The three green crosses are the locations used for the comparison between the UAC model and BOA method. (3) The four blue stars are the locations of the AERONET sites and the nearby blue crosses are the locations for the matched SeaWiFS data. (4) The pink triangle is the location of the SeaWiFS measurements used for the analysis of the BOA method.

#### **3.2. Some results**

higher than the satellite TOA values. Therefore, magnitudes of the aerosol scattering reflec‐ tance in the visible bands are very sensitive to the bias of the reflectance in the NIR bands using

Aerosol robotic network (AERONET) is a ground‐based optical aerosol monitoring network used to ensure the quality of aerosol products [29]. We obtained the AOT data from the AERONET project web page (http://aeronet.gsfc.nasa.gov). Data quality control is employed to eliminate errors caused by measurement conditions, such as clouds and biomass burning events [30]. There are a total of 554 measurement sites in the world. The site locations are distributed over islands, coastal areas, inland sites and mountains. The AOT was measured during the 2006 winter and 2007 autumn cruises over the ECS using a handheld multi‐band sun photometer (MICROTOPS manufactured by the Solar Light Company). A total of 17 stations during the winter cruise, and 18 stations during the autumn cruise were measured with much more measurements during the transit between stations. To compare values between SeaWiFS and AOT, a matched data set was assembled using a time window for the satellite overpass of ±2 h from the in situ measurement. The locations of the matched dataset

**Figure 3.** Some locations used in the paper. (1) Red dots represent the locations of matched MICROTOPS measured AOT and SeaWiFS data. (2) The three green crosses are the locations used for the comparison between the UAC model and BOA method. (3) The four blue stars are the locations of the AERONET sites and the nearby blue crosses are the locations for the matched SeaWiFS data. (4) The pink triangle is the location of the SeaWiFS measurements used for the

the BOA method. This problem can be significantly reduced by the UAC model.

**3. Aerosols monitored by satellites over the ocean**

**3.1. Data**

116 Aerosols - Science and Case Studies

are shown in **Figure 3**.

analysis of the BOA method.

Normally, the accuracy of the aerosol scattering reflectance depends directly on the value of epsilon. A small error of the epsilon will easily lead to an amplified error in the aerosol scattering reflectance. If the entire epsilon spectrum instead of one value at the NIR band is used to determine the two closest aerosol models, the results are more robust. The key advantage of the UAC model is that it finds suitable aerosol models by matching the epsilon spectrum and thus improves the accuracy of the aerosol reflectance. The UAC model was used to process the SeaWiFS image on March 27, 2007 and the epsilon image is shown in **Figure 4**. To produce this image, the Rayleigh scattering reflectance was computed using a multiple scattering approach to obtain the aerosol-water reflectance. The aerosol scattering reflectance and the epsilon spectra were obtained by the UAC model.

**Figure 4.** An epsilon image from SeaWiFS data on March 27, 2007 using the UAC model.

From the refined epsilon distribution in **Figure 4**, we can see that the high values of epsilon in the coastal regions are significantly reduced from around 1.4 to 1.05, with values closely in oceanic waters. As the spectra of the reflectance follow the Angstrom law, the structures of images in one band are similar to those in the other bands. The image of the aerosol scattering reflectance in Band 2 is selected and shown in **Figure 5**.

The *ρA* (*λ*) values in Band 2 vary from 0.0042 to 0.015 sr−1, with the mean value of 0.0092 sr−1 and the standard deviation of 0.0029 sr−1. The distributions of the image indicate that *ρA* (*λ*) varies significantly from one location to another, with the maxima in the northern part and the minima in the south. From the structures of the image in **Figure 5**, it is clear that the UAC model has almost removed the influence effects of the non-zero water-leaving reflectance in the NIR bands in the coastal regions, even for highly turbid waters.

The water-leaving reflectance in the NIR bands varies from 0 to 0.06 sr−1, almost six times higher than the aerosol reflectance. Under the BOA, up to six times extra values will be falsely added to the actual aerosol reflectance in the NIR bands, which obviously leads to the failure of the atmospheric correction procedure. In comparison, the UAC method makes use of the LUT, the magnitude variation in the ground reflectance has no effects in determining aerosol reflectance. The results show that the UAC model has clearly separated the aerosol scattering reflectance, working well from Case 1 to Case 2 waters.

**Figure 5.** An image of aerosol reflectance retrieved from SeaWiFS Band 2 on March 27, 2007 using the UAC model.

#### **3.3. Validation of the retrieval aerosols over the sea**

The globally distributed AERONET data can be used to validate the aerosol products derived from satellite ocean colour remote sensing data such as SeaWiFS [31]. The SeaWiFS reflectance was processed to obtain epsilon using the UAC model, and the results are shown in **Figure 6**. The AOT spectra were used to calculate the aerosol reflectance using the geometric angles of SeaWiFS, to obtain the epsilon values.

**Figure 6** shows that the two types of epsilon are relatively comparable. Both the magnitudes and the spectra shapes are similar to each other. AOT derived from MODIS and AERONET must be matched on space and time. Window size on space and time are set as 50 km × 50 km and 1 h in the anterior "http://www.iciba.com/" research. The correlation and regression coefficients show excellent agreement with AERONET measurements over the China Sea. Meanwhile, the relationship is all good and their correlation coefficients are greater than 0.9. Errors of 65% points are under ± 0.05 ± 0.05 *τ* based on NASA standard. Fine mode fraction (FMF) is validated over the ocean with the error of about 20%.

model has almost removed the influence effects of the non-zero water-leaving reflectance in

The water-leaving reflectance in the NIR bands varies from 0 to 0.06 sr−1, almost six times higher than the aerosol reflectance. Under the BOA, up to six times extra values will be falsely added to the actual aerosol reflectance in the NIR bands, which obviously leads to the failure of the atmospheric correction procedure. In comparison, the UAC method makes use of the LUT, the magnitude variation in the ground reflectance has no effects in determining aerosol reflectance. The results show that the UAC model has clearly separated the aerosol scattering reflectance,

**Figure 5.** An image of aerosol reflectance retrieved from SeaWiFS Band 2 on March 27, 2007 using the UAC model.

The globally distributed AERONET data can be used to validate the aerosol products derived from satellite ocean colour remote sensing data such as SeaWiFS [31]. The SeaWiFS reflectance was processed to obtain epsilon using the UAC model, and the results are shown in **Figure 6**. The AOT spectra were used to calculate the aerosol reflectance using the geometric angles of

**Figure 6** shows that the two types of epsilon are relatively comparable. Both the magnitudes and the spectra shapes are similar to each other. AOT derived from MODIS and AERONET must be matched on space and time. Window size on space and time are set as 50 km × 50 km and 1 h in the anterior "http://www.iciba.com/" research. The correlation and regression coefficients show excellent agreement with AERONET measurements over the China Sea.

the NIR bands in the coastal regions, even for highly turbid waters.

working well from Case 1 to Case 2 waters.

118 Aerosols - Science and Case Studies

**3.3. Validation of the retrieval aerosols over the sea**

SeaWiFS, to obtain the epsilon values.

**Figure 6.** The comparison of the epsilon spectra from SeaWiFS reflectance using the UAC model (Red lines) to those from the AOT measured by the MICROTOPS instrument (Black lines) in the ECS. Green lines represent abnormal epsilon spectra.

Some factors affect the results of the validation. As the AOT data of AERONET are retrieved from the ground based radiance measurements, the data quality of AOT can also be affected by instrument calibration and the measurement conditions [30]. The translation from AOT to the aerosol scattering reflectance is related to the parameters of the aerosol single albedo and its phase function, which are determined by the absorption, and the type and particle size distribution of aerosols. These parameters were not measured, and typical values were used in the computation of epsilon from the AOT.

#### **3.4. Spatial and temporal variations of aerosols over the ocean**

The seasonal mean of MODIS AOT over the China Sea from 2001 to 2006, shown in **Figure 7**, demonstrates a notable seasonal change in spatial distribution. The maximum AOT appears in spring with the minimum in summer. In winter, high AOT is distributed along the coast and decreases obviously with the distance far from continent. Maximum AOT has been observed in coastal region and the value of it is larger than 0.5, meanwhile, the minimum AOT appears in open ocean and value of it is 0.1. The distribution of FMF is similar, suggesting that human activity influences aerosols. AOT over the north of 25°N is all larger than 0.17 and it over the south of 25°N is smaller than 0.17. The distribution of AOT in spring is similar to that in winter. Because of dust prevailing, AOT increases obviously and is larger than 0.17 over the whole China Sea. The distribution in summer changes remarkably, with large values in the north and small in the south. Maxima appear over the Yellow Sea and Bohai Sea. Value of AOT in fall becomes the least in four seasons, and the pattern shows similarly structure of that in winter.

**Figure 7.** Seasonal distribution of AOT over the China Sea from 2001 to 2006. (a) winters; (b) spring; (c) summer; (d) fall.

Both AOT and FMF exist an annual cycle over the China Sea. The AOT appears maxima in spring with values larger than 0.4 and reaches minimum in summer with values less than 0.25. The FMF also has a periodic oscillation but an opposite tendency. The FMF reaches maxima in summer with values larger than 0.6 and gets minima in spring with values less than 0.5. The mean AOT and FMF during the 6 years (2001–2006) over the China Sea are shown in **Figure 8**. The AOT is the largest in April and the smallest in September, while the FMF is the largest in September and the smallest in April. Based on the above analysis, we can find that AOT and FMF exist in significant temporal variety with an annual cycle over the China Sea.

**Figure 8.** Monthly mean of the AOT (left panel) and the FMF (right panel) over the China Sea.

in winter. Because of dust prevailing, AOT increases obviously and is larger than 0.17 over the whole China Sea. The distribution in summer changes remarkably, with large values in the north and small in the south. Maxima appear over the Yellow Sea and Bohai Sea. Value of AOT in fall becomes the least in four seasons, and the pattern shows similarly structure of that in

**Figure 7.** Seasonal distribution of AOT over the China Sea from 2001 to 2006. (a) winters; (b) spring; (c) summer; (d)

Both AOT and FMF exist an annual cycle over the China Sea. The AOT appears maxima in spring with values larger than 0.4 and reaches minimum in summer with values less than 0.25. The FMF also has a periodic oscillation but an opposite tendency. The FMF reaches maxima in summer with values larger than 0.6 and gets minima in spring with values less than 0.5. The mean AOT and FMF during the 6 years (2001–2006) over the China Sea are shown in **Figure 8**. The AOT is the largest in April and the smallest in September, while the

winter.

120 Aerosols - Science and Case Studies

fall.

**Figure 9.** The latitudinal and longitudinal distribution of the AOT and FMF over the China Sea.

How about the spatial distribution of the AOT and FMF over the China Sea? The latitude-time and longitude-time diagrams of the AOT and FMF over the China Sea are shown in **Figure 9**. It represents a spatial characteristic related with the distance far from the coast and important industry regions, due to, which aerosols over the China Sea are mainly influenced by the continent source.

The AOT appears the highest value between 30°N and 40°N in the latitude direction, as there are main industry regions. The FMF also reaches the highest values in the same region. Both AOT and FMF get less towards the southern of the 30°N, due to lack of industry pollution. The human influence is very evident at longitude direction. China coast is around 120°E with the highest of the AOT and the FMF. Both the AOT and FMF gradually decrease away from the coast. The decrease tendency of AOT over the East China Sea has an affinity with the distance far from the continent, suggesting the AOT should be influenced by the continental sources.

#### **3.5. Correlation to meteorological conditions**

The aerosol's transport depends on wind, so seasonal change of wind may lead to the spatial and the temporal distribution of aerosols over the China Sea. The mean wind field at 850 hp in 2001–2006 is shown in **Figure 10**. The patterns are similar between winter and spring, demonstrating that northwestern wind control the north of 20°N and eastern wind control the south of 20°N. The eastern wind prevails at most of region in summer and fall, while southwestern wind exists in the region between 30°N–40°N. Meanwhile, rainfall is the high in summer and low in winter and spring.

**Figure 10.** Seasonal distribution of wind field at 850 hp over the China Sea. (a) winters; (b) spring; (c) summer; (d) fall.

The aerosols are sent from the continental to ocean by the northwestern wind in winter. High wind speed causes that aerosol is difficult to bank up onto the China Sea, slowing down the values of the AOT. Meanwhile, high wind speeds create big particles of marine aerosol, leading to the smallest of the FMF. In spring, northwestern winds carry dust aerosols to the China Sea by and wind speed becomes weaker, so that aerosols build up onto the China Sea. The AOT reach maxima with a small of the FMF. In summer, east wind prevails, so continental aerosol can't be transported onto the sea. The rainfall is the large, bringing down the big particles of aerosols. So, the AOT is the smallest with the largest of the FMF. In fall, the west winds control the region of the north of 30° N, so the AOT is larger than summer. Rainfall is next below in summer, so the FMF is smaller than that in summer, but higher than that in other seasons. Therefore, meteorological conditions play an an important role on distribution of the AOT and FMF.
