Current Issue

The earth’s atmosphere has been directly or indirectly affected by both human activities and natural change factors (IPCC, 2014). For example, human activities affect the atmospheric environment by emitting various kinds of particulate matter and gaseous substances through automobiles, heating, or agricultural or production activities (McDuffie et al., 2020; Reddington et al., 2019). As for the natural factors, sand storms in deserts increase dust particles in the atmosphere and volcanic eruptions can release various particulate or gaseous substances into the atmosphere (Liora et al, 2015). In the ocean, micro-organisms in water also act as a factor of climate change by releasing oxides (Cavicchioli et al., 2019; Vallina and Simo, 2007; Leck and Bigg, 2005; Charlson et al., 1987). These issues related to the atmospheric environment are not only specific to an independent local area, but also the adjacent and distant areas caused by intertwined effects. For example, long-range transported air pollutants including desert dust storms, biomass burning smoke, and industrial air pollution may occur internationally (Hatakeyama et al., 2017; Coulibaly et al., 2015). Under these circumstances, it is becoming more important to understand whether the cause of air pollution is from a local source or the contribution of the long-range transport from the outer boundary. Therefore, it is necessary to estimate the current states and evaluate the air pollution levels by multi-dimensional monitoring.

As an atmospheric environmental monitoring technology, satellite data has been widely used to study the temporal and spatial variables of the atmospheric environment over the past few decades (Kokhanovsky and de Leeuw, 2009; Lee et al., 2009; King et al., 1999). For example, the meteorological satellites such as the NOAA series, the Geostationary Operational Environmental Satellite (GOES), the Geostationary Meteorological Satellite (GMS), METEOSAT, and the Earth Observing System (EOS) series have been conducting continuous observation of the global atmosphere (Ackerman et al., 2019). The Earth observing satellites has stimulated the developments in various environmental remote sensing technologies and the production of analytical data. In addition, the analysis techniques using satellite data play an important role in their applications in various fields such as meteorology, air quality, ocean, land cover/land change, agriculture, and so on (Guo et al., 2015). Therefore, the purpose of this study is to review the characteristics of various Earth observing satellites providing environmental big data and processing techniques for obtaining environmental parameters. This study introduces the theoretical principles of characteristics and observation platforms currently used in global environmental monitoring. In addition, by collecting information on the global environment observed for a long time from satellites, the patterns of global environmental changes are also identified based on statistical analysis and comparative analysis results for each parameter.

The main purpose of earth observation is to enhance the understanding of the global environment and to establish a more effective earth observation system (Anderson et al., 2017). Focusing on this purpose, comprehensive studies of the ocean, atmosphere, land, climate, vegetation, diseases, and disasters have been carried out through many earth observation platforms (Dowman and Reuter, 2016). In order to maximize the utilization of Earth observing satellite, practical efforts such as presentation of data policies, collaborations among international researchers, operation of data centers, and holding of periodic science meetings have been deployed. For example, the Megacity (Zhu et al., 2012) or the Asian Brown Cloud (ABC) programs (Nakajima et al., 2007; Ramanathan et al., 2005) by the International Global Atmospheric Chemistry (IGAC) are international joint observation programs using satellite data for comprehensive analysis on environmental monitoring.

For a better understanding of the global environment, satellite observations are inevitable because satellites provide a wide range of data on the Earth’s environment. Representative examples of satellites include EOS, Environment Satellite (ENVISAT), and Advanced Earth Observing Satellite (ADEOS) series as listed in Table 1. These Earth observing satellites have been widely used for global environmental monitoring and output data have been rapidly increased in all fields such as geographic information, military information, resource exploration, environmental observation, and global climate change prediction. For example, the surface temperature of land and ocean, primary production, land cover, clouds, aerosols, water vapor, temperature, and forest fires have been observed for a long time, and the scientific information produced has played an important role in environmental protection and policy making (Dowman and Reuter, 2016).

More recently, new generation sensors aboard polar orbit satellites and geostationary orbit satellites play a large role in planetary monitoring. The capabilities of these sensors have been greatly improved in terms of spectral spectrum, radiometry, observation time, and scanning speed. For example, the new generation sensors onboard the most recently launched geostationary satellites, such as Advanced Baseline Imager (ABI) on the Geostationary Operational Environmental Satellite-R (GOES-R) series of satellites (Schmit et al., 2017, 2005), Advanced Himawari Imager (AHI) on the Himawari series (Bessho et al., 2016), and Advanced Meteorological Imager (AMI) on the Cheollian satellites (Kim et al., 2021; Jee et al., 2020), have three times more spectral channels, four times better spatial resolution, and five times faster scanning speeds than the previous meteorological sensors. The advanced sensors offer great opportunities for instant and continuous monitoring of acute and extensive disasters (e.g., dust storms, wind storms, and volcanic abruptions). Fig. 1 shows the wavelength range of the three sensors along with typical satellite sensors.

Fig. 1. 
Spectral band positions across the sensors alongside those of missions for atmospheric environment monitoring. The black line is solar transmittance and flux from LOWTRAN. AHI - Advanced Himawari Imager on Himawari 8/9 (Japan), AMI - Advanced Meteorological Imager on GEO-KOMPSAT2 (South Korea), MODIS - Moderate Resolution Imaging Spectroradiometer (United States).

The development of sensor technology has greatly enhanced the ability of satellite data applications. For example, ENVISAT is one of the representative Earth observing satellites and can obtain environmental observation data from 10 onboard sensors. Among these sensors, Scanning Imaging Absorption Spectrometer for Atmospheric ChartY (SCIAMACHY) is a spectrometer that measures sunlight with a wavelength resolution of (0.2 to 1.5 nm) in the ultraviolet-near infrared region (240 nm to 2,380 nm) which can measure atmospheric constituents. In addition, Ozone Monitoring Instrument (OMI), The TROPOspheric Monitoring Instrument (TROPOMI), etc. can observe various types of gases, aerosols, radiation, clouds, and cloud altitudes (Fig. 2).

Fig. 2. 
Spectral ranges for hyperspectral sensors including GEMS, GOME, OMI, SCIAMACHY, and TROPOMI.

One of the oldest earth observing satellites currently in operation is the EOS-AM1 (Terra) satellite, which was launched on December 18, 1999. It has been collecting various earth observing data and analysing the impact of human activities on global environmental changes. The EOS-PM1 (Aqua) satellite, launched in 2001, is the successor to the Terra Earth Observation Satellite. Both EOS satellites have been used to identify seasonal cycles of ecosystems in land and ocean, to detect seasonal changes in the atmosphere and clouds, and to calculate the daily surface temperature of the globe regardless of weather conditions. Among the various sensors mounted on the EOS satellites, the MODerate Resolution Imaging Spectroradiometer (MODIS) (Barnes et al., 2003; Pagano and Durham, 1993) is a multi-spectral radiometer that measures radiation energy in the visible to long-wave infrared wavelength region at intervals of about 1.478 seconds. The observation range of MODIS is between -55 degrees west to +55 degrees east, which corresponds to a range of about 2,330 km on the ground (cross track), and it can scan an area of about 10 km in the direction of the satellite (along track). The EOS satellites orbit the Earth’s polar orbit once for about 100 minutes, and it orbits about 14.4 times a day, so it takes about 2 days to scan the entire Earth. MODIS has a total of 36 channels, among which 2 channels (640 nm, 860 nm) with a resolution of about 250 m at nadir view, 5 channels with a resolution of 500 m (470 nm, 555 nm, 1,240 nm, 1,640 nm, 2,130 nm), and 29 channels with a resolution of 1 km. About 20 channels out of 36 measure in the spectrum of reflected sunlight, and the others measure in those of infrared.

Satellite observations have become a means to complement the existing computational numerical models by providing information on the status and behaviour of atmospheric aerosols and clouds. The aerosol-related physical quantity commonly used in satellites is the aerosol optical thickness (AOT), which means the amount of radiation attenuated by the aerosol particles in the atmosphere. In general, AOT is determined by analysing the radiative transfer process of the reflected light of the sun in the visible region. The following Eq. (1) is widely used for the radiative transfer process by the Earth’s atmosphere and surface in the visible wavelength range:

where, ρ_TOA , ρ_Aer, ρ_Ray , and ρ_Sfc are the reflectance observed by the satellite sensor at the top of the atmosphere, the reflectance by atmospheric aerosol, the reflectance by the molecules in the atmosphere, and the reflectance by the ground surface, respectively. T_Sun and T_Sat are the atmospheric transmittance corresponding to the path from the target point observed by the satellite to the sun and the satellite, respectively. r_Hem is the reflectivity due to the atmospheric hemisphere. λ, τ represent wavelength and AOT, respectively. Basically, the AOT is determined from the relationship between τ and ρ_Aer which can be acquired by deducting the molecular and surface reflectance terms from ρ_TOA. Based on this process, various types of aerosol retrieval algorithms were developed (Lee et al., 2009).

On the other hand, Eq. (1) uses scattered light, there are a few algorithms that use the differential absorption of the aerosols for ultraviolet (UV) or infrared (IR) channels. These methods use the various functions of difference or ratios derived from the sensitive and insensitive wavelengths to the absorption of aerosol, e.g.,

where I, L_abs and L_nabs are the derived index for absorbing aerosols, radiance of absorbing aerosols, and non-absorbing aerosols, respectively. It is worth noting that the index value is expressed in various ways depending on the algorithm (Li et al., 2021).

Additionally, the thermal infrared (TIR) spectrum is preferentially sensitive to the coarse-mode aerosol particles (Peyridieu et al., 2010). The TIR has been implemented for retrieving large aerosol particles, such as dust (Zhang et al., 2006) and volcanic ash aerosols (Wen and Rose, 1994). The satellite sensor captured thermal radiances consisting of two parts, the radiance emitted from the aerosol (e.g., dust, volcanic ash) cloud layer and the radiance emitted from the underlying surface. The radiative transfer process at thermal spectra can be approximated as:

where I_i is the observed radiance at the top of the atmosphere (TOA), t_i and ε_i are the transmissivity and emissivity of the cloud layer at wavelength i. T_s and T_c indicate the underlying surface temperature and the top temperature of the cloud layer. The B stands for the Plank function. B(T_s) and B(T_c) are the blackbody radiances at equivalent temperatures of T_s and T_c. The emissivity of the underlying surface is assumed to be 1 in Eq. (4). Both the transmissivity and emissivity of the cloud layer are determined by the optical depth τ_i and they are expressed as:

According to Eqs. (4), (5), and (6), the satellite observed radiance is strongly controlled by the underlying surface, cloud temperature, and optical depth of the aerosol layer. Therefore, the AOT can be inversely calculated if the TOA radiance has been obtained and T_s and T_c been fixed.

Table 2 summarizes the aerosol retrieval algorithms which have been developed according to the operating purpose and characteristics of each satellite sensor.

To remove (or minimize) the effect of surface reflection, the aerosol retrieval methods listed in Table 2 generally use background images (e.g., clearest images with very high atmospheric transmittance) or dark surface reflectance in the specific channels (e.g., near-infrared channels in seawater, blue and red channels in areas with dense vegetation, etc.). After removing the influence of surface reflection, the satellite signal is then converted into the physical quantity of aerosol using the pre-calculated look-up tables (LUTs).

One of the important environmental measurements of the EOS program is the observation of atmospheric aerosols and clouds. Earth observing satellites prior to MODIS were difficult to observe aerosols over bright surfaces, mainly due to the limitations of the number of observation channels. MODIS made it possible to observe aerosols with improved spectral resolution. One of the characteristics of the MODIS aerosol retrieval method is to determine the visible channel surface reflectance by using both background image and dark surface methods. The “Deep Blue” and “Dark Target” methods use a background surface reflectance and a change rate for each wavelength, respectively (Levy et al., 2013; Hsu et al., 2006; Remer et al., 2005). Another feature is that the aerosol signal is divided into fine mode and coarse mode. The aerosol load and size information are obtained using the calculated look-up tables assuming 11 particle size distributions (5 fine particle modes, 6 coarse particle modes) in 6 bands between 0.55 μm to 2.13 μm. MODIS aerosol products are processed into data (codename: MOD04, MYD04) with a spatial resolution of about 10 km at Level-2 and gridded data at one-degree latitude and longitude intervals at Level-3 (MOD08) format.

Recently, the MODIS aerosol retrieval algorithm has been updated to version 6, and different aerosol products according to the algorithm have been produced. The results of comparing the different MODIS aerosol products over the East Asian region were reported (Lee, 2018). Fig. 3 shows the mean AOT in East Asia observed by MODIS from 2001 to 2016. Statistical distribution of aerosol can be confirmed through this long-term satellite observation result. It clearly shows that the Asian aerosols are mainly caused by carbon particles generated by combustion and sand storms in dry areas (i.e. Gobi Desert, Taklamakan Desert, etc.), and pollutant particles generated in densely populated areas or industrial areas. Moreover, AOT in these areas appears to be in relatively large values.

Fig. 3. 
The averaged spatial distribution of AOT from different collections and retrieval algorithms for MODIS level 2 aerosol products during 2000-2019 (Terra) and 2002-2019 (Aqua), respectively. Note that AOT data from DB algorithm are available over land only.

Fig. 4 shows the long-term trend of AOT from 2001 to 2019. On a global scale, the pattern of AOT clearly shows a seasonal cycle that rises in summer and falls in winter, but a significant increasing or decreasing trend is not found (linear fit slope=1.996×10^-4, Y-intercept=0.170). On the contrary, the AOT over East Asia (10°S-60°N, 60°E-150°E) is larger than that of the global scale and is showing an increasing trend until 2012 (linear fit slope=1.879×10^-4, Y-intercept=0.251). After 2013, it shows a slight decrease (linear fit slope=-4.0×10^-4, Y-intercept=0.279). Mean AOT values during the whole period are 0.264±0.039 for East Asia and 0.180±0.015 for the globe, that the East Asian mean AOT is about 1.47 times larger than the global average. These regional imbalances and long-term trends in atmospheric aerosols found in East Asia are unique. Therefore, areas with a high load of atmospheric aerosols are known to have a high contribution to atmospheric hydrological circulation and radiation balance (Ramanathan et al., 2001). This mechanism is one of the uncertainty factors in climate change study and is also related to many other mechanisms of climate change that are still not fully understood.

Fig. 4. 
Time series of area averaged Terra/MODIS AOT over 2000-2019. Second order poly-fit equations are plotted as dashed lines.

The properties of absorbing or scattering sunlight by aerosol particles differ depending on the chemical composition and physical properties of the aerosol (Myhre, 2009; Forster et al., 2007). Aerosols also interact with cloud and hydrological circulation by acting as cloud condensation nuclei (CCN) and ice nuclei (Seinfeld et al., 2016). While the liquid water content (LWC) is constant, more CCN increases the cloud albedo (indirect cloud albedo effect) (Lohmann and Feichter, 2005) and decreases the precipitation efficiency (indirect cloud lifetime effects) (Koch and Del Genio, 2010; Hansen et al., 1997), reducing the global annual average of net radiation at the top of the atmosphere. However, these effects may be partially offset by evaporation of cloud particles due to aerosols (indirect effects) and more ice nuclei (glacial effects). In addition, the surface can be cooled due to the lack of incoming solar radiation, while the boundary layer can be warmed by absorption of solar radiation; interaction between the two systems can increase thermal stability and reduce cloud formation, thereby reducing rainfall. Menon et al. (2002) stated that the presence of pollution particles mainly composed of carbon components in the Asian atmosphere exhibits a cooling effect of the surface by scattering or absorbing sunlight, which is approximately three times larger than the greenhouse effect.

Atmospheric aerosols are directly or indirectly related to climate, and eventually affect terrestrial organisms. Specifically, both the clouds and aerosols can exert influences on the surface vegetation. On one hand, the presence and interaction of aerosols and clouds can block sunlight reaching the surface and thus reduce the amount of light energy required for photosynthesis of plants. On the other hand, photosynthesis and evaporation may rather increase with more scattered rays by aerosols and clouds. Therefore, it is critical to analyse the relationship between the aerosols and clouds in the Earth’s atmosphere with climate change and environmental change. Several indicators have been developed to quantify the respective properties of aerosols, clouds, and vegetations, including the Angstrom Exponent (AE), Cloud Effective Radius (CER), and Normalized Difference Vegetation Index (NDVI). The AE describes how the AOT depends on the wavelength of the light (Ångström, 1929). For τ₁ and τ₂ at two different wavelengths of λ₁ and λ₂, the AE is given by;

It is typically used as a relative indicator of aerosol size, with larger values indicating small particles, and smaller values indicating large particles. The CER is a weighted mean of the size distribution of cloud droplets which is typically used in cloud remote sensing (Hansen and Trevis, 1974). It is expressed as;

CER data from satellite remote sensing has been widely used for estimating aerosol-cloud interactions (Ross et al., 2018; Menon et al., 2008) and cloud feedback (Tan et al., 2019). NDVI is generally used to quantify vegetation density by measuring the difference reflection between near-infrared (which vegetation strongly reflects) and red light (which vegetation absorbs). It is calculated as the difference between near-infrared (NIR) and red (RED) reflectance (ρ_NIR and ρ_Red) divided by their sum.

NDVI ranges from -1.0 to 1.0. Values close to one represents green vegetation dominated; values close to zero are mainly caused by soil or rocks; negative values are primarily formed from clouds, water, and snow. Multi-temporal profiles of the satellite-based NDVI has been widely used to study their seasonal phenologies (Huete et al., 2002).

Fig. 5 displays the averaged AOT, AE, CER, and NDVI collected from the Terra/MODIS during 2000-2019. It highlights the heavy pollution over India and East Asia, biomass burning smokes over Southeast Asia, and large dust aerosols over desert regions. East Asia exhibits a high aerosol level, while a relatively low CER compared to other regions. These results prove that the interaction between aerosols and clouds has strong regional variations.

Fig. 5. 
Mean AOT, AE, CER, and NDVI during 2000-2019 from Terra/MODIS.

Given that aerosol can have a serious impact on regional-scale climate, a correlation analysis between aerosol and cloud, aerosol, and vegetation distribution is performed. In this analysis, pixel values with the same latitude and longitude in every year are used to derive a linear regression between the two selected variables. Such analysis can reveal the temporal association between those variables at a numerical level. Fig. 6 shows the results of correlation analysis between aerosol, cloud, and vegetation indices. Compared to Fig. 5, it is found that in light aerosol-loaded regions, the AOT has positive relationships with both CER and NDVI (Fig. 6), revealing a positive relationship between CER and NDVI as well; in heavy aerosol-loaded regions, the AOT exhibit negative relationships with both CER and NDVI (Fig. 6) and thus CER is uncovered to retain the positive relationship with NDVI. The above findings imply that an increase of AOT can lead to an increase of vegetation under the light-aerosol-loaded situation, but a decrease of vegetation under the heavy-aerosol-loaded situation. Whereas, an increase of clouds can constantly enhance vegetation density under both situations.

Fig. 6. 
Linear regression coefficients derived from the monthly (upper) AOT with CER and (lower) AOT with NDVI.

Due to the various air pollution events including sand dust storms occurring in dry areas, forest fires and biomass burning activities, and air pollutants emitted from densely populated/industrialized areas, an effective atmospheric environmental monitoring is required. Especially, air pollution in these East Asian regions has been reported as it had been rapidly increased until the early 2000s (Akimoto, 2003). The satellite observation provides visualized information on these atmospheric environmental phenomena by retrieval of the physical parameters, and provides information on the emission sources and the transport pathways of pollutants. By using past or currently operating earth observing satellites, information on aerosols and clouds as well as other environmental parameters can be obtained, which is useful for monitoring the atmospheric environment. Those earth observing satellites are expected to provide an opportunity to solve various environmental issues on earth.

Moreover, earth observing satellites are a collection of cutting-edge complex sciences and technologies of system planning, design, manufacturing, operation, and utilization. Unique environmental products have been produced from the various earth observing satellites, so users can find the data suitable for their purposes. Especially, recent geostationary satellites launched in Asia, Europe, and the United States are used to build a nationwide environmental monitoring network, producing scientific research results. Moreover, if research activities on design and data use for integrated environmental observation are actively carried out, it is expected that the ripple effect will be great for the research fields of local air pollution and global-scale climate change. In addition, it is expected that if technology research on disaster prediction and pollution forecasting is conducted, it will not only improve air pollution in Northeast Asia, but also encourage economic cooperation.