1.

Introduction

Nearly three million premature deaths are attributed to air pollution each year;¹ $\sim 90\%$ of the world’s population is in constant danger of exposure to high levels of particulate matter (PM) with an aerodynamic $\text{diameter}\le 2.5\mu \mathrm{m}$ (PM2.5).² Due to various factors including ongoing industrialization and urbanization with significant changes in climate and geography, especially over East Asia, South Korea had the worst air quality among the Organization for Economic Co-operation and Development (OECD) nations in 2013. Averaged annual exposure to PM2.5 over 3 years (2010 to 2012) was $29.09\mu \mathrm{g}/{\mathrm{m}}^3$, significantly greater than the OECD mean of $14.05\mu \mathrm{g}/{\mathrm{m}}^3$.³ In addition, PM10 levels (PM with a $\text{diameter}\le 10\mu \mathrm{m}$), which had been consistently improving since the 1990s, have deteriorated since 2013, while high concentrations of fine particles are becoming more common.

To address this deteriorating air quality, the South Korean government established the “Special Act on the Improvement of Air Quality in Seoul Metropolitan Area” and invested approximately $3 billion during phase 1 (2005 to 2014).⁴ Policies regulating air pollutant emissions have resulted in significant reductions in sulfur oxides, nitrogen oxides, volatile organic compounds (VOCs), and PM10 (Table 1). However, the concentrations of nitrogen oxides and PM2.5 remain high when compared with other developed nations.^5,6 Phase 2 is currently underway with a total budget projected to be approximately $4 billion over the next 10 years.

Table 1

Air quality conditions before and after phase 1 of the “Special Act on the Improvement of Air Quality in Seoul Metropolitan Area,” and phase 2 goals.

For air quality improvement policies to succeed, the relationships between sources and quantities of air pollutants in the atmosphere need to be more accurately understood. Moreover, optimum reduction quantities and efficient reduction scenarios require determination. Therefore, it is critical to distinguish between the contributions of domestic sources and foreign influxes to direct air quality improvement policies. Additionally, depending on climatic conditions, air pollutants emitted from surrounding nations directly impact the South Korean atmosphere with a lag time of 1 to 2 days.⁷ For example, recent increases in air pollution over South Korea have been closely attributed to air pollutant emissions from China.^8–11 Therefore, multilateral and international efforts have been proposed to determine the effects of long-range air pollutant transport and thus to improve air quality in East Asia. These efforts include the launching of joint research groups between South Korea and China, the operation of long-range transboundary air pollutant (LTP) projects among South Korea, Japan, and China,^12,13 and discussions at the Tripartite Environment Ministers Meeting and Korea–United States Air Quality Campaign.^14,15

The South Korean government is in the process of expanding its ground observation network used for monitoring air quality. It has also been performing episodic aerial monitoring to efficiently establish and enforce air quality policies, and to acquire fundamental scientific data. However, precise assessments are still limited due to deficient data from ground-based observations that do not sufficiently cover contiguous large target areas of interest. The development of advanced remote sensing technologies as well as the consistent monitoring of trace gases in the atmosphere directly contributes to overcoming these limitations, thereby improving our ability to analyze long-range transported air pollutants.¹⁶ Currently operational low Earth orbit (LEO) environmental satellites^16–21 are capable of monitoring air quality on a global scale; however, they are temporally limited to observing the same location only every 1 to 3 days. To consistently monitor deteriorating air quality over East Asia, the National Institute of Environmental Research (NIER), Republic of Korea, is planning to launch a satellite-based environmental observation instrument, the Geostationary Environment Monitoring Spectrometer (GEMS), in late 2019. GEMS is expected to perform quantitative analyses on air pollution phenomena in the East Asian region.^22,23

2.

GEMS Mission Overview

GEMS development was initiated in 2012 along with an advanced ocean color imager, the Geostationary Ocean Color Imager-II (GOCI-II).²⁴ Both sensors will be launched onboard GeoKOMPSAT-2B (GK-2B) as part of the GK-2 mission, which consists of two satellites (GK-2A/B):²⁵ the 2A satellite, which carries the Advanced Meteorological Imager (AMI) for weather surveys, and the 2B satellite, which carries GOCI-II and GEMS for ocean and atmospheric environmental monitoring.

The AMI, GOCI-II, and GEMS sensors each have unique spectral and spatial coverage with resolutions optimized for their respective missions. In addition to the major products obtained from each payload, additional products are expected to be utilized by integrating the GK-2A/B observation data based on their different spectral and spatial resolutions. Research to develop potentially new output categories, and to improve the accuracy of existing outputs via the integration of payload data, is currently underway.

GEMS is scheduled to be launched after October 2019. Along with the Tropospheric Emissions: Monitoring Pollution (TEMPO) of the United States’ National Aeronautics and Space Administration (NASA), and Sentinel-4 of the European Space Agency (ESA), GEMS is expected be an essential part of the GEO air quality constellation for monitoring global air quality.^22,23

2.1.

Spatial and Spectral Coverage

The region of interest for GEMS covers East Asia, corresponding to longitudes 75° E–145° E and latitudes of 5° S–45° N. Its sampling spatial resolution is $3.5\times 8\mathrm{km}$ (north–south and east–west, respectively) per pixel at Seoul, with spatial and temporal co-adding depending on product category.

GEMS provides trace gas concentrations through a spectroscopic algorithm that calculates concentrations based on the gas absorption cross section with regards to solar irradiance (Fig. 1). The spectrometer observes ultraviolet–visible wavelengths in the 300- to 500-nm range with a sampling spectral resolution of 0.2 nm, and a full-width at half-maximum (FWHM) of 0.6 nm, which is the sum of all three spectral sampling data sources; thus its operation is limited to the daytime.

Fig. 1

Optical thickness of atmospheric species of interest as a function of wavelength. Adapted from Ref. 19.

2.2.

Observation Modes

GEMS and GOCI-II will operate alternatively every 30 min to capture hourly images; GEMS covers a wider area, from east of Japan to Indonesia (Fig. 2). Its operational scenario includes a concentrated observation mode that allows enhancement of its signal-to-noise ratio (SNR) by reducing the east–west scan range during the mornings and evenings when solar radiation is weak. GEMS can utilize three observational modes: normal observation, East Asia observation, and local area observation. These observational modes can be utilized to closely monitor corresponding areas during instances of high air pollution (Table 2).

Fig. 2

Region of requirement (yellow box), GEMS field of regard (FOR) (red box), and GEMS FOV (green box) for the GEMS. The background satellite imagery data are obtained from the Blue MarbleNext Generation. (Image credit: Reto Stckli, NASA Earth Observatory.)

Table 2

Main features of the GEMS.

Operational observation modes		Observation cycle (min)	East–west scanning range (longitude based on Seoul)	Notes
General observation		60	75°E–145°E
Special observation	East Asia (EA) mode	60	110°E–140°E	Monitoring of East Asian region
	Enhanced East Asia (EEA) mode	60	115°E–130°E	Enhanced monitoring of East Asian region
				Operational mode for winter
	Local area (LA) mode	30	Based on ground command	For emergency situations on the Korean Peninsula

3.

Instrument Design and Performance

3.1.

Overall Design and Performance

In normal mode, GEMS observes its field of view (FOV; Fig. 2) from east to west by moving its scan mirror position a total of 701 times. At a given scan position, 37 images over a period of 63.5 ms are coadded onboard to enhance its SNR; coadding is possible up to 63 times (Fig. 3). However, GEMS is limited to only 30 min of observation per hour, and its allowable FOV decreases for concentrated observations to enhance SNR. Earth radiance values are given as

Eq. (1)

$$P{2}_{i,j}(DN)=\sum _{k=1}^{k=C\mathrm{O}N}{\{P{1}_{i,j}(DN)\}}_k,$$

where $i$ is the spectral position and $j$ is the ground position (north-south direction); $P{1}_{i,j}(DN)$ is the focal plane subsystem output with $14\text{bits}/\text{pixel}$; $P{2}_{i,j}(DN)$ is the Earth observation radiance value; $C\mathrm{O}N$ is the number of Earth observations (generally 37, maximum 63); and $DN$ is the mirror position (generally 1 to 701).

Fig. 3

Basic concept of GEMS operations.

The design of GEMS, including the light path, uses optical paths equal to those used in observing the Earth and Sun for spectral calibration (Fig. 4). The grating and reflector spectroscopic parts were designed in the form of an Offner relay to minimize distortion. One scan spatially covers 2048 pixels in a north–south direction (one pixel corresponds to 3.5 km at Seoul) and spectrally covers 1000 pixels (spectral resolution of 0.2 nm); scanned data are sent to ground control after each observation. Spatial coaddings are performed based on product category to enhance the SNR at the ground station, the Environmental Satellite Center (ESC), which eventually improves the accuracy of the level 2 data.

Fig. 4

Optomechanical subsystem of the GEMS.

3.2.

Spectral Performance

The GEMS bandpass function has been measured at selected wavelengths (Table 3). Using a laser with an accuracy of 0.006 nm, 10 images were captured at each wavelength with a sampling interval of 0.06 nm over the scan range of $\pm 1.8\mathrm{nm}$. The collected data were used to determine bandpass function with the wavelengths for the maximum bandpass as well as for the FWHM. The point spread function (PSF) was calculated at four selected wavelengths (300, 370, 430, and 500 nm) and three spatial positions. The PSF was characterized at the spectroscopic level.

Table 3

Spectral scale and bandpass test matrix.

Spectral scale center wavelength	Bandpass wavelength scan range (from center)	Bandpass wavelength sampling interval (step size)	Number of samples per wavelength
301.8	$\pm 1.8\mathrm{nm}$	0.06 nm	10 images
330.0
365.0
400.0
435.0
470.0
498.2

Signals at both extremes of the bandpass function are hidden by noise, making them difficult to measure. Thus, broadband characterization was performed with multiple laser sources and long pass filters. These data can also be utilized in the assessment of stray light performance with the PSF. Stray light can be also from out-of-band (i.e., wavelengths outside of 300 to 500 nm) and out-of-field, which is spatially affected in the north–south direction. To measure stray light, data from the middle range are collected using a proper laser source intensity (or exposure length) level to ensure that signals at middle wavelengths (370 and 430 nm) are not saturated. Even if the signal in the middle range becomes saturated, the laser source intensity can be set so that data from the two extremes are greater than the noise level and can be subsequently recorded with high SNRs. Similarly, ranges at the extremes can also be analyzed with high SNRs by using a stronger laser source. As shown in Fig. 5, a clean PSF without the influence of noise can be obtained by standardizing collected data to their corresponding laser source intensities.

Fig. 5

PSF results using the “stitching method”: (a) before PSF and (b) after PSF.

As opposed to methods involving adjustment of light source intensity, GEMS collects data by adjusting the laser exposure time. This method ensures uniformity of the laser source with respect to wavelength and makes standardization easier. In addition, adjusting the laser exposure time is more accurate than adjusting the intensity of the laser source.

3.3.

Calibration

GEMS covers a wavelength range of 200 nm (between 300 and 500 nm) with a spectral resolution of 0.6 nm; the accuracy of the final output is heavily affected by the slightest perturbations in the observations. Hence, GEMS is expected to validate and calibrate its observational data through methods such as solar observation and light-emitting diode (LED)-assisted linearity analysis. Representative on-orbit calibration activities include solar observations using diffusers of GEMS, dark imaging with the calibration wheel in closed position, utilization of internal LEDs, and the validation of linearity of the results (Table 4).

Table 4

Radiometric calibration activities.

Radiometric calibration activity	Frequency	Time
Dark imaging	Two times/day	Before and after daytime recording
Working solar diffuser observation	One time/day	Before midnight; solar elevation angle at 30 deg
Internal LED observation	One time/week	Between dark imaging and solar diffuser observation
Reference solar diffuser observation	Two times/year	Before midnight; solar elevation angle at 30 deg

3.3.1.

Working/reference diffuser calibration

GEMS monitors the Sun once a day to validate and calibrate any self-caused changes in the Earth’s radiance data. For this reason, GEMS is equipped with a calibration mechanism assembly (CMA), which has an open position for Earth observations, two diffuser positions, and a closed position to prevent incoming debris. GEMS also carries a working diffuser as well as a reference diffuser to correct the degradation of the working diffuser every 6 months. The diffuser is used for solar calibrations before midnight when the incident angle of sunlight on GEMS is $\sim 30\mathrm{deg}$. The diffuser transmission is affected by the solar incidence angle; thus, the solar calibration time during the year is determined at a constant incident angle (Fig. 6).

Fig. 6

Solar calibration geometry over a yearly cycle.

Observed solar irradiance [$F_{\mathrm{meas}}(i,j)$] is used to calibrate the Earth radiance observations from GEMS compared to existing reference values. $F_{\mathrm{meas}}(i,j)$ is expressed as

Eq. (2)

$$F_{\mathrm{meas}}(i,j)=\frac{k_{i,j}^r·C_{\text{photon}}^i(i,j)}{{\mathrm{BTDF}}_{\text{Diffuser}}(G_{i,j})·N_{\mathrm{Coadd}}·t_{\mathrm{int}}},$$

where $k_{i,j}^r$ is the pixel radiance calibration coefficient at column $i$, row $j$; $C_{\text{photon}}^i(i,j)$ is the measured signal in irradiance mode at column $i$, row $j$; ${\mathrm{BTDF}}_{\text{Diffuser}}(G_{i,j})$ is the bidirectional transmission distribution function (BTDF) of the on-board diffuser, which is a function of the goniometric angle; $N_{\mathrm{Coadd}}$ is the number of image coadditions; and $t_{\mathrm{int}}$ is the single frame integration time.

3.3.2.

Dark imaging and on-board LED source linearity calibration

Dark imaging is performed to analyze GEMS noise signals and to offset values when the CMA is in the closed position. The imaging is conducted twice daily (before and after Earth observations) and is used for the current dark correction during the observation. Dark calibration is performed by combining the frames of dark imaging and normal observation, which have the same integration times and numbers of images.

In addition, GEMS utilizes its internal LEDs once a week to check the homogeneity of CCDs and the linearity of the image. It collects LED images at 20 different times and interpolation is used in between. Each data element consists of a combination of five frames. The calibration wheel moves to the closed position during LED imaging.

4.

Main Products

GEMS is expected to provide a total of 16 level 2 (L2) products (Table 5), including aerosol optical depth (AOD), ozone (${\mathrm{O}}_3$), and trace gases in the atmosphere such as sulfur dioxide (${\mathrm{SO}}_2$), nitrogen dioxide (${\mathrm{NO}}_2$), glyoxal (CHOCHO), and formaldehyde (HCHO).

Table 5

Main L2 outputs of the GEMS unit.

Some GEMS products are provided as vertical column densities (VCD) by retrieving slant column densities and applying the air mass factor (AMF). The acquired data are expected to be used as feedback for the detection of clear sky pixels and calculation of surface reflectance to further contribute to the improvement of accuracy. The accuracy of the products retrieved is needed to be validated after the satellite launched.

5.

GEMS Ground Station

For the reception, processing, and application of GEMS data, the Korean Ministry of Environment constructed the ESC at the NIER and established a data processing system. The NIER is currently conducting an integrated air quality forecast;⁴² a significant improvement in the forecast accuracy is expected with the utilization of GEMS data.

GEMS data are simultaneously received at the Korea Aerospace Research Institute, which controls the satellite bus system, and the Korea Ocean Satellite Center, which is the ground control station for the GOCI-II and holds a mutual data backup relationship with its counterparts (Fig. 7). The raw data observed from GEMS and received via the reception system at the ESC produce the L1B data through the decomposition module, radiation calibration, and geometric calibration. The L2 data are provided in the form of outputs such as AOD (Fig. 8 and Table 6). To improve the accuracy of the main L2 outputs at the ESC, forecast data from the ocean color satellite GOCI-II and the atmospheric chemistry modeling data are collected. The production of external-data driven value-added products, including mean field and surface concentrations, is in preparation.

Fig. 7

Systematic diagram of the GK-2 ground control process.

Fig. 8

Structural diagram of the GEMS data processing system.

Table 6

Abbreviations of GEMS data processing systems.

6.

Application and Expected Impacts of GEMS

GEMS is expected to provide the hourly distribution of the product variables and simultaneously analyze their emission, transport, and sinks. Unlike those of LEO environmental satellites, the spatial and temporal distribution characteristics of GEMS enable continuous monitoring of ${\mathrm{O}}_3$, ${\mathrm{SO}}_2$, ${\mathrm{NO}}_2$, aerosols, and HCHO from a single location; it thus has a wide range of expected applications, including top-down pollutant emissions and long-range transport. In particular, it is anticipated that emission calculations and monitoring will be possible through the spatial and temporal observations of tropospheric ${\mathrm{O}}_3$, PMs, VOCs, and precursors,⁴³ and that these will be utilized in the quantitative analysis and transport pathway determination of long-range air pollutants. Furthermore, methods to calculate the surface concentrations of PM could be developed using the satellite-observed AODs,^44–47 which would be useful in the monitoring of PM distributions, and analyses of their trends and high PM concentration events.

In the field of climate change, GEMS is expected to contribute toward understanding active climate change responses, such as quantitative analyses related to climate change through the consistent monitoring of surface and atmospheric East Asian SLCP (e.g., ${\mathrm{O}}_3$, aerosols) concentrations. It is also likely to provide solutions to reduce SLCPs, which, unlike long-lived ${\mathrm{CO}}_2$, represent short-term reduction effects.

Furthermore, GEMS is anticipated to be used in the detection of sulfuric gases, as an indicator of volcanic eruptions, for the confirmation, prediction, and analysis of air pollutants (e.g., toxic substances), for understanding transport paths from chemical accidents, and the quantitative estimation of air pollutants from large wildfires.

In particular, the use of GEMS as input data in air quality models is expected to significantly improve forecast accuracies through data assimilation and the amelioration of initial and boundary condition uncertainties.^48,49 GEMS may also bring about quality improvement through the evaluation of air quality prediction models, accuracy enhancement of air pollutant emissions through bottom-up estimations, and through accuracy enhancement and improvement of air quality prediction intervals through assimilation of satellite data and forecast models.^49–51 It is also likely to improve the accuracy of air pollution forecasts and warnings through earlier estimation of high air pollutant concentrations. These examples of forecast and warning accuracy improvements will enhance public trust and increase air pollution avoidance rates; thus, improving the health and quality of life of citizens.

Lastly, along with TEMPO²² and Sentinel-4,²³ GEMS is expected to ensure broad and consistent monitoring of air quality in East Asia, North America, and Europe’s concentrated areas. In conjunction with scientific data for air pollutant emissions, transport, and degradation at the global scale, international cooperation among state governments and efficient air quality policy implementation will be achievable. It should be noted that globally spreading air pollution is impossible to mitigate with the efforts of only a few nations. Therefore, the three major geostationary environmental satellites, including GEMS, are anticipated to provide solutions for international cooperation and ensure efficient and effective air quality control.

7.

Conclusions

The South Korean government is developing the first geostationary environment spectrometer (GEMS) and planning its launch in 2019 to consistently monitor deteriorating air quality and establish and enforce efficient and cost-effective air quality policies. GEMS is expected to provide fundamental data on trace gases including ${\mathrm{NO}}_2$, ${\mathrm{SO}}_2$, HCHO, ${\mathrm{O}}_3$, and AOD. In addition, its output can be utilized in air quality prediction models and contribute toward increasing the accuracy of forecasts and warnings that are needed to prevent air pollution, thereby enhancing quality of life. Moreover, GEMS will potentially be able to contribute to international efforts and assist in the development of a clean atmosphere.