Preflight Evaluation of the Environmental Trace Gases Monitoring Instrument with Nadir and Limb Modes (EMI-NL) Based on Measurements of Standard NO₂ Sample Gas

Abstract

Hyperspectral observations are used to retrieve high-resolution horizontal distribution and vertical profiles of trace gases (O₃, NO₂, HCHO, and SO₂), thereby playing a vital role in monitoring the spatio-temporal distribution and transportation of atmospheric pollutants. These observations reflect air quality changes on global and regional scales, including China, thereby elucidating the impacts of anthropogenic and natural emissions on atmospheric composition and global climate change. The DaQi 02 (DQ02) satellite carries the Environmental Trace Gases Monitoring Instrument with Nadir and Limb modes (EMI-NL) onboard, which will simultaneously perform nadir and limb measurements of high-resolution ultraviolet and visible solar scattered light in the nadir and limb directions. Combined with the absorption of different trace gases in this wavelength band, this information can provide high-resolution horizontal and vertical distributions of trace gases. We examined the spectral measuring ability and instrument characteristics of both modules of EMI-NL by measuring different light sources and concentrations of the NO₂ sample gas. In the nadir module test, when the NO₂ sample gas concentration was 198 ppm and 513 ppm with scattered sunlight as the light source, the average relative errors of spatial pixels were 4.02% and 3.64%, respectively. At the NO₂ sample gas concentration of 198 ppm with the integrating sphere as the light source, the average relative error of spatial pixels was −2.26%. In the limb module test, when the NO₂ sample gas concentration was 198 ppm and 1000 ppm with the tungsten halogen lamp as the light source, the average relative errors of spatial pixels were −3.07% and 8.32%, respectively. When the NO₂ sample gas concentration was 198 ppm and 1000 ppm with the integrating sphere as the light source, the spatial pixel average errors were −3.5% and 8.06%, respectively. The retrieved NO₂ slant column density between different spatial pixels exhibited notable inconsistency in both modules, which could be used to estimate the stripe of spatial dimension. These results confirm the ability of EMI-NL to provide accurate spaceborne monitoring of NO₂ globally.

1. Introduction

Spaceborne hyperspectral remote sensing technology is widely used for monitoring global atmospheric pollution by measuring trace gases. Thus far, several key spaceborne instruments are used to monitor trace gases, including: the Total Ozone Mapping Spectrometer (TOMS), launched in 1978 onboard NIMBUS G [1]; the Global Ozone Monitoring Experiment (GOME), launched onboard the European Space Agency (ESA) ERS-2 satellite in 1995 [2]; the Scanning Imaging Absorption Spectrometer for Atmosphere Chartography (SCIAMACHY), launched in 2002 onboard the ESA Envisat satellite, which can achieve nadir, limb, and occultation observation function [3]; the Ozone Monitoring Instrument (OMI), launched onboard the National Aeronautics and Space Administration (NASA) Earth Observing System (EOS) Aura satellite in 2004 [4]; GOME-2 [5]; and the Tropospheric Monitoring Instrument (TROPOMI), launched onboard the Sentinel-5 Precursor satellite in 2017 [6]. All these spaceborne instruments mainly reveal information on atmospheric transportation, atmospheric chemistry, and air pollutant emissions [7,8,9,10,11,12,13,14,15] through hyperspectral measurements.

In recent years, China has initiated the active development and application of hyperspectral observation technologies. Of such initiatives, the DaQi-02 satellite (DQ02) stands out as a new operational spaceborne system as part of a series of atmospheric observation satellites in China. Conceptually, DQ02 is a comprehensive detection satellite, aimed at meeting environmental management needs, namely, acquiring information about global climate change and atmospheric composition. From a technical standpoint, DQ02 is equipped with five types of remote sensing instruments: atmospheric detection lidar, a wide-range hyperspectral greenhouse gas monitor, infrared hyperspectral atmospheric composition detection, an ultraviolet hyperspectral atmospheric composition detector, and a cloud and aerosol imager. By exploiting these instruments, DQ02 will achieve large-scale, continuous, dynamic, and all-day comprehensive monitoring of atmospheric elements, such as greenhouse gases, gaseous pollutants, clouds, and aerosols by combining active and passive remote sensing techniques. Quality-assured observations from DQ02 will strengthen the technical capabilities of spaceborne monitoring in China by providing timely quantitative information on the ecological environment, natural resources, and agriculture.

The Environmental Trace Gases Monitoring Instrument with Nadir and Limb modes (EMI-NL) is one of the five payloads carried by DQ02, equipped with independent nadir and limb observations, mainly used to monitor the horizontal and vertical distribution of atmospheric trace gases. An instrument of the same series but only with a nadir observation mode, EMI [16], was launched in May 2018 and has been thereafter applied in numerous studies focused on trace gas and cloud parameter retrievals [17,18,19,20,21,22,23]. In particular, EMI02 [24] was launched in September 2021 with an equator-crossing time of 10:30 local time, and the DaQi-01 satellite (EMI03) was launched in April 2022. As mentioned, the EMI-NL onboard DQ02 should be launched in the following years, and the series of payloads will provide continuous observations, thereby reflecting horizontal and vertical distribution information of global atmospheric trace gases such as NO₂, SO₂, and O₃.

Most previous studies on the trace gas monitoring capability of spaceborne instruments before launch have normally relied on a xenon lamp. A xenon lamp allows the quantification of a trace gas in a gas cell at different temperatures and pressures [25,26,27,28]. For instance, the measurements of NO₂ and O₃ in a gas cell were carried out for OMI using a xenon lamp and zenith-scattered sunlight [29]. In this context, the spectral measurement capability of EMI can be evaluated by quantifying NO₂ in a sample gas cell using scattered sunlight in laboratory conditions [30]. Furthermore, the NO₂ slant column density (SCD) can be retrieved by the DOAS [31] algorithm.

To evaluate the trace gas measuring ability of EMI-NL, we measured the absorption spectrum of standard NO₂ sample gas with different concentrations in the gas cell for nadir and limb modules using different light sources in laboratory conditions. Specifically, the measured spectrum retrieved by the DOAS algorithm was utilized in our study to (a) evaluate the trace gas monitoring ability of spatial pixels of the nadir and limb modules of EMI-NL, and (b) investigate the cross-track stripe phenomenon of the EMI-NL spatial dimension.

2. Data and Methods

2.1. EMI-NL Description

EMI-NL is used to obtain high-resolution ultraviolet and visible scattered sunlight in the nadir and limb directions. The high-resolution trace gas horizontal and vertical distribution information is acquired using the absorption of different trace gases at this band. In this way, the quantitative monitoring of global atmospheric trace gas distribution and transportation is realized. The in-orbit observation schematic diagram of EMI-NL is illustrated in Figure 1, while the main parameters of EMI-NL are summarized in

Table 1. EMI-NL instrumental properties.

	Nadir Module	Limb Module
Spectral channels	UV1: 300–400 nm, VIS1: 400–500 nm	UV1: 290–380 nm, UV2: 380–480 nm, VIS1:520–610 nm
Spectral resolution	≤0.6 nm
Telescope FOV	114° (cross-track)	≥4.5° (horizontal direction)
Spatial resolution	≤7 km (swath direction) × 7 km (flight direction)	≤2 km (tangent direction)
CCD detectors	UV: 1022 × 954 (spectral × spatial) pixels VIS: 1022 × 954 (spectral × spatial) pixels	UV: 1024 × 1024 (spectral × spatial) pixels VIS: 1024 × 1024 (spectral × spatial) pixels
Mass	90 ± 0.9 kg
orbit	Polar, sun-synchronous, ascending node equator crossing time: 13:30

.

2.1.1. Nadir Module

The EMI-NL nadir module measures at the wavelength range of 300–500 nm, including two spectral channels of ultraviolet 1 and visible 1 (UV1: 300–400 nm, VIS1: 400–500 nm), using a Littrow–Offner convex grating imaging spectrometer, with the spectral resolution of 0.6 nm. Compared with the exclusive nadir geometries of EMI [1] and EMI-2 [18], the spatial resolution of the EMI-NL nadir module is substantially improved by providing 7 km (swath direction) × 7 km (flight direction) measurements. Moreover, its nadir module is characterized by a wide instantaneous field of view (IFOV) of 114° with a ground coverage of 2600 km in cross-orbit direction, thereby offering nearly a daily global coverage by observations.

The optical system block diagram of the EMI-NL nadir module is illustrated in Figure 2a. As seen, the light scattered and reflected from the Earth’s atmosphere, or a surface is collected by the front telescope of the system and subsequently enters the relay optical system. The relay optical system splits the light using the color separation filter. Consequently, the light from each channel is reflected and converged into the corresponding Littrow–Offner imaging spectrometer. The imaging spectrometer has been designed using the Littrow–Offner structure, which facilitates the miniaturization adaptable for space technology. Finally, the dispersion in the spectrometer is imaged on the area-array CCD detector to obtain high spectral resolution and high spatial resolution spectral information.

2.1.2. Limb Module

The wavelength range of the limb module of EMI-NL is 210–610 nm, including three channels: UV1 (290–380 nm), UV2 (380–480 nm), and VIS1 (520–610 nm) with a spectral resolution of 0.6 nm, and the spatial resolution in the tangent direction is 2 km. Notably, the limb observation method is a new observation approach in atmospheric remote sensing. In this method, the sunlight reflected by atmospheric molecules, aerosols, clouds, and the Earth’s surface forms a 100 km thick limbic atmosphere around the Earth’s periphery. The limb module measures the radiance of the limbic atmosphere to ultimately retrieve the high-altitude vertical distribution of trace gases and atmospheric aerosols. The retrieval is based on analyzing the spectral radiation characteristics of the limbic atmosphere.

Figure 2b displays the optical block diagram of the EMI-NL limb module. The scanning mirror guides the light into the front optical system, the off-axis three-mirror telescope performs the telephoto imaging, and the color separation filter splits light, thereby ultimately forming three independent spectral detection channels. The limb spectral information is gathered into the slit of each channel’s imaging spectrometer. At the final stage, the condensed light of the front telephoto imaging system enters their channel’s imaging spectrometer, disperses, and images to the area array CCD detector, thereby retrieving spectral imaging information in this way.

2.2. Experimental Design

The measurement of standard NO₂ sample gas in a cell was carried out for EMI-NL under the standard atmospheric pressure and temperature of 20 °C. The measurement was performed in the laboratory at the experimental observation field of the Anhui Institute of Optics and Fine Mechanics (located at 117.18°E, 31.9°N). Due to this fact, four sets of measurements were performed in the experiment: (1) the EMI-NL nadir module was evaluated with the scattered sunlight and the NO₂ sample gas concentrations with the mixing ratios of 198 and 513 parts per million (ppm) respectively; (2) the EMI-NL nadir module measurements were performed with the integrating sphere as the light source with the NO₂ sample gas concentration of 198 ppm; (3) the EMI-NL limb module was measured using the tungsten halogen lamp with the NO₂ sample gas concentration of 198 and 1000 ppm, respectively; and (4) while the EMI-NL limb module was measured using integrating sphere with the NO₂ sample gas concentration of 198 and 1000 ppm, respectively. The NO₂ sample gas concentration of 198 ppm is close to the high concentration of NO₂ SCD in the actual atmosphere; higher concentrations (513 ppm and 1000 ppm) of the NO₂ sample gas are good for DOAS fitting and accurate evaluation of EMI-NL. The schematic of the experimental setup is illustrated in Figure 3. As seen, the scattered sunlight entered the laboratory from the quartz glass window, passed through the sample gas cell (the sample gas cell length = 8 cm), and was ultimately measured by EMI-NL. The integrating sphere light source directly passed through the sample gas cell and was also ultimately measured by EMI-NL. During the experiment, the N₂ sample gas was first flushed for a few minutes to remove other gases from the sample gas cell and was subsequently filled with the NO₂ sample gas at a 6 L/min rate. Note that the entire field of view of EMI-NL nadir and limb modules was tested in this experiment.

2.3. DOAS Fitting

The spectrum measured by flushing the gas cell with the N₂ sample gas was utilized as the reference spectrum, whereas the NO₂ SCD was retrieved by the DOAS algorithm using the previously measured NO₂ sample gas spectrum. In this study, the fitting interval of the visible band 430–470 nm was applied. All these relevant parameter settings are summarized in

Table 2. NO₂ DOAS fit settings.

Parameter	Data Source
NO2	NO2 at 298 K [32]
Polynomial	5th
Fitting interval	430–470 nm

. Zhang et al., (2018) [30] have previously outlined that by adding a ring cross section, one only negligibly affects DOAS spectral fitting results for laboratory sample gas measurement. On this basis, we did not add the ring cross section here. The DOAS algorithm can be formalized by Equation (1).

$$\ln \left[\frac{I_{{\mathrm{N}}_2}\left(\lambda \right)}{I_{{\mathrm{NO}}_2}\left(\lambda \right)}\right]={\sigma }_{{\mathrm{NO}}_2}\left(\lambda \right)S_{{\mathrm{NO}}_2}+P\left(\lambda \right)$$

(1)

I_N2(λ) and I_NO2(λ) are the measured spectral when the gas cell is filled with N₂ and NO₂ sample gas, respectively, σ_NO2(λ) is the NO₂ absorption cross section, S_NO2 is the NO₂ SCD, and P(λ) is the polynomial. Figure 4 shows an example of DOAS fitting in the nadir module of the 102nd pixel when the NO₂ sample gas concentration was 198 ppm and scattered sunlight was used.

2.4. Signal-to-Noise Ratio Estimation

The signal-to-noise ratio (SNR) is a key indicator reflecting the ability of imaging spectroscopy to acquire effective target information. In particular, the magnitude of SNR reflects the detection limit monitoring ability of EMI-NL, which directly affects the quality of level 2 product retrieval calculation. The noise in the SNR is usually characterized by the standard deviation in statistics, and the SNR is usually calculated by continuous measurement and statistics with a stable light source at a certain brightness under laboratory conditions. Thus, thorough evaluation and determination of SNR are key steps in spaceborne imaging spectrometry. This study used the measurement spectrum using tungsten halogen lamps in the laboratory and treated the average value of 100 repeated spectral measurements at a wavelength of λ as the signal, while the standard deviation was used as the noise. Lastly, the ratio of signal and noise was defined as the SNR [16] at the wavelength λ according to Equations (2)–(4) below:

$$\mathrm{SNR}=\frac{\mathrm{signal}}{\mathrm{noise}}.$$

(2)

$$\mathrm{signal}={\displaystyle \sum _{i=1}^{100}{I}_i\left(\lambda \right)}.$$

(3)

$$\mathrm{noise}=\sqrt{\frac{{\displaystyle \sum _{i=1}^{100}{\left(I_i\left(\lambda \right)-\overline{I\left(\lambda \right)}\right)}^2}}{99}}.$$

(4)

3. Results and Discussion

3.1. Results of the Nadir Module

The spatial row pixels 1–203 of the nadir module were measured in the experimental setup, as shown in Figure 3a. Note that the NO₂ SCD of the NO₂ sample gas was estimated based on the NO₂ volume mixing ratio and the length of the gas cell. At the concentrations of 198 ppm and 513 ppm, the NO₂ SCDs were found to be 3.92 × 10¹⁶ molecule/cm² and 1.01 × 10¹⁷ molecule/cm², respectively. Figure 5a demonstrates NO₂ SCD with fitting error and relative deviation of different spatial pixels when the flushed NO₂ sample gas was 198 ppm. As seen, the averaged NO₂ SCD of spatial pixels was (4.08 ± 0.14) × 10¹⁶ molecule/cm² (±0.14 × 10¹⁶ molecule/cm² is the averaged NO₂ SCD fitting error of spatial pixels), while the NO₂ SCD standard deviation of spatial pixels was 2.69 × 10¹⁵ (see

Table 3. NO₂ gas cell measurement results of the nadir module.

Reference Spectral	NO2 Gas (ppm)	Averaged NO2 SCD (molecule/cm2)	Standard Deviation of NO2 SCD	Averaged NO2 Fitting Error (molecule/cm2)	Standard Deviation of NO2 Fitting Error
Scattered sunlight	198	4.08 × 1016	2.69 × 1015	1.40 × 1015	4.67 × 1014
	513	1.05 × 1017	4.47 × 1015	1.59 × 1015	4.96 × 1014
Integrating sphere light	198	3.83 × 1016	2.27 × 1015	4.28 × 1014	7.86 × 1013

). Compared with the estimated NO₂ SCDs, the average relative error of spatial pixels is 4.02%. Figure 5b shows that NO₂ SCD when the cell flushed NO₂ sample gas was 513 ppm. The averaged NO₂ SCD of spatial pixels was (1.05 ± 0.02) × 10¹⁷ molecule/cm², and the NO₂ SCD standard deviation was 4.47 × 10¹⁵. Figure 3b shows the measurement setup using an integrating sphere. The NO₂ SCD of the spatial pixels 1–190 is displayed in Figure 6 when the sample gas cell was filled with the 198 ppm NO₂ sample gas. Note that the NO₂ SCD of other spatial pixels 191–203 was not plotted due to the large observation elevation angle resulting in the radiance of these pixels being too low. Moreover, the averaged NO₂ SCD of spatial pixels was (3.83 ± 0.04) × 10¹⁶ molecule/cm², and the standard deviation was 4.28 × 10¹⁴. Figure 6 also shows that the retrieved NO₂ SCD between different spatial pixels exhibited notable inconsistency. Note that the possible drivers behind this inconsistency have been previously elaborated in detail by Zhang et al., (2018) [30]. The inconsistency could be caused by the influence of the low transmittance of quartz glass, which may have been triggered by the incomplete correction of the instrument’s spatial pixels wavelength correction, slit function, dark current compensation, stray light correction, etc. Alternatively, the inconsistency could be driven by the stripe phenomenon of the two-dimensional CCD. In the experimental setup (1), the NO₂ sample gas concentrations were 198 and 513 ppm, while the variation trend of retrieved NO₂ SCD and relative deviation of different spatial pixels were fairly consistent, with the scattered sunlight applied. However, they were strikingly inconsistent when the integrating sphere was used as a light source. Consequently, the averaged NO₂ fitting error of integrating sphere was markedly lower than that of scattered sunlight when the NO₂ sample gas concentration was 198 ppm. This finding indicates that different spatial pixels responded differently to different light sources. In turn, this induces the emergence of different fitting errors, potentially triggered by the inconsistency of the light source.

3.2. Results of Limb Module

In the experimental setup, shown in Figure 3c, the spatial pixels 1–237 of the limb module were measured using the tungsten halogen lamp as a light source. When the NO₂ sample gas concentration was 198 ppm and 1000 ppm, the estimated NO₂ SCDs were 3.92 × 10¹⁶ molecule/cm² and 1.98 × 10¹⁷ molecule/cm², respectively. As shown in Figure 7a, when the flushed NO₂ sample gas is 198 ppm, the NO₂ SCD with fitting error and relative deviation of different spatial pixels are plotted. The average NO₂ SCD of spatial pixels was (3.64 ± 0.15) × 10¹⁶ molecule/cm², while the standard deviation was 2.48 × 10¹⁵, as seen in

Table 4. NO₂ gas cell measurement results of limb module.

Reference Spectral	NO2 Gas (ppm)	Averaged NO2 SCD (Molecule/cm2)	Standard Deviation of NO2 SCD	Averaged NO2 Fitting Error (Molecule/cm2)	Standard Deviation of NO2 Fitting Error
tungsten halogen lamp	198	3.64 × 1016	2.48 × 1015	1.48 × 1015	4.67 × 1014
	1000	2.06 × 1017	4.27 × 1015	1.65 × 1015	4.96 × 1014
Integrating sphere light	198	3.69 × 1016	8.32 × 1014	4.80 × 1014	2.75 × 1013
	1000	2.04 × 1017	1.70 × 1015	7.92 × 1014	1.10 × 1014

. Figure 7b shows the NO₂ SCD when the cell flushed NO₂ sample gas was 1000 ppm. Furthermore, the average NO₂ SCD of spatial pixels was (2.06 ± 0.02) × 10¹⁷ molecule/cm², and the NO₂ SCD standard deviation was 4.27 × 10¹⁵. The measurement using integrating sphere in the setup, shown in Figure 3d and Figure 8a demonstrate that NO₂ SCD of spatial pixels at the sample gas cell was filled with 198 ppm NO₂ sample gas. Moreover, the average NO₂ SCD of spatial pixels was (3.69 ± 0.05) × 10¹⁶ molecule/cm², and the standard deviation was 8.32 × 10¹⁴. Figure 8b shows the retrieved NO₂ SCD when the NO₂ sample gas was 1000 ppm. The average NO₂ SCD of spatial pixels was (2.04 ± 0.01) × 10¹⁷ molecules/cm², and the NO₂ SCD standard deviation was 1.7 × 10¹⁵. The same figure also indicates that the retrieved NO₂ SCD exhibited remarkable inconsistency between the different spatial pixels, while the fitting error of the integrating sphere was significantly smaller, compared to that of the tungsten halogen lamp. Moreover, at the NO₂ sample gas concentration was 198 ppm, regardless of the light source (the tungsten halogen lamp or the integrating sphere), the average NO₂ SCD of spatial pixels was underestimated by −3.07% and −3.5%, respectively. When the sample gas concentration was 1000 ppm, NO₂ SCD was overestimated by 8.32% and 8.06%, respectively. This may be caused by the characteristics of the instrument itself or the unstable control of the sample gas flushing rate. It should also be noted that such spatial pixel-dependent error can be regarded as the error evaluation and can therefore be utilized for correcting the NO₂ SCD stripe [33,34,35] in the actual in-orbit monitoring.

3.3. Results of SNR Estimation

Figure 9a–c shows the ground-based measured dark current, spectral, and SNR of the 92nd pixel of the VIS1 channel of the nadir module in the range of 430–470 nm, where the SNR fluctuated between 900 and 1400 with the integration time of 90 ms. Moreover, the nadir SNR was potentially underestimated at some wavelengths as the integration time of the measurement was somewhat short. Figure 9d–f shows the dark current, spectral, and SNR of the limb module in the 430–470 nm range of the 125th pixel of the UV2 channel. As can be seen, within the range of approximately 1000–1300 and the integration time of 0.2 s, the limb SNR meets the EMI-NL performance requirements.

4. Conclusions

This study used the measurements of different concentrations of NO₂ sample gas in a gas cell based on different light sources with EMI-NL in the laboratory to evaluate the trace gas monitoring capability and performance of nadir and limb modules of EMI-NL. The NO₂ SCD was retrieved by the DOAS algorithm using the measured spectrum and was compared with the estimated NO₂ SCD of the standard sample gas, while the NO₂ SCD cross-track stripe structure was obtained as well. By evaluating the monitoring ability of different spatial pixels of EMI-NL nadir and limb modules, one can acquire NO₂ SCD strips during spaceborne monitoring. Interestingly, the EMI-NL nadir and limb module pixels row pixels exhibited the same phenomenon in the experiment. More specifically, when the light source is the same, but the NO₂ sample gas concentration differs, the trends of retrieved NO₂ SCD and relative deviation for different spatial pixels are consistent. However, when the light source differs at the same NO₂ sample gas, the trends of retrieved NO₂ SCD and fitting error are inconsistent for different spatial pixels. This finding indicates that the response of spatial pixels depends on the choice of a light source, thereby inducing the inconsistencies in NO₂ SCDs and affecting the fitting error. Moreover, the experimental measurement of the limb module revealed an underestimation of NO₂ SCD when the NO₂ sample gas concentration was 198 ppm, while the NO₂ SCD was overestimated when the NO₂ sample gas concentration was 1000 ppm. Lastly, the measurement of the tungsten halogen lamp spectrum in the laboratory was used to quantify the SNR of the EMI-NL nadir and limb modules in the 430–470 nm band. The SNR evaluation demonstrated that the SNR of the nadir and limb modules exhibits variability within the range of approximately 900–1400 and 1000–1300, respectively. Overall, the SNR of the nadir module is potentially underestimated, and the integration time of the measurement could be adjusted to the most suitable time for future studies.

In general, EMI-NL performs well in laboratory measurements. The NO₂ SCDs relative error of most spatial pixels of the nadir and limb modules is within 10%, which can achieve accurate global monitoring of the NO₂ horizontal and vertical distribution. The results provide useful data for the in-orbit monitoring of EMI-NL and for the subsequent data retrieval algorithm. In future work, we could add more NO₂ sample gas concentration data points, especially those within the actual atmospheric NO₂ concentration range, so as to more clearly understand the relationship between the NO₂ SCD retrieved from the EMI-NL measured spectrum and that estimated from the NO₂ sample gas concentration, which can be used for the correction of underestimated and overestimated NO₂ sample gas concentrations in the EMI-NL limb module.