Background

Under the Paris Agreement, countries submit their pledges toward the temperature goal in the form of Nationally Determined Contributions (NDCs), and their progress will be evaluated every five years from 2023 (Global Stocktake, GST). The first GST suggested that our climate mitigation effort is insufficient to achieve Paris agreement goal, and thus, we need further actions [1]. Science-based approaches are necessary to track and guide our actions to the second GST in 2028. The research community has explored the ways to utilize greenhouse gas (GHG) data collected from various observation platforms (e.g., ground, aircraft, and satellites) in order to support the evaluation at GST [2]. Over the past decade, GHG remote sensing has significantly advanced, matured, and started playing a key role in collecting GHG data for science [e.g., 3–12], and for climate mitigation monitoring applications [e.g., 13–15].

In 2020, the Japan Aerospace Exploration Agency (JAXA) and ANA HOLDINGS INC. (ANAHD) jointly launched a new project named Greenhouse gas Observations of Biospheric and Local Emissions from the Upper sky (GOBLEU) (https://www.eorc.jaxa.jp/GOSAT/ANAexp/index.html). In support of subnational climate mitigation efforts, GOBLEU collects GHG and other relevant data from regular passenger aircraft and visualizes our climate mitigation effort progress. GOBLEU aims to achieve the objective by combining JAXA’s GHG remote-sensing technique and experiences built on the GOSAT mission and ANA’s world-leading air passenger carrier capacity. Just a few years back, the COVID-19 pandemic hit the airline industry hard due to the travel restrictions implemented globally [16]. The number of passenger flights in 2020 dropped by 60% from 2019 [17]. Airlines have not carried passengers at full capacity for over two years, including the Olympic and Paralympic Games 2021 in Tokyo. However, GOBLEU turns " ghost " flight operations into action for climate monitoring and mitigation by carrying a GHG monitoring instrument on board passenger flights and contributing to sustainability.

JAXA and ANAHD have designed and developed an innovative carry-on size instrument suite that can be installed on two window seats and collects data through the cabin window, just like a passenger. The GOBLEU instruments are based on the same remote-sensing observation technique developed and used by the current state-of-the-art GHG observing GOSAT (2009-on, 3) and GOSAT-2 (2018-on, 6) satellites. The instrument suite is designed to collect carbon dioxide (CO2) data as well as additional variables, such as nitrogen dioxide (NO2) levels as a tracer of CO2 from fuel combustion and solar-induced fluorescence (SIF) as an indicator of the plant production (carbon removals) of the terrestrial biosphere (e.g., forests and other vegetated areas) [18, 19]. This is expected to enhance the ability to monitor carbon emissions and removals from major key emission sectors and provide GHG information relevant to climate mitigation tracking at various decision making and climate action levels. Data to be collected by GOBLEU and the derived GHG information are complementary to common GHG information, such as emission inventories. These are typically developed and updated on a relatively low frequency (often annually) with high latency (more than a year or two). GOBLEU expects to provide timely GHG information by promptly collecting high-resolution GHG data and emission and removal estimates with greater information granularity. The high-spatial-resolution data should provide GHG information to stakeholders at different subnational levels (e.g., states/prefectures, cities, the private sectors, and citizens).

GOBLEU should also provide direct technical and scientific implications to the synergic use of remotely sensed GHG and AQ data that is planned by future space GHG observing missions, such as Japan’s Global Observing SATellite for Greenhouse gases and Water cycle (GOSAT-GW; planned launch 2024) [20] and Europe’s Copernicus Carbon Dioxide Monitoring mission (CO2M; planned launch 2026) [21]. As recent studies [22,23,24,25,26] demonstrated, simultaneously collecting CO2 and NO2 data should enhance our ability to quantify anthropogenic GHG emissions. However, it is important to note that the previous studies have been based on data from different satellite platforms based on certain spatial and temporal colocation criteria or a campaign flight.

The first GOBLEU proof of concept flight occurred on the 26th of October 2020 during the global COVID pandemic. This manuscript describes the instrument, the first flight, and the results. We also discuss the current challenges and limitations as well as our plans.

Methods

GOBLEU monitoring instrument suite

Figure 1 shows the GOBLEU monitoring instrument suite. Before the GOBLEU project launched, we evaluated the transmissivity of the cabin window using ASD FieldSpec4 (a wide-spectral range spectrometer). We confirmed that most cabin windows on ANA-operated aircraft transmit the reflected solar light spectra between 400 nm and 1650 nm sufficiently for the measurement, except for windows with an electronic shade. Given the spectral range, GOBLEU has chosen CO2 as the main target GHG. The instrument suite consists of three in-house customized imaging (grating) spectrometers coupled with 53 arrays of bundled optical fibers: Fiber-coupled NO2 imaging spectrometer (FNO2), Fiber-coupled SIF imaging spectrometer (FSIF), and Fiber-coupled Greenhouse gas imaging spectrometer (FGHG). Details of the spectrometers are summarized in Table 1. A common fore-optics relay is the solar light reflected on the Earth’s surface through the bundled optical fibers. Three instruments cover the spectral ranges in 420–490 nm for NO2, 670–780 nm for SIF, and 1560–1640 nm for CO2, with spectral sampling intervals of 0.03 nm, 0.05 nm, and 0.17 nm, respectively. These spectrometers are coupled with a 2-dimentional (2D)-CMOS camera (2048 × 2048) for NO2 and SIF and an 2D-InGaAs camera (640 × 512) for CO2 manufactured by Hamamatsu Photonics K.K. GOBLEU is a push broom hyperspectral imager, integrating spectroscopy and 2D-spatial mapping in one single system for each target species.

Fig. 1
figure 1

The GOBLEU monitoring instrument suite. The instrument suite consists of two modules: GHG observation (left) and NO2/SIF observation (right). These modules are connected by optical fiber from the window seat. The instrumental parameters, including the observation setting, differ between left-side and right-side seats. The instruments are labeled as “-L” or “-R” for their proper seat to avoid missetting

Full size image

CO2, NO2, and SIF imaging spectrometers are packed in two carry-on luggage-sized boxes (one for NO2 and SIF observation and one for CO2 observation, 320 mm (W) x 380 mm (D) x 600 mm (H) per pack). They weigh less than 30 kg and collect high-spatial-resolution remote-sensing spectra of NO2, SIF, and CO2.

Like passengers, these " passenger " instrument boxes are mounted on seats also with seatbelts. Thus, no modifications to the aircraft are needed. In theory, these instrument suites can take a cabin seat on either side of the aircraft; however, we must switch the setting depending on the side of choice and optimize the coordination. To avoid the missetting, the instrument suites are labeled as “-L” for left-side use and “-R” for right-side use. Figure 1 shows the instrument models for left-side seats. The typical spectra observed by the imaging suites, presented in Fig. 1, are also plotted in Fig. 2.

Table 1 A summary of instrument details
Full size table

The optical lens and optical fibers are made of quartz glass for wide-spectral coverage. Two packed instruments can be handily to mounted on a passenger economy (coach) class seat. The input optics, the passengers’ “eye”, combines a wide viewing lens and an inertial navigation unit, coupled with a Global Positioning System (GPS) signal receiver, to collect the solar radiation reflected by the Earth’s surface and identify the starting locations.

In addition, during taxing before and after the observation flight, a mechanical shutter shields the lens, and the dark spectra are acquired for reference. While the cabin pressure and temperature of passenger aircraft are well controlled, localized temperature gradients around the instrument suites should be avoided. Circulators are installed for each instrument suite. Both navigation and spectra data are simultaneously logged on a laptop computer through a USB-3.0 connection, and the collected data can be monitored in flight.

Fig. 2
figure 2

The collected spectra by FNO2 (top), FSIF (middle), and FGHG (bottom) during a ground function test period

Full size image

All the electronic power for the electronic devices, including camera image acquisition, is supplied by a mobile Li-ion battery (Fujikura BA-155, 155 Wh). The battery capacity needs to be less than 160 Wh to be on board an aircraft, according to a regulation by Japan’s Ministry of Land, Infrastructure, Transport and Tourism (MLIT). The power supply is installed in carry-on luggage-sized boxes, and this configuration is permissible on ANA flights. A single battery with the current observation configuration allows continuous NO2, SIF, and CO2 observation for at least three hours. For round-trip observation, these batteries can be switched with fully recharged packs at the airport, allowing an additional three hours of observations.

The retrieval principles

To retrieving NO2, SIF, and CO2, the differential optical absorption spectroscopy (DOAS) technique [27], the iterative maximum a posteriori differential optical absorption spectroscopy (IMAP-DOAS) technique [28] and full physics retrieval technique [e.g. 29–37] are planned to apply, respectively.

The NO2 absorption lines allocated between 420 and 490 nm are applied DOAS spectral fit. To process the DOAS fit for NO2 retrieval from GOBLEU, the QDOAS software is optimized for GOBELU analysis [38]. The detail of NO2 retrieval setting and data analysis are described in results section.

Solar induced Chlorophyll Fluorescence (SIF) is emitted between 600 nm and 900 nm with the peak at 685 nm [39]. To observe the SIF with remote-sensing, the isolated Fraunhofer lines (absorption features in the solar atmosphere) are used to retrieve the fluorescence emission intensity from O2 A-band observed by GOSAT [40,41,42,43,44]. In addition, SIF is retrieved from GOME-2, TROPOMI [45,46,47] both O2 A-band and B-band. To cover the isolated Fraunhofer lines allocated both O2 A- and B-bands for SIF retrieval, our instrument cover between 670 and 780 nm with 0.05 nm spectral sampling interval. SIF emission intensity from the collected GOBLEU spectra is planned to retrieval by optimizing the iterative maximum a posteriori differential optical absorption spectroscopy (IMAP-DOAS) technique [28] both O2 A- and B- bands regions.

In the case of space-based CO2 observation, 1.6 μm (weak CO2 absorption spectra) and 2.0 μm (strong CO2 absorption spectra) of CO2 bands are used as CO2 retrieval channels among GOSAT [29,30,31,32,33,34,35], GOSAT-2 [36, 37], OCO-2 [48] and OCO-3 [48]. These retrievals are coupled with strong and week CO2 channels and O2A channels to detect the effective optical path length. The transmittance of cabin window only allows to transmit week CO2 spectra. The CO2 retrieval from GOBLEU flight will be performed with full physics algorithm, its constructed with the simultaneous spectral fit between the observed week CO2, O2 A-band spectra and theoretical radiative transfer calculation. Additionally, IMAP-DOAS for CO2 [49] is also considered to first processing of CO2 enhancement in the region of interest.

Observation pattern

The passenger’s eye views the slant nadir through the cabin window, where the observation viewing angle is managed around + 65 º from the nadir to avoid the vignetting of the solar light reflecting the Earth’s surface and stray light from the cabin window and its frame. The pictures of GOBLEU imaging spectrometer suites mounted on the seat of an ANA aircraft are presented in Fig. 3. This configuration can provide wide spatial coverage. From one side of the aircraft (as shown in Fig. 4), the spatial coverage is around 50 km with an 11 km flight altitude and + 65 º viewing angle. It directly depends on flight altitude and viewing angles. The spatial coverage can be easily doubled by expanding to two-side (right and left side of the aircraft) observation with additional carry-on equipment. The ground sampling distance (GSD) is 100 m (along-track) by sub km (across-track) assuming a typical altitude of 11 km a. g. l., ground speed of 200 m/s and integration time of 500 milli-seconds. This configuration will support an increase in monitoring frequency and spatial coverage. In the case of double-sides observation, the coverage for across-track is up to around 100 km with 100 m along-track spatial resolution, which is ten times higher than current satellites.

Fig. 3
figure 3

The GOBLEU instrument suite onboard the ANA aircraft. The instruments are seated on cabin seats (top left panel), targeting the input optics to slant-view thorough the cabin window (top right panel)

Full size image
Fig. 4
figure 4

The GOBLEU observation configuration. “New passenger” (Monitoring instruments) onboard ANA passenger flights and slant-viewing to detect changes in NO2, SIF, and CO2 levels over the surface using cabin seats. A typical sampling distance in along-track is 100 m with 50 km across-track coverage. In the double-side case, the coverage is up to 100 km. The sampling size for across-track direction depends on the viewing angle from the aircraft, and it varies from sub-km to 2 km

Full size image

Flight route plan

The ANA flight nationwide route network that connects Japan’s megacities should allow our observations to focus on collecting data over those intense local emission areas where current satellites have had a challenge to collect data. As described earlier, our operation can be extended to three hours without replacing a battery. We can select any direction from Tokyo Haneda Airport in this time frame. The longest leg is northward from Tokyo Haneda Airport to Hokkaido Wakkanai Airport, which takes three hours. Southward, a 2.5-hour flight from Tokyo Haneda Airport to Kyushu-Nagasaki Airport is allowed. These flights cover the major Japanese megacities and large industrial areas. The main observation route, from Tokyo Haneda Airport to Fukuoka Airport, is illustrated in Fig. 5 coupled with the Emissions Database for Global Atmospheric Research (EDGAR) CO2 [50,51,52] emission map (left) and the one-year integrated TROPOMI SIF intensity [53] (right). The left panel in Fig. 5 illustrates the Tokyo Haneda to Kyusyu Fukuoka flight route and CO2 emission strength based on EDGAR CO2. These routes cover Japanese megacities such as Tokyo (pop: 140 M), Osaka (pop: 88.2 M), Nagoya (pop: 22.9 M), Fukuoka (pop: 15.4 M), Sendai (pop: 10.8 M), and Sapporo (pop: 19.5 M). Also, the major industrial areas in Japan are often located in coastal areas, such as Tokyo Bay, Nagoya Bay, Osaka Bay, and Setouchi Bay. Regular passenger flights also cover these areas. Normally, it takes two hours from Tokyo Haneda Airport to Fukuoka Airport (the southern mega city in Japan). The two-hour flight allowed us to collect soundings every 0.5 s (around 0.8 million observations per side in two-hour flight; 53 across-track x 2 images/sec x 2 h) from 130°E to 140°E in longitude and 33.5°N to 36°N in latitude (about a 900 km travel distance).

Fig. 5
figure 5

The observation flight route with an EDGAR CO2 emission map (left) and a TROPOMI SIF intensity map (right). Flight routes cover Tokyo, Osaka, Nagoya, Fukuoka, Sendai, and Sapporo in Japan and industrial area between Tokyo to Fukuoka

Full size image

Results

The first GOBLEU flight and data processing

On the 26th of October 2020, during the COVID-19 pandemic, the GOBLEU monitoring suite began its first trip from the Tokyo Haneda Airport to the Fukuoka Airport with an ANA regular flight (NH247). A summary of the first flight is shown in Table 2. This flight aimed to realize our observation concept; thus, only the NO2 module was activated. The main objective was to monitor NO2, CO2, and SIF via the cabin window using remote-sensing technology and operation on board a passenger aircraft. To focus on the proof of our observation concept (e.g., performed by one person for one instrument), the first observation test was performed using the NO2 instrument only with the minimum number of support staff. It also aimed to reduce the risk of Coronavirus infection during the COVID-19 pandemic.

The prototype instrument, with the same imaging spectrometer and input optics of the complete package shown in Fig. 1 but mounted on prototype packaging, was treated by an operator as carry-on luggage from the laboratory to Tokyo Haneda Airport (see Fig. 6). After passing through the security inspection including X-ray testing, the instrument was ready for the on-board commercial flight. The “passenger” instrument has been granted permission to be installed on the aircraft after conducting thorough radio interference checks and confirming with the airline that it will not cause any interference. With two pieces of carry-on luggage, any domestic observation flight from the Tokyo Haneda airport (central Japan) can be selected.

To secure the preparation time for the instrument on the cabin seat, we were preboarded ten minutes before regular boarding in the first trial. Then, the instrument was fixed with a seatbelt, as would be a typical passenger. As expected, the aircraft departed on time and took off smoothly. Once the captain turned the seatbelt sign off, we inspected the instrument and confirmed that there was no damage due to takeoff. During the flight, the operator constantly checked the computer display, which indicated two spectra in one second, real-time aircraft position and attitude, and file storage status. Also, the operator noted the weather conditions at several key locations, such as megacities and industrial areas.

Fig. 6
figure 6

Carry-on instrument with operator. The picture was taken just before the security gate of Haneda Airport Terminal 2. The instrument is brought as carry-on luggage for the operator

Full size image
Table 2 A summary of the first GOBLEU flight
Full size table

The weather was fine for the first flight. The operator could observe the ground over Nagoya, Osaka, and the industrial area over Setouchi Bay. After 1.5 h from takeoff from the Tokyo Haneda airport, the aircraft landed at Fukuoka airport. Since the GOBLEU package was firmly mounted in the cabin seat, we did not observe any damage to the package during the landing. At Fukuoka Airport, the instrument suite was “relaxed” and unmounted and left the aircraft as the last passenger from the flight.

The collected spectra were analyzed by using the differential optical absorption spectroscopy (DOAS) technique to retrieve NO2 data. The DOAS spectral fit was done for the 430–460 nm, which is optimal for NO2 retrieval from GOBLEU using the QDOAS software.

In this spectral fit, the differential slant column densities (dSCD) are retrieved with a reference for DOAS analysis, which spectra acquired over the rural area (no emission sources). The retrieved slant column density of NO2 was converted to the vertical column density of NO2 to consider the airmass factor (AMF)

Based on the visual inspection of simultaneously acquired visible camera images by GoPro 8, the highly cloud contaminated areas and the optimal reference area (no emission sources area) are identified. NO2 absorption cross sections were subsequently fitted to the differential optical depth as well as O2-O2, H2O, O3, the pseudo ring spectrum interjacences. Spectrally slowly varying signatures were accounted for including a 3rd order polynomial in the fit. The spectral calibration is also conducted in QDOAS processing coupling with a high-resolution solar spectrum, implemented in QDOAS [54]. As for the instrumental line shape function, the error function, which is based on the convoluted box car and gauss function, is selected for this analysis.

The AMF is defined as the ratio between slant and vertical column of a trace gases. The AMF inferred from the radiative transfer model with geographic and atmospheric model parameters. In this study, the linearized discrete ordinate radiative transfer model: LIDORT [55] are employed with the climatological NO2 profile, which provided from chemical transportation model results of the Tropospheric Chemistry reanalysis version 2 (TCR-2) database [56], surface albedo was set to the monthly climatology data of Lambertian-equivalent reflectivity (LER) derived from TROPOMI [57]. For the aerosol, Aerosol optical thickness (AOT) at 500 nm is applied the constant value of 0.21 coupled with gauss distribution model at1.5 km peak height. For other parameters such as O3 vertical profile are used US standard atmosphere.

Through the optimal fitting by QDOAS, the NO2\(\:{dSCD}_{obs}\), which is the difference between \(\:{SCD}_{obs}\) from the observation spectra and \(\:{SCD}_{ref}\) from the reference spectra, is derived and express the formulation in Eq. (1).

$$\:{dSCD}_{obs}={SCD}_{obs}-{SCD}_{ref}$$
(1)

To deriving the vertical column density (\(\:VCD\)) of NO2, the estimated \(\:{dSCD}_{obs}\) are converted to \(\:{VCD}_{obs}\) coupled with\(\:AMF\) both observation (\(\:{AMF}_{obs}\)) and reference scene (\(\:{AMF}_{ref}\)), and the reference of vertical column density (\(\:{VCD}_{ref}\)) by applying the Eq. (2).

$$\:{VCD}_{obs}=\frac{{dSCD}_{obs}+{VCD}_{ref}\times\:{AMF}_{ref}}{{AMF}_{obs}}$$
(2)

The model parameters used \(\:AMF\) in the calculation, such as surface albedo, vertical profiles of trace gases including NO2 and the aerosol scenario were kept constant throughout all observation points. These AMF calculations are conducted for each observation pixels both reference and target.

In this study, areas around 137.6\(\:^\circ\:\) of longitude and 35.0\(\:^\circ\:\) of latitude are selected as reference area and 1.35 × 1015 molec/cm2 is assumed as \(\:{VCD}_{ref}\) reference concentration, which was determined using climatological NO2 dataset based on monthly averaged TROPOMI observation data [58]. The differential approach cancels out the stratospheric NO2 concentration to the signal, making the measurements only sensitive to tropospheric absorption, under the assumption that the stratospheric NO2 filed has a negligible spatial and temporal variability during the time between the aquation of the reference spectrum and the measurements.

The error in the retrieved NO2\(\:{VCD}_{obs}\) originates from uncertainties in the calculated \(\:{dSCD}_{obs}\), \(\:{SCD}_{ref}\), and \(\:{AMF}_{obs}\). One assumes that the contributing uncertainties are sufficiently uncorrelated as they arise from nearly independent steps. Base on Eq. (1), the error of the NO2\(\:{VCD}_{obs}\) retrieval algorithm can be quantified based on the following error propagation method [59]:

$$\:{\sigma\:}_{{VCD}_{}}^{}=\sqrt{{\left(\frac{{\sigma\:}_{{dSCD}_{obs}}}{{AMF}_{obs}}\right)}^{2}+{\left(\frac{{\sigma\:}_{{dSCD}_{ref}}}{{AMF}_{ref}}\right)}^{2}+{\left(\frac{{SCD}_{obs}\times\:{\sigma\:}_{{AMF}_{obs}}}{{AMF}_{obs}^{2}}\right)}^{2}}$$
(3)

To estimate a typical \(\:{\sigma\:}_{{VCD}_{}}^{}\), Eq. (3) is considered with our flight data. The error in the DOAS fit, \(\:{\sigma\:}_{{dSCD}_{obs}}\), is a direct output of QDOAS for each fit. In parallel, \(\:{AMF}_{obs}\) is calculated for each observation point coupled with modeled parameters. For the analysis areas including Nagoya and Osaka, the first term in the error estimation equation is calculated as 4.8 × 1015 molec/cm2. (only in blanket in the first term.)

The second error term originates from the estimation of the NO2 residual amount in the reference spectra (\(\:{\sigma\:}_{{dSCD}_{ref}}\)), which is determined as the standard deviation of reference observations fits. In this assumption, the second term is calculated as 5.9 × 1015 molec/cm2. The calculated third term (\(\:{\sigma\:}_{{AMF}_{obs}}\)) has to be estimated based on the variation of modeled parameters.

It is too important to take the sensitivity of various parameter to AMF into account; however, it is beyond the scope of this study. Thus, the reference values, which were reported by [59], are applied in this estimation. In the literature, the relative error of \(\:{\sigma\:}_{{AMF}_{obs}}\) was based on” Heavily polluted” condition in [59]. In this case, \(\:{\sigma\:}_{{AMF}_{obs}}\) is assumed to be 0.6. To apply this assumption, the third term in the bracket is 6.8 × 1015 molec/cm2. Finally, a typical \(\:{\sigma\:}_{{VCD}_{}}^{}\) was estimated to be 1.0 × 1016 molec/cm2.

As shown in the following sections, we compared the retrieved NO2 results with other data, such as satellite NO2, a gridded emission inventory and ground-based in-situ observations.

The observed NO2 at a glance

Figure 7 shows the collected first high-definition NO2 map (Fig. 7a) showed good spatial pattern correspondence between both NO2 data collected by the state-of-the-art satellite NO2 instrument TROPOspheric Monitoring Instrument (TROPOMI) [60, 61] (Fig. 7b) and a high-resolution (1 km) gridded NOX total emission estimates (Fig. 7c) developed by the Ministry of the Environment (MOE) (hereafter, MOEJ) [62], notably at megacities, such as Nagoya (pop: 7.5 M) and Osaka (pop: 8.8 M). The observation duration for these areas is around 30 min for GOBLEU and less than 1 min for TROPOMI. The white arrows present the wind vector taken from the ERA5 hourly wind dataset (GOBLEU for UT 10:00 and TROPOMI NO2 for UT 12:00) [58, 63]. The map depicts the spatial pattern of NO2 over the two megacities. In addition, the emissions from industrial areas are also clearly captured. The general spatial pattern is consistent with what TROPOMI satellite NO2 observations captured approximately two hours after the GOBLEU flight.

GOBLEU and TROPOMI NO2 spatial patterns are compared. Two NO2 datasets (34°-37°in latitude and 132°-138°in longitude) were binned at a 0.05° and averaged. The values were finally aggregated latitudinally. The GOBLEU NO2 map (aggregated mean in 0.05º longitudinal range, see Fig. 8) also showed good spatial correspondence with NO2 data collected by TROPOMI (aggregated mean in 0.05º longitude range). While the observations were taken two hours apart, GOBLEU and TROPOMI NO2 data shared major longitudinal spatial patterns peaking at megacity locations. It is important to note that GOBLEU data show more spatial features in the NO2 field that seem to be co-located with large CO2 sources of emission indicated by the EDGAR CO2 inventory [50,51,52]. EDGAR CO2 inventory data aggregated mean in the 0.1º longitudinal range is also plotted in Fig. 8.

Fig. 7
figure 7

NO2 data collected during the first GOBLEU flight between Tokyo Haneda Airport and Fukuoka Airport (a) and from TROPOMI (b). The bottom panel (c) shows 1 km x 1 km total NOx emission values from the MOE inventory. The white arrows present the wind vector: GOBLEU for UT 10:00 and TROPOMI NO2 for UT 12:00, respectively

Full size image
Fig. 8
figure 8

0.05º latitudinally averaged NO2 distribution of both GOBLEU (GB in blue) and TROPOMI (TROP in red) with EDGAR CO2 inventory (Gray). Yellow shading area indicates megacities’ longitudinal range. The Nagoya and the Osaka area are shaded

Full size image

The NO2 spatial patterns over megacities

Looking closely at Nagoya and Osaka, the two maps showed different fine-scale spatial patterns. Figure 9 shows GOBLEU NO2 aggregated to 0.005º x 0.005º and TROPOMI NO2 aggregated in 0.05º x 0.05º with EDGAR CO2 aggregated in 0.1º x 0.1º. The GOBLEU NO2 map also depicts the point-wise intense NO2 concentration, which collocated with the power generation facilities. The differences are unsurprising because two observations were taken at different local times (10:30 for Nagoya and10:45 for Osaka). The observed NO2 concentration with different time presents the different concentration due to the diurnal cycles in emission and NOx-chemistry [64] as well as the wind magnitude/direction changes. However, they suggest the potential impact of observation time difference on the NO2’s performance as a marker for CO2 emissions and highlight the significant benefit of simultaneous CO2 and NO2 observations for accurately estimating CO2 emissions. Even a few hours’ difference in observation time (which we see in previous studies) could significantly impact our ability to estimate CO2 emissions using an NO2 marker. In addition, the high-resolution observation should help capture the activity level emissions signatures. GOBLEU provides 10 or 20 times finer spatial structures of NO2 concentrations than that of TROPOMI and the EDGAR inventory.

Fig. 9
figure 9

NO2 spatial patterns observed from GOBLEU (top) and TROPOMI (middle) during the first flight over Osaka (left) and Nagoya (right) during one flight. The boom row shows the spatial distributions of CO2 emissions from the EDGAR inventory. Note that the local times for GOBLEU and TROPOMI are not the same. While the two remotely sensed NO2 data share major spatial patterns, GOBLEU data captured fine-scale emission hot spots that might be diluted in TROPOMI data due to the spatial resolution

Full size image

The spatial correspondence between GOBLEU and TROPOMI was further examined using the 0.05º aggregated data (Fig. 10). The difference could be attributable to the NOx-chemical process, wind speed and direction for NO2 transportation, and diurnal emission changes of NO2 due to the two-hour observation time difference, regardless of the potential contributing factors. The comparison shows that NO2 concentration derived from GOBELU is almost similar range with that of well validated TROPOMI NO2.

Fig. 10
figure 10

Comparison of GOBLEU NO2 and TROPOMI NO2 in Nagoya (red filled circle) and Osaka (blue filled circle). The dotted lines represent the 1:1 line. The range of variation between GOBLEU and TROPOMI is similar but GOBLEU NO2 present a higher concentration than that of TROPOMI. It seems that GOBLEU has highly sensitivity for local emission changes due to higher spatial resolution than that of TROPOMI

Full size image

Temporal correlation of NO2

GOBLEU NO2 data’s performance as a CO2 emission marker was further examined by looking at spatial correlation with hourly NO2 data from ground-based monitoring sites (https://soramame.env.go.jp/station), acknowledging known challenges (see Fig. 11). Hourly data from 136 ground stations in Nagoya and 251 ground stations in Osaka are used for this analysis. These stations are designed to monitor major local emission sources within the cities, such as traffic and industries, mainly for air quality monitoring purposes. The correlation with these observational data should be able to evaluate the sensitivity of the remotely sensed data to local emissions.

As expected, we found that GOBLEU’s observations showed the best correlation at their local time (10:30 for Nagoya and 10:45 for Osaka). The level of the correlation went down significantly with time. Interestingly, while TROPOMI’s correlations with the ground-based data peaked at the observation time (12:30 JST) for Nagoya, the Osaka case showed two correlation peaks in the morning (10 a.m. local time) and afternoon (2 p.m. local time). While GOBLEU and TROPOMI share the basic remote-sensing principle, GOBLEU captures NO2 changes at the surface significantly better than TROPOMI. This is a caution for estimating emissions at a particular time of the day rather than obtaining an average emission over a certain period, using NO2 as an emission marker. This analysis only loosely examined the skill at representing spatial patterns of potential CO2 emissions. Accurately deriving emissions using these markers is yet another major challenge due to meteorology and chemistry.

Fig. 11
figure 11

Correlation analysis using remotely sensed NO2 data from GOBLEU (GB in blue), TROPOMI (TROP in red), and hourly surface NO2 data. The spatial correlation of remotely sensed and surface NO2 was calculated over Osaka (top left) and Nagoya (top right) as a metric to characterize spatial patterns of two remotely sensed NO2 levels. Hourly changed wind magnitude and direction are plotted in bottom panels. The vertical dotted lines indicate the local observation times for GOBLEU (blue) and TROPOMI (red). The gray line shows hourly NOx emission estimates taken from a gridded inventory developed by MOEJ. The correlation tends to be higher near the local observation times for GOBLEU and TROPOMI; however, the two remotely sensed NO2 data collected roughly 1.5–2 h apart clearly showed a different correlation with the surface NO2 over time. This suggests the significance of collocated NO2 observations. The GOBLEU case showed much larger changes, which might be attributed to the higher spatial resolution of the data compared to TROPOMI

Full size image

Discussions

The first proof of concept flight with a NO2 prototype instrument was successfully carried out. We proved and confirmed that our instrument collects science-quality useful spectra through the cabin window without modifying the aircraft. The airborne remote-sensing observation collects denser and finer spatial resolution GHG because the flight altitude of passenger aircraft (~ 11 km) is 1/60 of the current satellites (666 km) [3]. The follow-on instruments, a complete package of NO2, SIF, and CO2 modules, will be onboard both sides of the aircraft after the instrumental preparation and inspection. GOBLEU will simultaneously observe CO2 and SIF, in other words, the emission and removal makers. The SIF information as an indicator for an activity level of removals is informative to improve our knowledge of biospheric removals and provide this information.

We also plan to perform extensive evaluation and validation of the collected data using the upper-looking data collected by the Total Carbon Column Observing Network (TCCON) [65] and Collaborative Carbon Column Observing Network (COCCON) [66]. These observations are based on the sun-directed ground-based spectrometer. As for NO2, JAXA also prepared a ground-based instrument, PANDORA [67], for NO2 validation. In the future, we will coordinate the validation campaign for our datasets to set the ground-based instrument under the flight route.

Our GOBLEU observation plans to prototype the synergistic use of CO2 and NO2 data and improve anthropogenic emission estimates. By using the correlation between CO2 concentration and SIF intensity, GOBLEU also aims to estimate the carbon removals in the surrounding megacity area by the terrestrial biosphere. By observing emissions and removals, GOBLEU aims to collect actionable data for monitoring emissions from dominant source sectors, such as local traffic and industries, within cities and the removal impacts by the terrestrial biosphere. By visualizing the emission strength and reduction impact by utilizing new emission reduction technologies, our mitigation effort will be clearly depicted on the map. As also described in the results section, GOBLEU data has fine spatial resolution and is more detailed than that of state-of-the-art spaceborne imaging spectrometer instruments such as TROPOMI, OCO-3, and new geostationary air quality instruments: Geostationary Environment Monitoring Spectrometer (GEMS) [68]. Then, GOBLEU data can potentially validate these spaceborne datasets with finer spatial resolution data. Figure 12 presents the spatial distribution comparison between GOBLEU, TROPOMI, and OCO-3. GOBLEU and TROPOMI data are aggregated in 0.005º x 0.005º, in 0.05º x 0.05º, respectively. OCO-3 has a 1.29 km x 2.25 km spatial resolution. Comparing these plots, GOBLEU data depict the local pattern of NO2 concentration.

Fig. 12
figure 12

GOBLEU, TROPOMI, and OCO-3 images over Osaka. Emission proxy NO2 data over Osaka collected by GOBLEU NO2 (left, 13:15 LT; UTC + 9), TROPOMI NO2 (middle, 11:11 LT; UTC + 9), and OCO-3 CO2 (right, 11:05 LT; UTC + 9). These observations were conducted on the same day (Oct. 27, 2020.) but at different times. GOBLEU data depict the local pattern of NO2 concentration over Osaka. TROPOMI also suggests the high NO2 concentration area near Osaka Bay, but it is difficult to identify the individual sources by pixel. OCO-3 also indicated a high CO2 concentration area but did not correspond to the TROPOMI NO2 map. These results also encourage the simultaneous observation of CO2 and NO2, to understand the emission sources well

Full size image

While aircraft-based data are unique and valuable for GHG research [24], it is often challenging to increase data volume due to the cost. Our observations can be performed with two economy seats for instruments, the same as regular passengers. Taking a charter flight instead of a regular passenger flight increase the cost of one-flight up to several times. For example, two hours research flight could cost 2.4 K USD [https://www.jsforum.or.jp/other/zerog/]. Also, every day, passenger flights are scheduled to fly from one city to another, and seats can greatly increase the chance for observability. These situations are highly beneficial when applying for a regular passenger flight. The GOBLEU instrument has currently booked its seats on a regular flight, operated daily at noon (e.g., ten round-trip flights between Tokyo Haneda Airport and Fukuoka Airport from 9:00 JST to 17:00 JST). If we select multiple flights in a day, GOBLEU could be a “frequent flyer” to identify the diurnal variations. As described in the previous section, all the security and safety inspection including X-ray test, radio interference checks and confirming with the airline that it will not cause any interference are required and confirmed for GOBLEU. However, operator training will be required to install each module safely onboard. After optimizing the FGHG, FNO2 and FSIF instrumental package, regular observations once per month were started in the spring of 2023. During the COVID-19 pandemic, the seats for instruments were easily bookable. However, after the pandemic, the regular passenger returns on their flight, and the instruments must book months in advance. In this operation chain, selecting a flight with fine weather is difficult. To overcome this, frequent flights with instruments are desirable with simple operation. Also, there are regular flights between Tokyo Haneda Airport and Fukuoka Airport less than every hour, from early morning to late night every day. In addition, GOBLEU should have a chance to elucidate diurnal variations and weekly dependent emission patterns with high-frequency observations.

As indicated in Fig. 5, ANA has a nationwide flight route network that connects key capital and major airports in Japan, such as Tokyo Haneda Airport and Hokkaido Wakkanai Airport, Tokyo Haneda Airport, and Kyushu Nagasaki Airport, which are potential destinations for future GOBLEU flights. These routes cover the high-intensity SIF, which indicates the production level (carbon removal) of the terrestrial biosphere area, such as the near-Sendai forestry area (see Fig. 5). The intensity of SIF indicates the strength of emission remover.

The area under the flight route Tokyo Haneda Airport to Fukuoka Airport includes approximately 30% of Japan’s total CO2 emissions and about 59% of Japan’s Gross National Product (GDP) [69]. However, it is important to note that our airborne remote-sensing approach also shares the challenge due to cloud contamination. In the satellite observation case, around 80% of the data is cloud-contaminated despite the small foot-print size [70]. The data yield is expected to be impacted significantly in the summertime. In addition to anthropogenic CO2 emissions, monitoring of natural CO2 emissions and removals is a key to Japan’s GHG management [71]. 67% of Japan’s national land area is cover the forest [72].

Conclusion

Under the GOBLEU project, JAXA has developed a remote-sensing technique that can be operated on commercial passenger aircraft, by JAXA and ANAHD. The GOBLEU monitoring instrument suite is designed to collect CO2, NO2, and SIF data. The first trial flight with the NO2 instrument was carried out and collected NO2 data over major populated and industrialized areas. The NO2 data showed timely snapshots of NO2 spatial distributions, suggesting the utility of NO2 data as a proxy for anthropogenic CO2 emissions that should enhance emission estimations and attribution ability. GOBLEU instrument can be operated in almost any passenger aircraft without any modifications. While based on the same space-based observation technique as the current satellites, our airborne observation can collect denser and higher spatial resolution GHG data more frequently over key emission areas, such as Japanese megacities.

GOBLEU is a new challenge for GHG remote-sensing observation on a new platform and has the potential to open up a new field for passenger aircraft use. Nominally, passenger aircraft only carried passengers. To contribute to climate action by air company, the ANA Group is promoting ESG management that considers the Environment, Society, and Governance, aiming to realize a sustainable society and enhance corporate value. Under the slogan of “ANA Future Promise”, the ANA group started actions to reduce the carbon emissions from aviation, actively adopting Sustainable Aviation Fuel (SAF) for aviation as well as a recycle-based society. In addition to these activities, ANAHD is scoping extended aviation use for climate mitigation activities such as climate monitoring tools.

GOBLEU aims to monitor the climate mitigation effort over Japan’s intensive industrialized areas and contribute to the second GST scheduled in 2028. GOBLEU expects to provide timely GHG information by promptly collecting high-resolution GHG data and emission and removal estimates with greater information granularity. The high-spatial-resolution data should provide GHG information to stakeholders at different subnational levels (e.g., states/prefectures, cities, private sectors, and citizens) toward carbon neutrality under the Paris Climate Agreement. We also expect to expand the observation coverage overseas with domestic and international partnerships through enhancing internal cooperation.