Did Atmospheric CO₂ and CH₄ Observation at Yonagunijima Detect Fossil-Fuel CO₂ Reduction due to COVID-19 Lockdown?

Abstract

Synoptic-scale variabilities of atmospheric CO₂ and CH₄ observed at Yonagunijima (Yonaguni Island, YON, 24.47°N, 123.01°E) during winter (from January to March) in 1998–2020 were examined. The monthly mean variability ratios (ΔCO₂/ΔCH₄) based on correlation slopes within 24 h time windows showed a clear increasing trend, which is mainly attributed to the unprecedented increase in the fossil fuel-derived CO₂ (FFCO₂) emissions from China. A similar increasing trend of the ΔCO₂/ΔCH₄ ratio had been reported for the observation at Hateruma Island (HAT, 24.06°N, 123.81°E), located at approximately 100 km east of YON. Nevertheless, the absolute values for YON were 34 % larger than those for HAT. Additionally, the monthly average in February 2020 for YON showed no marked change, whereas that for HAT showed an abrupt considerable decrease associated with the FFCO₂ emission decrease in China presumably caused by the COVID-19 lockdown. Investigating the diurnal variations, we found that the local influences were larger at YON, especially during daytime, than at HAT. Using nighttime data (20-6 LST) and a longer time window (84 h), we succeeded in reducing the local influences and the resulting monthly mean ΔCO₂/ΔCH₄ ratio showed considerable similarity to that observed at HAT including the abrupt decrease in February 2020. These results convinced us that the ΔCO₂/ΔCH₄ ratio could be successfully used to investigate the relative emission strength in the upwind region.

1. Introduction

In December 2019, the novel coronavirus disease (COVID-19) was first reported in Wuhan, China, and rapidly spread across the whole country. The government of China imposed lockdown measures first in Wuhan on 23 January 2020 and then extended the area across the nation in February 2020. Since such lockdown measures strictly restricted socioeconomic activities, a considerable reduction in fossil fuel-derived CO₂ (FFCO₂) emissions was expected (Le Quéré et al. 2020). This situation raised the question of whether atmospheric observations could detect the COVID-19-induced FFCO₂ reductions.

Substantial reductions in air pollutants like nitrogen dioxide (NO₂) were observed during the lockdown period in China by using in situ and satellite measurements (e.g., Bauwens et al. 2020; Le et al. 2020). Nevertheless, the corresponding change in atmospheric CO₂ mole fraction was expected to be relatively small because of its rather long lifetime in comparison with the above air pollutants. Satellite retrievals of the CO₂ column average (XCO₂) over China were examined by several studies to detect the influence of the COVID-19 lockdown (Chevallier et al. 2020; Buchwitz et al. 2021), which concluded that the detection was challenging because the expected change of XCO₂ was less than the precision of the satellite observations.

Examining synoptic-scale variations in the atmospheric CO₂ and CH₄ observed at Hateruma Island (HAT, 24.06°N, 123.81°E), Tohjima et al. (2020) found that the ΔCO₂/ΔCH₄ variability ratio showed a marked decrease in February 2020 compared with the former 9 year (2011–2019) average. Because HAT is in the downwind region of continental Asia during the late autumn-early spring season because of the East Asian monsoon, correlative elevations of the CO₂ and CH₄ mole fractions are often observed (Tohjima et al. 2010). A previous study revealed that the gradual increase of the ΔCO₂/ΔCH₄ ratio was attributed to the unprecedented increase in the FFCO₂ emissions from China (Tohjima et al. 2014). Consequently, Tohjima et al. (2020) concluded that the observed ΔCO₂/ΔCH₄ decrease in February 2020 was related to the COVID-19 lockdown in China. Furthermore, they evaluated the decrease in the FFCO₂ emissions from China to be approximately 32 ± 12 % and 19 ± 15 % for February and March, respectively, from a comparison of the observed and simulated ΔCO₂/ΔCH₄ ratios. Nevertheless, the change in the ΔCO₂/ΔCH₄ ratio associated with the COVID-19 was considerably faint and was detected only in single-site observations, limiting the certainty of the estimate of emission change. Thus, to enhance the certainty, an independent observation is required.

Yonaguni Island (YON, 24.47°N, 123.01°E), where atmospheric CO₂ and CH₄ have also been observed by the Japan Meteorological Agency (JMA) since 1997 and 1998, respectively, is located at approximately 100 km west of HAT. The observed seasonal cycles and trends for HAT and YON were almost identical (Zhang et al. 2007) because of the similarity of the geographical positions. Thus, expectedly, the CO₂ and CH₄ observations at YON can be used to constrain the relative emission strengths from China. In this study, we examined whether the ΔCO₂/ΔCH₄ ratio observed at YON showed similar variations, especially an abrupt decrease in February 2020, as the observations at HAT did.

2. Data and method

2.1 Yonagunijima

Yonagunijima (Yonaguni Island: YON), located at the western end of the Ryukyu Archipelago, is the westernmost inhabited island of Japan. The island has a shape of an almond (longer in the east–west direction), an area of approximately 29 km² with the highest point being 231 m, and a population of approximately 1700. There are mountainous areas in the southern and western parts of the island, covered by subtropical forests. JMA built a local meteorological station to the east of the north settlement and approximately 800 m inland from the north coast in 1989. The station is surrounded by grassland and sugarcane fields. The observation at YON is generally influenced by the air masses from the Pacific region during summer and from the Asian continent during winter. Such seasonality in the air mass transport also influences the observation at HAT. Note that no clear diurnal patterns in wind direction and wind speed were observed at YON.

2.2 CO₂ and CH₄ measurement systems

The atmospheric CO₂ and CH₄ measurement systems used at YON are briefly described here; the details were presented elsewhere (e.g., Watanabe et al. 2000; Tsutsumi et al. 2006). The sample air was drawn by a pump from an air intake located at the top of a mast near the station building (20 m above ground level) and was delivered to the CO₂ and CH₄ measurement systems.

The CO₂ in the air sample was measured by using non-dispersive infrared analyzers (NDIRs). The analytical precisions of the NDIRs were better than ±0.02 µmol per mol (ppm). The CH₄ mole fractions were measured using an NDIR during January 1998 to December 2007 and using a gas chromatograph equipped with a flame ionization detector (GC/FID) after January 2008. The precisions of the atmospheric CH₄ measurement systems were better than ±10 nmol per mol (ppb) for the NDIR system and ±5 ppb for the GC/FID system.

The CO₂ and CH₄ measurement systems were calibrated by introducing working standard gases from high-pressure cylinders, which were prepared at the laboratory in JMA (Watanabe et al. 2000; Matsueda et al. 2018). The CO₂ and CH₄ mole fraction values at YON were reported on the WMO scale. Note that although the CO₂ and CH₄ mole fractions at HAT were reported on the NIES original scales, it was confirmed from the WMO/IAEA Round Robin comparison experiment (e.g., Zhou et al. 2009) that the differences between the WMO and NIES scales were kept within ±0.15 ppm.

2.3 Analytical methods of the variability ratio

The time series of CO₂ and CH₄ at YON often showed correlative synoptic-scale variations. To extract the variability ratios from such correlative variations, we used the same method as Tohjima et al. (2014, 2020) used to calculate the variability ratios for the observation at HAT. Since the details of the methods are given in the literature, here, we give only a brief description of the methods.

First, using a set of the hourly CO₂ and CH₄ data within a certain time window, we computed a correlation coefficient (R), standard deviations for both CO₂ and CH₄ (σ_CO2 and σ_CH4), and a correlation slope by using a reduced major axis method (Hirsch and Gilroy 1984). Shifting the time window by 1 h, we repeated the above computations for the entire data set. Then, setting the criteria of the correlation coefficient (R) and the CO₂ standard deviation (σ_CO2), we discarded the correlation slopes with R and σ_CO2 less than the criteria. Finally, using the selected correlation slopes, we calculated the monthly average and the moving averages of the variability ratio. Although the time window was set to 24 h in Tohjima et al. (2020), we also tried a longer time period of up to 120 h in this study. Note that we used the same R = 0.7 and σ_CO2 = 0.1 ppm for the criteria as were used in a previous study (Tohjima et al. 2020).

3. Results and discussion

3.1 Synoptic-scale variations of the atmospheric CO₂ and CH₄

The time series of the hourly CO₂ and CH₄ mole fractions at YON from January to February 2020 and those observed at HAT are plotted in Fig. 1. Both CO₂ and CH₄ time series show similar synoptic-scale variations with periods of several hours to several days. The synoptic-scale variations observed at YON were almost the same as those observed at HAT except for phase differences; the temporal variations at YON generally proceed a few hours ahead of those at HAT. Additionally, the CO₂ short-term variations observed at YON seem to be slightly larger than those at HAT.

Fig. 1.

Time series of the hourly CO₂ (red line, left Y-axis) and CH₄ (blue line, right inverse Y-axis) mole fractions observed at YON from February to March 2020. The gray lines represent corresponding hourly mole fractions observed at HAT.

We calculated the average diurnal cycles of CO₂ and CH₄ at YON and HAT for 3 months (January–March) to clarify the differences (Fig. 2). The average diurnal cycle is calculated as the average deviation from the corresponding daily means for the individual hours. Although CO₂ diurnal cycles showed a decline during daytime at both sites, the amplitudes of the decline for YON are more than twice as large as those for HAT. The declines during daytime are probably attributed to the photosynthetic CO₂ sequestrations of the biosphere on the islands. These results suggest that the local CO₂ fluxes more strongly influence the atmospheric observation at YON than at HAT. The larger local influence at YON may be attributed to the fact that the station at YON is located in a more inland area of the larger island and partially that the sampling inlet is placed at a lower position.

Fig. 2.

Average diurnal cycles of (a) CO₂ and (b) CH₄ at YON (red lines) and HAT (black lines) for 3 months (January: triangles; February: circles; March: squares). The average diurnal cycle is calculated as the average deviation from the daily means for the individual hours.

Meanwhile, compared with the analytical precision of the CH₄ measurements [±2 ppb for HAT (Tohjima et al. 2002)], the ranges of the average CH₄ diurnal cycles of ±1.5 ppb and ±1 ppb for YON and HAT, respectively, are relatively small. These results suggest that the influence of the local sources on the observed atmospheric CH₄ variations is negligible for both sites.

3.2 Temporal change in the monthly mean ΔCO₂/ΔCH₄ ratio

Monthly mean ΔCO₂/ΔCH₄ ratios based on the observation at YON in January, February, and March from 1998 to 2020 are plotted in Fig. 3a together with the ΔCO₂/ΔCH₄ ratios for HAT, reported in a previous study (Tohjima et al. 2020). As the figure shows, the ΔCO₂/ΔCH₄ ratios for YON are 39 mol mol⁻¹ on average larger than those for HAT (averages of 156 mol mol⁻¹ for YON and 117 mol mol⁻¹ for HAT). Additionally, the increasing ratios determined by linear regression for the entire period (1998–2020) are larger for YON (3.2 ± 0.3 mol mol⁻¹ yr⁻¹) than for HAT (1.9 ± 0.2 mol mol⁻¹ yr⁻¹). The smaller increasing trend at HAT may be related to the fact that the ΔCO₂/ΔCH₄ ratio for HAT reached a plateau after 2011 in association with the stagnant FFCO₂ emission increase from China (see Fig. 2 of Tohjima et al. 2020). The increasing rate for HAT decreased from 2.7 ± 0.4 mol mol⁻¹ yr⁻¹ in 1998–2010 to −0.6 ± 0.7 mol mol⁻¹ yr⁻¹ after 2010, whereas that for YON for the later period was 3.7 ± 0.7 mol mol⁻¹ yr⁻¹.

Fig. 3.

(a) Monthly mean ΔCO₂/ΔCH₄ ratios based on the observation at YON (red closed symbols) and HAT (black open symbols) for January (triangles), February (circles), and March (squares) from 1998 to 2020. The ΔCO₂/ΔCH₄ ratios for HAT are taken from Tohjima et al. (2020). A 24 h time window was used to calculate the ΔCO₂/ΔCH₄ ratios both for YON and HAT. (b) Same as (a) but only nighttime data (20-6 LST) and an 84 h time window were used to calculate the ΔCO₂/ΔCH₄ ratios for YON.

Considering the relatively close geographical locations of YON and HAT, it is quite difficult to imagine that the air masses transported from continental Asia induce such considerably large differences between YON and HAT. Thus, the larger short-term variability probably caused by the local CO₂ fluxes would explain the larger ΔCO₂/ΔCH₄ ratios for YON. The average diurnal cycles of CO₂ at YON (Fig. 2) show rather stable values during nighttime but rapid decreases and increases during daytime, which might indicate an enhancement of the short-term variability during daytime. To investigate the difference in the shortterm variability of CO₂ for YON between daytime and nighttime, we computed the frequency distributions of the standard deviations of the CO₂ data within the 24 h time windows for the daytime (7-19LST) and nighttime (20-6LST). The result showed that the peak of the frequency distribution was located at a larger standard deviation for the daytime data than for the nighttime data (supplementary Fig. S1). Accordingly, there is a possibility that the daytime data might enlarge the ΔCO₂/ΔCH₄ ratios for YON.

3.3 Reduction of local influences

To reduce the local influences on the observations at YON, we calculated the monthly ΔCO₂/ΔCH₄ ratio by only using nighttime data (supplementary Fig. S2). Although the resulting monthly means decreased by approximately 20 mol mol⁻¹ on average, those values were still approximately 19 mol mol⁻¹ larger than the values for HAT, suggesting the local influence was still not sufficiently reduced. To further reduce the local influences, we used a rather long time window for the correlation analysis. As was discussed in the previous sections, the local influences appeared in the diurnal cycles, whereas the synoptic-scale variations usually had a longer timescale of several days. Thus, the longer time window could reduce the influence of the local CO₂ fluxes on the total CO₂ variations. Hence, we calculated the monthly ΔCO₂/ΔCH₄ ratios for YON for different time windows up to 120 h and examined the root mean square (RMS) of the differences of the monthly ΔCO₂/ΔCH₄ ratios between YON and HAT (supplementary Fig. S3). The RMS showed a minimum value when an 84 h time window was used. The result using the 84 h time window is plotted in Fig. 3b. The modified ΔCO₂/ΔCH₄ ratios are similar to the ΔCO₂/ΔCH₄ ratios for HAT; the average difference (YON–HAT) is −2 mol mol⁻¹. The increasing rates determined by the linear regression were 3.7 ± 0.7 mol mol⁻¹ yr⁻¹ during 1998–2010 and 0.1 ± 0.8 mol mol⁻¹ yr⁻¹ during 2011–2020, which are also consistent with those for HAT.

Moreover, the modified ΔCO₂/ΔCH₄ ratio for YON showed a substantial decrease in February 2020. As is described in the Introduction, Tohjima et al. (2020) attributed the decrease in the monthly ΔCO₂/ΔCH₄ ratio in February and March 2020 from the preceding 9 year averages to the decrease in the FFCO₂ emissions from China associated with the COVID-19-induced lockdown. The monthly averages of the modified ΔCO₂/ΔCH₄ ratio for YON and the previously reported corresponding values for HAT are listed in Table 1 for comparison. Note that the duration of the time window also influenced the ΔCO₂/ΔCH₄ ratios for HAT as suggested by Tohjima et al. (2020) and the February ΔCO₂/ΔCH₄ ratios for HAT decreased to 22 mol mol⁻¹ when the 84 h time window and the nighttime data were used. Even if the influence of the duration of the time window is considered, the decreases in the ΔCO₂/ΔCH₄ ratio for YON agree well with those for HAT within the uncertainties both for February and March. Although the decrease in March for YON is unclear in comparison with that for HAT, using a 120 h time window enlarges the decrease to 10 mol mol⁻¹, which suggests that it is difficult to completely reduce local influences.

Figure 4 shows the temporal variation in the 30 day moving average of the modified ΔCO₂/ΔCH₄ ratio for YON from January to March 2020. In the figure, the preceding 9 year (2011–2019) average of the 30 day moving average for YON with the range of the uncertainties (1σ) and the 30 day moving average of the ΔCO₂/ΔCH₄ ratio for HAT from January to March 2020, which were shown in Fig. 3 of Tohjima et al. (2020), are also depicted. Compared with the preceding 9 year average, the moving-averaged ΔCO₂/ΔCH₄ ratio for YON shows a rapid decrease between January and February, a bottom in the middle of February, and an asymptotic increase toward the 9 year average. Such variations are generally consistent with those for HAT. Additionally, the pattern of variations in the moving-averaged ΔCO₂/ΔCH₄ ratio is similar to the estimated change in the FFCO₂ emissions from China based on Le Quéré et al. (2020) which is also depicted in Fig. 4.

Fig. 4.

(Top, left Y-axis) Temporal variations in the 30 day moving average of the modified ΔCO₂/ΔCH₄ ratio for YON (red) and HAT (blue) from January to March 2020. The ΔCO₂/ΔCH₄ ratios for YON are based on the night-time data and an 84 h time window to reduce the local influences (see text). The ΔCO₂/ΔCH₄ ratios for HAT were taken from a previous study (Fig. 3 of Tohjima et al. 2020). The gray line with vertical bars represents the preceding 9 year (2011–2019) average of the 30 day moving average for YON with the range of the uncertainties (1σ). (Bottom, right Y-axis): the estimated temporal change in the FFCO₂ emissions from China based on Le Quéré et al. (2020).

From the above results, we concluded that the atmospheric CO₂ and CH₄ observations at YON could detect the signals related to the COVID-19 lockdown in China as the change in the ΔCO₂/ΔCH₄ ratio. This conclusion supports the idea that the atmospheric ΔCO₂/ΔCH₄ ratio is effective in evaluating temporal changes in the relative emission strengths in the upwind source regions.

4. Summary and conclusion

To detect the signal associated with the COVID-19 lockdown, we examined the synoptic-scale variability ratio of the atmospheric CO₂ and CH₄ (ΔCO₂/ΔCH₄) observed at YON during the period from 1998 to 2020 by applying the analytical approach of Tohjima et al. (2020). Being different from the results observed at HAT (Tohjima et al. 2020), the ΔCO₂/ΔCH₄ ratio was approximately 34 % larger than that for HAT, and the ΔCO₂/ΔCH₄ ratio for February 2020 did not show a marked decrease. Examining the diurnal variations of CO₂ and CH₄ at YON, we found that the local fluxes, probably air-to-land biosphere exchange, enhanced the CO₂ variability especially during daytime (7–19 LST).

Using the nighttime (20-6 LST) data and a longer time window of 84 h, we were able to effectively reduce the local influences on the ΔCO₂/ΔCH₄ ratios; the resulting monthly ΔCO₂/ΔCH₄ ratios showed considerable agreement with those for HAT, including a marked decrease in February 2020 in comparison with the preceding 9 year (2011–2019) averages. Additionally, the 30 day moving average of the ΔCO₂/ΔCH₄ ratio abruptly decreased between January and February 2020, reached the bottom in the middle of February, and gradually returned to the level of the former 9 year average in March. Such a decline pattern is similar to the change in the FFCO₂ emissions from China estimated on the basis of the study of Le Quéré et al. (2020).

As is described in the Introduction, detecting the decrease in the atmospheric CO₂ related to the COVID-19 outbreak is still challenging (Chevallier et al. 2020; Buchwitz et al. 2021). The change in the ΔCO₂/ΔCH₄ ratio at HAT associated with the COVID-19 outbreak is also very faint. Thus, the fact that a decrease in the ΔCO₂/ΔCH₄ ratio in February 2020 was observed at YON convinced us of the correctness of the previous result reported by Tohjima et al. (2020). This result also demonstrates the usefulness of utilizing adjacent monitoring sites like YON and HAT to confirm the detection of such faint signals as shown in this study. Both sites are located at the downwind side of the Asian continent that is the biggest CO₂ emitter in the world and hence the most important area to be monitored. Monitoring with observations at YON and HAT would continue to provide valid information on emission changes over the continent.

Data Availability Statement

The time series of the atmospheric CO₂ and CH₄ mole fractions at YON are available via the website of WDCGG (World Data Centre for Greenhouse Gases). (https://xml.kishou.go.jp/)

Supplements

Supplement 1 shows the frequency distributions of the standard deviations for the daytime and nighttime CO₂ data for YON. Supplement 2 shows the same figure as Fig. 3, but only nighttime data are used to calculate the monthly average of the ΔCO₂/ΔCH₄ ratios for YON. Supplement 3 shows the root mean square (RMS) of the differences of the monthly ΔCO₂/ΔCH₄ ratios between YON and HAT against the duration of the time window used for the calculation for YON.

Acknowledgments

We are grateful to many staff members of the Japan Meteorological Agency for their work in the long-term observations of atmospheric CO₂ and CH₄ at YON. This study was financially supported by funds provided by the Environment Research and Technology Development Fund (JPMEERF21S20802).

References

•  Bauwens,  M.,  S.  Compernolle,  T.  Stavrakou,  J.-F.  Müller,  J.  van Gent,  H.  Eskes,  P. F.  Levelt,  R.  van der A,  J. P.  Veefkind,  J.  Vlietinck,  H.  Yu, and  C.  Zehner, 2020: Impact of coronavirus outbreak on NO2 pollution assessed using TROPOMI and OMI observations. Geophys. Res. Lett., 47, e2020GL087978, doi:10.1029/2020GL087978.
•  Buchwitz,  M.,  M.  Reuter,  S.  Noël,  K.  Bramstedt,  O.  Schneising,  M.  Hilker,  B. F.  Andrade,  H.  Bovensmann,  J. P.  Burrows,  A.  Di Noia,  H.  Boesch,  L.  Wu,  J.  Landgraf,  I.  Aben,  C.  Retscher,  C. W.  O'Dell, and  D.  Crisp, 2021: Can a regional-scale reduction of atmospheirc CO₂ during the COVID-19 pandemic be detected from space? A case study for East China using satellite XCO₂ retrievals. Atmos. Meas. Tech., 14, 2141-2166.
•  Chevallier,  F.,  B.  Zheng,  G.  Broquet,  P.  Ciais,  Z.  Liu,  S. J.  Davis,  Z.  Deng,  Y.  Wang,  F.-M.  Bréon, and  C. W.  O'Dell, 2020: Local anomalies in the column-averaged dry air mole fractions of carbon dioxide across the globe during the first months of the coronavirus recession. Geophys. Res. Lett., 47, e2020GL090244, doi: 10.1029/2020GL090244.
•  Hirsch,  R. M., and  E. J.  Gilroy, 1984: Methods of fitting a straight line to data: Examples in water resources. J. Amer. Water Resour. Assoc., 20, 705-711.
•  Le,  T.,  Y.  Wang,  L.  Liu,  J.  Yang,  Y. L.  Yung,  G.  Li, and  J. H.  Seinfeld, 2020: Unexpected air pollution with marked emission reductions during the COVID-19 outbreak in China. Science, 369, 702-706.
•  Le Quéré,  C.,  R. B.  Jackson,  M. W.  Jones,  A. J. P.  Smith,  S.  Abernethy,  R. M.  Andrew,  A. J.  De-Gol,  D. R.  Willis,  Y.  Shan,  J. G.  Canadell,  P.  Friedlingstein,  F.  Creutzig, and  G. P.  Peters, 2020: Temporary reduction in daily global CO₂ emissions during the COVID-19 forced confinement. Nat. Climate Change, 10, 647-653.
•  Matsueda,  H.,  K.  Tsuboi,  S.  Takatsuji,  T.  Kawasaki,  M.  Nakamura,  K.  Saito,  A.  Takizawa,  K.  Dehara, and  S.  Hosokawa, 2018: Evaluation of a new methane calibration system at JMA for WCC Round Robin experiments. Pap. Meteor. Geophys., 67, 57-67.
•  Tohjima,  Y.,  T.  Machida,  M.  Utiyama,  M.  Katsumoto,  Y.  Fujinuma, and  S.  Maksyutov, 2002: Analysis and presentation of in situ atmospheric methane measurements from Cape Ochi-ishi and Hateruma Island. J. Geophys. Res., 107, ACH 8-1-ACH 8-11.
•  Tohjima,  Y.,  H.  Mukai,  S.  Hashimoto, and  P. K.  Patra, 2010: Increasing synoptic scale variability in atmospheric CO₂ at Hateruma Island associated with increasing East-Asian emissions. Atmos. Chem. Phys., 10, 453-462.
•  Tohjima,  Y.,  M.  Kubo,  C.  Minejima,  H.  Mukai,  H.  Tanimoto,  A.  Ganshin,  S.  Maksyutov,  K.  Katsumata,  T.  Machida, and  K.  Kita, 2014: Temporal changes in the emissions of CH₄ and CO from China estimated from CH₄/CO₂ and CO/CO₂ correlations observed at Hateruma Island. Atmos. Chem. Phys., 14, 1663-1677.
•  Tohjima,  Y.,  P. K.  Patra,  Y.  Niwa,  H.  Mukai,  M.  Sasakawa, and  T.  Machida, 2020: Detection of fossil-fuel CO₂ plummet in China due to COVID-19 by observation at Hateruma. Sci. Rep., 10, 18688, doi:1038/s41598-020-75763-6.
•  Tsutsumi,  Y.,  K.  Mori,  M.  Ikegami,  T.  Tashiro, and  K.  Tsuboi, 2006: Long-term trends of greenhouse gases in regional and background events observed during 1998–2004 at Yonagunijima located to the east of the Asian continent. Atmos. Environ., 40, 5868-5879.
•  Watanabe,  F.,  O.  Uchino,  Y.  Joo,  M.  Aono,  K.  Higashijima,  Y.  Hirano,  K.  Tsuboi, and  K.  Suda, 2000: Interannual variation of growth rate of atmospheric carbon dioxide concentration observed at the JMA's three monitoring stations: Large increase in concentration of atmospheric carbon dioxide in 1998. J. Meteor. Soc. Japan, 78, 673-682.
•  Zhang,  X.,  T.  Nakazawa,  M.  Ishizawa,  S.  Aoki,  S.  Nakaoka,  S.  Sugawara,  S.  Maksyutov,  T.  Saeki, and  T.  Hayasaka, 2007: Temporal variations of atmospheric carbon dioxide in the southernmost part of Japan. Tellus B, 59, 654-663.
•  Zhou,  L. X.,  D.  Kitzis, and  P. P.  Tans, 2009: Report of the fourth WMO Round-Robin reference gas intercomparison, 2002–2007. 14th WMO/IAEA Meeting of Experts on Carbon Dioxide, Other Greenhouse Gases and Related Tracers Measurement Techniques.  Laurila,  T. (ed.), GAW Rep. 186, WMO, Genova, Switzerland, 40-43.