Carbon dioxide variations in the upper troposphere and lower stratosphere from GOSAT TANSO–FTS TIR profile data

Carbon dioxide (CO2 ) variations in the upper troposphere and lower stratosphere (UT and LS; pressure = 300– 70 hPa) were investigated with profile data derived from the thermal infrared region (Band 4: 5.5–14.3 μm) of the Thermal And Near-infrared Sensor for carbon Observation (TANSO)–Fourier Transform Spectrometer (FTS) onboard the Greenhouse gas Observing SATellite (GOSAT). The vertical profile data (Level 2; version 1) of CO2 mixing ratios were obtained by bias correction using the in situ aircraft data. Data with sufficient sensitivity were used for the analysis by selecting values with 5 higher degrees of freedom. The analysis period is over four years from January 2010 to December 2013. The increase of CO2 mixing ratios obtained with the bias correction in the UT and LS was 1.8–2.4 ppmv year−1 from 2010–2013, which is consistent with a previous observational study and almost the same as for the lower troposphere (LT). The seasonal variations in the UT and LS show that a maximum peak exists one or two months after the peak in the LT, and has an amplitude of 4–5 ppmv, which is less than that in the LT. The inter-annual variations observed in the tropical UT appear 10 to be affected by ENSO events; higher (lower) CO2 mixing ratios were observed during La Niña (normal/El Niño) periods. Intra-seasonal variations over the Asian summer monsoon region were associated with both vertical and horizontal motions owing to deep convection and monsoonal anticyclonic circulation, respectively.

vertical mixing ratio distributions of CO 2 and CH 4 can be retrieved (Saitoh et al., 2009(Saitoh et al., , 2016. Given that the resolution of the spectra is 0.2 cm −1 higher than that of conventional satellite equipment, except for band 1, the atmospheric mixing ratio data are of high accuracy. For example, the accuracy is about 4 ppmv for XCO 2 (Yoshida et al., 2011). Moreover, the high-resolution spectra allow high-resolution vertical data to be obtained by TIR.
The TANSO-CAI sensor observes clouds, aerosols, and surface conditions during the day. From the TANSO-CAI observa-100 tional data, it can be determined whether there is cloud over a wide range, including the field of view (instant field of view, IFOV, which is approximately 10.5 km in diameter) of the TANSO-FTS sensor (Ishida and Nakajima, 2009;Ishida et al., 2011). If there is an aerosol or a cloud within the IFOV, then the retrieval processing of CO 2 , CH 4 , and H 2 O mixing ratios is not conducted on the SWIR and TIR spectra. However, optically thinner clouds are not detected by the TANSO-CAI; therefore, the accuracy of the TIR profile data has a bias with respect to cloud frequency, especially for thinner cloud. 105 This study used Level 2 vertical profile data (version 1; latest version) retrieved from the GOSAT TIR spectrum acquired over the clear sky for both day and night for four years from 1 January 2010 to 31 December 2013. The retrieval process for the vertical mixing ratio distribution of CO 2 is explained in detail in Saitoh et al. (2009Saitoh et al. ( , 2016. The pressure levels are 28 layers from the surface to 0.1 hPa. The present study defined the UT-LS region as 287.30-90.85 hPa, and the number of retrieved layers is 9 to 14 in Table 1 of Saitoh et al. (2016). The CO 2 mixing ratios data were averaged in a 5 • box for the longitude 110 and latitude. The algorithm was described in detail by Saitoh et al. (2016), and the altitude distribution data for 28 layers were specifically derived. The algorithm has a sensitivity peak in the upper troposphere.
2.2 Bias correction and data pre-processing Saitoh et al. (2016) determined the TIR CO 2 profile in the UT-LS region using Continuous Measurement Equipment (CME) observations of the Comprehensive Observation Network for TRace gases by AIrLiner (CONTRAIL) (Machida et al., 2008), 115 and the accuracy of the CO 2 measurements was less than 0.2 ppmv (Machida et al., 2008(Machida et al., , 2011. It was found that the CO 2 mixing ratios in the UT-LS had a maximum bias of 2.3 ppmv between the equator and 40 • N in the boreal summer and 2.4 ppmv between 20 • N and 40 • N in the boreal spring during the four years from 2010 to 2013. In summary, Saitoh et al. (2016) found that the CO 2 profile data had consistent negative biases of 1.0%-1.5% as compared with the CME CO 2 data in the LT and MT regions (736-287 hPa). Saitoh et al. (2017) found that the CO 2 profile data at the CME altitude range (approximately 11 120 km) was less than 0.5 ppmv at lower latitudes and 1.0 ppmv at middle and high latitudes. In addition, after applying the bias correction coefficient based on a comparison with CONTRAIL, both the trends and seasonal fluctuations showed good agreement with the Nonhydrostatic Icosahedral Atmospheric Model (NICAM)-Transport Model (TM) (Niwa et al., 2011(Niwa et al., , 2012 after inverse analysis. The present study conducted a bias correction at the retrieved layers (287.30-90.85 hPa) for each month Saitoh et al. (2016Saitoh et al. ( , 2017. In addition, data with a higher degree of freedom (DOF), with more than two times the standard deviation from the average of each month, were used for the analysis, because we intended to use data that were not constrained by a priori information.
Furthermore, anomalous data that were more than five times the standard deviation from the average of each month were not used for the analysis.

130
The trend was derived from the slope of a linear least-square fit defined by equation (1) from the monthly mean time-series of CO 2 mixing ratios from January 2010 to December 2013.
, where a and b are the slope and intercept, respectively. The trend was calculated at each pressure level and latitude at 5 • intervals. Because the unit of a is ppmv month −1 , the trend is a multiplied by 12 months.

135
The seasonal cycle was derived from a harmonic analysis (Fourier time-series analysis), similar to Nakazawa et al. (1991), of the monthly mean time-series subtracted from the linear trend defined by equation (1). Seasonal cycle components were extracted from the observed time-series data: , where A i and B i are the amplitude parameters of the sine and cosine terms, respectively. t is the elapsed time since January 140 2010. i is the order and k was set to three, which satisfactorily expresses the sub-seasonal and annual cycles.
For analysis of the inter-annual CO 2 variations, we used the anomaly data from the four-year monthly mean derived from equation (2), after subtracting the trend for each month and pressure and latitude.
Meteorological data, including the temperature and horizontal wind field, were obtained by ECMWF reanalysis version 5 (ERA5) ().  Fig.1 show data from latitudes where the data were bias-corrected.
The trend at 250 hPa was around 2.0 ppmv year −1 .
150 Table 1 shows a summary of the trend at several pressure levels during the analysis period.  (Sawa et al., 2012). For example, the growth rate at Mauna Loa was 2.22 ppmv year −1 during the analysis period (c.f. Thoning et al., 1989). The increase of 1.82-2.08 ppmv/yr in the UT-LS is similar to that near the surface (2.0-2.9 ppmv/yr) (WMO Greenhouse Gas Bulletin, 2012, 2013, 2014, and thus it appears that CO 2 is well mixed from the LT to UT. At 500 hPa (Fig.2d), a high CO 2 mixing ratio appears at higher latitudes between 40 • N and 65 • N in March and April. Prior to that, CO 2 mixing ratios increase from November. In the Northern Hemisphere, the peak month of the seasonal variation becomes later approaching the equator. At low latitudes in the Southern Hemisphere (around 30 • S), the peak is in September.

165
The seasonal amplitude is approximately 8 ppmv at higher latitudes and this decreases at lower latitudes. The peak of the seasonal cycle around the equator is June. These features are similar to that at the near-surface, which reflects seasonal controls on vegetation in the Northern Hemisphere (e.g. Keeling et al., 1996).
At 250 hPa, the maximum CO 2 mixing ratios occur in May at lower latitudes (30 • S to 30 • N), which is one or two months earlier than at mid-latitudes at 500 hPa. The CO 2 mixing ratio increases in January or February and decreases from June.

170
The seasonal amplitude at lower latitudes is approximately 5 ppmv, which is smaller than at lower altitudes. Furthermore, the amplitude at higher latitudes (north of 30 • N) is smaller than that at lower latitudes.
At 150 hPa, there are two maximum CO 2 mixing ratio peaks at 30 • N from April to May and at 20 • S from May to July. The southern peak at 20 • N extends southward with each month. The northern peak at 30 • N is independent of the southern peak and extends northward, but only until June. Similar to 250 hPa, the CO 2 mixing ratios at lower latitudes were higher than those 175 at higher latitudes.
At 100 hPa, the CO 2 mixing ratio peak at 15 • N occurred in May and June. The higher mixing ratio extended to the Southern Hemisphere across the equator, and it appears that this extension causes the seasonal cycle at lower latitudes in the Southern Hemisphere. The amplitude was approximately 4 ppmv. In summary, a seasonal CO 2 cycle was observed in the UT, but the amplitude was small. The peak was located at the lower latitudes and occurred earlier than the peak at the lower altitudes (150 180 and 250 hPa).

185
Seasonal variation is clearly evident at mid-and high latitudes in the Northern Hemisphere, especially below the MT (300 hPa). High mixing ratios of CO 2 occur at mid-and high latitudes and extend from the south up to the equator between March and May. After June, the CO 2 mixing ratios at mid-and high latitudes decreased until September, and then increased gradually from October to the following May. At lower latitudes, high CO 2 mixing ratios from the LT to UT were observed from April to June, which are related to the location of upward motion in a Hadley cell in the Northern Hemisphere (Fig. 3). 190 There are high CO 2 mixing ratios in the UT in the equatorial region, which is the Tropical Tropopause Layer (TTL); the pressure range is 150-70 hPa. This high CO 2 mixing ratio region is isolated from the MT throughout the year, except for during March-June, when high CO 2 mixing ratios extend from the LT to UT and lower latitudes (equator to 30 • N) due to upward vertical motion, such as Hadley cells. This feature is consistent with the seasonal cycle in the UT (250 hPa) shown in Figure   2c. 195 During the boreal winter (austral summer) season (December-February), the lowest mixing ratio was observed up to 200 hPa along the upward vertical velocity zone located to the south of the equator, which can reach a higher altitude than during the northern summer. During the boreal summer when the annual cycle of CO 2 is at a minimum, the upward motion was active over the Asian and American monsoon regions (15 • N-45 • N), where the CO 2 mixing ratio gradually decreased in the MT and UT.

200
Given that the source of CO 2 , particularly anthropogenic CO 2 , is at lower altitudes at mid-latitudes in the Northern Hemisphere, the high CO 2 mixing ratio is transported from mid-latitudes to lower latitudes in the LT by dynamical circulation, such as Ferrel and Hadley circulations. This suggests that the high CO 2 mixing ratio is transported from lower altitudes during the boreal spring to early summer, and remains in the equatorial region where there are no anthropogenic sources or sink of CO 2 .
Furthermore, the CO 2 mixing ratios in the equatorial region of the UT reached a minimum around September, which means 205 that the aforementioned transport from mid-latitudes to lower latitudes by general circulation, such as Ferrel and Hadley circulations, takes 1-2 months. In the LS above the tropopause at approximately 350 K, the CO 2 mixing ratio decreased gradually with increasing altitude. In the Northern Hemisphere, the CO 2 mixing ratio vertical gradient varied with season. The gradient in the Southern Hemisphere was steeper than in the Northern Hemisphere and, for example, the CO 2 mixing ratio decreased from 385 ppmv at 250 hPa to 370 (375) ppmv at 100 hPa at 60 • S (60 • N) in January. This suggests that the CO 2 mixing ratio 210 in the UT and LS depends on vertical and horizontal transport in the troposphere. Figure 4 shows a time-latitude plot of CO 2 mixing ratios at 100, 150, 250, and 500 hPa from January 2010 to December 2013, using de-trended data at each latitude.

Year-to-year variation
At 500 hPa, the seasonal variation at northern high latitudes (50 • N) is large, as shown in Fig. 2. The highest CO 2 mixing 215 ratio was observed in April at high latitudes and extended to lower latitudes in each year. In April 2013, this extension was larger than for the other years. Note that the bias correction was not applied to the data at high latitudes and, therefore, the bias remains at high latitudes. The minimum CO 2 mixing ratios were located at latitudes higher than 60 • N and at around 15 • N from June to November. The southern minimum extended to the Southern Hemisphere from September to April. In 2011, the minimum CO 2 mixing ratio near the equator was larger than for the other years. 2012 were normal years in terms of ENSO and, therefore, the upward motion was weaker than in La Niña and El Niño years (e.g. Oort and Yienger, 1996;Yun et al., 2021). The high CO 2 mixing ratios extended to higher latitudes in both hemispheres, however, the northern maximum extended further than the southern maximum from April to June.
At 150 hPa, two maximum CO 2 mixing ratios were located at 35 • N and the lower latitudes of 15 • N/S. The southern maximum shows larger year-to-year variations than the northern maximum. The peak at 35 • N occurred in April, which suggests 230 that the high mixing ratio at 150 hPa could have resulted from transport from the LT at mid-latitudes. The 150 hPa pressure level corresponds to the base of the TTL at lower latitudes, and the deep convection zone in the tropics.
At 100 hPa, the higher mixing ratios are located at 15 • N in May-June each year and extend to the Southern Hemisphere in June, and also the mid-latitudes in the Northern Hemisphere. The extension to higher latitudes was more prominent in the Northern Hemisphere than in the Southern Hemisphere. The 100 hPa pressure level in the tropics is in the TTL (within 25 • N/S), 235 but is located in the LS in extratropical regions. This suggests that high CO 2 mixing ratios gradually enter the LS at 100 hPa during the boreal spring and early summer, and the seasonal cycle is then influenced by dynamic processes (i.e., vertical transport due to deep convection associated with monsoons), because CO 2 is a long-lived species in the UT and stratosphere. Given that the zonal averaged vertical wind in the tropics was weaker during the La Niña period than in normal years and El Niño periods (not shown), it is possible that air with higher CO 2 mixing ratios preferred to remain in the MT and, as a result, 250 there was less air transport to the UT-LS during the La Niña period. This again shows that the CO 2 mixing ratio in the UT-LS depends on the magnitude of upward motion.

Intra-seasonal variation
The Asian summer monsoon region is an important area of stratospheric and tropospheric (ST) exchange processes. In particular, the intra-seasonal variations over this region are active and reflect the seasonal effects of the Asian monsoon from June 255 to September. Xiong et al. (2009) found that the strong transport of atmospheric pollutants from LT to UT in Asian during the monsoon season by AIRS CH 4 observation. The intra-seasonal variations of trace gases have been investigated to understand ST processes (e.g. Park et al., 2007Park et al., , 2008Park et al., , 2009. Intra-seasonal variations in carbon monoxide (CO) as a tropospheric tracer and ozone and HCN as a stratospheric tracer both reflect the effects of vertical transport via deep convection and horizontal advection caused by anticyclones associated with the Asian monsoon.

260
The intra-seasonal variations of CO 2 mixing ratios varied from year-to-year, especially during the boreal spring to autumn seasons. Figure 6 shows the time-latitude section of CO 2 mixing ratios at 250 hPa over the Asian summer monsoon region averaged between 60 • E and 120 • E from the equator to 45 • N from 1 April to 31 October of each year. The data were de-trended and are a seven-day running average.
There are CO 2 mixing ratio variations over durations of a few weeks in the UT, and high mixing ratios were observed from 265 May to June, especially in May 2010 and May-June in 2013. Given that the high CO 2 mixing ratios were transported from the LT by deep convection, as evident from the seasonal variation, it appears that the intra-seasonal variation was also related to deep convection. After June, the CO 2 mixing ratios started to decrease gradually from high latitudes, except in August 2012. Figure 7 shows a horizontal section of CO 2 mixing ratios at 250 hPa from 16 June to 1 September 2012, with 15-day intervals obtained from the 14-day mean. The high mixing ratios of CO 2 were mainly located on the southwestern side of the Tibetan

270
Plateau high (i.e., the Asian monsoon high) in June ( Fig.7(a)). The high mixing ratio in the UT was caused by vertical transport due to deep convection associated with the Asian Monsoon. The maximum mixing ratio area moved northward of the Tibetan Plateau high in July when the Tibetan Plateau high extended to the east and strengthened ( Fig.7(b,c)). At the same time, the CO 2 mixing ratio on the eastern side of the study area decreased gradually. There is an east-west change in mixing ratio.
In August, the high-mixing ratio area (>393 ppmv) was diminished, and lower CO 2 mixing ratios were transported hori-275 zontally from the northeastern part of the study area to lower latitudes along the Asian monsoon high. In September when the Tibetan Plateau high weakened, the CO 2 mixing ratio recovered in the western part of the study area. CO 2 mixing ratios on an intra-seasonal timescale are influenced by vertical transport due to deep convection and horizontal advection caused by horizontal wind fields in the UT.

Summary and conclusions
280 CO 2 mixing ratio variations in the UT and LS were investigated with the GOSAT TANSO-FTS TIR Level 2 dataset (version 1) on intra-seasonal to inter-annual timescales. The data were bias-corrected based on Saitoh et al. (2016Saitoh et al. ( , 2017 and excluded smaller DOF values. The level 2 data were converted to a 5 • N box grid and monthly mean values, except for the intra-seasonal analysis. The increase of CO 2 in the UT-LS was 1.8-2.4 ppmv year −1 , which is consistent with that obtained from in situ aircraft-285 obtained data by CONTRIAL (e.g., Machida et al., 2008). The increase in the tropical Southern Hemisphere in the MT and UT (250 and 500 hPa) was slightly (approximately 0.5 ppmv year −1 ) larger than that in the Northern Hemisphere.
In terms of the seasonal cycle, the maximum CO 2 mixing ratio was in May-June in northern mid-latitudes, and one or two months later than in the LT and MT. This suggests that vertical transport and mixing between the LT and UT occurs on a 9 https://doi.org/10.5194/acp-2022-46 Preprint. Discussion started: 18 February 2022 c Author(s) 2022. CC BY 4.0 License.
timescale of one or two months. The amplitude of the seasonal cycle in the UT was 4-6 ppmv, which is ≤50% of that in the LT 290 and MT. This is due to the absolute mixing ratio decreasing with altitude and, to a lesser extent, mixing with a low CO 2 mixing ratio air mass.
Inter-annual variations were clearly evident in the tropics, and appear to be affected by ENSO events. At 150 hPa, higher (lower) CO 2 mixing ratios occurred during La Niña (normal/El Niño) periods. This suggests that the higher CO 2 mixing ratios in the UT during the La Niña periods were induced by weaker vertical transport than during normal/El Niño periods.

295
In terms of intra-seasonal variations, CO 2 mixing ratios over the Asian summer monsoon region were variable on a timescale of a few weeks, which reflect vertical and horizontal transport owing to deep convection and monsoonal anticyclonic circulation, respectively. Vertical transport increased the CO 2 mixing ratios in the upper layer for short durations (a few days) over a relatively small area. Horizontal transport associated with the Tibetan Plateau anticyclone caused the CO 2 mixing ratios to vary between the UT and LS on a sub-continental scale.   same as (a), but for the 14-day averages from 1, July, 16 July, 1 August, 16 August, and 1 September, respectively.