Measurement report: Regional characteristics of seasonal and long- 1 term variations in greenhouse gases at Nainital, India and Comilla, 2 Bangladesh 3

12 Emissions of greenhouse gases (GHGs) from the Indian subcontinent have increased during the last 20 years along with 13 rapid economic growth, however, there remains a paucity of GHG measurements for policy relevant research. In northern 14 India and Bangladesh, agricultural activities are considered to play an important role on GHGs concentrations in the 15 atmosphere. We performed weekly air sampling at Nainital (NTL) in northern India and Comilla (CLA) in Bangladesh from 16 2006 and 2012, respectively. Air samples were analyzed for dry-air gas mole fractions of CO 2 , CH 4 , CO, H 2 , N 2 O, and SF 6 , 17 and carbon and oxygen isotopic ratios of CO 2 (δ 13 C-CO 2 and δ 18 O-CO 2 ). Regional characteristics of these components over 18 the Indo-Gangetic Plain are discussed compared to data from other Indian sites and Mauna Loa, Hawaii (MLO), which is 19 representative of marine background air. 20 We found that the CO 2 mole fraction at CLA had two seasonal minima in February‒March and September, 21 corresponding to crop cultivation activities that depend on regional climatic conditions. of δ 13 C-CO 2 at NTL (0.96 ± 275 0.16‰) and CLA (0.85 ± 0.19‰) were larger than that at CRI (approximately 0.6‰). These results suggested that the atmospheric CO 2 mole fraction of NTL and CLA were strongly influenced by photosynthesis of local plants in summer and their respiration in winter, and other anthropogenic emission which were moderated at the other sites by the influence of the oceanic air. Also, small episodic peaks of the atmospheric CO 2 mole fraction and isotopic ratio of δ 13 C-CO 2 of CLA at the beginning of each year was influenced by the biomass burning for heating in the close region, which is considered to be inland area from the site according to the air trajectory analysis.

minima in September, the carbon isotopic signature suggested that photosynthetic CO2 absorption by crops cultivated in each 23 season contributes differently to lower CO2 mole fractions at both sites. The CH4 mole fraction of NTL and CLA in  October showed high values (i.e., sometimes over 4,000 ppb at CLA) mainly due to the influence of CH4 emissions from the 25 paddy fields. High CH4 mole fractions sustained over months at CLA were a characteristic feature in the Indo-Gangetic Plain, 26 which were affected by the both local emission and air mass transport. The CO mole fractions at NTL were also high and 27 showed peaks in May and October, while CLA had much higher peaks in October-March due to the influence of human 28 activities such as emissions from biomass burning and brick production. The N2O mole fractions at NTL and CLA increased 29 in June-August and November-February, which coincided with the application of nitrogen fertilizer and the burning of 30 biomass such as the harvest residues and dung for domestic cooking. Based on H2 seasonal variation at both sites, it appeared 31 that the emissions in this region were related to biomass burning in addition to production from the reaction of OH and CH4. 32 The SF6 mole fraction was similar to that at MLO, suggesting that there were few anthropogenic SF6 emission sources in the 33 district. 34 The variability of CO2 growth rate at NTL was different from the variability in the CO2 growth rate at MLO, which is 35 more closely linked with the El Niño Southern Oscillation (ENSO). In addition, the growth rates of the CH4 and SF6 mole 36 fractions at NTL showed an anticorrelation with those at MLO, indicating that the frequency of southerly air masses strongly 37 influenced these mole fractions. These finding showed that rather large regional climatic conditions considerably controlled 38 interannual variations in GHGs, δ 13 C-CO2, and δ 18 O-CO2 through changes in precipitation and air mass. 39

Introduction 42
The mole fraction of many greenhouse gases (GHGs) in the atmosphere, including CO2, CH4, and N2O, has been 43 increasing worldwide in recent years. As for CO2, rapid increases in CO2 emissions from developing countries contribute 44 strongly to acceleration of the growth rate of its mole fraction (Friedlingstein et al., 2019). For instance, anthropogenic CO2 45 emission of India has increased in 2017 it reached to 2.45 GtCO2 yr -1 which was the third highest in the world (Muntean et al., 46 2018). Therefore, South Asian region must be important to evaluate GHG in the future. Patra et al. (2013) calculated the CO2 47 0.263‰ (IAEA, 1993, Kim et al., 2015. Additionally, corrections for N2O bias and δ 17 O-CO2 showed by Brand et al. (2010) 162 were made to obtain final isotope ratios. 163

Back trajectory analysis 184
To determine the sources of regional air masses affecting the stations (NTL and CLA), we calculated backward air 185 trajectories using the Meteorological Data Explorer (METEX) system (Zeng and Fujinuma, 2004)  (http://db.cger.nies.go.jp/metex/index.html). METEX uses three-dimensional wind speed (horizontal and vertical wind) 188 estimated from the European Centre for Medium Range Weather Forecast (ECMWF) analyses on a 0.5° × 0.5° mesh to 189 calculate 72-h trajectories. We use 1940m for NTL and 30m for CLA as the starting height. We referred the altitude data when 190 we evaluated the effects of GHGs emissions sources near the surface. 191 The ratio of air mass from south per year was calculated by the frequency of the air mass from south side of Indian 192 Ocean on the flask sampling date in each year with reference to the 72-h backward air trajectories data calculated by METEX. 193

Data analysis method for short-term and long-term 194
Mean values for every 10 days were calculated from the weekly data and were used to calculate the long-term trend 195 and smoothing fitting curve. Because sampling interval is not punctual and we sometimes had missing data, we decided to use 196 10 days average to calculate the trend curve. The value of the missing period was supplemented with an interpolated values 197 from the previous and following data of the missing period for calculating the continuous long-term trend and smoothing fitting 198 curve. 199 Long-term trends of the mole fractions were calculated based on the idea of Thoning et al. (1989) with a cut-off 200 frequency of 667 days (0.5472 cycles yr -1 ) for a FFT filter. The smoothing fitting curve was made for an FFT filter with a cut-201 off frequency of 50 days (7.3 cycles yr -1 ). 202 We defined and expressed seasonal component by a "Δ" term (e.g., ΔCO2) which was calculated by subtraction of the 203 long-term trend curve from 10 days mean of real data. Also, we defined and expressed short-term variations by a "d" term 204 (e.g., dCO2), which were characterized by the deviation of 10 days mean of real data from the smoothing fitting curve. Figure  205 2(c) shows how such components were calculated. Growth rates of mole fraction of observed gases were calculated using the 206 long-term trends. 207 3. Results and discussion 208

Overview of GHGs mole fractions at both sites 209
Basically, the air masses over the Indian subcontinent were transported from the Indian Ocean region during summer 210 (monsoon season) and from the inland during winter. Air mass trajectories are shown for our sampling sites and related sites 211 in Figure 3. In the case of anthropogenic GHGs, except CO2, their mole fractions at CLA generally showed relatively low 212 values when the air mass came from the ocean, while the mole fractions were relatively high when the air mass came from 213 inland. On the other hand, mole fractions of GHGs at NTL overall did not show relatively low values, even if the air mass 214 came from the Indian Ocean region (i.e., south-eastern wind) because the air mass from Indian Ocean was strongly affected 215 by local GHGs emissions while passing over the Indo-Gangetic Plain. However, the CO2 mole fraction changed not only due 216 to transport but also due to the photosynthetic sink strength of terrestrial ecosystems and cultivated crops. 217 Annual mean GHG mole fractions at NTL and CLA are summarized in Table 1 (Lin et al. 2015). Note that there is no data 220 for CLA in 2010, however the annual CO2 mole fraction at CLA is usually only 1-2 ppm higher than at NTL. This seemed to 221 be due to the influence of photosynthesis at both sites. Generally, the CO2 mole fractions at NTL and CLA decreased strongly 222 (typically twice a year) due to photosynthesis of local crops, making the annual CO2 mole fractions lower than at other sites 223 despite the likelihood that anthropogenic emission are high in this area. 224 On the other hand, the annual mean mole fractions of CH4, CO, H2, and N2O at NTL and CLA (Table 1) were almost at 225 the highest levels on the Indian subcontinent due to the influence of strong emission sources. For example, the annual mole 226 fractions of NTL and CLA were 50-470 ppb for CH4, 30-200 ppb for CO, and 0-5 ppb for N2O higher compared to other 227 Indian sites (e.g., CRI [Bhattacharya et al., 2009], HLE, PON, and PBL [Lin et al., 2015]). In this region, high CH4 and N2O 228 emissions were possible from paddy fields and cultivated areas. Also, much CO is considered to be produced by biomass 229 burning in this region. As for H2, the mole fraction at CLA was higher than those at other Indian sites, however, it was relatively 230 low at NTL compared to other sites such as CRI (Bhattacharya et al., 2009), PON, and PBL (Lin et al., 2015), but similar to 231 HLE, which is located on a higher mountain. In the case of the SF6 mole fraction, it has smaller regional differences, suggesting 232 there are no remarkable SF6 sources near the measurement sites. Below we describe in detail the characteristics of sources and 233 sinks of each component (CO2, δ 13 C-CO2, δ 18 O-CO2 CH4, CO, H2, N2O, and SF6) at NTL and CLA on the Indo-Gangetic Plain 234 in terms of seasonal variations, amplitudes, and growth rates. 235 3.2 CO2 and δ 13 C-CO2 236 3.2.1 CO2 mole fraction and growth rate variations 237 Figure 4 shows the time series of the atmospheric CO2 mole fraction and the isotopic ratio of δ 13 C-CO2 at our sampling 238 sites (NTL and CLA) together with data from CRI on the west coast of India and MLO in Hawaii. The CO2 mole fractions at 239 NTL and CLA in August-October were characteristically lower (approximate 10-20 ppm) than the mole fractions observed 240 at CRI and MLO. The CRI and MLO sites are representative of CO2 mole fractions in the Southern and Northern Hemisphere, 241 respectively, for the period of the southwest monsoon season (June-September). On the other hand, the δ 13 C-CO2 at NTL and 242 CLA were inversely correlated with the CO2 mole fractions, and generally the values at both sites were higher than at MLO 243 and CRI. 244 Air masses at NTL and CLA in August-October passed over the Indo-Gangetic Plain and the southeast area of India,245 respectively, while the air masses of CRI were transported from the Indian Ocean region (Fig. 3). Thus, it was suggested that 246 the air mass from the Indian Ocean in August-October prevailing over CRI was hardly influenced by anthropogenic emission 247 and photosynthesis over the Indian subcontinent, whereas CO2 mole fractions over NTL and CLA seemed to be influenced 248 during these season by the sources and sinks on the Indo-Gangetic Plain and the south/east areas of the Indian subcontinent. 249 Such transport characteristics must affect the annual average and growth rates in the CO2 molar ratio and δ 13 C-CO2 in addition 250 to their seasonal variations. 251 We show the CO2 growth rates observed at NTL, CLA, and MLO in Figure 5(a). Mean CO2 growth rate at NTL 252 (approximately 2.1 ppm yr -1 during 2007-2020) and CLA (approximately 2.9 ppm yr -1 during 2013-2020) were similar to 253 other sites (e.g., MLO). However, variations of the calculated growth rates were greater than those at MLO. The range was 0-254 5 ppm yr -1 in the case of NTL, and CLA had higher variability than NTL because local sink and source influences affected the 255 concentration more than remote sites such as MLO. In general, Pacific sites such as MLO and Japanese remote sites in the 256 Northern Hemisphere showed a relationship between CO2 growth rates and the ENSO index (e.g., Keeling, 1998). This 257 relationship is often explained from the viewpoint of a global temperature anomaly, which has a strong relationship with the 258 ENSO index. On the other hand, the variability at NTL has no associations with the variability in the CO2 growth rate at MLO 259 and the ENSO index ( Fig. 5[b]). Both growth rates seemed to be slightly inversely correlated with each other from 2007 to 260 2015. However, since then, similar relatively high growth rates have been observed for both sites around 2015-2016 and 2018-261 2019, indicating that overall, the CO2 growth rate at NTL is less correlated with the CO2 growth rate at MLO and the ENSO 262 index. 263 It is well known that the Indian Ocean Dipole controls meteorological conditions such as air mass transportation and 264 precipitation patterns on the Indian subcontinent (e.g., Saji et al., 1999, Ashok et al., 2004, Hong et al., 2008. Such changes 265 in regional climatic pattern could affect the CO2 uptake flux by plants in the surrounding area and the atmospheric movement, 266 leading to a change in the CO2 growth rate. However, we did not find a simple relationship between DMI and CO2 growth rate 267 at NTL (Fig. 5[b]). Here we have shown that the pattern of CO2 growth rate in this region is different from the global pattern 268 seen in places like MLO, but the relationship between local climatic factors and changes in CO2 sinks and emissions is likely 269 to be complex, and further study is needed to interpret the differences. 270  (Table 2) at NTL (22.1 ± 3.9 ppm) and CLA (20.3 ± 5.7 ppm) were much larger than those at other Indian 274 sites (CRI, 15 ppm; HLE, 8.2 ppm; PON, 7.6 ppm; PBL, 11.1 ppm). Also, the annual amplitudes of δ 13 C-CO2 at NTL (0.96 ± 275 0.16‰) and CLA (0.85 ± 0.19‰) were larger than that at CRI (approximately 0.6‰). These results suggested that the 276 atmospheric CO2 mole fraction of NTL and CLA were strongly influenced by photosynthesis of local plants in summer and 277 their respiration in winter, and other anthropogenic emission which were moderated at the other sites by the influence of the 278 oceanic air. Also, small episodic peaks of the atmospheric CO2 mole fraction and isotopic ratio of δ 13 C-CO2 of CLA at the 279 beginning of each year was influenced by the biomass burning for heating in the close region, which is considered to be inland 280 area from the site according to the air trajectory analysis. 281

Seasonal variation and its characteristics 271
As shown in Figure 4 (a) and (b) and Figure 6(b) and (d), the seasonal variation pattern at CLA has two lower seasons 282 in CO2 and two higher seasons in δ 13 C-CO2 in February-April and July-October. Similarly, in the case of NTL, we sometimes 283 observed relatively low mole fractions of CO2 in February-March and September, and higher δ 13 C-CO2. In general, in many 284 cases including at MLO, only a summer minimum CO2 mole fraction is observed, while a minimum in February-March is not 285 usually observed. 286 Twice-yearly decreases in the CO2 mole fraction have also been observed at several Indian sites such as Dehradun 287 masses at these sites must be mainly transported from the ocean or from areas other than the Indian subcontinent during these 296 periods. 297 The characteristic CO2 seasonal variation on the Indo-Gangetic Plain (including NTL and CLA) is very likely to be 298 related to CO2 uptake by regional vegetation. Generally, in the case of Uttar Pradesh state located in the center of the Indo-299 Gangetic Plain, rice and other summer plants (maize, millets, etc.) are planted mainly in June-July and harvested in October-300 November, while large areas of wheat are sown in October-December and harvested in March-April. Therefore, relatively 301 low CO2 mole fractions observed in those periods are considered to be due to CO2 uptake by plants cultivated in each season 302 near NTL. 303 In Bangladesh, rice, being the staple food, is cultivated three times a year in some regions. Usually rice is grown twice 304 (Aus and Amon rice) from April-October (including the monsoon season), however, often rice is also cultivated (Boro rice) in 305 the winter season from November-April (SID/MP, 2018). Other agricultural products include maize, jute, and vegetables in 306 the summer season, and small amount of wheat in the winter season. Therefore, we concluded that the observed lower CO2 307 mole fractions in July-October and February-March were influenced by CO2 uptake by local plants (mainly rice). Especially 308 at CLA, the lower mole fraction in February-March was clear and a strong contribution from CO2 uptake from Boro rice was 309 estimated. As another viewpoint on CO2 seasonal variation, we observed that the CO2 maximum in May was not so high, while 310 the CO2 mole fraction in December was higher. Because precipitation in Bangladesh is stronger than in the north Indian region, 311 the duration of rice cultivation over summertime is also longer than in north India. Therefore, the contribution of plant uptake 312 to the CO2 mole fraction in the atmosphere at CLA over the summer season is likely to be relatively large compared to that at 313

NTL. 314
Thus, the decreases in the CO2 mole fractions in February-March and September in NTL and CLA were estimated to 315 be caused by photosynthesis of plants cultivated in each season over the Indo-Gangetic Plain. NTL and CLA indicated this 316 more clearly compared with other Indian sites due to the proximity to the source region. Figure 7(a) shows the relationships 317 between the annual mean CO2 mole fraction and δ 13 C-CO2 in 2010 and 2012. The slope between the CO2 mole fraction and 318 δ 13 C-CO2 showed -0.050 and -0.054‰ ppm -1 which indicated that the spatial variability of the atmospheric CO2 mole fraction 319 (e.g., a lower mole fraction at NTL than at MLO and CRI) basically occurred due to CO2 exchange between the atmosphere 320 and terrestrial biosphere. 321 Furthermore, we examined the relationship of the CO2 mole fraction and carbon isotope ratio, because there are some 322 seasonal differences in the species cultivation. On the Indo-Gangetic Plain, rice (especially in Bangladesh) and wheat 323 (especially in North India), as C3 plants, are cultivated in January-March, while C4 plants (e.g., maize, sugarcane, sorghum 324 and Bajra (Pearl millet) in addition to rice are cultivated on the Indo-Gangetic Plain and in Bangladesh in June-September 325 (DAC/MA, 2015; SID/MP, 2018; DES/MAFW, 2019). We calculated the end member of the isotope value for absorbed CO2 326 by using intercept values of the "Keeling plot" between the reciprocal of the CO2 mole fraction and the ratio of δ 13 C-CO2 327 obtained from two continuous datasets of air samples, which has > 1 ppm difference in CO2 mole fraction and > 0.05‰ in 328 δ 13 C-CO2. Since in this study two datasets had 1-week intervals, we assumed that the difference in CO2 and δ 13 C between two 329 datasets would include broader influences of photosynthetic activities from relatively large areas on the Indo-Gangetic Plain. 330 We found that the intercept values of NTL and CLA showed differences in January-March and June-September ( Fig.  331   7[b]), which appeared to reflect the differences in the contributions of C3 and C4 plants in this region. In June-September, we 332 found relatively heavier intercept values at both NTL (-25.0 ± 2.4‰) and CLA (-23.5 ± 4.1‰), suggesting that C4 plants partly 333 contributed to the CO2 absorption (or emission) in this season, while in January-March, the end member showed -29.0 ± 4.3‰ 334 (NTL) and -28.3 ± 4.0‰ (CLA), which were similar to the general C3 plant (rice or wheat). If we assume the value for C4 plant 335 to be -12 to -14‰, the contributions of C4 plant in NTL and CLA were approximately 25 ± 5% and 31 ± 9%, respectively. 336 According to database (DAC/MA, 2015; SID/MP, 2018; DES/MAFW, 2019) for crops area in Uttar Pradesh district, the area's 337 ratio of C4 plants (e.g., maize and sugarcane) to C3 plants in the summer season was approximately 26% in 2012, which was 338 a similar proportion as estimated by the C isotope ratio. In the case of Bangladesh, despite there being no recent data reported, 339 according to data in 2008, the area for maize was approximately < 10% compared to the rice area. However, based on the 340 recent C isotope ratio, it appears likely that more maize has been cultivated. 341

δ 18 O-CO2 342
In general, δ 18 O-CO2 is related to that value of water in plants and soil, because oxygen atom of CO2 can be exchanged 343 with oxygen atom of H2O in plant and bacteria cells during photosynthesis and soil respiration. Plants and soil water mainly 344 originate from rainwater in the study region, however, in the case of the agricultural area, water is often introduced by irrigation 345 systems using river and groundwater. In many cases, photosynthesis produced relatively heavier δ 18 O-CO2 than soil respiration 346 because δ 18 O-H2O in plant becomes heavier than soil water due to plant transpiration. 2010]). During the rainy season, due to the so-called "amount effect", δ 18 O-H2O in rain will decrease with an increase in the 365 amount of precipitation (e.g., Rozanski et al., 1993). However, in the Indian region it has been reported that seasonal changes 366 in the origin of moisture strongly affected the δ 18 O-H2O (Sengupta andSarkar, 2006, Tanoue et al., 2018). In winter (i.e., when 367 there is less rain), moisture comes from the west or north. Therefore, the northern area of the Arabian Sea and the western land 368 area supply moisture, which has a higher δ 18 O-H2O. However, the air mass in the summer monsoon season (mainly June-369 September) comes from the southern part of the Arabian Sea and sometimes passes over the Bay of Bengal carrying much 370 moisture. The value of δ 18 O-H2O in the moisture in the air mass decreases with the process of raining along the air trajectory. 371 In the post-monsoon season (mainly October-December), some portion of moisture comes from the Pacific, Bay of Bengal, 372 and the inland area (Tanoue et al., 2018). 373 In the winter monsoon season (mainly February-May), δ 18 O-H2O in rain was reported to be approximately 0-1‰ (vs 374 VSMOW). During the winter monsoon season, there is little precipitation, so plant cultivation utilizes irrigation systems using 375 river and groundwater. River and groundwater usually show not so large seasonal variation in δ 18 O and have a close value to 376 the annual mean of δ 18 O-H2O in rain, such as -6 to -8‰ (Kumar et al., 2019). According to the variation of δ 18 O-CO2, in winter 377 its value was approximately 2‰ (vs VPDB-CO2; VPDB-CO2 scale is fairly close to the scale of CO2 equilibrated with 378 VSMOW water as mentioned in section 2.3), which was higher than that of rain and other water reservoirs, suggesting that 379 δ 18 O-H2O in plants and soil must become higher due to transpiration during dry and relatively warm conditions in winter. 380 Based on the fact that during the summer monsoon season, δ 18 O-CO2 decreased from 1 to -2‰ with a decrease of δ 18 O-381 H2O from 0 to -10 or -15‰ in the rain, the range of variation in δ 18 O-CO2 was approximately one third or one fifth that of rain. 382 Because land water may come from both rain and irrigation systems, the real ranges of δ 18 O in soil water and plant water are 383 likely to be smaller than in the case of rain only. Furthermore, because CO2 from soil respiration contributes more in the rainy 384 season, a balance between photosynthesis and respiration CO2 will, in general, have a small effect on the seasonal variation. 385 As for the annual trend of δ 18 O-CO2 shown in Figure 8 the monthly mean of the precipitation, the correlation coefficient (R 2 ) between the monthly mean δ 18 O-CO2 at CLA and the 396 monthly mean of precipitation increased to be 0.4 or 0.5). Therefore, the amount of precipitation partly contributes to the 397 regional level of δ 18 O-CO2. However, it must be influenced not only by precipitation but also by seasonal changes in air flow 398 patterns and rain systems, as explained above, as well as by the water reservoir situation, soil water content at that time, and 399 photosynthesis in the region. The CH4 mole fractions at NTL and CLA are illustrated in Figure 9(a). We detected high CH4 mole fractions at NTL 408 and CLA, where they sometimes exceeded 2,100 and 4,000 ppb, respectively, showing that the Indo-Gangetic Plain region 409 had relatively strong CH4 emissions. The seasonal amplitude of the CH4 mole fraction, especially at CLA (486 ± 225 ppb; 410 Table 2) was much larger than the those of other Indian sites such as NTL (114 ppb Reasonably good correlations were seen between short term components in variations of CH4 and CO in January-March, 443 April-June, and October-December. Ratios of dCH4 to dCO showed ranges such as 0.64-0.80 ppb ppb -1 in NTL and 1.85-444 1.98 ppb ppb -1 in CLA, as shown in Figure.  ratios of CH4 to CO in biomass burning such as crop residue burning, firewood burning, and biogas burning were 0.04-0.90 447 ppb ppb -1 . Therefore, the ratios observed in these seasons could suggest a strong influence on CH4 and CO emissions from 448 biomass burning (such as crop residue burning), despite the other large CH4 emissions such as paddy fields and waste treatment, 449 which will increase the ratio, especially at CLA in July-September. 450 As a result, it is evident that annual CH4 mole fractions at the sites used in this study on the Indo-Gangetic Plain are 451 enriched by various CH4 sources, depending on the season. Generally speaking, because April-June is a dry and hot season, 452 CH4 decomposition processes will proceed, decreasing its mole fraction at both sites. 453 The variability in the CH4 growth rate in the trend line at NTL was different to the variability at MLO (Fig. 9[b]), which 454 may be influenced by regional climatic condition, including the Indian Ocean Dipole. Because the frequency of air mass 455 transportation from the south increased if the Indian Ocean Dipole was often activated, the air mass passed over the Indo-456 Gangetic Plain (which has strong CH4 emissions), reaching NTL with a high CH4 mole fraction. The difference between the 457 variability in the CH4 growth rate between NTL and CLA may also be explained by the above hypothesis. If the frequency of 458 air mass transportation from the south increased by the activation of Indian Ocean Dipole (e.g., in 2015) because the air mass 459 was directly transported from the Indian Ocean with a relatively low CH4 mole fraction, the CH4 mole fraction at CLA would 460 become relatively low compared to a usual year ( Fig. 9[b]). On the other hand, as mentioned previously, in 2015-2017, even 461 in high Indian Ocean Dipole mode, Bangladesh had relatively high precipitation which could strengthen CH4 production from 462 rice paddy fields and other aquatic environments. This potential situation well-matched the high CH4 mole fraction in summer 463 and the high growth rate at CLA during 2016-2017. 464

CO 465
High annual CO mole fractions at both NTL and CLA (Table 1)  In addition, the seasonal amplitude of the CO mole fraction ( On the other hand, the annual mean CO mole fraction at NTL gradually decreased approximately by 50 ppb for 10 490 years (2006-2015; Fig. 10[a]). Especially, the monthly mean CO mole fraction in November of each year (i.e., the highest 491 level in the year) at NTL decreased by 120 ppb during that period. This suggests that the amount of harvest residues burnt 492 decreased, the ratio of incomplete combustion in car engines was improved, or the type of fossil fuel for cooking changed from 493 biofuel to natural gas. Such decreasing trends in the CO mole fraction level were also detected by Pandey et al. (2017) who 494 reported total-column CO levels during 2003-2014 over the Indo-Gangetic Plain. However, the CO mole fraction level at NTL 495 appeared to increase slightly from 2015. Although the reason for the increase is unclear from this study only, CO emissions 496 from car exhaust were recently estimated to have increased (EC-JRC/PBL, 2016). Therefore, further monitoring is important. 497 The trend in the CO mole fraction and its inter-annual variability at NTL was similar to those in CH4 at NTL (Fig. 9[b] 498 and Fig. 10[b]). The mole fractions of CO and CH4 at NTL tended to be slightly higher when the air mass passed over the 499 Indo-Gangetic Plain, where there are strong sources of both CO and CH4. In 2015 and 2017, a large positive Indian Dipole 500 Mode occurred, in addition to El Nino in 2015. Therefore, we observed more frequent southern winds, causing higher CH4 501 and CO mole fractions at NTL. However, at CLA, southern wind will decrease the mole fraction of CO. Thus, temporal 502 variations of both CO and CH4 mole fractions in both sites must be strongly controlled by meteorological conditions as well 503 as source strength. 504

H2 505
Mole fractions, growth rates, and seasonal variations of H2 at both sites are shown in Figure 11(a-d). It was found 506 that CLA, especially, showed a higher mole fraction than the other sites. Novelli et al. (1999) reported that the main sources 507 of H2 were combustion (fossil fuel combustion and biomass burning) and photochemical sources such as the oxidation of CH4 508 and non-CH4 hydrocarbons (NMHCs), which account for 90% of the total source. The other 10% is attributed to emissions 509 from volcanoes, oceans, and nitrogen fixation by legumes. Therefore, we have to assume that there are some emission sources 510 at CLA. 511 On the other hand, H2 is removed from the troposphere by reacting with OH and by deposition and oxidation at 512 surface soil. The amounts of sources and sinks for H2 in the global budget were estimated to be equal, resulting in a near-513 equilibrium state (Novelli et al., 1999). The strengths of H2 removal in the atmosphere over the Indian subcontinent do not 514 differ greatly by region according to Yashiro et al. (2011), whereas the strengths of H2 sources may differ by region (Price et 515 al., 2007). Lin et al. (2015) reported that H2 mole fractions at Indian sites were influenced by biomass burning and were 0-40 516 ppb higher than those at regional background sites (e.g., eastern Kazakhstan and central China). Figure 11 December at NTL, and the maximum in November-January and the minimum in June-August at CLA, which were different 519 from the averaged seasonal variation in the Northern Hemisphere, which showed the maximum in March-April and the 520 minimum in August-September (Novelli et al., 1999). 521 Because the burning of biomass (such as harvest residuals and dung) appeared to be actively carried out on the Indo-522 Gangetic Plain (including at NTL) during April-May and at CLA during November-February, H2 production must, therefore, 523 increase during these seasons. Furthermore, since higher CH4 mole fractions at NTL and CLA were observed during August-524 September and September-October due to strong paddy field emissions at those times, H2 production from CH4 degradation 525 can also increase. Figure 11(e) and (f) show short-term variable components (such as dCO and dH2, and dCH4, and dH2) at 526 both NTL and CLA during those periods, and that they had positive correlations. These figures may suggest some relationship 527 between H2 emission with biomass burning, and between photochemical reactions between OH and CH4, respectively. 528 Furthermore, the minimum H2 in June-August was influenced by a fresh air mass from the Indian Ocean which is only 529 minimally affected by anthropogenic emission. 530 As mentioned above, the H2 mole fraction level at CLA was higher than that at NTL. The amplitude of the seasonal 531 variation of the H2 mole fraction (Table 2) at CLA showed 70.4 ± 42.2 ppb, which was also larger than the amplitudes at other 532 Indian sites such as Nainital (50 ppb), CRI (50 ppb) (Bhattacharya et al., 2009), HLE (22 ppb), PON (16 ppb), and PBL (22 533 ppb) (Lin et al., 2015). These tendencies were consistent with the results of Price et al. (2007), which indicated a larger H2 534 emission area around the Eastern Indo-Gangetic Plain, such as at CLA, than on the Western Indian subcontinent. Thus, our 535 observation and previous studies both indicated that the Indian subcontinent had relatively strong H2 sources. 536 3.7 N2O 537 Garg et al. (2012) reported that the agricultural sector accounted for approximately 75% of the total N2O emission in India 538 in 2005, including around 49% from nitrogen fertilizer use. In particular, they reported that northern India (the Indo-Gangetic 539 Plain) has the highest N2O emission in India because nitrogen fertilizer was applied to extensive paddy fields, was denitrified, 540 and N2O was produced and emitted to the atmosphere. Ganesan et al. (2013) reported that the N2O mole fraction at Darjeeling 541 (north-eastern Indian site) was enhanced due to air mass transportation from the Indo-Gangetic Plain. The annual mean N2O 542 mole fraction at NTL (Table 1) appeared to be almost the same as at Darjeeling sites in North India and was higher than at 543 another two Indian sites (CRI [Bhattacharya et al., 2009] and HLE [Lin et al., 2015]) and at MLO (Fig. 12[a]). 544 Thompson et al. (2014) estimated that the N2O emissions of the Eastern Indo-Gangetic Plain, including CLA, were 545 higher than those of the Western Indo-Gangetic Plain. This is supported by our observation results that show that the N2O 546 annual mean mole fraction during 2013-2019 at CLA on the Eastern Indo-Gangetic Plain was 1-2 ppb higher than at NTL on 547 the Western Indo-Gangetic Plain (Table 1), and the seasonal amplitude of the N2O mole fraction (Table 2)  seemed to have relatively higher mole fractions than the sites in this study. As for the seasonal variation in the N2O mole 553 fraction at NTL, a higher mole fraction was seen in May-September (Fig. 12[c]). Generally, nitrogen fertilizer was frequently 554 applied to paddy fields in May-September in northern India. Delhi and reported that the flux increased immediately after the application of nitrogen fertilizer to the fields. Therefore, high 556 N2O levels and increases in the N2O mole fraction at NTL in May-September were influenced by the enhancement of the N2O 557 flux due to the denitrification of nitrogen fertilizer in paddy fields. 558 The N2O mole fraction at CLA increased in November-February (Fig. 12[d]) and such seasonal variation was almost 559 identical to the seasonal variation in CO at CLA. The seasonal component in the N2O mole fraction (ΔN2O = deviation of N2O 560 mole fraction from the long-term trend) at CLA showed positive correlations (R 2 = 0.81-0.88) with that of the CO mole fraction 561 (ΔCO) each year (Fig. 11[e]). Also, their ratio (ΔN2O/ΔCO) showed 0.013-0.015 ppb ppb -1 , which was same (0.015 ppb ppb -562 1 ) as the ratio of total N2O and total CO emissions in Bangladesh from the EDGAR v4.3.2 database (EC-JRC/PBL, 2016). 563 Although such seasonal variation is likely to be partly related to the lower mixing height in the winter season, variations in 564 N2O emission flux must affect the seasonal variations in the mole fraction. In general, the CO mole fraction was influenced by 565 biomass burning in this season. Because many inventory data showed that biomass burning produced both N2O and CO, N2O 566 may be affected partly emitted from biomass burning. However, the emission ratios of N2O to CO are fairly variable with an 567 dung burning is one of major N2O sources among many kinds of biomass burning in India, its contribution was also possible. 572 On the other hand, nitrification and denitrification processes of nitrogen fertilizer in rice paddy soil are considered 573 to be major causes of N2O emissions in this region (EDGAR v4.3.2), however, the emission rate appeared to have seasonal 574 variation. Related to the irrigation system, the N2O flux was thought to be larger in alternating wet and dry conditions than 575 After the summer monsoon (from October), the water level in the paddy field intermittently changed with the situation. 578 Therefore, relatively a higher N2O emission rate likely occurred during the winter season, when rice (Boro rice) was still grown, 579 enhancing the N2O mole fraction in the winter season. Further observations of high frequency variations of both N2O and CO 580 mole fractions will contribute towards precisely evaluating the N2O emission sources at this site. 581 The N2O growth rates at NTL and CLA were similar to that of MLO (Fig. 12[b]), however, the variations in the N2O 582 growth rate at both NTL and CLA were larger than that of MLO during 2016-2020. The variation in the N2O growth rate 583 showed a similar pattern to the growth rates of CO and H2 (Fig. 9[b] and Fig. 10[b]), indicating that the sources of these gases 584 had basically common characteristics. 585

SF6 586
SF6 is mainly emitted artificially from factories and urban areas (Olivier et al., 2005). Ganesan et al. (2013) reported 587 that the SF6 emission at Darjeeling (northeastern Indian site) was considerably weak. Our results also showed that SF6 mole 588 fractions at NTL and CLA were almost the same as the background SF6 mole fraction (e.g., MLO in Fig. 13[a] and other sites 589 such as HLE, PON, and PBL [Lin et al., 2015]). In addition, the annual amplitudes of the SF6 mole fraction at Indian sites 590 (HLE, PON, and PBL) were 0.15, 0.24, and 0.48 ppt, respectively, which were almost within the same range (0.15-0.23 ppt) 591 as at NTL and CLA (Table 2). These results suggested that there was no large SF6 source on the Indo-Gangetic Plain. 592 Figure 13(c) and (d) show that the seasonal variations of the SF6 mole fraction at NTL and CLA decreased in summer 593 (NTL: July, CLA: June-August), which was the same variation as those detected at PON and PBL (Lin et al., 2015). In the 594 summer season, air masses from the south via the Indian Ocean prevailed in the NTL and CLA regions, as shown in Figure 2. 595 Generally, the SF6 mole fraction in the Southern Hemisphere was lower than that in the Northern Hemisphere (Geller et al., 596 1997). Thus, the seasonal variation in the SF6 mole fraction was explained by the frequency of air mass transportation from 597 the south. 598 Figure 13(b) shows the interannual variability of the SF6 growth rate at NTL, CLA, and MLO and southern air mass 599 contribution at NTL and CLA. The variability in the SF6 growth rate at NTL was different to the variability at MLO, and in 600 fact we could see an anticorrelation between them. In the case of CLA, an anticorrelation was not so clear because of a relatively 601 shorter data record. The decrease in the growth rate at NTL seemed to have a relationship with the increase in the frequency 602 of southern air mass transportation. This indicated that the growth rate of the SF6 mole fraction at NTL may be controlled by 603 the regional climatic condition though the transportation process. Because SF6 had weaker sources in Northern India, the 604 variation in its trend could be explained more clearly by the influence of the air mass movements. 605 As mentioned above, anticorrelation in the growth rates between MLO and this region was also seen in CO2 and CH4. 606 Therefore, we must take into consideration the influence of the variation in large-scale atmospheric circulation to the GHG 607 mole fraction and trends in their growth rates in the Indian region. 608

Conclusions 609
We characterized GHGs and related gases over the Northern Indian region using air samples collected weekly at On the Indo-Gangetic Plain, rice, wheat, other cereals, and millet are cultivated in the respective seasons corresponding 616 to the change between wet and dry climatic conditions. Therefore, seasonal variations in the atmospheric CO2 mole fraction 617 were strongly influenced by the crop CO2 sink at that time. In general, low CO2 mole fractions in the winter season in the 618 Northern Hemisphere were not observed, however, we observed relatively lower mole fractions during January-March in this 619 region, especially at CLA. In Bangladesh, rice is grown even in the winter season. The δ 13 C-CO2 signature showed C3 plants 620 (e.g., rice and wheat) affected the CO2 mole fractions in the winter season, while in the summer season the δ 13 C-CO2 signature 621 showed C4 plants (corn, sugar cane etc.) contributed some portion. 622 The seasonal variations in δ 18 O-CO2 showed almost the same variation as that in the δ 18 O in local rain. Effects of the 623 amount of precipitation and the origin of moisture, appeared to affect δ 18 O in local rain and CO2. As a result, δ 18 O in CO2 was 624 affected by the climatic variation related to the amount of precipitation, which was enhanced during 2015-2017. These facts 625 are also consistent with the explanation that CO2 exchange by photosynthesis (and respiration) by land biomass strongly 626 affected CO2 seasonality in mole fraction. 627 At both sites, higher CH4 mole fractions were observed than were recorded at other Indian sites. Especially, higher 628 mole fractions than 4000 ppb were recorded at CLA, where rice paddy fields covered the area. Rice cultivation was one of 629 major emission sources in this region. Because CH4 production activities increased after rice planting, we observed the highest 630 peak in September-October at both sites and a small peak in spring at CLA. A large amount of precipitation during those 631 seasons is likely to have affected the CH4 production rate of rice paddy fields through soil anaerobic conditions and, as a result, 632 increased the atmospheric CH4 mole fraction. Air mass transport also influenced seasonal variation and the variability of its 633 growth rate. Beside emissions from rice paddy fields, we identified the relationship between biomass burning and the CH4 634 mole fraction in a season other than September-October, when biomass burning occurred frequently. In addition, enteric 635 fermentation and wastewater handling were large emission sources in this region. The large number of sources appeared to 636 increase the average CH4 mole fraction in this region. 637 CO was strongly related to biomass burning activities at both sites. The mole fraction was high in the dry season and 638 after crop harvesting. At CLA in winter, a higher mole fraction was observed together with a high N2O mole fraction, which 639 may suggest some link to biomass burning as a N2O source. The CO level gradually decreased throughout the observed period. 640 CO emissions must, therefore, be reduced by various technical progresses including automobile emission and industrial 641 combustion efficiency improvements. 642 We observed higher N2O levels in the crop season (i.e., the rainy season) from May-September at NTL, but much 643 higher levels in the winter season at CLA. N2O is known to be mainly emitted from soil though nitrogen fertilizer applications 644 to rice fields and crop lands in this region. However, for CLA, we estimated seasonal variations in the emission rate due to the 645 water level in the rice paddy field, because intermittent irrigation in winter generally produces more N2O than continuously 646 flooded conditions in the rainy season. 647 H2 showed some relationship to both CO and CH4 mole fractions. We found that CO had a good correlation with H2 in 648 the biomass burning season, indicating some H2 contribution from biomass burning. On the other hand, in the season when the 649 CH4 mole fraction was high, the H2 mole fraction was also relatively high compared to CH4, suggesting that chemical reactions 650 of CH4 and H2 may contribute some portion of the H2 mole fraction. 651 SF6 showed consistent mole fractions with other Indian sites. Seasonal variations were strongly related to the southern 652 air mass frequency, because the SF6 mole fraction in the southern region was relatively low. 653 We found that the interannual variabilities in CH4, SF6 and also partly in CO2, growth rates at NTL were anticorrelated 654 with those at MLO, which is located in the Pacific. Growth rates for many GHGs are known to be influenced by El Nino events 655 for many reasons (e.g., hot climate, dry conditions on a global scale). However, in the Indian region, growth rates of some 656 GHGs seemed to be more affected by the regional climate condition such as the Indian Ocean Dipole, which usually affects 657 air circulation and precipitation in the Indian region. In the case of CLA, although the data duration was insufficiently short, 658 growth rates of CO2, CH4, and SF6 changed differently from those at MLO, which could be partly explained by the climatic 659 variations due to the Indian Ocean Dipole. Because CLA is located relatively close to the ocean, sometimes the variation was 660 thought to be different from that at NTL. 661 These findings have not been reported previously. In this study, long-term records of GHGs data at NTL enabled a 662 long-term analysis. These findings suggested that the mole fractions of GHGs and their emissions on the Indian subcontinent 663 could change with climatic conditions in this region in the near future, in addition to changes in anthropogenic activities 664 relating to GHG emissions and countermeasure for the emissions. Therefore, long-term GHG monitoring should be continued 665 and the effectiveness of countermeasures for reducing GHG emissions on the Indian subcontinent, including the Indo-Gangetic 666 Plain, should be evaluated. 667

Data availability 668
We will add digital object identifiers (DOIs) to weekly flask sampling data of Nainital and Comilla and those data on 669 our website (http://db.cger.nies.go.jp/portal/geds/atmosphericAndOceanicMonitoring) by 2021. 670

Conflicts of Interest 671
The authors declare no conflicts of interest.    and (c) diagram of the calculation method for "Δ" term (e.g., ΔCO2) which was calculated by subtraction of the long-term 889 trend curve from 10 days mean of real data and "d" term (e.g., dCO2) which was characterized by the deviation of 10 days 890 mean of real data from the smoothing fitting curve.