Diurnal and seasonal variability of CO2 and CH4 concentration in a semi-urban environment of western India

Amongst all the anthropogenically produced greenhouse gases (GHGs), carbon dioxide (CO2) and methane (CH4) are the most important, owing to their maximum contribution to the net radiative forcing of the Earth. India is undergoing rapid economic development, where fossil fuel emissions have increased drastically in the last three decades. Apart from the anthropogenic activities, the GHGs dynamics in India are governed by the biospheric process and monsoon circulation; however, these aspects are not well addressed yet. Towards this, we have measured CO2 and CH4 concentration at Sinhagad, located on the Western Ghats in peninsular India. The average concentrations of CO2 and CH4 observed during the study period are 406.05 ± 6.36 and 1.97 ± 0.07 ppm (µ ± 1σ), respectively. They also exhibit significant seasonal variabilities at this site. CH4 (CO2) attains its minimum concentration during monsoon (post-monsoon), whereas CO2 (CH4) reaches its maximum concentration during pre-monsoon (post-monsoon). CO2 poses significant diurnal variations in monsoon and post-monsoon. However, CH4 exhibits a dual-peak like pattern in pre-monsoon. The study suggests that the GHG dynamics in the western region of India are significantly influenced by monsoon circulation, especially during the summer season.


Study area
The study area (Sinhagad; denoted as sng: 18° 21′ N, 73° 45′ E, 1600 m above msl) is a semi-urban location in the Western Ghats, India. This region is positioned at a distance of 30 km south-west from the city of Pune and 200 km east from the coastline of the Arabian Sea in Maharashtra, India. The basic climatology is presented in Fig. 1. The outgoing longwave radiations (OLR), on a monthly scale for the summer and winter seasons, are shown in Fig. 1a,b, respectively. The corresponding circulations are depicted by arrows. The windrose diagrams for CO 2 and CH 4 are also shown (Fig. 1c,d).

Results
Seasonal variation of CO 2 and CH 4 . Figure 2a,c shows the monthly mean and standard deviation (SD; shaded region) of CO 2 and CH 4 concentrations, respectively. The annual mean concentration of CO 2 is 406.05 ± 6.36 (µ ± 1σ) ppm. CO 2 is maximum (427.2) in May-2015 and a minimum (399) in September-2014. This leads to a seasonal amplitude of ~ 28 ppm. A comparison of the seasonal amplitude of other sites, global (Seychelles-sey, Mauna Loa-mlo), and Indian sites (Kaziranga-knp, Ahmedabad-amd, Shadnagar-sad, Cabo de Rama-cri), are shown in Fig. 2b. sey and mlo data are taken from ESRL-NOAA, knp data is obtained under Metflux India 25 project. amd and sad seasonality are taken from Chandra et al. 22 and Sreenivas et al. 26 , cri data is taken from World Data Centre for Greenhouse Gases (WDCGG). The global sites sey and mlo are mostly oceanic, hence possess smaller seasonal variation. In contrast, knp is a forest site of north-east India. It shows a larger seasonality of ~ 25 ppm, with a minimum during pre-monsoon and post-monsoon (Fig. 2a). Ahmedabad is an urban site in western India and has a CO 2 seasonality of 26 ppm. Shadnagar is a semi-urban site in central India with seasonality of 16 ppm. cri is a coastal region on the west coast of India. The mean seasonal amplitude of cri is 20 ppm with a minimum in monsoon and maximum in February-March. Among all these sites, sng shows maximum seasonal amplitude with pre-monsoon maximum and post-monsoon minima. Mean values of CO 2 for different seasons are 403.34 ± 5.71, 402.87 ± 6.03, 409.72 ± 4.33, 417.06 ± 5.11 ppm during the monsoon, postmonsoon, winter, and pre-monsoon, respectively. Mean CO 2 increases about 6.85 ppm from post-monsoon to winter and again increases about 7.34 ppm from winter to pre-monsoon.
The annual mean value of CH 4 over the study region is 1.97 ± 0.07 (µ ± 1σ) ppm. CH 4 concentration is minimum (1.863 ppm) in July-2014 and maximum (2.037 ppm) in October-2014 (Fig. 2c). The seasonal pattern over cri is very similar to the sng. The sng CH 4 shows 2.29 times and 3.62 times more seasonality than global sites sey and mlo (Fig. 2d). Whereas sad shows more seasonal amplitude of ~ 240 ppb than sng (~ 174 ppb). While cri seasonal amplitude, 168 ppb, is very close to sng seasonal amplitude, 174 ppb. The average values of CH 4 in different seasons are 1.903 ± 0.0412, 2.024 ± 0.0567, 1.995 ± 0.0629, 1.966 ± 0.0466 ppm in monsoon, post-monsoon, winter, and pre-monsoon, respectively. These clusters are mean trajectories of the air mass. Their percentage contribution to the total, calculated for different seasons over the study period at surface level, is presented in Fig. 3a-d. Figure 3a reveals that sng receives almost 84% of wind from the Arabian sea due to south-west monsoon flow. During the post-monsoon season, the wind blows from the Indian sub-continent. Therefore, the post-monsoon wind carries the contaminated air from the continental region to the sng site. During pre-monsoon time, sng receives 50% air mass from the Arabian Sea and 50% from the Indian continent. So, the observed maximum CO 2 concentration during pre-monsoon may be a local phenomenon, not a large scale transport.
Influence of vegetation. The normalised difference vegetation index (NDVI) is widely used as an index of vegetation cover of a given region 27,28 . We have plotted CO 2 and CH 4 as well as NDVI to investigate their relationship. The monthly climatology of CO 2 , CH 4, and NDVI are shown in Fig. 4a,b. Additionally, monthly climatology (2000-2010) of sector-wise CH 4 emission from Carbon-Tracker (CT) is plotted in Supplementary Fig. 3. It is quite clear that agriculture and waste management practices are the dominant sector of CH 4 over the study region. Hence, the monthly-climatology of CH 4 emission from agriculture and waste is also shown in Fig. 4b. Moreover, fossil fuel and biospheric emission of CO 2 and their residual is also plotted in Fig. 4a.
An inverse correlation is found between CO 2 and NDVI (Fig. 4a). NDVI time series reveals the growth of vegetation starts from the monsoon. Also, the growth rate is higher during the monsoon season than the non-monsoon season. The NDVI data clearly shows an enhanced vegetation cover from August and a concurrent decrease of CO 2 in our study region. Increased vegetation cover increases the rate of photosynthesis, which helps in decreasing CO 2 concentration. Further, NDVI reduces from post-monsoon to winter and pre-monsoon months, and CO 2 concentration consequently rises. This result is also supported by Sreenivas et al. 26 , who found a negative correlation between NDVI and CO 2 concentration at sad for 2014. Moreover, residual flux (biosphere + fossil) is  www.nature.com/scientificreports/ high positive (positive sign denotes CO 2 added to the atmosphere) during June-July-August, though atmospheric CO 2 concentration is low (Fig. 4a). CH 4 emission from the agriculture and waste (AW) sector of CT consists of enteric fermentation, animal waste management, wastewater and landfills, and rice agriculture. Emission from the AW sector is high during monsoon. The co-occurrence of high NDVI and AW sector emission suggest that rice agriculture is a dominant part of AW sector emission. In comparison, low surface CH 4 concentration is observed in monsoon.

Influence of planetary boundary layer (PBL).
The planetary boundary layer (PBL) is the lowermost layer of the troposphere, where temperature and wind speed plays an essential role in its height variation. The boundary layer can mix the GHG emitted at the ground level up to a certain height and reduce its concentration near the ground. So, seasonal changes in the boundary layer may affect the ground concentration of GHGs. Monthly PBLH is computed by averaging the hourly data and compared with CO 2 and CH 4 concentrations. Monthly PBLH is observed to be minimum (maximum) during the monsoon (pre-monsoon). Seasonal PBLH during monsoon, post-monsoon, winter, and pre-monsoon is 754. 8, 1136.45, 1213.72, and 1420.08 m, respectively. The influence of PBLH on CO 2 and CH 4 is shown in Supplementary Fig. 2a We find two cases: Effect of meteorology in different seasons. A biplot analysis is carried out for each season to identify the interdependency of several meteorological parameters such as wind speed (WIND), wind direction (dir), www.nature.com/scientificreports/ outgoing longwave radiation (olr), planetary boundary layer (PBL), 2 m-air temperature (t2m), soil temperature in layer 0-7 cm (stl1) and soil temperature in layer 7-28 cm (stl2) with GHGs ( Fig. 5a-d). In the two-dimensional space of two leading principal components, we used the biplot technique 29 to describe the PCA result. The two axes in the biplot represent the first two principal components, and the arrow vectors describe the variables in this space. Supplementary Fig. 8a-d shows the scree plot for monsoon, post-monsoon, winter, and pre-monsoon, respectively. Scree plot is the plot of eigenvalues organised from largest to smallest. Here scree plot is shown in terms of the percentage of explained variance. It is to be noted that in each season, the first two PCs (PC1 and PC2) are dominant components; hence, biplot analysis is carried out for each season to identify the interdependency of several meteorological parameters and GHGs. The length of an arrow represents the variance, and the cosine between two arrows represents the linear correlation between the two variables. All the variables are scaled to unit variance before performing PCA. The variables that are better explained by the two principal components will be longer and closer to the unit circle. Acute and obtuse angle represents positive and negative correlation, respectively, while a right angle implies a lack of correlation. An anti-correlation between CH 4 and wind speed (Fig. 5a) is found in monsoon. The wind rose diagram ( Fig. 1d) also supports this finding. The prevailing south-westerly wind in monsoon is associated with low CH 4 values in the wind rose diagram. A positive correlation between CO 2 and wind speed is found. This interplay between CO 2 and wind is discussed further in the following section. The association of CO 2 , CH 4 with wind is reduced in post-monsoon (Fig. 5b). While a positive correlation between CO 2 and CH 4 is evident in the winter months (Fig. 5c). The correlation coefficient value between CO 2 and CH 4 in winter is 0.52 (n = 7108). The close association of CH 4 -PBL and CO 2 -soil temperature (both layers 1 and 2) is the dominant feature in pre-monsoon (Fig. 5d).
Influence of prevailing meteorology. Correlation coefficients (R) between wind speed and CO 2 (R CO2 ) during monsoon, post-monsoon, winter and pre-monsoon are 0.51 (n = 118), 0.15, − 0.02 and − 0.28, while for CH 4 are (R CH4 ) − 0.57 (n = 118), − 0.3, − 0.02 and − 0.27 respectively. A good inverse correlation between GHG and wind speed suggests that with an increase in wind speeds, GHG concentrations would decrease. In contrast, a weaker correlation would suggest regional/local transport plays some role 30,31 . Strong wind, especially during the monsoon season ( Supplementary Fig. 1c) is likely to dilute the GHG concentration. This is validated for A strong negative correlation between CH 4 and wind presents dilution of CH 4 due to intrusion of southern hemispheric clean air with a strong south-westerly wind of monsoon, schematically shown in Fig. 1a. Anthropogenic signature on GHG's probability distribution. To investigate the anthropogenic and biospheric signature on GHGs, we have partitioned the CO 2 and CH 4 concentration for the day (07:00-18:00 LT) and night hours (20:00-06:00 LT) for the entire study period. Supplementary Fig. 4a,b shows the probability distribution (PD) of CH 4 and CO 2 concentrations during the daytime and nighttime, respectively. Supplementary Fig. 4b shows that the PD of CO 2 is narrow (broad) during the night (day) time. Mean daytime and nighttime CO 2 concentrations are 404.6 ± 7.8 ppm (µ ± 1σ) and 407.42 ± 5.93 ppm, respectively. On the other hand, the CH 4 concentration in the daytime and nighttime are practically the same. The respective mean values are 1.974 ± 0.078 ppm and 1.968 ± 0.07 ppm. We have also calculated the skewness (S) and kurtosis (K) of these distributions. The lower skewness S CO 2 = 0.04 for nighttime distributions than that of the daytime distribution S CO 2 = 0.16 implies that the nighttime distribution is more symmetric. The same is the case for CH 4 , for which the values are 0.37 and 0.97, respectively. This means the nighttime distributions are more constrained. For CO 2 , the kurtosis values for both day (0.52) and nighttime (1.20) are much lower than those obtained for a normally distributed curve, which is 3. This may imply that the extreme values are less relative to the normally distributed curve, but compared to daytime, the nighttime emissions are characterised by a slightly more number of extreme values. However, CH 4 shows the opposite behaviour, since the kurtosis value for daytime (2.7) is higher than that of the nighttime (− 0.24).
The probability distribution of CO 2 and CH 4 of day and nighttime data has also been carried out for different seasons. Supplementary Figs. 5a-d and 6a-d show the results. As found earlier, the monsoon season daytime PD www.nature.com/scientificreports/ is characterised by a broad peak, but the nighttime PD is relatively narrow. The nighttime mean (Supplementary Fig. 5a) is right-shifted, as there is practically no sink of CO 2 . The post-monsoon season shows a broader spectrum for both the period ( Supplementary Fig. 5b), indicating an increase in the nighttime source of CO 2 . The PDs for the winter (DJF) and the pre-monsoon season (MAM) are quite broad, and they show similar characteristics ( Supplementary Fig. 5c,d). Another feature of these distributions is the range of daytime CO 2 : the monsoon season has a range of 385-410 ppm, and the post-monsoon season 385-415 ppm. At the same time, the winter season shows a range of 402-425 ppm and the pre-monsoon season 405-435 ppm. Throughout the monsoon, the mean CO 2 concentration is 400.22 ± 5.48 ppm during the day, whereas an elevated CO 2 level, 406.57 (~ 6.35 ppm more than daytime) with low SD, is a vital feature of the nighttime variability (Table 1). This difference is also noticeable through the post-monsoon (ON), but the difference of mean day and night CO 2 concentration gets decreased. During the winter and pre-monsoon (DJF and MAM) the difference during the day and night CO 2 concentration almost vanishes.  www.nature.com/scientificreports/ In comparison, CH 4 does not show any significant daytime and nighttime variation in most seasons except winter. Figure 4b and Table 1 reveal a significant seasonal variation of CH 4 , but day-night variation in intraseasonal timescale existed only in winter. Wintertime day and night mean CH 4 concentration differs by 15 ppb. High mean CH 4 in daytime indicates the source, and higher SD represents diversity in source processes of CH 4 than night. This is also reflected in the S and K values of methane; the daytime values are high (S = 2.58, K = 11.33) for the winter season (DJF). Similarly, the pre-monsoon season (MAM) also shows relatively higher values (S = 2.28, K = 9.90). This means that methane concentrations in this region remain high from November through March due to enhanced emission and/or reduced loss due to the reduction in the OH radical 32 . Figure 6a-d show the diurnal cycle of CO 2 and CH 4 over the sng site averaged over a seasonal cycle. During the monsoon season (Fig. 6a), the diurnal pattern of CO 2 remains high in the early morning and then steadily decreases due to increased photosynthetic activity and becomes minimum around 13:00 LT. In the post-monsoon season (Fig. 6b), the minimum value is shifted to 10:00 LT. For the winter season and pre-monsoon (Fig. 6c,d), the patterns are very different; the maximum and minimum values are not well defined, and the diurnal pattern is somewhat linear. A large deviation from the monsoonal-pattern during the winter and pre-monsoon strongly indicates a weakening biospheric role and increased anthropogenic activities driving the diurnal behaviour of CO 2 concentration in these seasons. On the other hand, the diurnal pattern of CH 4 during the monsoon is not well defined. The patterns, however, are quite different for the other seasons, as illustrated in Fig. 6b-d; the minimum in the early hours and the maxima around 10:00 LT. The pre-monsoon season also gets second maxima around 19:00 LT. Figure 6a-d shows the seasonal variation of the diurnally averaged Planetary Boundary Layer Height (PBLH) in association with CH 4 and CO 2, respectively. Table 2 shows the amplitude, i.e., the difference between the diurnal minima and the maxima for different seasons. The table indicates that the diurnal variation of CO 2 is low during the winter or pre-monsoon time (1.98 and 2.75 ppm, respectively). The variability is increased during post-monsoon (4.1 ppm) and obtains maximum amplitude (10.01 ppm) during the monsoon. Moreover, it is noted that PBL height attains its maximum value around 14:00-15:00 LT for almost every season while the time of lowest CO 2 is different for different seasons. CO 2 reaches a minimum of around 10:00 LT during post-monsoon (Fig. 6a,b), shifted to 12:00-13:00 LT during monsoon. This shifting may be related to the amount of vegetation

JAS (monsoon) ON (post-monsoon) DJF (winter) MAM (pre-monsoon)
CO 2 (ppm) 10 www.nature.com/scientificreports/ around the site. Figure 4a suggests that NDVI (a proxy of vegetation) is high during October-November, which may lead to enhance photosynthesis during the noon hours (11:00-12:00 LT). Some interesting features are observed during the period 00:00-06:00 LT. CO 2 levels remain somewhat constant for the monsoon and post-monsoon periods. Constant levels at night during monsoon and post-monsoon give evidence of continuous but weak sources such as plant and soil respiration. CH 4 shows a maximum (minimum) diurnal amplitude (Table 2) of 62.05 (5.3) ppb during winter (monsoon). The monsoon to post-monsoon transition phase experiences the maximum increase in CH 4 amplitude (around 443%). On the other hand, the pre-monsoon to monsoon transition registers a modest decrease (~ 87%) in CH 4 diurnal amplitude.

Discussion and conclusions
The seasonal amplitude of CO 2 , is high over sng as compared to knp (forest site), amd (urban site) and sad (semiurban site) of India. The seasonality of knp-CO 2 is mostly driven by the biosphere. Pre-monsoon rainfall in knp enhances Leaf Area Index (LAI), which in turn increases CO 2 assimilation during daytime 11 hence reducing the atmospheric CO 2 concentration. While, during monsoon, though LAI is high, occasional overcast conditions reduce photosynthetically active radiation (PAR) from reaching the canopy, reducing the CO 2 uptake. Simultaneously, sad shows enhanced CO 2 concentration in pre-monsoon months due to higher temperature and solar radiation 26 and minimum in monsoon mostly driven by enhanced photosynthesis with the availability of higher soil moisture. CO 2 mixing ratio over cri is highest in February-March, due to increased heterotrophic respiration and anthropogenic activity in northern India 33 . The high seasonal amplitude of sng is characterised by low CO 2 in monsoon and post-monsoon and elevated CO 2 during pre-monsoon season. The steady growth of CO 2 during the dry season (November to May) indicates a decreasing trend of vegetation uptake in the neighbouring regions (Fig. 4a). A sharp increase in mean value (410-417 ppm) during the pre-monsoon period could be attributed to enhanced solar radiation. Higher temperature enhances CO 2 photosynthesis during daytime and respiration during the nighttime 34 . In that case, the diurnal amplitude (maximum-minimum) of CO 2 should be high, but during pre-monsoon, this amplitude becomes negligible (discussed in diurnal variation of GHG section). Soil respiration and biomass burning also act as a source of CO 2 into the atmosphere. With the advancement of monsoon season, the CO 2 concentration steadily reduces mainly due to the CO 2 uptake by the biosphere. Additionally, the reduction in temperature further decreases the leaf and soil respiration 35,36 . Moreover, NDVI (a proxy of vegetation) is increasing (Fig. 4a) during monsoon months. CH 4 concentrations over monsoon Asia (including China) show higher values during the wet seasons (JAS and ON) and low values during dry periods (DJF and MAM) driven by agricultural practices, i.e., paddy fields as well as large scale transport and chemistry 37,38 . Like the 'background' region, mlo in the Pacific Ocean, we have also observed low methane concentrations during the summer months (JAS, Fig. 2c) though the mechanism is not the same as that of mlo. In our case, low concentration is controlled by strong monsoon circulation though surface emission (from AW sector, Fig. 4b) is high. Low surface CH 4 concentration instead of high local emission is also found by Guha et al. 39 . They suggest the intrusion of southern hemispheric clean air with monsoonal south-westerly wind is responsible for low surface CH 4 concentration. Therefore, maximum CH 4 concentration is found during post-monsoon when south-westerly current is decreased.
In comparison, the second maximum of CH 4 emission is observed in February-March-April with very low NDVI. Hence, emission from wastewater and landfills, enteric fermentation, and animal waste management plays a dominant role in CH 4 emission during February-March-April. It is found that boundary layer dynamics is not sufficient for the seasonal change of CO 2 and CH 4 levels. In a nutshell, the tropospheric CH 4 concentration in this region is determined by the following processes: a balance between the local to regional scale surface emission, destruction by the OH radicals at the hemispheric scale, and the regional monsoon circulation. Meanwhile, a low concentration of CO 2 instead of high positive residual flux (biosphere + fossil) indicates that monsoon flow brings cleaner air, which lowers the average concentration of atmospheric CO 2 over sng as observed for CH 4 . Hence, we found a strong negative correlation between wind speed and CH 4 . But interestingly, a positive correlation is evident between CO 2 and wind speed in monsoon.
Monsoon rainfall frequently comprises wet and dry spells of precipitation over a period of 10-90 days, widely known as monsoon intraseasonal oscillation (ISO). 10-20 days 40 and 20-60 days 41,42 are two dominant modes of ISO. Cross equatorial low-level jet (LLJ, surface south-westerly wind) is a dominant feature of monsoon. LLJ also shows intraseasonal oscillation in association with monsoon ISO 43 or precisely with north/north-eastward propagation of deep convection 44 , but with a lag of about 2-3 days. Valsala et al. 45 also show the interplay between monsoon ISO and net biosphere CO 2 flux. OLR is considered a proxy for the deep convection and is used for precipitation estimation [46][47][48] . A lag correlation analysis is carried out between filtered (10-60 days band passed) wind vs. filtered OLR and filtered CO 2 vs. filtered wind (see "Supplementary section"). A maximum correlation is observed between OLR and wind when OLR leads the wind by 2-3 days.
In contrast, CO 2 shows a strong positive correlation with the wind, with wind lagging 1-2 days. Hence, the positive correlation between CO 2 and wind may arise due to the response of monsoon intraseasonal oscillation. A strong monsoon circulation brings cleaner air, which reduces the CO 2 and CH 4 both, but CO 2 is modulated by biospheric uptake. The biosphere uptake is further modulated by monsoon intraseasonal oscillation. Consequently, we found a positive relation between CO 2 and wind as a response to monsoon.
A higher SD of CO 2 histogram during the daytime indicates a broader spread with respect to the nighttime distribution, which is characterised by a lower SD. So, the broadness of the CO 2 distribution function during the daytime is caused by a diverse source/sink of CO 2 . With the development of the boundary layer, CO 2 gets mixed vertically. As the day progress, the photosynthetic CO 2 sink reduces the CO 2 concentration, which is moderated by the increase in PBLH. While anthropogenic sources of CO 2 and plant respiration are also active during the day, a broader CO 2 distribution spectrum is yielded during different hours of the day. The narrow PD for CO 2 www.nature.com/scientificreports/ in the nighttime is suggesting the dominating role of CO 2 release by respiration and anthropogenic activity. The difference between daytime and nighttime CO 2 distribution is evident in monsoon and post-monsoon only. Moreover, the diurnal variation of CO 2 is also most prominent in these seasons. Daytime CO 2 minima (around noon), a constant value of CO 2 during the night (00:00-06:00 LT), different daytime and nighttime CO 2 histogram are the key features in monsoon and post-monsoon season. In contrast, the diurnal variation of CO 2 in winter and pre-monsoon diminishes. Though, daytime PBLH maximum is more (> 2000 and > 2500 m) during winter and pre-monsoon (Fig. 6c,d), which indicates a strong mixing. This clearly shows that the monsoon and post-monsoon season get their CO 2 share mainly from the active biosphere. In contrast, the other two seasons get from the degradation of the biosphere and anthropogenic activities. Boundary layer dynamics are ineffective when vegetation is less. Moreover, the close association of soil temperature at level1 and 2 with CO 2 (Fig. 5d) implies that soil respiration is a dominant part of pre-monsoon CO 2 . A similar pattern in CH 4 histogram during daytime and night time implies that the source and transport processes of CH 4 remain more or less invariant (note that the CH 4 sink by OH is a slow process, with a time scale of 1 year or longer in summer over the tropics; Patra et al. 17 ). Diurnal variation of CH 4 (Fig. 6a-d) shows morning CH 4 develops with the advent of PBLH other than monsoon. Such a pattern suggests, trapped CH 4 in the neighbouring valley due to a stable boundary layer of the previous night becomes available at our site (top of a hill) with the rise in the boundary layer in the morning hours. So, we get a morning peak in CH 4 concentration. As PBLH grows beyond the site elevation, CH 4 drops due to mixing with a larger area. Winter is characterised by a small peak in CH 4 levels (Fig. 6c) at the evening (around 19:00 LT), which further develops and emerges as a dominant peak in pre-monsoon. Hence, a close association of PBL and CH 4 is observed in biplot in premonsoon (Fig. 5d).

Data and methodology.
Climatology of the study area. The mean monthly variation of relative humidity (RH in %) and temperature (°C) from NCEP-FNL reanalysis dataset over sng is shown in Supplementary  Fig. 1a,b during the period 2014-2015. Temperature over sng varies from ~ 25 to ~ 31 °C. Relative humidity (RH) was maximum during south-west monsoon (June-July-August-September, JJAS) season of > 75%, and the minimum occurred during winter (December-January-February, DJF) of about < 50%. At sng, the wind speed at 850 hPa (data source: ERA-Interim) varies between 1 and 12 ms −1 . Maximum wind speed occurred mainly from the south-west direction during the Indian summer monsoon (ISM) months, JJAS, which originated from the Arabian Sea. In winter, the winds are mostly from the northeast direction, originated from the Indian subcontinent ( Supplementary Fig. 1c,d). Figure 1b shows the location of the study area with the mean outgoing longwave radiation (shaded) and mean wind (1000 hPa) flow in vector form. Figure 1a depicts the south-westerly monsoon flow from the ocean to land with enhancing convection (low OLR) over the Indian sub-continent. Figure 1b illustrates an opposite flow pattern during the winter associated with suppressing convection (high OLR). So, it is evident that our study area experiences a strong seasonally reversing of the wind flow from summer to winter. The wind rose diagram shows south-westerly wind is associated with low CO 2 and CH 4 concentration (Fig. 1c,d). The interplay between wind and GHG concentration is discussed further in "Influence of prevailing meteorology" section. GHG analyser. Continuous air sampling was done through a fast greenhouse gas analyser (model: LGR-FGGA-24r-EP) from a 10 m meteorological tower. It is based on enhanced off-axis integrated cavity output spectroscopy (OA-ICOS) technology 49 . This instrument is able to provide CH 4 , CO 2, and H 2 O concentration simultaneously with high temporal resolution (up to 1 Hz). The sensor was calibrated using a zero air cylinder having known CO 2 , CH 4 concentrations. The ' dry values' of CO 2 and CH 4 mixing ratios, corrected for water vapour, are reported in this paper. The CO 2 and CH 4 data integrated for 100-s intervals are presented here. The analyser has 0.3 ppb, 0.05 ppm, and 5 ppm precision of CH 4 , CO 2, and H 2 O when operating in the 0.01 Hz frequency. Moreover, we take 15-min average CO 2 and CH 4 measurements for further analysis. The site has been operational from July-2014 to November-2015. There were several data gaps in between, with an opening from 3-May-2015 to 9-July 2015 (longest gap), due to instrument maintenance. This gap is filled with weekly flask samples data 24 obtained from the same site. CO 2 and CH 4 concentration data have been plotted on diurnal and monthly time scales. The year was divided into four different seasons, i.e., monsoon (July-August-September), post-monsoon (October-November), winter (December-January-February), and pre-monsoon (March-April-May).
Due to the unavailability of AWS in the study area, no in-situ meteorological data were available; instead, we use different kinds of reanalysis data as mentioned later. Kaziranga (knp) CO 2 data. The Metflux India flux observational site Kaziranga National Park (knp) is a semi-evergreen forest located in the north-eastern state of Assam. The CO 2 concentration over the forest is measured at the height of 37 m using an enclosed path CO 2 -H 2 O infrared gas analyser (LI-7200, LI-COR, USA) at frequency of 10 Hz. The high-frequency data are processed using the EddyPro software and averaged in the time interval of 15 min. The details of the study area and instruments can be found in 11 .

Moderate-resolution imaging spectrometer (MODIS). The MODIS was launched in December 1999
on the polar-orbiting NASA-EOS Terra platform 50,51 . It has 36 spectral channels covering visible, near-infrared, shortwave infrared, and thermal infrared bands. In the present study, we have used 5-km spatial resolution having 16-day temporal resolution NDVI (Normalized difference vegetation index) data. We got the dataset from MODIS (Product-MOD13C1) official website ("https ://modis .gsfc.nasa.gov/data/datap rod/mod13 .php"). The NDVI is a normalised transform of the near-infrared (NIR) to red reflectance ratio (RED) and calculated using the following equation  ERA5. ERA5 is the latest version of reanalysis produced by ECMWF. ERA5 is produced using 4D-Var data assimilation in ECMWF's Integrated Forecast System. A temporal resolution of 1 h and a vertical resolution of 137 hybrid sigma model levels. The 37 pressure levels of ERA5 are identical to ERA-Interim 57 . ERA5 assimilates improved input data that better reflects observed changes in climate forcing and many new or reprocessed observations that were not available during the production of ERA-Interim. ERA5-Land provides the land component of the model without coupling to the atmospheric models. It uses the Tiled ECMWF Scheme for Surface Exchanges over Land with revised land-surface hydrology (HTES-SEL, CY45R1). It is delivered at the same temporal resolution as ERA5 and with a higher spatial resolution of 0.1° × 0.1°. 2 m air temperature, soil temperature level 1 (0-7 cm), and soil temperature at level 2 (7-28 cm) is used.
NCEP FNL re-analysis. The NCEP FNL (final) operational global analysis data are on 1° × 1° grid prepared operationally every six hour. This product comes from the Global Data Assimilation System, which continually gathers observational data. The time series of the archive is continually extended to a near-current date but not preserved in real-time (http://rda.ucar.edu/datas ets/ds083 .2/). The key aim of these re-analysis data is to provide compatible, high-resolution, and high-quality historical global atmospheric datasets for use in weather research communities 58,59 . Air temperature and RH are averaged over the area 18-19° N and 73-74° E.