A high concentration CO₂ pool over the Indo-Pacific Warm Pool

Abstract

Anthropogenic emissions have produced significant amount of carbon dioxide (CO₂) in the atmosphere since the beginning of the industrial revolution. High levels of atmospheric CO₂ increases global temperature as CO₂ absorbs outgoing longwave radiation and re-emits. Though a well-mixed greenhouse gas, CO₂ concentration is not uniform in the atmosphere across different altitudes and latitudes. Here, we uncover a region of high CO₂ concentration (i.e. CO₂ pool) in the middle troposphere (500–300 hPa) over the Indo-Pacific Warm Pool (IPWP, 40° E–140° W, 25° S–25° N), in which the CO₂ concentration is higher than that of other regions in the same latitude band (20° N–20° S), by using CO₂ satellite measurements for the period 2002–2017. This CO₂ pool extends from the western Pacific to the eastern Indian Ocean. Much of the CO₂ pool is over the western Pacific Ocean (74.87%), and the remaining lies over the eastern Indian Ocean (25.13%). The rising branch of Walker circulation acts as a “CO₂ Chimney” that constantly transports CO₂ released from the natural, human-induced and ocean outgassing processes to the middle and upper troposphere. The CO₂ pool evolves throughout the year with an average annual trend of about 2.17 ppm yr⁻¹, as estimated for the period 2003–2016. Our analysis further reveals that La Niña (El Niño) events strengthen (weaken) the CO₂ pool in the mid-troposphere. The radiative forcing for the CO₂ pool suggests more warming in the region and is a grave concern for global warming and climate change.

Introduction

Carbon dioxide (CO₂), a major greenhouse gas (GHG), concentration has been steadily rising in the atmosphere since the mid-nineteenth century. The global warming due to high levels of GHGs might increase surface temperatures over 1.5 °C above the pre-industrial levels by 2030^1,2. This accelerated warming is worrisome, as it leads to more frequent and severe extreme events such as heatwaves^3,4, floods^5,6, and changes in tropical cyclone activity^7,8 and rainfall patterns^9,10 with devastating economic and environmental consequences¹¹. The atmospheric CO₂ levels climbed up to 412 ppm in January 2020, with an average global trend of 2.11 ppm yr⁻¹ during the period 2003–2016^12,13,14.

A thorough understanding of the dynamics, evolution, fate and human influence on atmospheric CO₂ is essential to enact effective carbon mitigation policies. The spatial and temporal variations of atmospheric CO₂ show distinct annual, seasonal and latitudinal gradients¹⁵. There are substantial differences in CO₂ concentrations between continents and oceans, which are dependent on the sources of emissions. Large-scale atmospheric circulations, weather systems and jet streams distribute CO₂ around the globe¹⁵. Northern Hemisphere (NH) is known for higher levels of atmospheric CO₂ than that of the Southern Hemisphere (SH). In NH, seasonal changes in CO₂ are primarily driven by the result of metabolic activity of terrestrial plants and soils¹⁶. The warmer climate alters the seasonal CO₂ cycle¹⁷. The effect of rising atmospheric CO₂ increases radiative forcing, which leads to higher sea surface temperature (SST)¹⁸. These also cause a reduction in surface-to-deep ocean transport of CO₂ and a reduction in oceanic carbon, which might increase the concentration of atmospheric CO₂¹⁹.

An oceanic region enclosed by 28 °C or higher SST isotherm is known as a warm pool²⁰. These are the regions with high precipitation, strong atmospheric convection and surface wind convergence²¹. The warm pool regions of western tropical Pacific Ocean and the tropical Indian Ocean constitute the Indo-Pacific Warm Pool (IPWP)²². Previous studies suggest that warmer SST in IPWP strengthens the upwelling branch of Hadley circulation and weakens (strengthens) its downwelling branch in NH (SH). Furthermore, due to the rapid warming of Indian Ocean, the westward extension of IPWP shifts the walker circulation westward, which decreases subsidence over eastern Africa and makes the region drier^23,24.

IPWP has a significant role in transporting surface emissions to higher altitudes over the tropics. Such a system, which is a source of heat and moisture, prevails throughout the year regardless of seasons in West Pacific and East Indian Ocean (EIO) in the tropics. IPWP shows meridional and zonal variability that can influence global atmospheric circulation, and onset, intensity and duration of climate modes such as El Niño Southern Oscillation (ENSO)²⁵. As compared to the oscillations of the eastern edge of western Pacific Warm Pool (WPWP), it is observed that South and North Warm Tongues are dominated by the annual SST cycles and are influenced by the El Niño onset, but East Cold SST Tongue is totally dominated by the El Niño onset and shows no distinct annual cycle²⁶. Yan et al.²⁷ studied the temperature and size variations of WPWP, and found that changes in solar irradiance, ENSO events and global warming could have modified the distribution of SST and the size of IPWP. The warm pool in Indian Ocean has a stronger annual cycle than that of Pacific²⁸. Furthermore, SST in Indian Ocean is rising at a faster rate than in other tropical ocean regions^29,30.

During El Niño, warmer than normal subsurface water in the tropical Pacific replaces the cooler CO₂-rich water³¹. It reduces or reverses the CO₂ released by tropical oceans and increases the uptake of global-oceanic CO₂³². For instance, surface measurements show a reduction in CO₂ outgassing during the 2015–2016 El Niño event³³. Therefore, oceans act as CO₂ sink during El Niño and thus, they reduce CO₂ concentrations in the atmosphere when the tropical SST is warmer¹⁷.

This study is organized as follows: we analyze the anomalously high values of mid-tropospheric CO₂ over IPWP using satellite data. We examine possible reasons for the high CO₂ region and its radiative forcing. We also investigate the influence of ENSO on the CO₂ concentrations in the mid-troposphere. We supplement our analysis using buoy measurements to assess changes in near-surface CO₂, which has a critical role in contributing to the high CO₂ region.

Results

Global distribution of atmospheric CO₂

Figure 1a shows a region of high CO₂ concentration between 30° N and 60° N and comparatively lower concentration in the rest of the region. The largest seasonal difference is observed in the 45°–60° N region (3.21 ppm) and lowest in 0°–30° N (1.32 ppm), as shown in Supplementary Figure S1. North America and Europe exhibit the highest CO₂ concentration in 45°–60° N, whereas the largest CO₂ concentration over Asia is found in 30°–45° N. The highest difference is observed between the regions 30°–60° N and 0°–30° N, with its peak in 2010 (1.81 ppm, Supplementary Figure. S1). There is an increase of 5.8% in global CO₂ emissions, with a peak of 33 billion tonnes in 2010 due to the continued growth of developing countries and economic recovery in developed nations³⁴, which is also the reason for the highest difference in 2010.

Figure 1

Atmospheric CO₂ and Sea Surface Temperature. Annual averaged (a) mid-tropospheric CO₂ (b) HadISST Sea Surface Temperature (SST) for the period 2002–2017. SST contours of 28 °C and 29 °C are overlaid. These maps are generated using Cartopy⁶⁴ 0.18.0 (https://scitools.org.uk/cartopy).

The average of CO₂ over land regions during 2003–2016 is 388.52 ± 0.93 ppm, whereas it is 387.95 ± 1.27 ppm over the oceans. Also, there is a difference of 1.13 ± 0.1 ppm between land regions of the north and south hemispheres, and 0.63 ± 0.06 ppm between the oceans of both hemispheres. Higher emission regions show higher CO₂ concentrations (e.g. NH), but lower over the carbon sink regions (e.g. the Atlantic minimum). Note that atmospheric circulation also plays a key role in the redistribution and mixing of CO₂¹⁵. As illustrated in Fig. 1a, the southern Atlantic and Pacific Oceans have the low CO₂ zones among the regions, which can be attributed to the presence of large carbon sinks, atmospheric subsidence and absence of CO₂ sources there¹². The anomaly between the global ocean CO₂ and low CO₂ regions of the Atlantic and Pacific Oceans shows differences larger than 1 ppm, although it decreases after 2010 (Supplementary Figure. S1). The decreasing tendency in CO₂ anomalies at these low CO₂ regions indicates that the minimum zones are shrinking and it could be the repercussions of deforestation in the Amazon rainforest and climate change³⁵.

The regional temperature exhibits a positive feedback on the atmospheric concentrations of CO₂. A major region of heat source in the ocean, the IPWP, acts as a medium to distribute CO₂ in the atmosphere to different altitudes¹². The global distribution of SST is shown in Fig. 1b.

A high CO₂ pool over the tropical ocean

It is interesting to note a zone of high CO₂ concentration in the mid-troposphere over IPWP, although CO₂ concentrations over tropical regions are relatively lower than that in the mid-latitudes of NH³⁶. Hereafter, it is referred to as “the CO₂ pool”, which is shown in Fig. 2a. The CO₂ pool stretches from the tropical East Indian Ocean to the tropical central Pacific Ocean, with a major part over the latter region.

Figure 2

CO₂ pool and its driving factors. (a) Annual-averaged (2002–2017) mid-tropospheric CO₂ with the CO₂ pool region. (b) Annual and latitude (20° S – 20° N) averaged vertical velocity for the period 2002–2017 and (c) Annual-averaged (2002–2017) Sea Surface Temperature (SST) in the Indo-Pacific region. Here SST contour of 28° C represents the Indo-Pacific Warm Pool (IPWP). These maps are generated using Cartopy⁶⁴ 0.18.0 (https://scitools.org.uk/cartopy).

Figure 2b shows the region-averaged (20° S–20° N) vertical velocity for the period 2002–2017. The positive values between 60° E and 150° W indicate the ascending arm of Walker circulation and the negative values between 120° W and 90° W show its descending branch in Pacific Ocean. The upwelling air over IPWP (Fig. 2c) brings high CO₂ from the surface to mid-troposphere, which makes high CO₂ concentrations there. However, the sinking air over the eastern Pacific Ocean brings low CO₂ to the mid-troposphere from higher altitudes. The descending arm of Walker circulation over the western Indian Ocean brings lower CO₂ concentration in these regions.

Variability of Indo-Pacific CO₂ pool

Figure 3a shows the seasonal variability of CO₂ over the tropics, overlaid with SST, which indicates the temporal changes in IPWP. Regions of high CO₂ concentrations align with the 28–29 °C SST contours in most months. The zonal and meridional extensions of IPWP are also mimicked by the CO₂ distributions. For example, IPWP extends more to the southern hemisphere in winter, which is also followed by CO₂. Similarly, the zonal growth of IPWP is partly responsible for the high concentration of CO₂ that extends to the eastern Pacific in summer, and even to the tropical Atlantic Ocean.

Figure 3

Variability of the CO₂ pool over the Indo-Pacific Warm Pool (IPWP). (a) Seasonal average (DJF: December, January and February, MAM: March, April and May, JJA: June, July and August, SON: September, October and November) mid-tropospheric CO₂ (ppm) over the tropical oceans for the period 2002–2017. Black contours represent the seasonal average of Sea Surface Temperature (SST) in °C for the period 2002–2017. These maps are generated using Cartopy⁶⁴ 0.18.0 (https://scitools.org.uk/cartopy). (b) Latitudinal variability of annual averaged mid-tropospheric CO₂ over different regions (with 95% confidence interval) such as the Pacific Ocean Minimum (POM), Atlantic Ocean Minimum (AOM), AFE (Africa and Europe, averaged between longitudes 10° E and 45° E), CO₂ pool over Western Pacific Ocean (WPO) and CO₂ pool over East Indian Ocean (EIO). (c) Yearly variations of the area of CO₂ pool and contributions of WPO and EIO are indicated. Note that the y-axis starts from 3.5 × 10⁷ km².

The seasonal variability of CO₂ is attributed to the changes in its sources, sinks, horizontal winds and vertical transport. The CO₂ pool over the eastern Indian Ocean shows more visible seasonal changes than that in the western tropical Pacific Ocean. During October–November, the CO₂ pool is one of the high CO₂ concentration regions over the oceans, higher than that over the northern hemisphere. Supplementary Figure S1 shows the interannual variability of CO₂ pool with an average linear trend of 2.17 ppm yr^-1; indicating a continuous increase of CO₂ over the region.

The monthly vertical velocity (Pa s⁻¹) averaged from 1000 to 300 hPa is shown in Supplementary Figure S2. The upward movement of air favors the transport of CO₂ to higher altitudes, whereas the sinking of air suppresses CO₂ to lower altitudes. Latitude-wise distribution of CO₂ over different regions is shown in Fig. 3b. All regions exhibit two major peaks, whereas West Pacific Ocean (WPO) and EIO have additional peaks in the tropics representing the CO₂ pool. As the CO₂ pool lies over the tropics, it extends to both hemispheres with a noticeable seasonal variability. The western Pacific CO₂ pool spreads over both hemispheres, although the Indian Ocean shows skewness towards the northern hemisphere. To quantify the minimum zones of CO₂ over the Atlantic and Pacific Oceans, we selected two regions with coordinates 0°–20° S, 50° W–15° E (AOM) and 0°–15° S, 125°–80° W (POM). Figure 3b shows higher CO₂ in the northern hemisphere, about 3.5 ppm more than in AOM. Similarly, Africa and Europe, averaged between longitudes 10° E and 45° E (AFE) show high CO₂ in NH high latitudes.

To assess the spatial coverage of CO₂ pool, we have calculated the year-wise area, as discussed in Methods section. Figure 3c shows the yearly total area of CO₂ pool from 2003 to 2016. The average area of the CO₂ pool during 2003–2016 is about 5.82 × 10⁷ km², and the largest area is estimated for 2010 (6.04 × 10⁷ km²). The yearly percentage contribution of WPO and EIO to CO₂ pool indicates that WPO is more prominent in size, roughly three-fourths of the total area of CO₂ pool in all years.

Influence of climate modes on CO₂ pool

During the ENSO events, changes in the Walker circulation have a profound influence on the distribution of CO₂ in the mid-troposphere³⁷. Jiang et al.³⁸ reported that the mid-tropospheric CO₂ levels are elevated over the central and repressed over the western Pacific Oceans during the El Niño events. The change in Walker circulation during El Niño favors CO₂ transport over the central Pacific Ocean, whereas the downwelling air over WPO limits the CO₂ transport to higher altitudes. We have selected two regions to quantify the differential response of CO₂ over the eastern and western Pacific Oceans to the ENSO events. The coordinates of these regions are R1: 10° S–10° N, 85°–160° E and R2: 10° S–10° N, 170°–70° W.

Liu et al.³⁹ showed high CO₂ emissions during El Niño events due to associated fire activities in south Asia. However, the ENSO composites of the detrended, deseasonalized and latitude-averaged vertical velocity (Supplementary Figure S5) reveal that its negative anomaly (downward motion of air) inhibits the upward transport of high CO₂ concentration at R1 and thus, make lower CO₂ in mid-troposphere there. Similarly, positive anomalies of vertical velocities are present over R1 during La Niña events. This upward-moving air transports high CO₂ to the mid-troposphere from the surface, which gives rise to positive anomaly of CO₂ over IPWP during strong La Niña events (Fig. 4d). As compared to R1, a contrasting effect is present over R2 owing to the opposite influence of vertical velocity during the La Niña and El Niño events.

Figure 4

Impact of El Niño and La Niña events. El Niño composites of deseasonalized and detrended (a) Sea Surface Temperature (SST) and (b) mid-tropospheric CO₂ during the period 2002–2017. La Niña composites of deseasonalized and detrended (c) Sea Surface Temperature (SST) and (d) mid-tropospheric CO₂ during the period 2002–2017. R1 and R2 are the selected regions. These maps are generated using Cartopy⁶⁴ 0.18.0 (https://scitools.org.uk/cartopy).

The yearly averaged distribution of CO₂ and SST show high values across the western central Indian Ocean that extends to the southeast Arabian Sea¹³. The region where the warm pool lies is also under the influence of ENSO events (Fig. 4a, c). The response of atmospheric CO₂ to the ENSO events is identified using the averaged data of respective months (Fig. 4b, d). For instance, the positive SST anomaly over R1 during La Niña corresponds to relatively higher CO₂ and the negative SST anomaly reciprocates with lower CO₂ in the mid-troposphere. Similarly, the response of R2 to La Niña and El Niño reiterates the positive relationship between SST and CO₂ anomalies. Furthermore, the northeast Atlantic Ocean also has a similar response, which replicates R1; indicating the influence of climate modes on SST and CO₂ in all oceanic regions.

The interannual variability of CO₂ anomalies at R1 and R2 is shown in Fig. 5a,b. We have considered only the strong El Niño (December 2006, November 2009–February 2010, June 2015–March 2016) and La Niña (September 2007–March 2008, July 2010–February 2011, October 2011–December 2011) events here. During El Niño and La Niña periods, anomalies of CO₂ to corresponding SST differences are nearly proportionate, but opposite over the regions R1 and R2. This shows the CO₂ response to a warming ocean. During the 2015–2016 El Niño, a strong feedback is observed over R1, and CO₂ anomalies reached − 1.1 ppm. During the neutral and weak ENSO events (ENSO index within ± 1), R1 and R2 show similar CO₂ peaks; suggesting that those peaks could be the result of the variability in large-scale CO₂ emissions.

Figure 5

Response of CO₂ anomalies to ENSO and comparison of AIRS CO₂ with surface and buoy measurements. (a), Detrended and deseasonalized month-wise CO₂ anomalies (ppm) over the regions R1 and (b) for R2 during the period 2002–207. Blue indicates mid-tropospheric CO₂ during La Niña months (CO₂^Niña, ENSO index <  − 1 ), Orange is mid-tropospheric CO₂ during El Niño months (CO₂^Niño, ENSO index >  + 1) and pale green indicates CO₂ during neutral months and months with ENSO index within ± 1 (CO₂^neutral). Regions R1 and R2 are marked in Fig. 4d. Mid-tropospheric CO₂ during La Niña and El Niño events is shown with 95% confidence intervals and are indicated as error bars (grey). (c) Detrended and deseasonalized measurements from Mauna Loa observatory (MLO), ∆pCO₂ (μatm) from the buoys Chuuk K1, TAO 0°, 165° E, TAO 8° S, 165° E and TAO 0, 170° W representing the western Pacific ocean (WPO) and Stratus 85° W 20° S, TAO 0°, 110° W, TAO 0, 125° W and TAO 0, 140° W represent the eastern Pacific oceanic regions. Detrended and deseasonalized AIRS CO₂ (at 95% confidence level) over R1 as marked in Fig. 4d is indicated as R1 here. The periods of El Niño and La Niña are shaded.

CO₂ observations at ocean surface

The near-surface measurements of atmospheric CO₂ show an immediate and prominent response to changes in carbon sources than CO₂ in mid-troposphere due to the proximity advantage of surface measurements (Fig. 5c). A study on the influence of Pacific Ocean on atmospheric CO₂ using measurements from Mauna Loa observatory (MLO) shows that most El Niño events correspond to an immediate decrease in atmospheric CO₂ within a month or two⁴⁰, which can be also found in Fig. 5c. The subsequent rise in CO₂ in the following months can be attributed to the influence of declined CO₂ intake by the global biosphere, increased plant and soil respiration (0.6 ± 1.01 gigatons C in Africa) and enhanced fire emissions (0.4 ± 0.08 gigatons C in tropical Asia) during the El Niño events^33,39. Anomalies computed from the satellite observations at R1 and surface measurements at MLO show similar variability in CO₂ with some lag/lead as illustrated in Fig. 5c. The magnitude of CO₂ peaks is often very high for MLO and higher for buoy measurements located at the central and eastern tropical Pacific Ocean (EPO) than that of satellite observations. The shallow warm water in IPWP acts as a barrier between the cooler water and atmosphere, and restricts CO₂ venting there, which is evident from the lower ΔpCO₂ (difference between surface seawater pCO₂ and atmospheric pCO₂) values at WPO than that of EPO (Fig. 5c). There is a reduction in CO₂ outgassing from EPO and WPO during El Niño events (e.g. ΔpCO₂ ≈ 0 μatm at EPO), but it enhanced during La Niña events (e.g. ΔpCO₂ > 70 μatm at EPO and WPO), although not all ENSO events have a notable influence on CO₂ concentrations in the atmosphere. These differences are probably due to the complex ocean–atmosphere–biosphere coupled interactions⁴⁰. The increased CO₂ venting during La Niña events contributes to the higher CO₂ in the mid-troposphere. The outgassed CO₂ together with other emissions from land regions are transported to higher altitudes by the vertical winds (see Supplementary Figure S4).

Impact of CO₂ pool on global climate

Figure 6a depicts the trend in yearly-averaged mid-tropospheric CO₂ from 2003 to 2016. Regions north of 60° N show the highest trend values, particularly over the northernmost regions (greater than 2.3 ppm yr⁻¹). The eastern Atlantic and Pacific Oceans show higher trends than their regional counterparts. The average annual trend for the CO₂ pool is 2.17 ppm yr⁻¹ (Supplementary Figure S1), which is higher than that of other oceanic regions, e.g. the average trend of CO₂ over AS is 2.13 ppm yr⁻¹ for the period 2003–2016¹³. Nevertheless, the average trend over the region within 20° S–20° N and 50° E–160° W, without the CO₂ pool criteria, is similar to that of the global average CO₂ trend (2.11 ppm yr⁻¹). The CO₂ minimum zones over the oceans, AOM and POM, exhibit a high positive trend of CO₂ (2.14 ± 0.02 ppm yr⁻¹ and 2.12 ± 0.02 ppm yr⁻¹ respectively). To assess the impact of increased levels of CO₂ on climate, we also calculate the Radiative Forcing (RF) of CO₂ in 2016 with respect to that of 2003.

Figure 6

Radiative forcing. (a) Annual trend of mid-tropospheric CO₂ (ppm yr⁻¹) during 2003–2016. Stippling indicates significant trends at 95% confidence level. (b) Radiative Forcing (RF; W m⁻²) of CO₂ in 2016. Mid-tropospheric CO₂ in 2003 is the reference CO₂ concentration for the calculation of RF here. These maps are generated using Cartopy⁶⁴ 0.18.0 (https://scitools.org.uk/cartopy).

Figure 6b shows the spatial distribution of RF across the regions. High RF is found in regions where CO₂ concentrations are relatively higher and the average RF at R1 is 0.37 ± 0.01 Wm⁻². The Atlantic Ocean and WPO exhibit high RF and annual CO₂ trends (ppm yr⁻¹) than those in other oceanic regions. These incessant increase of CO₂ in the atmosphere exacerbate global warming and this would make it more difficult to keep the warming below 2 °C^2,41.

Discussion

A region of high CO₂ concentration is present over IPWP in the tropics. Temporal (monthly, seasonal and annual) analysis of the CO₂ data over the region shows that it is a permanent feature. The monthly distribution of SST and mid-tropospheric CO₂ are congruent, particularly over IPWP. The CO₂ pool shows an average annual trend of 2.17 ppm yr⁻¹, which is comparatively higher than that over other oceanic regions. Several studies have discussed ocean warming in the context of increased GHGs^42,43. Weller et al.⁴⁴ identified GHG forcing as the major cause of the observed increase in IPWP intensity and size, which produced an increase in ocean heat content and high sea level rise in the twentieth century. Our analysis confirms that the changes in Walker circulation can alter CO₂ distribution in the mid-troposphere. The transport of surface emissions, both natural and anthropogenic, controlled by atmospheric circulation have distinct seasonal, intra-seasonal and interannual variability. Our assessment suggests that La Niña conditions enhance atmospheric CO₂ over IPWP, whereas El Niño has a negative effect on it. Apart from these, strong winds in the middle and upper troposphere transport and mix these high concentrations of CO₂ to the higher altitudes during the periods of ENSO.

Immediate and sharp responses of these climate modes are also well captured by the near-surface and moored buoy CO₂ measurements. Chatterjee et al.³³ analyzed the influence of 2015–2016 super El-Niño on atmospheric CO₂ using the column-averaged measurements from Orbiting Carbon Observatory-2 (OCO-2) and CO₂ measurements from buoys. They showed an initial drop in atmospheric CO₂ during the onset of El Niño due to the suppression of upwelling in the tropical Pacific, which reduced the outgassing of CO₂ from the ocean to atmosphere. The reduction in CO₂ exchange is accountable for the weak response of CO₂ anomalies in R2 during the El Niño period, as shown in Fig. 5b.

Furthermore, Indian Ocean Dipole (IOD)⁴⁵ has a comparable effect on mid-tropospheric CO₂ as that by ENSO¹³. That is, mid-tropospheric CO₂ exhibits a negative (positive) anomaly during the strong positive (negative) IOD over EIO. During the strong negative IOD in 2016, EIO had the highest contribution to the total area of CO₂ pool (29.20%) compared to that in other years. A possible reason for this is that the 2015–2016 super El-Niño had already reduced the size of CO₂ pool over the Pacific Ocean followed by the strongest negative IOD (2016), which favored the uplift of CO₂ to the mid-troposphere^46,47. In addition, the combination of strong La Niña and negative IOD favors the upward transport of CO₂. For instance, a similar event occurred in 2010 and henceforth, the largest CO₂ pool over IPWP is observed in that year; reiterating the influence of climate modes on the regional and vertical distribution of CO₂. The atmospheric CO₂ is gradually increasing, even in the lower CO₂ regions over the Pacific and Atlantic Oceans and over the CO₂ pool. The difference between these low CO₂ regions and the global mean shows a negative trend, which implies that the CO₂ concentration in the minimum pool is increasing faster than that in other regions. Our analyses also give new insights on the global distribution of CO₂ in the mid-troposphere and its variability, particularly over the oceanic regions as they are largest carbon sinks on the Earth. Understanding the connection between IPWP and CO₂ pool is beneficial to tackle the complex ocean–atmosphere interactions and thus, to mitigate global warming and climate change triggered by CO₂ in the atmosphere.

Methods

We have used the Atmospheric Infrared Sounder (AIRS) CO₂ data, which is available from September 2002 to February 2017 with a horizontal resolution of 2.5° × 2°. AIRS is the first hyperspectral Infrared spectrometer on board the Earth Observing System (EOS) Aqua, operating at wavelengths ranging from 3.7 to 15.4 μm^48,49. Vanish Partial Derivative Method is used to retrieve the mid-tropospheric CO₂ utilizing a set of 15 μm spectral channels that have peak sensitivities to CO₂ between 500 and 300 hPa^15,50. Vanishing Partial Derivatives method utilizes the general property of multivariate total differentials to isolate the contribution of individual minor gases. It minimizes the difference between the observed cloud-cleared and calculated radiances with multiple iterations until the radiance residuals of geographical variables are minimized. Since these satellite data are available only up to 2017, our analysis is performed for the period 2002–2017.

Monthly SST data are obtained from the Hadley Centre Sea Ice and SST dataset (HadISST)⁵¹, with a 1° × 1° horizontal resolution. HadISST is available from 1871 to date, but we have opted for the same period as that of AIRS CO₂. Based on Seabold and Perktold⁵², CO₂ and SST data are detrended and deseasonalized to remove trend and seasonality. The average of satellite measurements over different regions are expressed as the mean ± standard deviation. We use the Ocean Niño Index (ONI) to assess the variability in CO₂ during ENSO. Composites of CO₂ and SST are computed for different phases of ENSO. Meyers and O'Brien⁴⁰ indicated that all the ENSO events do not lead to change in CO₂. Henceforth, we have considered only strong El Niño and La Niña events with the criterion of the 3-month season, with ONI values greater than 1 for El Niño and less than -1 for La Niña months. Among the selected El Niño events, 2015 is the equatorial Pacific and rest are central Pacific El Niño events.

The vertical velocity at pressure levels from 1000 to 300 hPa with a horizontal resolution of 25 × 25 km are taken from the ECMWF Reanalysis version 5 (ERA5) data⁵³. The monthly vertical velocity, ω (Pa s⁻¹), is scaled by -1 so that positive values indicate upward motion of air masses and negative values indicate downward motion. Longitudes of the regions selected for the meridional average, other than AOM and POM, are 70°–120° E for EIO, 120° E–160° W for WPO and 10°–45° E for representing land region (Africa and Europe). For this analysis, a standard latitude range of 45° S–60° N is selected. For the calculation of the area of CO₂ pool, we have selected the grid points enclosed over the region within 20° S–20° N and 50° E–160° W, where the anomaly of CO₂ concentration is greater than its standard deviation but less than three times its standard deviation.

$$CO_{2} \, pool = \sigma < CO_{2}^{20^\circ S - 20^\circ N,\; 50^\circ E - 160^\circ W} < 3\sigma$$

where, σ is the standard deviation of CO₂ over the region 20° S to 20° N and 50° E to 160° W.

We have also considered the atmospheric and surface seawater partial pressure of CO₂ (pCO₂) measured by the open ocean moored buoys. ΔpCO₂ is calculated by subtracting atmospheric pCO₂ from surface seawater pCO₂. We have combined ΔpCO₂ from Tropical Atmosphere Ocean (TAO) mooring at 0°, 165° E⁵⁴, TAO at 8° S, 165° E⁵⁵, Chuuk K1 mooring at 7.46° N, 151.90° E⁵⁶ and TAO at 0°, 170° W⁵⁷ to represent the western Pacific, whereas Stratus at 85° W, 20° S⁵⁸, TAO at 0°, 110° W⁵⁹ and TAO at 0°, 125° W⁶⁰, TAO at 0°, 140° W⁶¹ to represent the eastern Pacific based on the mooring locations. Monthly averaged CO₂ data from the Mauna Loa is also used to supplement the satellite data⁶². It is detrended and deseasonalized based on the method described by Seabold and Perktold⁵².

Since the CO₂ pool is an area of high concentrations (e.g. 14 years annual average > 385 ppm in the region), we examine the contribution of CO₂ to regional warming there. Therefore, we have estimated the RF of CO₂ using the formula⁶³,

$$\Delta {\text{F}} = \left( {{5}.{\text{35 W m}}^{{ - {2}}} } \right){\text{ ln}}\left( {{\text{C}}/{\text{C}}_{{\text{O}}} } \right)$$

where, ∆F is the RF for CO₂ in W m⁻², C is the CO₂ concentration based on which RF is calculated. C_O is the reference CO₂ concentration. Conventionally, the pre-industrial level of CO₂ (280 ppm) is taken as C_O. However, we consider CO₂ levels in 2003 as C_O to look at the changes in the recent decade. We present the grid-wise radiative forcing and linear trend of CO₂ to assess the impact of increased CO₂ levels.

The statistical significance of trend analysis and ENSO composites of vertical velocity is examined using the two-sided student’s t test. The difference between El Niño and La Niña composites is calculated and tested for its statistical significance. Grid points with p-values less than 0.05 is considered statistically significant at a 95% confidence level and marked in Supplementary Figure S4. For the ENSO composite analysis of mid-tropospheric CO₂ and SST, the statistical significance is tested at 90% significant level and the statistically significant points are marked in Supplementary Figure S5. The uncertainties indicated in Fig. 3b, 5 and Supplementary Figure S1 are at 95% confidence intervals.