1. Introduction

Anthropogenic change in atmospheric composition is an Essential Climate Variable (ECV) in the World Meteorological Organisation (WMO) Global Climate Observing System (GCOS). Carbon dioxide, CO₂, is described as “the dominant human produced greenhouse gas which has increased by about 50% since pre-industrial times due to the proliferation of fossil fuel combustion”. GCOS presents CO₂ measurements by the Scripps Institution of Oceanography (SOI) since 1958 and National Oceanic and Atmosphere Administration (NOAA) since 1982, from the Mauna Loa (MLO, 20°N) observatory, as evidence for the continuing global greenhouse change due to fossil fuel emissions.

International research into atmospheric change is coordinated by the Global Atmospheric Watch (GAW) program. GAW distinguishes 30 global stations from 400 + regional stations. The global sites have more international involvement, and for key components, including CO₂, provide ‘background’ data with minimal regional influence (S1 Text). MLO, with Kennaook-Cape Grim, 42°S (CGO), provide the strongest scientific support and facilities for intensive atmospheric composition programmes and campaigns in their respective hemispheres. As is the case in GCOS, the CO₂ at MLO has often been used to represent the global CO₂ trend; we will demonstrate increasing influence of Northern Hemisphere emissions complicated by interhemispheric exchange at low latitude sites. All data used in this manuscript are third party data that are publicly accessible. The Data Availability statement provides links to the data with further information and guidance for accessing the CO₂ data provided in S3 Text.

Under the United Nations Framework Convention on Climate Change (UNFCCC), recent CO₂ research has largely focussed on verification of national emissions based on associating growth rate changes to regional air-surface exchange. But the present study, with focus on anomalous behaviour in the 2009–2015 period, relies primarily on changes in the total mass of CO₂ in the atmosphere. We focus primarily on measurements made in CSIRO’s Global Atmospheric Sampling LABoratory (GASLAB). This choice is based on the following:

• Extensive trace gas monitoring of background (mainly marine boundary layer) air collected at 6 locations that span 71° of extra tropical Southern Hemisphere (SH) latitudes, plus MLO and Alert (ALT, 80°N).
• Ultra-high precision in CO₂ and its stable carbon isotope ratio results from pioneering small sample methodology for both parameters, maintained with minimal change since 1992.
• Pioneering “CGO same air” inter-comparability/verification with NOAA measurements have been routinely conducted since 1992.
• Radon²²² verification of CGO air mass history dates from 1992 and is available at MLO since 2004.

An introductory perspective is provided in Fig 1 showing 1992–2021 mean CO₂ absolute differences from a CGO baseline, using the 8 GASLAB sites plus 22 (mainly Northern Hemisphere (NH)) sites from the NOAA flask network plotted as a function of latitude. Data and method are provided in S1 Text. Where there is co–monitoring, there is agreement between the laboratories. Significance is attached to the lower envelope of the NH residuals described by a dashed red line, as indicating NH sites least affected by regional emissions; NH sites that show large positive variation in mean enhancements are consistent with persistent regional influences of power generation. On average, background sites at latitudes less than 30°N have lower residuals than the NH average, which we link to the large interhemispheric exchange flux. Sites above the Arctic Circle (66°N) are higher than average which we link to seasonality in south to north within hemisphere transfer of mid-latitude fossil emissions.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Average 1992-2021 CO₂ difference from a CGO baseline.

NOAA sites (red circles) and GASLAB sites (blue crosses, connected by a blue line) from the CGO baseline as a function of latitude (site details are in S1 Text). The dashed red line is a schematic representing a lower envelope of NOAA Northern Hemisphere data. I tried to replace the caption to Fig 1. but your system will not allow me to make changes to crossed text. All the changes to the all the figure captions have been uploaded. Please insert them in the appropriate places.

https://doi.org/10.1371/journal.pclm.0000682.g001

A simple calculation highlights the importance interhemispheric exchange for interpreting the observations, particularly in the Southern Hemisphere (SH). For example, with 95% of the current ~10 PgC yr^–1 of fossil CO₂ emitted north of the Intertropical Convergence Zone (ITCZ), and with a first order allocation of 2/3, 1/3 respectfully, of terrestrial and ocean sinks to this region, more than 2.6 Pg C yr^–1 arrives in the SH, mainly via equatorial upper troposphere exchange (S2 Text). This atmospheric flux dominates the magnitude of estimated ocean uptake south of the ITCZ (~1.5 Pg C), while terrestrial uptake in this region, ~ –0.9 Pg C yr^–1, is offset ~+1 Pg C yr^–1 by equatorial wildfires, implying minimal net terrestrial influence in the SH data.

Francey and Frederiksen [1] introduced an index, , (the strength of the 300 hPa westerly wind in the dominant Pacific westerly duct) as a measure of the eddy exchange between Northern (NH) and Southern (SH) hemispheres. The index was introduced to explain the unprecedented boreal winter–spring jump in concentration difference, C_MLO–C_CGO, in 2009–2010. Subsequently, Frederiksen and Francey [2] have defined indices that characterize the mean NH Hadley cell convective (the 300 hPa vertical velocity in pressure coordinates across the Pacific,) and advective (200 hPa meridional wind, ) transport between the NH and SH. The transport indices were used to describe the unusual late boreal spring and summer–autumn interhemispheric CO₂ exchange in 2015–2016. It was found in these studies that the MLO–CGO CO₂ difference has significant correlation with the index for Dec–May (peaking in Feb–Apr) and anticorrelation with the and indices for Jun–Nov (peaking in Jun–Aug). The exchange indices and their correlations with MLO–CGO difference are detailed in S2 Text.

Significant CO₂ interhemispheric transfer occurs when the sub–annual transport indices and CO₂ interhemispheric partial pressure difference coincide. The interhemispheric partial pressure difference is dominated by the large seasonality in NH mid–latitude forests. These factors are CO₂ specific, which means that representations of CO₂ interhemispheric exchange by other trace species, or by long term mean trace gas differences, are unlikely to be reliable indicators of the exchange. In other words, the conventional multi-year averaging and smoothing that reduces the atmospheric impact of the sub-annual eddy and mean interhemispheric exchange will underestimate this important exchange flux which dominates SH air-surface exchange and can influence low latitude NH data like that at MLO.

The plan of this article is as follows. In Section 2, we document the highly systematic nature of the variation and anomalous behaviour in the global atmospheric CO₂ and its stable isotope composition since 1992; Fig 2 (monthly time resolution) and Fig 3 (annual time resolution) summarize the findings. Section 3 presents a comparison of global interannual CO₂ background variation with El Nino–Southern Oscillation (ENSO), then with the ENSO variation suppressed with Fig 4 depicting the results. In Section 4, the important differences between studies using growth rates compared to carbon amount are addressed with the results presented in Fig 5. Section 5 then examines the roles of changes in interhemispheric transport exchange in the annual behaviour in CO₂ with the results shown in Fig 6. The global significance of the data is discussed in each case. Section 6 presents a summary and discussion of the results and our conclusions. All results and discussion are effectively model independent.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Evidence of the 2010-2015 anomaly in CGO raw data and MLO-CGO difference.

a. Individual GASLAB CO₂ (C), left axis, dark colour) and δ¹³CO₂*CO₂ (δ. C), right axis, light colour) from Mauna Loa (MLO, red) and Cape Grim (CGO, blue), including quadratic fits (extrapolated dashed curves) to each data set. The ellipse identifies data below the CGO quadratic. b. The 12-month running mean of C_MLO – _CCGO. Ellipses highlight the 2009–2015 anomalya. Individual GASLAB CO₂ (C), left axis, dark colour) and δ¹³CO₂*C (δ. C, right axis, light colour) from Mauna Loa (MLO, red) and Cape Grim (CGO, blue), including quadratic fits (extrapolated dashed curves) to each data set. The ellipse identifies data below the CGO quadratic. Fig 2b. The 12-month running mean of C_MLO – C_CGO. Ellipses highlight the 2009–2015 anomaly.

https://doi.org/10.1371/journal.pclm.0000682.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Global consistency of NH and SH annual residuals from MLO and CGO quadratic trends attributed to fossil emissions.

Step plots distinguish MLO (NH, red dots) and CGO (blue) from xy plots for other sites in each hemisphere (SH, blue/green; ALT orange dots). a. C residuals, with latitude offsets included in the legend. b. δ.C residuals. No normalisation offset is required for δ.C plots with ALT the only exception (see text). Shading highlights the persisting CO₂ anomalies below zero in 2009–2015.

https://doi.org/10.1371/journal.pclm.0000682.g003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Annual residuals for CGO (blue), MLO (red dot) and ALT (purple dash).

a. Comparison of the annual C residuals (Cres) in ppm on the left axis, with ENSO variation, using SOI* = SOI for the previous year (right axis, black). b. Variability of C* (left axis), the C residual adjusted for SOI influence using the slope of a linear regression through CGO C residuals and SOI* variation. Detrended GCP terrestrial (green) and ocean (light blue) flux interannual variations (right axis, see text).

https://doi.org/10.1371/journal.pclm.0000682.g004

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Comparison of anomaly (shaded) in amount versus growth rate.

a. Time series of annual residuals from hemispheric MLO and CGO baselines for different combinations of MLO and CGO or SPO data with the 2009–2015 CO₂ anomaly, shown by shading. b. Corresponding annual growth rates, calculated from the same data, with the anomaly not clearly discernible. Growth rates are plotted halfway between the two years defining them as their difference.

https://doi.org/10.1371/journal.pclm.0000682.g005

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Indices for N to S atmospheric transport.

a. Negative annual average NAO, indicating within NH mixing. b. Mean interhemispheric transport indices (Jun–Nov), with negative ω300P (pink, left axis) tracking the strength of convection in the NH Pacific Hadley cell, and negative v200P (blue, right axis) the North–to–South wind near the top of the NH Hadley cell. c. Eddy interhemispheric transport.a. Negative annual average NAO, indicating NH mixing. b. Mean interhemispheric transport indices (Jun–Nov), with negative (pink, left axis) tracking the strength of convection in the NH Pacific Hadley cell, and negative .

https://doi.org/10.1371/journal.pclm.0000682.g006

2. Sampling, measurement and CO₂ data

The CO₂ anomalies defined in this report are detected against two major systematic influences:

1. The 3–decade CO₂ (C) increase, and stable carbon isotope ratio (δ) decrease (represented by the conservative quantity δ.C [3]), are closely represented by quadratic functions. The quadratics are Taylor expansions of the exponential changes anticipated from the accumulation of annually increasing [4], long–lived [5] fossil CO₂ emissions. S2 Text includes the evidence that results are insensitive to quadratic, exponential or running mean representations of the trend. Over the three decades since 1992, we measure SH background CO₂ mixing ratios in dry baseline air rising from 354 ppm to 409 ppm, while reported fossil emissions totalled 219 PgC [4], monotonically increasing from 8 to 10 ppm yr^–1. Around 3.9 PgC of anthropogenic emissions, released predominantly at mid NH latitudes, have corresponded to a 1 ppm increase in the SH mixing ratio. The SH background dry CO₂ mixing ratios (marginally lower than in the NH), when converted to atmospheric mass, provide lower limits to global CO₂ amount ranging from 724 to 838 PgC. Quantifying variation in anthropogenic forcing of around 10 PgC yr^–1 against this background requires exclusion of regional influence and the maintenance of high precision, which we examine using residuals from the fossil fuel trend.
2. The next largest systematic variation, ~ ±1 ppm in CO₂ around the annual fossil trend, is consistent with a lagged response of equatorial terrestrial plants to ENSO forcing. Both inversion modelling of global CO₂ [6,7] and bottom–up ecosystem modelling [8], including wildfires [9], involve flux estimates that imply a sensitivity of ~2 PgC ppm^–1. We confirm that this interannual variation is characterised by similar phase and amplitude globally.

Consideration of major atmospheric circulation patterns explains the two different sensitivities to emissions. At the NH latitudes, emissions are entrained into Ferrel and Hadley Cell circulations with ample opportunity for modification by air–surface exchange prior to interhemispheric transport. In contrast, the ~ 2 PgC ppm^–1 sensitivity reflects forcing in equatorial regions. Specifically, it refers to the Intertropical Convergence Zone (ITCZ), where the dominant transport is vertical convection, with minimal opportunity for air surface exchange prior to global distribution. On this basis, the ~ 2 PgC sensitivity is not appropriate to describe non tropical air–surface exchange. This is supported by the sub-annual isotopic signatures in CGO and MLO data discussed in Section 2.2.

3. Role of ENSO in variation in CO₂ residuals

There is a ubiquitous ENSO influence on atmospheric CO₂ due to climate interaction with the terrestrial biosphere. We show that subtracting out the contribution to CO₂ variability from ENSO in fact strengthens the anomaly signal. Of particular interest is the period after the 2008–2009 Global Financial Crisis (GFC). This overlaps the time of the 2010 record central Pacific El Niño [10] through to the strong 2015–2016 El Niño focused on the east Pacific and stretching to the centre [11].

Thoning et al. (1989) [12], using MLO data, found a negative correlation coefficient of between growth rate and the Southern Oscillation Index (SOI) with the SOI minimum occurring approximately 5 months before a maximum in growth rate. However, maximum CO₂ growth rate occurs well before the consequent maximum in C residuals (S4 Text) and the peak magnitudes of January to December average SOI correlations with monthly C residuals occur at just over a year (13 months for the NH and 14 for the SH). Table 1 shows the correlations of calendar year annually averaged residuals for both MLO and CGO with annually averaged negative SOI or positive Niño3.4 indices. Here, the SOI is the standardized Tahiti minus Darwin sea level pressure and the Niño3.4 index measures the average sea surface temperature anomaly in the area between 5^oS–5^oN, 120^oW–170^oW. There is little correlation between CO₂ residuals and SOI in the same year but the response to the previous year SOI (–SOI*) is strong. We use negative SOI to emphasize the similarity between C residuals. (Note: the precautionary use of correlation coefficients from 1995, to minimize the possible influence of Pinatubo, proved academic).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Correlation coefficients of CO₂ residuals at CGO and MLO with ENSO indices, SOI and Niño3.4, at the same year and previous year (bold).

https://doi.org/10.1371/journal.pclm.0000682.t001

The stronger delayed correlation at MLO compared to CGO is consistent with unusual interhemispheric transport in 2009–2010, discussed in relation to Fig 4. The overall similarity between SOI and CO₂ amount residual variation is displayed in Fig 4a. It reveals, for example, that SOI linked CO₂ variation based on growth rates and previously attributed to variation in 1994–2007 Asian anthropogenic emissions reduction [13], is instead explained by climate forcing. The similarity also supports the lack of reduced atmospheric response to COVID in 2019–2020 [14].

We define C^*, shown in Fig 4b, for which the influence of SOI is subtracted from C_res using the linear regression line of the annual carbon residuals versus the negative SOI from the previous year. The regression slope ± standard error is 0.037 ± 0.009 ppm per the SOI unit (S3 Text).

Multiyear negative anomalies occur in 1993–1994 and in 2009–2015. In the earlier period they are similar at MLO and CGO and insignificant at ALT; in the later period ALT and CGO anomalies are large but are small at MLO until 2015. We attribute the earlier event to the Pinatubo volcanic explosion at 18°N, normally linked to net uptake due to ocean cooling [15]. Surprisingly, the isotopic data suggest a significant terrestrial contribution, such as the suggested photosynthetic stimulation of equatorial forests because of light scattered by particulate matter [16].

In Fig 4b, a perspective on the 2009–2015 C anomaly comes from comparison with the annual variation in the detrended Global Carbon Project (GCP) modelled global ocean and terrestrial biosphere exchange variations [4], plotted on the right axis. These are scaled by the 3.9 PgC ppm^–1 to describe the three–decade SH change due to accumulated annual fossil emissions and their transport from the NH. The anomaly area below zero in Fig 4b is 2.7 ppm which converts to a –11.5 PgC total anomaly. The GCP modelling demonstrates the volatility of the terrestrial exchange compared with that of oceans, thus also supporting the improbability of the SH oceans contributing significantly to the anomaly. In Fig 4b, the mean magnitude in CGO annual residuals between 2009 and 2015 is negative with value –0.47 ± 0.075 PgC yr^-1. This magnitude significantly exceeds measurement uncertainties of less than ~0.1 PgC yr^-1 (S1 Text). With CGO variation statistically indistinguishable from that at other extra-tropical SH sites (Fig 3), this provides a major constraint on the global carbon budget. We note that a gap in MLO 2003 data resulted in no annual average value that year (Table A of S3 Text). In determining the statistical significance of residuals from the MLO 3-decade trend, an interpolated 2003 annual concentration was employed.

The absence of a similar decrease in the MLO C residuals around 2010 is consistent with reduced interhemispheric exchange, temporarily enhancing the build–up of the NH fossil emissions. This influence of interhemispheric exchange on low latitude NH sites is consistent with an examination of NOAA data residuals from the CGO quadratic in Fig 1.

4. Residual anomalies and total growth rates

In Fig 5a we reinforce the robust nature of the 2009–2015 anomaly by demonstrating that annual residuals from the 3–decade MLO and CGO quadratics are insensitive to different combinations of site and analysing laboratory. Averaging residuals from individual NH and SH sites and using CSIRO or NOAA data, effectively reproduce the same interannual variability and persisting 2009–2015 reduced atmospheric background CO₂ levels.

In Fig 5b we show that growth rates calculated from annual differencing of CO₂ concentration are also in robust NH and SH agreement, but significantly, show no clear evidence of the Fig 5a carbon amount anomaly, and this agrees with earlier growth rate studies where the anomaly has not been reported [17]. Fundamental dynamical differences between growth rates and carbon residual amount are presented in S4 Text together with a detailed explanation. Essentially annual growth rates, being differences between annual values, have greater scatter than the smoother residual amounts which also have delayed response in their peaks and troughs. We also note that if annual residuals are calculated from monthly growth-rate data obtained using residuals from a 650-day trend curve [12] then the sub-annual interhemispheric exchange will not be accurately captured.

5. Roles of atmospheric CO₂ transport

The first–order analysis of a recent annual global CO₂ budget, with ~10 PgCyr^-1 emissions in the Northern Hemisphere and with sinks distributed according to land/ocean area, indicates a cross–hemisphere exchange of ~ 3 PgC year^–1. Excluding fossil emissions, this flux exceeds any reported hemispheric annual air–surface exchange. Thus, interhemispheric exchange variation is a critical determinant of CO₂ global distribution.

The main paths for NH to SH interhemispheric exchange of CO₂ are by mean convective and advective transport in late boreal spring through summer to early autumn and by eddy transport in late winter and spring; they were introduced in Section 1 and detailed in S2 Text.

Fig 6 explores the relationship between CO₂ residuals and anomalies in interhemispheric exchange. Fig 6 shows the yearly variability of the Dec–May index, the Jun–Nov and indices as well as the Jan–Dec annual North Atlantic Oscillation (NAO) index. Here, the NAO index is determined by the strength of the leading rotated empirical orthogonal function of the NH 500 hPa geopotential height (S3 Text).

The period of the 2010–2015 CO₂ anomaly is indicated by shading. The NAO index is a measure of large scale within–NH mixing. During the positive phase of the NAO (and the Arctic Oscillation, AO) the more zonal flow provides an impediment to within NH mixing. In contrast during the negative phases the more meridional wave–like circulation enhances mixing. (For example, the correlation, between CO₂ at ALT and the NAO, for three–month averages starting each month from December to June). At MLO corresponding NAO correlations for three-month averages are negative starting from January to April, but quite small, with the most significant for April-June of . The signs of all indices in Fig 6 are chosen so that increased uplift and North–to–South mixing are upwards excursions on the graphs. A striking feature of the circulation in Fig 6 is the weakness of NH convective uplift () and mean North–to–South transport () of the NH Hadley circulation, leading into the commencement of the CO₂ anomaly. Indeed, the mean interhemispheric transport () remains relatively weak until 2016, when it increases considerably, while the uplift () reaches a maximum there, at the end of the anomaly period. At the start, in 2010, the interhemispheric eddy transport of CO₂ into the SH () is near zero leading to a particularly weak total, mean plus eddy, exchange. This is also the time of maximum CO₂ mixing between the high and low latitudes of the NH. The interhemispheric eddy exchange increases during the La Niña of 2011 and then decreases until 2016. Variation in interhemispheric exchange is expected to result in complementary changes in CO₂ concentration in each hemisphere, as observed.

6. Discussion and conclusions

Early in the 3–decade record, examination of residuals from the exponential fossil fuel trends in CO₂ have clarified the global atmospheric impact of the Pinatubo volcano, with δ.C indicating a significant contribution from the terrestrial biosphere. At the end of the record, we note the absence of negative residuals in 2019–2020, providing no evidence for a reported reduction in emissions accompanying the COVID–19 pandemic [14].

In between, the characteristics of an unprecedented anomaly found in global baseline CO₂ concentration between 2009 and 2015 have been analysed. The anomaly manifests as a negative excursion in annual concentration about the exponential growth of CO₂, matched by a positive excursion in stable carbon isotope ratio (Fig 3). The anomaly is most clearly and uniformly defined in the SH, including in the CGO data, which are rigorously selected for insignificant terrestrial influence, and isotopically exclude net ocean exchange.

Mole fraction variations that are the result of interdependent emissions and uptake in the terrestrial biosphere are effectively suppressed by annual averaging that eliminates seasonality and by a delayed ENSO correction to suppress interannual climate forcing. The variation in the interannual residuals reveal anomalous behaviour between 2009 and 2015 of –0.47 ± 0.075 PgC yr^-1 compared to sampling and measurement uncertainties estimated at less than ~0.1 PgC yr^-1.

We have cross checked the anomaly in the MLO, CGO, MLO–CGO and average of MLO and CGO data through two further means. Firstly, the residual concentration for the difference and average have been checked against constructing the residuals from the differences and averages of the separate residuals for MLO and CGO. The results are extremely close with total and detrended correlations greater than 0.97 in all cases. Secondly, the calculations have been repeated for the decadal (high-pass) anomalies from the 10-year running mean with total and detrended correlations for the four cases essentially 0.94 or greater.

For the MLO–CGO CO₂ difference there is considerable cancellation of the exponential growth curves, and the trend is more closely just a moderate linear trend. In this case the anomaly stands out unambiguously in the 12-month running mean of the raw data (Fig 2b) and is also evident in the monthly raw data.

We have noted the reasons why the anomaly is not easily seen in growth rates (Fig 5 and S4 Text) and expect this is why it has not previously been reported on.

The complex processes determining the global CO₂ distribution makes attribution of the causes of the anomaly difficult, and this has not been our aim. However, we have examined factors known to be significant in the global carbon cycle, including circulation and transport variations in the anomaly period bracketed by the 2010 record central Pacific El Niño and the 2015–2016 major east to central Pacific El Niño. We have noted the weak interhemispheric mean and eddy transport in 2010. In the period prior to the CO₂ anomaly both the convective uplift ( in Fig 6b), in the NH Pacific branch of the Hadley Cell, and the mean North-to South transport ( in Fig 6b) take relatively low values. As well, the eddy transport ( in Fig 6c) from the NH to SH is negligible in 2010. In this year, the negative phase of the NAO (Fig 6a) is also a maximum resulting in strong meridional mixing in the NH and reduction of CO₂ at ALT (Figs 3 and 6). During the following period, the mean interhemispheric transport () stays weak until 2016, when it and the uplift () increase considerably.

Of possible significance is that the anomaly period corresponds to a transition of the global cycle when the photosynthetic dip in NH seasonal CO₂ no longer falls below the mean SH concentration because of the increased fossil fuel emissions in the NH. That is, a SH to NH positive partial pressure difference no longer exists for part of the year. A further complicating factor may be increased uncertainty in global (NH) fossil emissions prior to the Conference of Parties in 2016 (COP21) which led to adoption of more uniform methodologies to estimate and integrate national global emissions.

The anomaly provides a constraint on conventional CO₂ transport models and, also a strong reason for further causal studies with such models. Our data driven findings throw direct light on some recent carbon cycle studies overlapping the anomaly period while others may benefit from re-examination with CO₂ transport models that can reproduce the anomaly. In Section 3, the data for the anomaly in Fig 4 is tabulated for convenience.

The 2010–2015 anomaly in SH background CO₂ is a large, robust, and systematic feature of global significance. It requires further analysis with CO₂ transport models and causal explanation to verify and improve global carbon modelling that informs future human response to anthropogenic climate change and ocean acidification. The SH anomaly offers an unprecedented opportunity in recent global CO₂ studies, to assess global emissions and interhemispheric exchange influences.

Supporting information

Acknowledgments

The data used in this study are the result of longterm institutional support from the Australian Bureau of Meterology, The Australian Antarctic Division, The Australian Institute of Marine Science, The US National Oceanic and Atmosphere Administration (NOAA), Environment and Climate Change, Canada. Staff in GASLAB and at the Cape Grim Baseline Air Pollution Station deserve special mention, along with those at other observing sites. Dr. Chin Wan of NOAA provided advice on the use of NOAA data.