Influence of El Niño on atmospheric CO₂ over the tropical Pacific Ocean: findings from NASA’s OCO-2 mission

Abstract

Space-borne observations of CO₂ from the Orbiting Carbon Observatory-2 are used to characterize the response of the tropical atmospheric CO₂ concentrations to the strong El Niño event of 2015–2016. Correlations between atmospheric CO₂ growth rate and the El Niño Southern Oscillation have been well known; however, the magnitude of the correlation and the timing of the responses of oceanic and terrestrial carbon cycle remain poorly constrained in space and time. Here we use space-based CO₂ observations to confirm that the tropical Pacific Ocean does play an early and important role in modulating the changes in atmospheric CO₂ concentrations during El Niño events – phenomenon inferred but not previously observed due to lack of high-density, broad-scale CO₂ observations over the Tropics.

El Niño Southern Oscillation, or ENSO, is the dominant mode of tropical climate variability on interannual to decadal timescales (1–5) and is correlated with large inter-annual variability in global atmospheric CO₂ concentrations (6–19). Studying the response of the carbon cycle to this natural climate phenomenon is critical to understand and quantify the sensitivity of the carbon cycle to climate variability, and by extension to climate generally (20). Although the ENSO cycle originates in the equatorial Pacific, its impact on the carbon cycle is felt globally due to its regional teleconnections (22–23) and influences on atmospheric and ocean circulation, precipitation, temperature, and fire emissions (1, 24–26). Partitioning the response of the constituent components of the carbon cycle to a complete El Niño event has been challenging because of the limited number of CO₂ observations over the tropical land and ocean regions.

Observations of atmospheric CO₂ from space provide a global view of the carbon cycle that can be used to describe phenomena that have been previously pieced together from sparse in situ data. NASA’s Orbiting Carbon Observatory-2 (OCO-2) mission was successfully launched on July 2, 2014 and started providing science data in early September 2014 (70). Within the first two years of operation of the OCO-2 mission, a major El Niño (warm phase of the ENSO) occurred (27–30). We provide an approach for studying the temporal sequence of El Niño-induced changes in global CO₂ concentrations using observations from the OCO-2 mission that are validated with the Tropical Atmosphere Ocean (TAO) mooring CO₂ data. We see a response from the tropical Pacific Ocean during the early stages of an El Niño event and a lagged (and much larger) terrestrial signal as the El Niño reaches maturity.

El Niño and the global carbon cycle

Correlations between the atmospheric CO₂ growth rate and El Niño activity have been reported since the 1970s (6–8, 31–32), although the magnitude and timing of the responses of the ocean and terrestrial components remain poorly constrained (33). Here, the word terrestrial includes both changes in biospheric productivity (respiration and photosynthesis) as well as biomass burning (fires). Following previous strong El Niño events (for example, the 1982–1983 and 1997–1998 El Niño events), methods for measuring the atmospheric CO₂ response to ENSO were based on in situ atmospheric CO₂ observations at a handful of surface stations that transect the tropical Pacific, including Mauna Loa, Christmas Island and American Samoa (8, 34) as well as shipboard transect measurements (12, 35–36). The annual growth rate of atmospheric CO₂ measured at these remote stations and other sites around the globe show remarkable correlation with ENSO indices, with a rapid increase in atmospheric CO₂ associated with the late stage of an El Niño event (19, 37). The ocean response to El Niño events is based on studies looking at in situ observations, for example, surface ocean pCO₂ observations from ships of opportunity (12), moorings (38–39), or targeted field campaigns during El Niño events (9–10, 40–41), and a variety of mechanistic ocean models (24, 52, 54, 61, 65, 67).

The overall increase in the release of CO₂ to the atmosphere during strong El Niño events has been attributed to a decrease in biospheric uptake of CO₂ (e.g., due to drying of tropical land regions and an increase in plant and soil respiration) combined with enhanced fire emissions. In recent years, this has led to a growing body of literature (42–49) concluding that ENSO-mediated variability in tropical net land primary productivity is what primarily influences the atmospheric CO₂ growth rate. A handful of studies (25, 50–51) have disputed any consistent or coherent response from the land component during El Niño events, thus highlighting the high level of uncertainty and disagreement within the carbon cycle community.

The El Niño-CO₂ signature should have a significant tropical Pacific Ocean component as well, with opposite sign to the terrestrial response (10, 13, 33). During strong El Niño events, there is a large-scale weakening of the easterly trade winds and suppression of eastern equatorial Pacific upwelling (indicated by a deeper thermocline) that reduces the supply of cold, carbon-rich waters to the surface (Fig. 1). This reduces the usual strong outgassing of CO₂ from this region (52–67), typically on the order of ~0.4–0.6 PgC yr⁻¹ to the atmosphere, by ~40–60% during an El Niño event (9–12, 33, 36, 60, 73). If net fluxes were to remain constant elsewhere, these substantial net air-sea CO₂ anomalies should lead to a reduction in the growth rate of atmospheric CO₂, at least during the early stages of El Niño.

Fig. 1.

Schematic of the mechanistic differences between normal (A) and El Niño (B) conditions and associated carbon response over the tropical Pacific Ocean. Warm ocean surface temperatures are denoted in red and cooler waters in blue. During El Niño conditions, easterly trade winds weaken and westerly wind bursts occur. In association with the shift in wind regimes, the western tropical Pacific warm pool moves eastward and the slope of the thermocline flattens in the central and eastern tropical Pacific. This suppresses upwelling of cold, carbon-rich waters in the central and eastern tropical Pacific, reducing the magnitude of CO₂ outgassing into the atmosphere. Also shown are changes in atmospheric convection, wherein convection shifts eastward in response to eastward displacement of western tropical Pacific warm pool waters.

Understanding these variations in atmospheric CO₂, the timing of these variations and the underlying processes that cause them have been of great interest within the carbon cycle community (1, 10–13, 15, 20, 33, 50). Integrating information from ocean- and atmosphere-based estimates, and modeling studies, we now know that it is the combined and opposite effect of ocean and terrestrial responses, which contribute to El Niño-related variations in atmospheric CO₂ (33). What remain controversial though are the timing of the ocean response and a precise quantification of its role. This is of crucial importance because typically the interannual variability (IAV) in the growth rate of atmospheric CO₂ is used to constrain the climate sensitivity of land carbon fluxes (ϒ_LT) (20–21); however, if a component of the IAV is being modified by ocean fluxes, then these inferences of ϒ_LT need to be reconsidered.

Because of the few surface CO₂ monitoring stations over the center of action (i.e., tropical Pacific Ocean), it has been challenging to directly observe the timing and changes in flux of CO₂ from the ocean to the atmosphere that affect the atmospheric CO₂ growth rate during an El Niño event. Efforts to analyze the data from distant measurement locations tend to identify the enhanced CO₂ fluxes from the terrestrial carbon cycle, which dominate during the later stages of El Niño. The high-density, broad-scale observations of CO₂ from OCO-2 provide a valuable tool to partition the ocean and terrestrial carbon cycle responses to El Niño.

Time series of X_CO2 anomalies during the 2015–2016 El Niño

OCO-2 observations describe the column-averaged CO₂ dry air mole fraction (X_CO2). More details regarding the OCO-2 mission, data features, X_CO2 retrievals, etc. are provided in the Supplementary Materials, and are available in (69) and (70) while validation of X_CO2 via comparisons to a ground-based network are provided in (71).

El Niño events are identified by warm sea surface temperature anomalies in precise regions of the tropical Pacific Ocean, with the most commonly used being the Niño 3.4 region (5°S–5°N, 170°W–120°W). Figs. 2A and 2B show the trend in X_CO2 anomaly (90) for the Niño 3.4 region and its temporal evolution relative to two ENSO indices (97), including the Oceanic Niño Index (ONI - derived from sea surface temperature anomalies in the Niño 3.4 region) and the Southern Oscillation Index (SOI - derived from observed sea level pressure differences between Tahiti and Darwin, Australia). The 2015–2016 El Niño began around March 2015 and reached its peak over the Central Pacific between November 2015 and January 2016 (30). The X_CO2 anomaly (Fig. 2B) shows two distinct periods over the entire El Niño event: (a) Development phase of El Niño (Spring-Summer 2015) – we argue that the negative X_CO2 anomaly is due to a reduction in local CO₂ outgassing from the tropical Pacific Ocean, and (b) Mature phase of El Niño (Fall 2015 onwards) – we argue that the positive trend in X_CO2 anomaly reflects an increase in atmospheric CO₂ concentrations due to terrestrial sources (i.e., combination of reduced vegetation uptake across pan-tropical regions and enhanced biomass burning emissions from SE Asia and Indonesia). The time series in Fig. 2B shows the space-based CO₂ dataset documenting the response of the carbon cycle (both oceanic and terrestrial) during an entire El Niño event, capturing both the development and the mature phase and the transition between those two. The timing of the OCO-2 launch was extremely fortuitous in this regard.

Fig. 2.

OCO-2 observes the response of the carbon cycle for an entire El Niño event. Temporal evolution of (A) the 2015–2016 El Niño as captured by the ONI and the SOI indices, (B) X_CO2 anomalies and associated uncertainties in the Niño 3.4 region, (C) ΔpCO₂ from the TAO 0°, 170°W mooring, and (D) the CO total column anomalies in the Niño 3.4 region.

Deriving the X_CO2 anomalies require observations taken by both NASA’s OCO-2 and the Japan Aerospace Exploration Agency’s (JAXA) Greenhouse Gases Observing Satellite (GOSAT) (68) mission. The short OCO-2 record makes it impossible to fit a long-time series and calculate anomalies, and hence data from the GOSAT mission (operating since January 2009) was utilized to generate the X_CO2 climatology. The OCO-2 team retrieved X_CO2 from the first 7 years of the GOSAT observations using the same retrieval algorithm that generated the OCO-2 data product (90). Continuous global coverage from these two missions allows us to stitch together a long-time series of X_CO2 over remote regions, such as the tropical Pacific Ocean (Figs. S1–S2). However, utilizing two data sources, i.e., GOSAT and OCO-2, can incur errors in the analyses due to changes in the two instruments, their observing strategies and sampling density. Fig. 2B also illustrates the corresponding uncertainty in our analyses. The uncertainty is calculated using an ensemble technique (Section C in Supplementary Materials) and further brings out the two phases in the time series of the Niño 3.4 X_CO2 anomaly – ±0.3 ppm uncertainties during the El Niño development phase with both the upper and lower bounds below the zero line, and larger uncertainties of ±0.5 ppm during the mature phase of the El Niño event. These larger uncertainties during the latter stages of the El Niño illustrate the challenge in attributing the changes in X_CO2 anomalies to the competing, and often opposing, signals from the ocean and the terrestrial components of the carbon cycle.

Attributing the two observed phases of X_CO2 anomalies to the ocean and the terrestrial response

Our argument for the two observed phases in the X_CO2 anomaly time series is supported by complementary data sources. The ocean response is corroborated by sea surface pCO₂ observations from an in situ network of autonomous CO₂ systems on the TAO moored buoy array (9, 38, 72). Data are not directly comparable to atmospheric X_CO2 as they describe CO₂ variations at the ocean surface. The trend of the difference between the sea surface and atmospheric CO₂ (ΔpCO₂), however, does capture typical El Niño signatures. For example, Fig. 2C illustrates data from one of the moored buoys in the Niño 3.4 region (0°, 170°W), which shows decreasing ΔpCO₂ over the spring and near-zero ΔpCO₂ by December 2015. A suppression in the upwelling of CO₂-rich waters caused by weakening of the easterly trade winds leads to a reduction in the surface ocean carbon content, which in turn leads to a decline in the magnitude of sea-to-air CO₂ fluxes. The flux estimates at this buoy location are 1.35 ± 0.21 (1σ) gC m⁻² month⁻¹ during the November 2014 to February 2015 period (i.e., non El Niño conditions) that gradually decrease to 0.087 ± 0.083 (1σ) gC m⁻² month⁻¹ between November 2015 and February 2016 (i.e., El Niño conditions). This indicates a near-total shutdown of sea- to-air flux during Boreal Winter 2015–2016 relative to the neutral 2014–2015 Boreal Winter. Previous studies focusing on the tropical Pacific Ocean have reported flux reductions of ~40– 60% over the entire basin (9–12, 33, 36, 60, 73). Atmospheric transport model calculations with a prescribed set of flux patterns and comparing to the observed X_CO2 anomalies (Section A in Supplementary Materials) suggest a flux reduction of ~26–54%.

While these numbers are roughly similar, we do recognize the limitation in comparing flux estimates from one point (namely the TAO location at 0°, 170°W) to flux estimates for the entire Niño 3.4 region and/or the tropical Pacific Ocean from previous studies. Large-scale changes in the physical and biogeochemical dynamics during El Niño events result in significant spatial and temporal variability in the surface pCO₂ distributions (12, 62, 65). Additionally, these spatial variations and their seasonal progression are uniquely tied to each El Niño event; thus, different flavors of El Niño events and/or shifts in the El Niño phenomena (86–88) will influence the evolution of the seasonal cycle of pCO₂ and air-sea CO₂ fluxes over the region. For the 20152016 El Niño event, the TAO buoy at 0°, 170°W lay closest to the edge of the warm pool and registered the first response to the onset of El Niño conditions. As observations from other TAO locations (92) are becoming available, it is evident that in the eastern part of the basin there was an overall suppression of the outgassing CO₂ source but with large variability in pCO₂. Data synthesis and modeling work with these and other in situ observations are ongoing to quantify the exact magnitude of ocean CO₂ fluxes over different tropical Pacific regions during the 2015– 2016 El Niño.

The second phase in the X_CO2 anomaly time series is driven by the terrestrial component of the carbon cycle, and the transport of this signal to the remote Niño 3.4 region. The anomalous increase in CO₂ can be attributed to a combination of terrestrial sources, including a reduction in the global biospheric uptake, increases in soil and plant respiration and enhanced fire emissions. In fact, the impact of enhanced fire emissions and their regional progression was a well-studied feature following the strong 1997–1998 El Niño (26, 43, 74–76). For the 2015–2016 El Niño event, strong correspondences between X_CO2 from OCO-2 and the carbon monoxide (CO) total column anomalies from the Measurements of Pollution in the Troposphere (MOPITT) instrument on the NASA Terra platform, are evident over the tropical Pacific Ocean, especially during Fall 2015 (Fig. 2D). We conjecture that these CO total column anomalies are representative of the emissions from the 2015–2016 Indonesian peat fires (77–80), which were advected into the tropical Pacific region. El Niño-related changes in the Walker circulation (i.e., westerly winds) and the slightly more southern than normal positioning of the Inter Tropical Convergence Zone (ITCZ) (81) may allow emissions from the Indonesian peat fires to carry over into this region (Fig. S4). It is interesting to note from Figs. 2B and 2D that the positive increase in X_CO2 anomaly actually leads the fire signals by 1–2 months. This indicates that the release of carbon flux resulting in an increase in CO₂ concentrations is only partially pyrogenic; reduced vegetation uptake due to droughts is a significant contributor, and quite possibly the initial cause of the increase in X_CO2 anomaly.

Isolating the observed negative X_CO2 anomaly to an ocean signal

The time dependence of the X_CO2 anomalies during the 2015–2016 El Niño indicate that the initial decrease in atmospheric CO₂ is due to suppression of upwelling in the tropical Pacific. This early negative response is subsequently offset by a large positive anomaly due to the terrestrial component. Assuming no significant interannual changes elsewhere in the global ocean, we can further confirm our argument by a comparison of the X_CO2 anomaly in the Niño 3.4 region with the global X_CO2 anomaly (Fig. 4A). By differencing the far-field effect from the local signal, the influence of the reduction in CO₂ outgassing from the tropical Pacific Ocean is clearly visible during the onset phase of El Niño. The peak reduction registered over the Niño 3.4 region relative to the global X_CO2 anomalies is 0.35 ppm in June 2015, which occurs a couple of months after the initiation of the El Niño event. Lag correlation of the Niño 3.4 X_CO2 anomalies against the ONI index indicate that the highest positive correlation occurs when the concentration-related anomalies lag the SST-related anomalies by 1–2 months (93) (Fig. S8). The time lag relationship can be precisely quantified during the onset phase of El Niño, but it is much more difficult to interpret during the succeeding El Niño stages when any reduction in CO₂ from decreased equatorial upwelling is masked by the signal from terrestrial processes. Thus, if it were not for the reduction in outgassing from the ocean, the impact from the terrestrial sources would likely be larger. Our analysis confirms the findings from (13) that the slowdown of atmospheric CO₂ increase during the early stages of an El Niño is indeed related to the decreased sea-to-air flux of CO₂ in the tropical Pacific Ocean. The coverage from the OCO-2 mission has enabled us to verify this hypothesis and monitor its temporal evolution using real atmospheric CO₂ observations.

Fig. 4.

Difference in X_CO2 anomalies between the Niño 3.4 region and (A) the globe, (B) the tropical Atlantic Ocean, (C) the subtropical Pacific Ocean from September 2014 to May 2016. Definitions of the regions are provided in Fig. 3. In Panel (A), we see a robust pattern of negative X_CO2 anomaly between Niño 3.4 and the globe that is largest in 2015 and well synchronized with the onset phase of El Niño. In Panels (B) and (C), non-zero differences between Niño 3.4 and the other ocean basins indicate that the Niño 3.4 trend is not reproducible in other ocean basins; thus, allowing us to attribute the negative anomaly in Fig. 2B to a reduction in local CO₂ outgassing over the tropical Pacific Ocean.

The early stage negative X_CO2 anomaly is unique to the tropical Pacific Ocean and is not influenced by global, terrestrial or large-spatial scale fluxes. Due to the large interhemispheric gradients in CO₂, typical variability in tropical CO₂ concentrations can be an aliasing of terrestrial processes occurring at higher latitudes. In order to confirm that the recovered ocean signal in the X_CO2 anomaly is unique to the tropical Pacific Ocean, we examined three other ocean regions - the subtropical North Pacific (20°–30°N, 120°–170°W), the subtropical South Pacific (20°–30°S, 120°–170°W) and the tropical Atlantic Ocean (5°N–5°S, 5°–35°W). Fig. 3 shows the specific regions (aside from Niño 3.4) that we have analyzed, and each of which assist us to reject alternative hypotheses. Non-zero differences in X_CO2 anomalies between these and the Niño 3.4 region (Fig. 4) indicate that the trend observed over the tropical Pacific Ocean is distinct from other ocean basins. This makes intuitive sense from our mechanistic understanding as well - while large impacts of ENSO on the sea-to-air CO₂ flux in the tropical Pacific Ocean are expected, studies have shown minute and delayed influence of the ENSO modes on the variability of carbon fields in the tropical Atlantic Ocean (66, 82–83).

Fig. 3.

Schematic showing the specific ocean basins (Niño 3.4, N. Pacific, S. Pacific and Trop. Atlantic) that were analyzed in this study. X_CO2 anomalies are calculated for these different ocean basins, and subsequently compared to the X_CO2 anomalies from the Niño 3.4 region. Each of these regions was considered to accept/reject a specific hypothesis that could potentially bias the observed trend in the Niño 3.4 X_CO2 anomalies. After rejecting these hypotheses, we conclude that the negative X_CO2 anomaly observed over the Niño 3.4 during the onset phase of El Niño 2015–2016 is unique and has to be driven by local changes in the ocean fluxes.

Perspective

The strong El Niño in 2015–2016 caused a reduction in the magnitude of CO₂ outgassing from the tropical Pacific Ocean. These changes, albeit of varying magnitude, extended over a large portion of the tropical Pacific, and impacted the large-scale modulation of the physical processes responsible for the CO₂ efflux from this region. Almost all observing networks (i.e., OCO-2, TAO, etc.) were aided by the strength of this signal. However, OCO-2 provided a more comprehensive view of the tropical Pacific Ocean signal than previous observing networks given its: (a) greater coverage and more frequent sampling than in situ networks, and (b) improved resolution and precision than earlier space-based instruments. For example, GOSAT, like OCO-2 is sensitive to the total CO₂ column, but has lower precision (2 ppm single sounding random error for GOSAT vs. 0.5 ppm for OCO-2) and lower sampling density (100× less soundings). The immediate next step will be to fold in these observations into an inverse modeling framework (13, 15, 50, 56) to infer the underlying net fluxes between the ocean and atmosphere and the terrestrial biosphere and the atmosphere. This would help establish the real benefit of OCO-2, especially against the backdrop of previous studies that had to rely on sparse atmospheric constraint to infer changes in CO₂ surface fluxes during El Niño events.

Based on OCO-2 data alone, however, we cannot quantitatively discriminate the relative roles of reduction in biospheric activity uptake due to a warmer and drier climate in 2015 versus enhanced fire emissions. While we can quantify the temporal response of the ocean versus the terrestrial component and qualitatively observe the gradients in the response of different tropical Pacific Ocean regions (Fig. 5), it is much more challenging to discriminate the contribution of fire emissions and the delayed response of the terrestrial biosphere to El Niño-induced changes in weather patterns. The impact of ENSO is typically felt by the terrestrial biosphere over several months to a year after the actual event. Studies on both progressions of droughts (84) and fires (26) during an El Niño cycle have shown a hysteresis in the Earth system’s response to changes in temperature and precipitation patterns. Analyses using ancillary data sources such as solar- induced fluorescence (SIF), bottom-up model simulations and inverse modeling calculations are typically necessary to quantify the partitioning of the terrestrial carbon fluxes (reduction in biospheric uptake vs. increase in fire emissions) as has been pursued in a companion study (85).

Fig. 5.

Time evolution of the X_CO2 anomalies (ppm) averaged over 5°S–5°N. The x-axis represents longitude and the y-axis shows the time progressing from top to bottom in months. The 2015–2016 El Niño event and its onset and mature phases are highlighted to show the distinct responses observed over the tropical Pacific Ocean. The grey dashed lines capture the boundaries of the Niño 3.4 region. During the onset phase (i.e., March – July 2015), perceptible gradients are observable from the far western Pacific to the central Pacific (consistent with the increasing flux from west to east) along with high variability in the X_CO2 anomalies in the central Pacific. We also notice that the X_CO2 anomalies are smaller over the eastern Pacific, which is consistent with surface seawater pCO₂ data collected on the TAO buoys (92). The transition from the ocean to the terrestrial signal happens between July and October 2015. Towards the latter stages of the El Niño event (i.e., November 2015 and later), the terrestrial signal dominates the observed trends in X_CO2 likely masking any underlying ocean signal.

Our study provides a short-term perspective on the potential of CO₂ observations from space for unraveling more complex relationships of carbon sources and sinks in the future. A longer time series of observations will enable testing more hypotheses such as the possibility of regionally dependent gradients in air-sea CO₂ fluxes in the tropical Pacific, or adding data to support biogeochemical theories at previously inaccessible scales. From a long-term perspective, such information will improve our process-based understanding, inform our current suite of mechanistic models, and ultimately, better constrain future carbon cycle projections.

Concluding remarks

The strong El Niño event of 2015–2016 provided us with an opportunity to study how the global carbon cycle responds to changes in the physical climate system. With the high-resolution (both spatial and temporal) observations available from OCO-2, we are able to directly: (a) observe the strong correlations that exist between atmospheric CO₂ concentrations and the El Niño signal, and (b) track the development of the atmospheric CO₂ anomaly as it switches from a negative phase (i.e., due to a reduction in CO₂ outgassing from the tropical Pacific Ocean) to a strong positive phase (i.e., due to a reduction in biospheric uptake and increased fire emissions). The most important contribution of the space-based OCO-2 mission is the ability to observe and monitor carbon cycle phenomena at high-density over large spatial scales, which has not been possible from the existing in situ network.

The complexity of the El Niño – CO₂ signature illustrates that it is a multifaceted system with contributions from many regions and processes. Understanding and predicting its behavior requires separating out the many terrestrial and marine regions that contribute (1, 33) and identifying both the geophysical (3, 27, 30) and the biological (10, 59, 89) phenomena that respond in their own unique ways. However, the impact on the carbon cycle is unified through the global mixing of CO₂ in the atmosphere - OCO-2 makes a unique contribution by providing both the global coverage and fine surface spatial detail; alongside the in situ CO₂ network of moorings and shipboard measurements provide the long-term climate-quality record of atmospheric and ocean CO₂ observations and serves to validate the OCO-2 observations and model products. We emphasize that this diverse observing portfolio is necessary, and the complementary information provided by these observing systems will likely prove critical in understanding the partitioning of carbon fluxes during the 2015–2016 El Niño, the relative contribution of ocean vs. land to the global atmospheric CO₂ growth rate, and the sensitivity of the carbon cycle to climate forcing on interannual to decadal timescales.

One Sentence Summary.

NASA’s OCO-2 mission provides a first-hand look at the space-time evolution of tropical atmospheric CO₂ concentrations in response to the 2015–2016 El Niño