Research Article

The Orbiting Carbon Observatory-2 early science investigations of regional carbon dioxide fluxes

See allHide authors and affiliations

Science  13 Oct 2017:
Vol. 358, Issue 6360, eaam5745
DOI: 10.1126/science.aam5745

Structured Abstract


Earth’s carbon cycle involves large fluxes of carbon dioxide (CO2) between the atmosphere, land biosphere, and oceans. Over the past several decades, net loss of CO2 from the atmosphere to the land and oceans has varied considerably from year to year, equaling 20 to 80% of CO2 emissions from fossil fuel combustion and land use change. On average, the uptake is about 50%. The imbalance between CO2 emissions and removal is seen in increasing atmospheric CO2 concentrations. In recent years, an increase of 2 to 3 parts per million (ppm) per year in the atmospheric mole fraction, which is currently about 400 ppm, has been observed.

Almost a quarter of the CO2 emitted by human activities is being absorbed by the ocean, and another quarter is absorbed by processes on land. The identity and location of the terrestrial sinks are poorly understood. This absorption has been attributed by some to tropical or Eurasian temperate forests, whereas others argue that these regions may be net sources of CO2. The efficiency of these land sinks appears to vary dramatically from year to year. Because the identity, location, and processes controlling these natural sinks are not well constrained, substantial additional uncertainty is added to projections of future CO2 levels.


The NASA satellite, the Orbiting Carbon Observatory-2 (OCO-2), which was launched on 2 July 2014, is designed to collect global measurements with sufficient precision, coverage, and resolution to aid in resolving sources and sinks of CO2 on regional scales. Since 6 September 2014, the OCO-2 mission has been producing about 2 million estimates of the column-averaged CO2 dry-air mole fraction (Embedded Image) each month after quality screening, with spatial resolution of <3 km2 per sounding. Solar-induced chlorophyll fluorescence (SIF), a small amount of light emitted during photosynthesis, is detected in remote sensing measurements of radiance within solar Fraunhofer lines and is another data product from OCO-2.


The measurements from OCO-2 provide a global view of the seasonal cycles and spatial patterns of atmospheric CO2, with the anticipated year-over-year growth rate. The buildup of CO2 in the Northern Hemisphere during winter and its rapid decrease in concentration as spring arrives (and the SIF increases) is seen in unprecedented detail. The enhanced CO2 in urban areas relative to nearby background areas is observed with a single overpass of OCO-2. Increases in CO2 due to the biomass burning in Africa are also clearly observed. The dense, global, Embedded Image and SIF data sets from OCO-2 are combined with other remote sensing data sets and used to disentangle the processes driving the carbon cycle on regional scales during the recent 2015–2016 El Niño event. This analysis shows more carbon release in 2015 relative to 2011 over Africa, South America, and Southeast Asia. Now, the fundamental driver for the change in carbon release can be assessed continent by continent, rather than treating the tropics as a single, integrated region. Small changes in Embedded Image were also observed early in the El Niño over the equatorial eastern Pacific, due to less upwelling of cold, carbon-rich water than is typical.


NASA’s OCO-2 mission is collecting a dense, global set of high-spectral resolution measurements that are used to estimate Embedded Image and SIF. The OCO-2 mission data set can now be used to assess regional-scale sources and sinks of CO2 around the globe. The papers in this collection present early scientific findings from this new data set.

El Niño impact on carbon flux in 2015 relative to 2011, detected from Greenhouse Gases Observing Satellite (GOSAT) and OCO-2 data.

OCO-2 uses reflected sunlight to derive Embedded Image and SIF. This shows OCO-2 Embedded Image data over North America from 12 August 2015 to 26 August 2015.


NASA’s Orbiting Carbon Observatory-2 (OCO-2) mission was motivated by the need to diagnose how the increasing concentration of atmospheric carbon dioxide (CO2) is altering the productivity of the biosphere and the uptake of CO2 by the oceans. Launched on 2 July 2014, OCO-2 provides retrievals of the column-averaged CO2 dry-air mole fraction (Embedded Image) as well as the fluorescence from chlorophyll in terrestrial plants. The seasonal pattern of uptake by the terrestrial biosphere is recorded in fluorescence and the drawdown of Embedded Image during summer. Launched just before one of the most intense El Niños of the past century, OCO-2 measurements of Embedded Image and fluorescence record the impact of the large change in ocean temperature and rainfall on uptake and release of CO2 by the oceans and biosphere.

Large fluxes of carbon dioxide (CO2) between the atmosphere, land biosphere, and oceans (1) occur within Earth’s carbon cycle. The exchange varies seasonally, with net carbon uptake into the terrestrial biosphere during the growing season, especially in the Northern Hemisphere. In the fall and winter, photosynthesis declines in the mid- and high latitudes, and plant respiration exceeds photosynthesis, returning CO2 to the atmosphere. Continuing emissions of CO2 from fossil fuels adds carbon to the atmosphere, mostly in the Northern Hemisphere (2). The uptake of CO2 from the atmosphere into the land and oceans constitutes between 20 and 80% of CO2 emissions from fossil fuel and land use change (3, 4) and is ~50% on average. The balance—that is, the fraction of anthropogenic carbon release not reabsorbed by the Earth System—is referred to as the “airborne fraction.” It is manifest in the rising burden of atmospheric CO2, whose concentration is increasing 0.50 to 0.75% each year [2 to 3 parts per million (ppm) per year increase in the atmospheric mole fraction] (5).

Subtle geographic and temporal variations in atmospheric concentrations of CO2, of fractions of a ppm to several ppm of the ambient ~400 ppm background, reflect the underlying uptake and release of carbon. These variations provide clues to the underlying mechanisms that drive differences in the airborne fraction. Measurements of the increasing inventory of carbon in seawater indicate that almost a quarter of the CO2 emitted by human activities is being absorbed by the ocean (6), where it contributes to ocean acidification. Mass balance demands that another quarter of the CO2 emitted by human activities must be absorbed by processes on land. The identity and location of these sinks are less well understood. Some studies have attributed this absorption to tropical (7) or Eurasian temperate (8) forests, whereas others indicate that these areas are just as likely to be net sources as net sinks of CO2 (9). The efficiency of these natural land and ocean sinks also appears to vary dramatically from year to year (3). Because the identity, location, and processes controlling these natural sinks are not well constrained, it is not clear how they will respond in the future (7). Understanding these mechanisms and their dependence on climate and atmospheric CO2 levels is central to understanding how the carbon cycle may amplify or mitigate future climate change (3, 7, 1013).

Measuring CO2 from space with OCO-2

The international network of ground-based in situ greenhouse gas measurement stations provides a long-term and precise (~0.07 ppm) record of the atmospheric CO2 concentration at ~147 locations across the globe (14, 15). Few measurements are obtained in tropical regions, in urban settings, or in Asia. The NASA satellite, the Orbiting Carbon Observatory-2 (OCO-2), which was launched from Vandenberg Air Force Base in California on 2 July 2014, is designed to collect global measurements with sufficient precision, coverage, and resolution to aid in resolving sources and sinks on regional scales. After completing a series of spacecraft check-out activities and orbit-raising maneuvers, on 3 August 2014, it joined the front of the Afternoon Constellation (A-Train) (16), which consists of six satellites orbiting at an altitude of 705 km. In this 98.8-min orbit, OCO-2 samples at a local time of about 1:30 p.m., and it has a set of 233 orbit paths that repeat in 16-day cycles. The OCO-2 sampling strategy repeats in 32-day cycles. Since 6 September 2014, the OCO-2 instrument has been routinely returning almost 1 million soundings each day over the sunlit hemisphere. This measurement concept was developed in the late 1990s, but this type of data has only been collected since 2014. The OCO-2 mission is a replacement for the original OCO instrument and spacecraft that were lost in a failed launch in February 2009; only small modifications to replace obsolete parts and to adapt to a different launch vehicle were required.

The OCO-2 spectrometers collect eight spatially resolved radiance spectra of reflected sunlight in three narrow wavelength bands three times per second, with resolving power (Δλ/λ) of ~17,000 (Fig. 1) (1719). The oxygen A-band (centered at ~765 nm) provides a sensitive measure of the atmospheric path length and is thus an accurate indicator of clouds and surface elevation. The radiance at two distinct CO2 absorption bands (1.61 and 2.06 μm) provides sensitivity to CO2. The column-averaged atmospheric CO2 dry-air mole fraction (Embedded Image), or the total column of carbon dioxide normalized by the column of dry air, is derived from these spectra using a physics-based retrieval method (2022). The sensitivity of the measurement is fairly uniform throughout the troposphere and lower stratosphere and varies with solar geometry and surface (23). Details of the instrument calibration and observatory operation are reported in Crisp et al. (19), and the data processing strategy is described by Eldering et al. (24).

Fig. 1 OCO-2 detects sunlight that has traveled through the atmosphere and is reflected back to space.

The sunlight is partially absorbed by the O2 A-band (A) and the weak and strong CO2 bands centered near 1.61 (B) and 2.06 μm (C).

A set of eight measurements are made along a narrow ground track as the spacecraft travels about 2.25 km along its orbit track, providing spatial resolution of <3 km2 per sounding. Embedded Image is retrieved only when there is sufficient light (solar zenith angles less than 85°) and when there are no optically thick clouds (25). OCO-2 returns roughly 2 million Embedded Image estimates each month that pass quality screening (Fig. 2), for a yield of about 6% of the total soundings.

Fig. 2 Maps of OCO-2 Embedded Image.

Maps of OCO-2 Embedded Image (bias corrected with quality flags applied) over 32-day periods in (A) March/April 2015, (B) June/July 2015, and (C) March/April 2016. The measurement area of each sounding has been exaggerated for visibility on a global scale.

Systematic biases in the OCO-2 Embedded Image estimates must be minimized to accurately resolve the small spatial and temporal variations in this quantity. The bias correction, designed to address known systematic errors, takes advantage of the high density of soundings gathered over regions with minimal variability in Embedded Image (such as contiguous data collected in small areas and over the Southern Hemisphere oceans) (26, 27). To relate the OCO-2 Embedded Image estimates to the standard CO2 scale set by the international CO2 in situ network, we use the transfer standard provided by a specially designed ground-based network of atmospheric observatories that comprise the Total Carbon Column Observing Network (TCCON) (28). Extensive comparisons have been made between OCO-2 satellite measurements coincident with the TCCON measurements (29, 30). After correcting biases, the OCO-2 Embedded Image retrievals have median differences [for collections of >100 soundings] of less than 0.5 ppm and root-mean-square differences that are typically below 1.5 ppm (30). Simulation studies conducted before the launch of OCO estimated that large climate anomalies, like the 2003 European summer drought, which created a carbon anomaly of 0.5 gigatons C (31), would be detected by OCO, whereas scientific questions about smaller changes (on the order of 0.01 gigatons C) would not be addressed by OCO (32). The systematic biases from OCO-2 are consistent with the prelaunch design studies. As discussed in Chatterjee et al. (33), concentration changes of 0.5 ppm regionally can be detected with the OCO-2 measurements.

The interpretation of OCO-2 measurements is enhanced by measurements from the other Earth-observing satellites. The availability of measurements of trace gases such as nitrogen dioxide (NO2) from the Ozone Monitoring Instrument (OMI), carbon monoxide (CO) from the Measurements of Pollution in the Troposphere (MOPITT) experiment, and many products from the Moderate Resolution Imaging Spectroradiometer (MODIS), have been used to disentangle the influence of complex and variable processes that contribute to the global carbon cycle (3437).

Measuring plant fluorescence from space with OCO-2

Measurements of solar-induced chlorophyll fluorescence (SIF) from satellites offer insight into terrestrial gross primary productivity (GPP), the gross uptake of CO2 through photosynthesis (38). The SIF signal, a small amount of light emitted during photosynthesis, is detected in remote sensing measurements of radiance within solar Fraunhofer lines. Retrieval methods were developed in recent years with the Japanese Greenhouse Gases Observing Satellite (GOSAT) Thermal and Near-Infrared Sensor for Carbon Observation–Fourier Transform Spectrometer (TANSO-FTS), Global Ozone Monitoring Experiment-2 (GOME-2) onboard MetOp-A, and OCO-2 measurements, and its potential for quantifying GPP is being assessed (3842). Although the SIF signal is quite small—enhancements are typically less than 2% of the reflected sunlight (43)—the high signal-to-noise spectra from OCO-2 enable precise SIF measurements at high spatial resolution (37). Typically, the random component of the retrieval error varies between 0.3 and 0.5 Wm−2 μm−1 sr−1 (15 to 25% of typical peak values of SIF) in the 757-nm fitting window (44), but the errors are substantially reduced by a factor of Embedded Image if single retrievals (from individual soundings) are binned to gridded maps (n is the number of soundings per grid cell) at certain temporal averaging domains. In a companion paper in this issue, Sun et al. (37) describe OCO-2 SIF characteristics in detail and illustrate mechanistic connections between SIF and GPP. They show that when OCO-2 data are compared with GPP from flux tower measurements, well matched in spatial scale, they have correlation coefficients ranging from 0.89 to 0.99, with similar slopes for three different biomes. Earlier studies that used sparse data sets that had to be interpolated over time indicated biome-specific linear relationships.

Observing the carbon cycle from space with OCO-2

Maps of the Embedded Image data collected over three 32-day periods illustrate the most robust features of the carbon cycle (Fig. 2). The March/April 2015 distribution is characterized by enhanced Embedded Image in the Northern Hemisphere. During winter, uptake by plants is minimal while the breakdown or decay of plant material continues. As a result of this, together with the continual emissions from fossil fuel burning (e.g., elevated Embedded Image over China, Europe, and the southeast United States), Embedded Image reaches a seasonal maximum in the Northern Hemisphere during April just before temperatures increase enough to reawaken the biosphere from the low activity of winter. As illustrated in Fig. 3, active photosynthesis is manifest in the strong signal of plant fluorescence across the vegetated springtime Northern Hemisphere.

Fig. 3 The OCO-2 SIF retrieval at 757 nm on 1° by 1° grid for the spring (i.e., the mean of April-May-June for 2015 and 2016).

By June/July 2015, the distribution of Embedded Image and fluorescence observed by OCO-2 has changed dramatically compared with March/April (Fig. 2 for Embedded Image). Although fossil fuel emissions continue, the uptake of CO2 by the terrestrial biosphere [e.g., figure 1 of (37)] has removed a large amount of CO2 from the atmosphere over much of the Northern Hemisphere; the latitudinal gradient of Embedded Image has reversed. In 2015, OCO-2 observations indicate that the springtime drawdown began in Europe and propagated eastward across Asia and North America over the months of May and June. In some regions, Embedded Image declined by 7 ppm in only 1 month. Movie 1 illustrates the CO2 from an atmospheric analysis, a product that merges OCO-2 observations with a high-resolution global model using a technique called data assimilation (45). The incorporation of OCO-2 observations corrects errors in the model’s prediction of atmospheric CO2 concentrations, and the model provides additional information about the vertical distribution of the gas and fills gaps in cloudy, data-sparse regions. Spanning March 2015 through the end of July 2015, the rapid reductions in CO2 concentrations in the Northern Hemisphere during June are evident in movie 1, as are the complex pathways that transport CO2 through the atmosphere, across oceans and continents.

By March/April 2016, the distribution of Embedded Image is similar to the year before, but with an increased concentration of ~3.5 ppm globally. A substantial fraction of this increase reflects emissions from fossil fuel burning. Direct evidence for these emissions includes the 1- to 3-ppm enhancements in Embedded Image over the regions with intense industrial activity. For example, Schwandner et al. (46) observe a local, persistent enhancement of 4 to 6 ppm in Embedded Image between OCO-2 measurements across the Los Angeles basin and the measurements that extend to the desert region to the north. The basin provides an ideal setting for such analyses, with a large urban population and accompanying emissions, and mountains to the north of the city, which trap air in the basin and provide a clear demarcation from the background region to the north. This is illustrated in Fig. 4 and in Schwandner et al. (46). More broadly, the spatial enhancements of Embedded Image due to the burning of fossil fuel across the Northern Hemisphere are illustrated in Fig. 5. Hakkarainen et al. (34) combined OCO-2 Embedded Image measurements with space-based observations of NO2 from the OMI instrument, as well as the Open-Source Data Inventory for Anthropogenic CO2 (ODIAC) emissions inventory (47). Using cluster analysis, they identified Embedded Image enhancements clearly linked to fossil fuel combustion, which are shown in Fig. 5.

Fig. 4 OCO-2 Embedded Image measurements across the Los Angeles basin and into the desert north of Los Angeles taken during one overpass on 8 September 2015.

The measurement swath is 10 km across. Two special sets of measurements taken for validation purposes—one at the California Institute of Technology and one at the Armstrong Flight Research Center—are also displayed.

Fig. 5 Maps of the OCO-2 Embedded Image anomaly (mean in each grid box of the daily anomaly from the regional median) in 1° by 1° cells between September 2014 and April 2016.

The anomalies are only plotted for the regions identified as clusters of enhancements due to fossil fuel burning in (34).

Another large signal seen in the OCO-2 data is the effect of seasonal biomass burning in Africa on the Embedded Image concentrations (Fig. 6). CO2 accounts for more than 90% of annual global fire carbon emissions in current emission inventories (48), and fire emissions are typically enhanced during El Niño periods. These emissions have typically been estimated from models rather than direct observations. Uncertainties in the extent of the burned area, the biomass density within the burned area, and the fraction of biomass emitted as CO2, CO, and other species compromise the accuracy of the estimates (4951). Top-down constraints on pyrogenic CO2 could therefore provide a much-needed check on fire emissions estimates.

Fig. 6 Maps of OCO-2 Embedded Image over sub-Saharan Africa for the beginning of the biomass burning seasons of 2015 and 2016, showing the rapid regional increase in Embedded Image.

The data have been averaged to 2° by 2° bins each month, after bias correction and quality screening were applied.

OCO-2 measurements were used to estimate the CO2 emissions from Indonesian fires in 2015 (35). Indonesia experienced an exceptional number of fires in 2015 due to El Niño–related drought and slash-and-burn agricultural practices. Emissions databases such as the Global Fire Assimilation System (GFASv1.2) and the Global Fire Emission Database (GFEDv4s) estimated the CO2 emission to be ~1100 megatons CO2 between July and November 2015. Heymann et al. (35) analyzed OCO-2 Embedded Image observations collected over Indonesia during this period using two different modeling approaches. They estimate pyrogenic CO2 emissions near 731 ± 271 megatons CO2. This estimate is 37 and 31% lower than those in the GFASv1.2 and GFEDv4s emissions databases. Interestingly, the OCO-2 based estimates are consistent with pyrogenic CO2 emissions estimates based on CO measurements from the MOPITT instrument on the Terra platform and fire radiative power estimates from Terra and Aqua MODIS (692 ± 213 megatons CO2) (52). Hakkarainen (34) also clearly sees the enhanced Embedded Image from biomass burning in his anomaly analysis, although their results are aggregated over time, so the seasonality is not reported. The Northern Hemisphere African biomass emissions peak in January each year (48, 53) and have a duration of 4 to 5 months. Figure 6 illustrates the growth phase of that cycle for 2 years of OCO-2 measurements.

Time series of the OCO-2 Embedded Image estimates clearly show the seasonal cycles and the latitudinal differences in those seasonal cycles that are similar to the record collected by ground-based networks. Figure 7 shows weekly average Embedded Image values for the South Pacific, the ocean around Hawaii, and a region over Europe and Asia. The South Pacific has a relatively flat seasonal cycle, because there are limited emissions and limited uptake by the terrestrial biosphere in this region. Around Hawaii, there is a stronger seasonal cycle, because it is influenced by the Northern Hemisphere springtime removal of CO2 from the atmosphere by the terrestrial biosphere and the growth in atmospheric CO2 in the winter, when human emissions are not balanced by natural removal mechanisms. Over Europe and Asia, there is a similar seasonal cycle, but the springtime Embedded Image reductions are more rapid, because these measurements are where the terrestrial biosphere is active.

Fig. 7 Time series of weekly OCO-2 Embedded Image averages for regions around Hawaii, the Southern Pacific Ocean, and EuroAsia, showing the contrast of the seasonal cycle in the Northern and Southern Hemispheres that is clearly observed by OCO-2.

The data have bias correction and quality screening applied.

The impact of the 2015–2016 El Niño on the carbon cycle

The massive 2015–2016 El Niño contributed to the anomalously large Embedded Image growth rate. The OCO-2 mission started approximately 6 months before the beginning of the El Niño. The 3 ppm global increase in Embedded Image recorded during this El Niño is one of the largest ever observed (4, 54), consistent with previous research that has shown that global CO2 increases anomalously during and in the year after large El Niños (5559).

Diagnosis of the specific mechanisms responsible for the large CO2 growth rates (e.g., the relative importance of changes in the ocean, the humid tropics, and the semiarid tropics), has been challenging due to a lack of observations of CO2 in those regions (60). Data from the OCO-2 mission thus provides a window into the response of the ocean and land carbon cycle to this large-scale climate perturbation (7). Leveraging the broad coverage of OCO-2 data, Chatterjee et al. (33), Liu et al. (36), Heymann et al. (35), and Sun et al. (37) report on the quantification of CO2 emissions sources and insights into the carbon cycle response to El Niño. These studies examine the role of ocean outgassing, drought, and fire as contributors to the increased growth rate of atmospheric CO2.

Chatterjee et al. (33) used OCO-2 Embedded Image observations to study the temporal evolution of Embedded Image anomalies over the tropical Pacific Ocean. Using a combination of data from OCO-2, the TAO (Tropical Atmosphere Ocean) moored buoy array network (61) and MOPITT, they identify two distinct phases in the response of atmospheric CO2—an early response driven by reduction in CO2 outgassing from the tropical Pacific Ocean followed by a lagged and much larger response driven by increased fluxes from the tropical land. To further elucidate the relationship between regional climate forcing and tropical biosphere carbon response, Liu et al. (36) contrast the 2015 carbon responses to 2011, which was a weak La Niña year with near-average temperature and precipitation over the tropical continents. They quantify net biosphere exchange (NBE) (i.e., the combined effects of respiration, fire, and GPP) for 2015 and 2011, respectively, by assimilating Embedded Image observations from OCO-2 and GOSAT into the NASA Carbon Monitoring System Flux (CMS-Flux) inversion system. To further partition the NBE into GPP, biomass burning, and residual respiration carbon fluxes, they optimized GPP and biomass burning fluxes with SIF from GOSAT (62) and CO observations from MOPITT (63), respectively.

The impacts of El Niño on the carbon cycle are complex (33, 36): Temperature and rainfall changes in Southeast Asia, Africa, and South America are distinct, resulting in diverse carbon cycle impacts. Embedded Image decreased over the tropical Pacific Ocean, but flows of carbon were larger to the tropical atmosphere over all three continents. Over South America, dry conditions reduced GPP, resulting in a net increase in the flux of carbon to the atmosphere. Over Africa, higher atmospheric temperatures drove increased respiration (Reco) but near normal GPP, increasing carbon flux to the atmosphere. Southeast Asia experienced higher temperatures and dry conditions, increased vulnerability to fire from land use, and increased emissions of CO2. See Liu et al. (36) for additional discussion.

OCO-2 measurements in context of other remote sensing data

OCO-2 is not the first instrument to measure CO2 from space, but its data have unique characteristics relative to existing data sets. Space-based measurements of CO2 have been made in the thermal infrared beginning in the early 2000s with the Atmospheric Infrared Sounder (64) and are now being made by several other instruments. However, extracting surface source/sink information from these measurements has been largely unsuccessful, owing to their low sensitivity to near-surface CO2, which provides the most information on surface exchange (65, 66). The SCanning Imaging Absorption spectroMeter for Atmospheric CHartographY (SCIAMACHY) instrument made near-infrared measurements of column CO2 from 2002 to 2012, although with relatively coarse spatial resolution (30 × 60 km2) and lower sensitivity (4 to 8 ppm) (6770). The Japanese GOSAT mission launched in 2009 (71, 72) was the first mission whose primary goal was to measure greenhouse gases (carbon dioxide and methane) from space. The GOSAT mission has fostered considerable international scientific collaboration, leading to a deeper understanding of the utility of total column CO2 measurements from space. The GOSAT CO2 observations have formed the backbone of a number of important scientific studies. The primary limitation of the GOSAT measurement scheme is its low sounding density, with a single, 85-km2 measurement per 250 km, resulting in fewer than 1000 cloud-free soundings each day.

The CO2 seasonal cycle has also been studied with SCIAMACHY and GOSAT data (e.g., 73, 74-76). The GOSAT measurements have been used to characterize a number of relatively large disturbances to the carbon cycle, including reduced carbon uptake in 2010 due to the Eurasia heat wave (77), larger than average carbon fluxes in tropical Asia in 2010 due to above-average temperatures (78), and anomalous carbon uptake in Australia (79). Parazoo et al. (62) used GOSAT Embedded Image and SIF estimates to better understand the carbon balance of southern Amazonia. Ross et al. (80) used GOSAT data to obtain information on wildfire CH4:CO2 emission ratios. Buchwitz et al. (67) provide an excellent overview of the SCIAMACHY and GOSAT remote sensing data sets.

The OCO-2 measurements have a higher spatial resolution than GOSAT and SCIAMACHY and include a larger number of measurements per day. OCO-2 was designed as a sampling mission, not a mapping mission, so it only samples a small fraction of the globe each day. Although it would be desirable to have high-precision measurements over the whole globe daily, current limitations in remote sensing create a trade-off in sampling coverage and measurement precision (81), and OCO-2 has been designed to have sparse sampling and high precision (18, 82). This trade-off allows OCO-2 to capture the data required for assessing regional fluxes of CO2 across the globe (32, 82). Additionally, OCO-2 Embedded Image high precision allows the detection of small changes in regional concentrations (33), including the observations, from a single overpass, of gradients across cities that the measurement path happens to cross (46).

Similarly, SIF has been derived from SCIAMACHY (41, 83), GOME-2 (84) and GOSAT measurements (39, 84) as well as OCO-2. Although these data sets span a longer period of time than OCO-2, the OCO-2 SIF product has a smaller footprint (<3 km2 for OCO-2 versus 2400 km2 for GOME-2, 1800 km2 for SCIAMACHY, and 85 km2 for GOSAT) and a higher signal-to-noise ratio (44, 85). As discussed in detail in Sun et al. (37), the OCO-2 SIF data provide higher single-sounding precision then the other data sets, with reduced spatial coverage of the globe. These characteristics are valuable for improving our mechanistic understanding, because the OCO-2 data spatial resolution is well matched to ground-based measurements and the scales of heterogeneity in many ecosystems.

Looking forward, NASA’s current plan calls for continued development of the space-based CO2 and SIF record with OCO-3, which will be deployed on the International Space Station (ISS) no earlier than the fall of 2018. Although the OCO-2 and OCO-3 instruments are very similar (the core spectrometer used by OCO-3 was the flight spare for OCO-2), differences in the OCO-2 and ISS orbits and observing capabilities will further enhance the value of simultaneous measurements from these two sensors. In particular, from its near-polar, sun-synchronous orbit, OCO-2 can sample most of the globe but can only measure Embedded Image and SIF at ~1:30 p.m. local time. In contrast, although the moderate inclination of the ISS orbit restricts OCO-3 coverage to ±51° latitude, the orbit precesses in time, enabling Embedded Image and SIF observations from dawn to dusk. Finally, although OCO-2 can collect targeted measurements over only one to two sites each day, OCO-3 will use its fast-pointing mechanism to acquire thousands of measurements over up to a hundred 70 km by 70 km targets each day. Combining the OCO-2 and OCO-3 data sets will therefore enable carbon cycle investigations that require uniform sampling of the globe, as well as sampling of the diurnal cycle or compact sources, such as megacities.

The combined OCO-2/OCO-3 climate data record will provide a valuable baseline for the Geostationary Carbon Cycle Observatory (GeoCARB) mission, which was recently selected by the NASA Earth Ventures Program (86). GeoCARB will be NASA’s first greenhouse gas sensor in geostationary orbit. If all goes as planned, GeoCARB will be ready for launch no earlier than 2021. It will deployed at 85°W longitude, where it can produce continuous global maps of Embedded Image, Embedded Image, and XCO over the North and South American continents. Internationally, a number of near-infrared CO2 measurements are beginning or planned, including China’s TanSat, which was launched in December 2016 (87, 88), the Japanese GOSAT-2, planned to launch in 2018 (89), and the French Space Agency’s (Centre National d’Etudes Spatiales) MicroCARB, with a planned launch in 2020 (90).

The record of SIF measurements will also expand greatly in the future. Sensors that will return SIF measurement include OCO-3, TROPOspheric Monitoring Instrument (TROPOMI) (91), ESA FLEX (92), GOSAT-2, MicroCARB, and GeoCARB. As with CO2, these new missions will both extend the SIF record in time and provide new capabilities, such as sampling over a range of daylight hours and with a range of spatial resolutions.


The dense, global, Embedded Image, and SIF data sets from GOSAT and OCO-2 are being combined with data from MODIS, OMI, and MOPPIT and used to disentangle the processes driving the carbon cycle on regional scales. The accompanying reports in this collection use these data to discriminate the impacts of fossil fuel emissions, fires, and the 2015–2016 El Niño on the atmospheric CO2 budget. A longer data record is needed to document the carbon cycle’s response as the tropical climate relaxes back to its background state. An even longer record will be needed to fully characterize the interactions between the present-day carbon cycle and climate system. This information is crucial for the development and validation of improved coupled carbon-climate models for predicting the carbon cycle’s response to a warming climate. Fortunately, as OCO-2 completes its 2-year prime mission and begins its first extended mission, the spacecraft and instrument remain healthy, and data products with improved accuracy and coverage are in development.

References and Notes

Acknowledgments: Retrieved Level 2 OCO-2 Embedded Image (version v7Br) data used in this study are archived in a permanent repository at NASA’s Goddard Space Flight Center’s Earth Sciences Data and Information Services Center (GES-DISC) and are also available at NASA’s Jet Propulsion Laboratory ( Part of the research described in this paper was carried out at the Jet Propulsion Laboratory, California Institute of Technology, under a contract with the National Aeronautics and Space Administration. The movie was created by B. Weir, L. Ott, S. Pawson, H. Mitchell, and G. Shirah at Goddard Space Flight Center and the Scientific Visualization Studio. B.W., L.O., and S.P. were supported by the NASA Carbon Monitoring System and the OCO-2 Science Team NASA Research Opportunities in Space and Earth Sciences (ROSES) projects. K. Yuen assisted with figure production. J.H. and J.T. were supported by the Academy of Finland Inversion Algorithms and Quantification of Uncertainties in Atmospheric Remote Sensing (INQUIRE) (grant number 267442) and Carbon Balance under Changing Processes if Arctic and Subarctic Cryosphere (CARB-ARC) (grant number 285630) projects.

Stay Connected to Science

Navigate This Article