Whole-atmosphere CO₂ growth rate from satellite observations

GRESO (Growth Rate from Satellite Observations) averages OCO-2 data between 50°S and 50°N, where coverage is good all year; CO₂ added or removed further poleward mixes into this band within weeks to months, so the result tracks the whole atmosphere.

1. We average the good OCO-2 measurements from each 16-day orbit cycle and date each average at the mean time of its measurements.
2. We remove the seasonal cycle and smooth the result over about four months.
3. Monthly growth is the rise of this curve over each month, given as a rate per year; the chart shows its 5-month running average.
4. Annual growth is the rise from 1 January to the next 1 January.

The 12-month value in the annual table is worked out the same way, over the latest 12 months of data.

By default (“OCO-2 + GOSAT mean”), the annual chart and table show GOSAT alone for 2010–2014 and the average of OCO-2 and GOSAT from 2015.

Method paper: Pandey, S., Miller, J. B., Basu, S., Liu, J., Weir, B., Byrne, B., Chevallier, F., Bowman, K. W., Liu, Z., Deng, F., O'Dell, C. W., & Chatterjee, A. (2024). Toward low-latency estimation of atmospheric CO₂ growth rates using satellite observations: Evaluating sampling errors of satellite and in situ observing approaches. AGU Advances, 5, e2023AV001145.

Each annual value has an uncertainty (1σ, one standard deviation) of ±0.07 to ±0.11 ppm a year. It combines three sources of error: differences between subsets of the satellite data, the amount of smoothing, and gaps in where and when the satellite can look. GOSAT's uncertainty combines smoothing uncertainty with a fixed 0.08 ppm a year; it does not include the satellite-subset component.

Monthly uncertainty is roughly ±0.7 ppm a year; its full assessment is still in progress. Each year ends on the same 1 January that starts the next, so errors in neighbouring years partly cancel. Don't treat them as independent when you average several years.

NOAA's annual uncertainty is the one given in its annual file.

Each month's value is smoothed with data from the months around it, so the newest months are early estimates that change as later data arrive. A month becomes provisional after three more months of data and final after six. Provisional values can still shift, by about 0.7 ppm a year after 3 months and 0.24 ppm a year after 5 months [revision sizes from an earlier pipeline run; recompute on this release]. In this release, April 2026 and earlier are final, May–July 2026 provisional and August–September 2026 early. A year's value is final once data through February of the next year are in, usually by April.

OCO-2 measures the average CO₂ in the whole column of air below it, so its global average follows the total CO₂ in the atmosphere. NOAA measures air at the surface, at remote ocean sites, and is the calibrated reference for the long-term rise. Over many years the two agree, but single years can differ, partly because CO₂ from land takes time to reach remote sites and the upper atmosphere. The two can differ enough to matter for the global carbon budget (Pandey et al. 2025, Nature Communications).

• OCO-2 (NASA's Orbiting Carbon Observatory-2): Lite XCO₂ (column-average CO₂) data from NASA GES DISC, version 11.2r through 2023 and version 11.3r from 2024, up to 30 September 2026, then preliminary data for October 2026 from the OCO-2 team's forward processing, which the public data will replace.
• GOSAT (Japan's Greenhouse gases Observing SATellite): ACOS GOSAT v11 Lite data from NASA GES DISC, April 2009 to October 2026.
• NOAA: the global surface average from NOAA's Global Monitoring Laboratory (Lan, Tans and Thoning): its monthly CO₂ file and annual growth-rate file, both of 9 October 2026. We calculate the monthly NOAA line ourselves, from the trend in the monthly file.
• Emissions: fossil fuels and land use. The yearly totals through 2024 are from the Global Carbon Budget 2025 (Friedlingstein et al. 2026). For the monthly line, each year is spread over the months with Carbon Monitor data (Liu et al. 2020), and the seasonal cycle is removed. From 2025 the values are a first estimate (“low-latency” in the files): Carbon Monitor's fossil emissions scaled to match the Global Carbon Budget, plus land-use emissions continuing their recent trend.
• Climate stripes: differences from the 1991–2020 average, leaving out 2011, for sea-surface temperature in the Niño-3.4 region of the Pacific, where El Niño is measured, and for temperature, air dryness and rainfall over tropical land. Air dryness is measured by the vapour-pressure deficit (VPD): the difference between how much water vapour the air could hold at its temperature and how much it holds. The land rows come from ERA5, the Copernicus Climate Change Service's reconstruction of past weather (Hersbach et al. 2020), and the Niño-3.4 row from NOAA's ERSST v5 sea-surface temperatures (Huang et al. 2017). The Niño-3.4 row uses this average rather than NOAA's official Oceanic Niño Index, so it can read higher.
• Units: 1 part per million (ppm) of CO₂ in the whole atmosphere is 2.124 petagrams of carbon (PgC, Global Carbon Budget convention). A petagram is a billion tonnes, and with the oxygen included 1 PgC is 3.664 billion tonnes of CO₂. So 1 ppm a year is 2.124 PgC, or 7.8 billion tonnes of CO₂ (GtCO₂), a year.

Contact: Sudhanshu Pandey, sudhanshu.pandey@caltech.edu

Pandey, S., O'Dell, C. W., Frankenberg, C., Payne, V. H., & Wennberg, P. O. (2026). Whole-atmosphere CO₂ growth rate from satellite observations, release v2026.11 [web page]. OCO Science Applications and Partnerships Center, Caltech. [page DOI].