Estimating atmospheric CO2 from reflected sunlight

Column-averaged atmospheric CO2 (XCO2) sourced from NASA’s ACOS (May 2009 to May 2014) and OCO-2 (October 2014 to March 2016), de-noised and spatially filled in with Fixed Rank Kriging.

NASA JPL’s Orbiting Carbon Observatory (OCO-2) does not actually measure the amount of CO2 in the atmosphere. Instead it carries a Fourier Transform Spectrometer (FTS) that measures the frequency of light (radiance spectra) from the Sun reflected off the surface of the Earth. This is similar to the approach used by Japan’s GOSAT satellite. The raw spectral data retrieved by GOSAT was used by NASA JPL in their ACOS project to test the algorithms that would later be used for OCO-2.

The network of calibrated ground-based solar-viewing Fourier transform spectrometers (FTS) in the Total Carbon Column Observing Network (TCCON) operate in a similar way, measuring radiance spectra from the ground looking up at the Sun. Data from TCCON is used to calibrate and verify data from ACOS and OCO-2.

Radiance spectra alone cannot provide accurate measurements of CO2. Below is a collection of quoted material (edited for clarity) describing how XCO2 is derived from radiance spectra, other measured variables, and prediction models.

How is XCO2 calculated?

Observations from OCO-2 will be integrated with those of other instruments that fly aboard the Aqua and Aura spacecraft. Among these measurements are the temperature, humidity, and CO2 retrievals from Atmospheric Infrared Sounder (AIRS), the cloud, aerosol and ocean colour observations as well as carbon source and sink measurements from the Moderate Resolution Imaging Spectroradiometer (MODIS), Cloud-Aerosol Lidar And Infrared Pathfinder Satellite Observations (CALIPSO), and the CH4 and CO retrievals from Tropospheric Emission Spectrometer (TES).

Source: Measurement Approach

Once the OCO-2 measurements are calibrated and geo-located, they are analysed with a “full-physics” remote sensing retrieval algorithm to derive estimates of column averaged dry-air CO2 mole fraction (XCO2) and other geophysical properties. This retrieval algorithm is a modified version of the one used to process GOSAT TANSO-FTS data as part of the NASA Atmospheric CO2 Observations from Space (ACOS) project.

Source: OCO-2 Data Product User’s Guide, Operational L1 and L2 Data Versions 7 and 7R Version F

The Science Validation system must link OCO-2 XCO2 data at footprint resolution to the point-resolution measurements from the National Oceanic and Atmospheric Administration (NOAA) in situ CO2 data record, the World Meteorological Organization (WMO) Standard, which serves as the de facto foundation for all current and future scientific investigations of atmospheric CO2 and the global carbon cycle.

TCCON provides the transfer standard between the OCO-2 space-based XCO2 and the WMO CO2 standard. FTS XCO2 will be calibrated to the WMO CO2 standard via over-flights of the TCCON validation sites by aircraft or balloons carrying instruments that acquire in situ CO2 measurements tied to the WMO standard.

The averaging kernels of the ground-based FTS systems and the OCO-2 instrument are calculated using a radiative transfer model, and so factor into all comparisons of XCO2 measurements from these two types of spectrometers.

Source: OCO-2 Science Validation Plan Version 1.0 Rev A

Total column abundances are retrieved from spectra measured with the TCCON instruments using a non-linear least-squares spectral fitting algorithm (GFIT), which scales an a priori profile to produce a synthetic spectrum that achieves the best fit to the measured spectrum.

The GFIT CH4, N2O, CO and HF a priori profiles are generated from MkIV FTS balloon profiles (Toon, 1991). The profiles are shifted up or down in altitude depending on the tropopause pressure for local noon on that day.

Source: Calibration of the Total Carbon Column Observing Network (TCCON) using aircraft profile data

The Forward Processing code assumes that the ECMWF forecast temperature profile is correct up to an overall offset. Similar to temperature, the water vapour profile from the ECMWF forecast is assumed to be correct, up to an overall scale factor. The a priori surface pressure value is taken from the meteorological data file.

Source: OCO-2 Level 2 Full Physics Retrieval Algorithm Theoretical Basis Version 1.0 Rev 4

Prior and first-guess meteorological variables are taken from a short-term forecast from the ECMWF model. The forecast length is between zero and nine hours. Details of the ECMWF model can be found at The model is on 137 vertical levels and has a horizontal resolution of 0.25º in latitude and longitude. ECMWF forecast files are downloaded daily.

The L2 retrieval software utilises both the L1B and ECMWF resampled products generated earlier in the processing pipeline. The L1B file provides radiometrically corrected spectra and geolocation information. The ECMWF resampled product provides pressure, temperature and specific humidity profiles derived from ECMWF forecasted or reinterpreted products.

The L2 software also utilises additional static data sources:
• ABSCO tables—tabulated absorption coefficients providing gas-optical-properties information
• Solar-spectrum information
A priori CO2 profiles
• MERRA a priori—aerosol optical depth data for the world interpreted on a monthly basis from MERRA products
• Aerosol Properties—scattering and optical properties of scattering particles, derived from MERRA data
a priori surface properties
a priori wavelength grid offset
• Residual empirical orthogonal function parameters
a priori values for atmospheric components not coming from other products
– Fluorescence
– Cox-Munk windspeed, Lambertian albedo
– Zero Level Offset
a priori covariance values for all quantities

Source: OCO-2 Level 2 Full Physics Retrieval Algorithm Theoretical Basis Version 2.0 Rev 2

To compare two XCO2 observations properly, the retrievals must be computed about a common a priori profile, and the effect of smoothing must be taken into account by applying the averaging kernels.

The ACOS XCO2 values are adjusted downward in the winter and upward in the summer, which has the effect of reducing the overall seasonal cycle of the ACOS-GOSAT retrieval.

apriori-tccon-acos-gosat averaging-kernels-tccon-acos-gosat
Source: A method for evaluating bias in global measurements of CO2 total columns from space

The a priori profile of CO2 is derived from a multi-year global run of the LMDZ model. The monthly zonal mean is calculated from the model in 10º latitude bands, separately for land and ocean surfaces. An offset is added to all the model values to make the global average surface value approximately equal to the measured value from GLOBALVIEW-CO2; this offset is to be updated monthly to reflect the increasing concentration of CO2.

The CO2 prior covariance submatrix naturally has the most impact on retrieval of XCO2. Currently a single CO2 prior covariance matrix is used for all retrievals. This covariance has been constructed by assuming a root-mean-square (rms) variability of XCO2 of 12 ppm, which is an estimate of global variability. Variability as a function of height is assumed to decrease rapidly, from ~ 10% at the surface to ~1% in the stratosphere. The covariance among altitudes in the troposphere was estimated based on the LMDZ model, but the correlation coefficients were reduced arbitrarily to ensure numerical stability in taking its inverse. The total variability embodied in this prior covariance matrix is unrealistically large for most of the world.

Source: OCO-2 Level 2 Full Physics Retrieval Algorithm Theoretical Basis Version 1.0 Rev 4

The a priori profile of CO2 is taken from the same algorithm used by TCCON. As stated in that work: “The CO2 a priori profiles are from a climatology based on the GLOBALVIEW dataset, and change based on the time of year and the latitude of the site. Stratospheric CO2 profiles are generated from the age of air relationship derived by Andrews et al. [53] (2001).” A key parameter is the height of the tropopause, which is taken from the ECMWF first-guess temperature profile.


Source: OCO-2 Level 2 Full Physics Retrieval Algorithm Theoretical Basis Version 2.0 Rev 2


The final XCO2 values produced by ACOS and OCO-2 (as seen in the video above) are based on data from many sources including the TCCON ground stations, accurate measurements from CO2 detectors around the world (on towers, boats, planes, balloons), weather prediction models, and statistically optimal spatial prediction. This is the best method we have to determine the global sources and sinks of CO2.