Skip to Main Content
Skip Nav Destination

Measuring atmospheric carbon dioxide from space

6 October 2014
If we can account for the confounding effects of aerosols, satellite-based measurements of CO2 will improve our understanding of Earth’s carbon cycle.

Atmospheric scientists have recently launched space-based instruments with the potential to provide global measurements of greenhouse gas concentrations, including carbon dioxide (CO2). The global coverage should improve the accuracy of carbon flux models, which are heavily dependent on CO2 measurements, and help to answer lingering questions about Earth’s sources and sinks of carbon and their interaction with the atmosphere.

We know that only about half of the CO2 emitted by humans stays in the atmosphere. The rest is absorbed by the land and ocean. Little is known about how and where this absorption occurs, especially over land. Carbon flux models, previously limited by sparse ground-based measurements of CO2, will theoretically receive exponentially more data, and thus improve the accuracy of their conclusions.

The improvement will depend on the accuracy space-based measurements. Specifically, an accuracy of better than 0.5% for space-based measurements is needed to gain more information about the carbon cycle than ground-based measurements alone can confer. Reducing uncertainties in satellite measurements of CO2 is therefore critical.

One way to measure CO2 from space is to detect sunlight reflected off Earth’s surface and determine how many CO2 molecules lie along the “light-path” (see figure). Satellites such as Japan's Greenhouse Gases Observing Satellite (GOSAT) and NASA'S Orbiting Carbon Observatory (OCO) make use of that method.

The basic concept of using reflected sunlight to retrieve the concentration of atmospheric CO2.

The basic concept of using reflected sunlight to retrieve the concentration of atmospheric CO2.

Typically, a so-called weak CO2 absorption band located around 1.6 μm is used in conjunction with a so-called strong CO2 absorption band at 2.0 μm to deduce the average amount of CO2 in the column of air seen by the instrument’s sensors. The single value of CO2 in the column of air is known formally as the column-averaged dry-air mole fraction of carbon dioxide, or XCO2.

Unfortunately, the CO2 bands are contaminated when even very thin layers of clouds and aerosols lie in the path of the instrument’s sensors. The contamination can lead to significant errors in determining XCO2—a deal-breaker when seeking an accuracy of better than 0.5%. If clouds and aerosols are present, they can scatter the reflected sunlight in different directions, thereby drastically altering the length of the light-path seen by the sensor, and throwing off the calculation of CO2 molecule concentration in the column.

One way to avoid such errors is to filter out scenes containing clouds and aerosols by using information in the measured spectra themselves. The filters are created using various parameter thresholds, and often employ different bands at other wavelengths, such as the O2 A-Band at 760 nm. However, filtering out data is never entirely effective at eliminating contaminated scenes. We need other strategies to avoid inducing measurement errors.

One helpful method is to incorporate a simple parameterization of clouds and aerosols in the CO2 retrieval algorithm. Most existing algorithms contain some kind of method to gain information about clouds and aerosols. The approach often entails adding one or more particle “types” to the algorithm along with parameters describing the particles. The variables are intended to represent typical clouds and aerosols found in the atmosphere. Adding cloud and aerosol parameters to the algorithm is imperfect, however, because multiple cloud and aerosol solutions often exist for a given observation.

Worse yet, so-called smart CO2 retrievals may actually incur biases from trying to account for clouds and aerosols when none are present! For ideal, very clear scenes the addition of cloud and aerosol parameters may be doing more harm than good. Indeed, a comparison of the CO2 algorithm’s aerosol parameters with those measured by more accurate ground-based instruments (for example, NASA's Aerosol Robotic Network) revealed that the aerosol information retrieved by the CO2 algorithms was not well correlated with the true aerosol information for a particular scene.

One way of dealing with clouds and aerosols is simply to avoid retrieving information about them. These so-called nonscattering retrievals assume no scattering or absorption by clouds and aerosols, and thus may avoid introducing unwanted biases and errors into the CO2 retrieval algorithm. Previous work has shown that nonscattering retrievals have acceptably small errors over the cloud-free ocean. Nonscattering retrievals are also desirable because of their high computational efficiency. Not having to solve complex equations relating to scattering and absorption by clouds and aerosols greatly reduces the computational time needed for the algorithm to find a solution.

Additionally, future space-based CO2 measurement missions will collect orders of magnitude more data than today’s satellites—more data, even, than can be analyzed at this time. Using a simplified, computationally cheaper algorithm would enable all measurements to be studied. Currently, it seems that nonscattering retrievals can perform well, but only if highly filtered and properly corrected for cloud and aerosol-related biases. Why nonscattering retrievals are initially less accurate, even for very clear scenes, remains unknown. Analyzing other varieties of nonscattering retrievals may also provide clues to answering this question.

Scientists are pursuing other approaches to cloud and aerosol parameterization. One approach is to use different assumptions about the size, shape, and location of the particles. Another is to use information from external data sources, such as the Modern Era Retrospective-Analysis for Research and Applications (MERRA), a global data reanalysis system that contains information about aerosols. The datasets can be used to select more appropriate types of aerosols to include in the algorithm using simple, generic types.

Further research must determine the best methods for handling clouds and aerosols in CO2 retrievals. Without that effort, uncertainties in measurements will persist and dramatically limit what GOSAT, OCO, and future satellite missions can tell us about how carbon—the fourth most abundant element in the universe—cycles through Earth's environment.

Robert Nelson is a graduate student in the department of atmospheric science at Colorado State University in Fort Collins.

Close Modal

or Create an Account

Close Modal
Close Modal