Mitigating fundamental biases in CO2 retrievals from space-based measurements

With now over two years of high-resolution near-infrared observations of carbon dioxide from the Greenhouse Gases Observing SATellite (GOSAT), tighter constraints on the sources and sinks of CO₂ based on these measurements are imminent. However, these constraints typically rely on inversion models, models which require a realistic estimate of both random and systematic errors in the retrievals of column CO₂ (XCO₂). It is well known that errors in spectroscopy as well as in the characterization of the instrument, for instance in terms of radiometric or spectral calibration, can lead to systematic errors in the XCO₂ retrievals of the order of several ppm. In the absence of spectroscopy or instrument errors, however, most retrieval algorithms will still suffer from fundamental biases, primarily as a result of the inability to distinguish thin clouds or aerosols from the underlying surface reflectance. These biases can also be of the order of several ppm in XCO₂, large enough to cause systematic errors in the flux inversions if not properly taken into account. In this presentation, we quantify these fundamental retrieval biases for the Atmospheric CO₂ Observations from Space (ACOS) retrieval algorithm. Even though the biases depend to some extent on the particular retrieval algorithm employed, we show that even in a simple analytical retrieval these biases are still present. We then explore modifications to a typical retrieval that can aid in mitigating these biases, with the ultimate goal of either their elimination or, at worst, an accurate error characterization. Only then can there be high confidence in the results of CO₂ flux inversions utilizing these space-based measurements.