Channel Selection for CO 2 Retrieval Using Near Infrared Measurements

1. Channel Selection for CO2 Retrieval Using Near Infrared Measurements Le Kuai1, Vijay Natraj1, Run-Lie Shia1, Susan Kulawik2, Kevin Bowman2, Charles Miller2, Bill Irion2, Yuk Yung1 1. Division of Geological and Planetary Sciences, California Institute of Technology 2. Jet Propulsion Laboratory, California Institute of Technology Abstract Understanding changes in the concentrations, global sources and sinks, dynamics and other processes that control the variability of atmospheric carbon dioxide (CO2) has emerged as one of the principal challenges of 21st century Earth system science. Satellite observations of atmospheric CO2 are poised to revolutionize our understanding of global carbon cycle science by providing unprecedented spatiotemporal resolution and coverage. Here, we carry out an information content analysis for CO2 retrieval from near infrared (NIR) bands (GOSAT). Since NIR measurements are sensitive to changes in the lower troposphere, CO2 sources and sinks near the surface are more detectable using NIR measurements. GOSAT will measure reflected sunlight in three NIR spectral regions (the 0.76 μm O2 A-band and two CO2 bands at 1.61 and 2.06μm). Information analysis provides a method to choose the first 200 out of 1016 channels in each CO2 band with the highest information content. Retrieval using these 200 channels in each band could provide as much as 90% information as a retrieval using all 1016 channels in each band. The advantage of using NIR CO2 bands to do the retrieval is that they are less sensitive to the temperature than the thermal IR band (AIRS or TES). The weighting function for AIRS and TES are both peaked in the middle and upper troposphere. However, the NIR CO2 bands are most sensitive to the CO2 near the surface (Fig. 1), where most of the CO2 sources and sinks are present. Therefore, CO2 retrieval using NIR measurements would provide complementary information to that from AIRS or TES. Definition of information content (H) and degree of freedom (ds) Sa: the a priori covariance matrix; Sξ: the measurement error covariance matrix; K: the Jacobian; A: the averaging Kernel; λi: the eigenvalues of a) 1.61 μm b) 2.06 μm Fig. 4 The first 60 channels selected for both bands (CO2) are from either P branch or R branch. Fig. 1 Weighting functions for CO2 have peak near surface. a) 1.61 μm CO2 band; b) 2.06 μm CO2 band The channels with the highest information content (IC) are from those with intermediate absorption (Fig.2). This is because, for very weak channels, the absorption is too low and no signal is received. On the other hand, for the saturated channels, the absorption is too high so that there is no sensitivity to the signal. Fig. 2 also shows that the channels with high information for temperature are different from those for CO2. In both clear sky and high aerosol optical depth (AOD) and high cloud optical depth (COD) cases, high information content channels are almost the same. Thus, the selected channels could be applied to retrieval under the different scenarios. The increase in the degrees of freedom (DOF) and information content (IC) in unclear sky case is due to backscattering by aerosols (Fig. 3). Some channels have DOF close to 1. (see the right panels of Fig. 3.) It is clear from Fig. 4 that the channels selected by IC have no bias towards P or R branch. Fig. 5 shows the first 40 channel ranked by IC for CO2 and their IC for temperature. Those channels that have most IC for CO2 but least IC for temperature would be chosen. 1) Clear sky case 2) High AOD, high COD case The simultaneous retrieval using all 2032 channels in both 1.61 μm CO2 band and 2.06 μm CO2 bands provide 1.67 DOF and 5.9 bits of IC. Fig. 6 shows that the retrieval using the first 200 channels in each band would have 1.55 DOF and 5.35 bits of IC. It is about 90% of the information provided by the retrieval using all channels. The channels were selected by ranking the IC from highest to lowest in each band. Fig. 5 First 40 Channels ranking by CO2 IC in red and their temperature IC in blue. Conclusions We apply information analysis to select channels for CO2 retrievals using NIR measurements. It is found that the intermediate channels provide most information. The high DOF and IC for high AOD and high COD scenarios are due to backscattering by aerosols. The channels selected for the clear sky case are consistent with the channels for other scenarios. Using the first 200 channels (in order of IC) in each band for retrieval gives more than 90% information as compared to retrieval including all 1016 channels in each band. Applying the same technique to choose channels most sensitive to T, P, aerosol and other parameters will narrow down the choice of channels to those most sensitive to CO2 and least sensitive to other parameters. Fig. 2 First 20 channels with highest IC. Red: most sensitive to CO2; blue: most sensitive to temperature. Left is clear sky case and right is the cloudy sky case. Fig. 6 The information of the retrieval increases with the number of channels selected from each band. The increase of information is very little from 200 channels to 1016 channels. Fig. 3 Channels with highest information content (CO2) are from the channels with intermediate radiance in both bands. Left is clear sky case and right is the cloudy sky case.