Channel Selection for Hyper Spectral CO₂ Measurement at the Near-infrared Band

The remote sensing of CO₂ with the near-infrared sunlight can detect the source and sink information of atmospheric CO₂ on the earth surface, which can be used in the research of global carbon cycle. The designing hyper spectral CO₂ instrument, which will be carried by TanSat to be launched in 2015, measures CO₂ column concentration using the near-infrared band. The instrument incorporates three bands with center wavelength of 0.76 μm, 1.6 μm and 2.06 μm. The spatial observing resolution is 1 km and the highest spectral resolution is 0.03 nm with the window width of 40 nm. Broad band and high resolution are a challenge for instrument manufacturing, as well as for observation processing including radiative transfer forward calculating and retrievals. The methods of degree of freedom (DOF) and information content are introduced. The CO₂ information content of channels at the near-infrared band is analyzed based on the above methods. The top 20 to 100 high information content channels are selected, which are then used in a retrieval experiment based on full physical retrieval algorithm. Results show that the selected 20 channels provide as much as 74.6% of the total channel information content. There are exactly 10 channels located at P-branch and R-branch of 1.6 micron band respectively, which indicates that two absorption branches are both equally important. The CO₂ retrieval error using the selected 20 channels only is 0.3×10^-6 larger than retrievals using all the channels at 1.6 μm band. After the convergence of retrieval is achieved, the spectrum residual distribution shows relatively smaller residuals in high information content absorption channels and larger residuals in low information content channels. Therefore, the high information content channels control the retrieval progress.The relationship between information content and channel number is also investigated. First, information content increases with increasing channels amount to 60, but the trend becomes slow after that. The relationship between CO₂ retrieval errors and high information channels amount is similar. The weak and strong CO₂ absorption bands near 1.6 μm and 2.06 μm have different high information content channel distribution calculated using CO₂ DOF and information content method. The high information content channels within 2.06 μm band are located at lines of moderate absorption radiance, and the distribution at two branches of 2.06 micron is asymmetry.It should be noted that the optical depth of aerosol is lower in the retrieval experiment, and cloud (thin cirrus) is also not included. Due to the disturbance of the backscatters of atmospheric aerosol and cirrus to radiation observed by satellite at near-infrared bands, the impact of cloud and aerosol to channel selection needs further investigations.