近红外CO_2高光谱探测仪通道选择

Channel Selection for Hyper Spectral CO_2 Measurement at the Near-infrared Band

近红外波段(1.6μm)遥感可探测大气CO_2含量信息,应用于碳循环研究中。宽波段、高分辨率不但对仪器研制是一个挑战,而且巨大的数据量对观测的正演、反演也是一个挑战性课题。该文应用自由度及信息量分析法,对近红外高光谱波段中探测通道进行CO_2信息量分析,选择前20~100个高信息量的CO_2探测通道,并进行了反演模拟测试。结果表明:前20个高信息量通道占所有通道总信息量的76.4%,仅用所选的前20个通道进行反演,与所有通道参加反演的结果相比,误差增加0.3×10^(-6);通道数增至60时,信息量增加,通道数再增加,信息量则增加不显著;CO_2反演误差存在相似的关系。在高CO_2信息量分布上,弱吸收性质的1.6μm波段和强吸收性质的2.06μm波段表现出不同特点。

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.