利用地基红外高光谱辐射数据可以反演得到高时间分辨率的边界层大气温度廓线。 目前的AERIoe最优化反演算法相比于传统的“剥洋葱”算法有较大的改进, 且对初值的依赖程度较低。 但AERIoe算法中正则化算子的选择对反演结果的稳定性和反演时间有重要影响。 目前主要采用经验的方法选择正则化算子, 迭代步数较多, 耗费大量的计算时间。 提出了利用L曲线方法代替经验法选取正则化算子的改进方案, 以提高AERIoe方法的反演速度。 改进后的算法通过绘制解范数和残余范数的二维曲线图, 取其拐点作为最优的正则化参数, 相比于传统的经验法有着更好的理论基础。 采用2011年美国大气辐射测量计划中SGP站点的晴空大气红外辐射数据进行反演实验。 结果表明, 利用该方法得到的反演结果具有很好的稳定性、 收敛性和精度。 相比于经验的方法, 利用L曲线方法获得的正则化算子反演温度廓线时的收敛速度更快, 迭代步数较少, 可以节约大量的计算时间; 在反演精度方面, L曲线方法在边界层中上层的反演精度更高, 1～3 km高度上温度廓线的RMSE值提高了大约0.2 K。

The thermodynamic profiles of Planetary Boundary Layer could be retrieved by using ground-based hyper-spectral infrared radiance. The AERIoe algorithm has a better performance at the dependency of initial profiles than the “onion peeling” method which was originally applied in the Atmospheric Emitted Radiance Interferometer. The regularization parameter is the key to the AERIoe algorithm, and the strategy for choosing the regularization parameter in the retrieval algorithm is based on the empirical method, which requires too much time for computation while the empirical method needs many iteration steps. A L-curve criterion is proposed to calculate the regularization parameter in AERIoe algorithm. The L-curve criterion is based on a log-log plot of corresponding values of the residual and solution norms, and the optimal regularization parameter corresponds to a point on the curve near the “corner” of the L-shaped region. Therefore, the L-curve criterion has better theoretical basis than the traditional empirical method. The result of retrieval experiment using the observed data collected at the SGP site of the year 2011 shows that, the L-curve method has a good performance in terms of stability, convergence and accuracy of the retrieval. Compared with empirical method, L-curve algorithm converges more quickly which saves much computation time when retrieving the temperature profiles. When considering the retrieval accuracy, the L-curve method has a better behavior at the middle and top of the boundary layer, with an improvement of 0.2 K of RMSE at the altitude of 1～3 km than the empirical method. Therefore, the L-curve algorithm has a better performance compared with the empirical method when choosing the regularization parameter in the retrieval of temperature profiles using the ground-based hyper-spectral infrared radiance.