Harrison L, Michalsky J
Appl Opt. 1994 Aug 1;33(22):5126-32. doi: 10.1364/AO.33.005126.
Optical depth retrieval by means of Langley regression is complicated by cloud transits and other time-varying interferences. An algorithm is described that objectively selects data points from a continuous time series and performs the required regression. The performance of this algorithm is compared by a double-blind test with an analysis done subjectively. The limits to accuracy imposed by time-averaged data are discussed, and an additional iterative postprocessing algorithm is described that improves the accuracy of optical depth inferences made from data with time-averaging periods longer than 5 min. Such routine algorithms are required to provide intercomparable retrievals of optical depths from widely varying historical data sets and to support large networks of instruments such as the multifilter rotating shadow-band radiometer.
通过兰利回归法进行光学深度反演会因云层过境和其他随时间变化的干扰而变得复杂。本文描述了一种算法，该算法可从连续时间序列中客观地选择数据点并进行所需的回归分析。通过双盲测试将该算法的性能与主观分析进行了比较。讨论了时间平均数据对精度的限制，并描述了一种额外的迭代后处理算法，该算法可提高从时间平均周期超过5分钟的数据中得出的光学深度推断的准确性。需要此类常规算法来提供对来自广泛不同历史数据集的光学深度的可相互比较的反演结果，并支持诸如多滤光片旋转阴影带辐射计等大型仪器网络。
