Intercomparison of Ground-Based Cloud Retrieval Algorithms

Wednesday, 17 December 2014
Chuanfeng Zhao, Qianqian Wang and Min Lv, Beijing Normal University, College of Global Change and Earth System Science, Beijing, China
This study summarizes nine different cloud retrieval algorithms and discusses their advantages and limitations. Potential uncertainties have also been discussed. These retrievals are mainly based on observations from millimeter radar, lidar, spectral radiation, radiosondes and so on.

The cloud retrieval algorithms are classified into three types: physical retrieval algorithm, statistical parameterization algorithm, and optimal iteration method. Analyses indicate that physical retrieval algorithms are theoretically accurate. However, assumptions used in these methods make it challenging to obtain highly reliable results. To improve the physical retrieval algorithm, more reasonable assumptions should be used. Compared to other two types of retrieval algorithms, empirical parameterization methods are simple and can be easily applied. However, these methods are generally based on limited cloud samples for certain types of clouds and locations, and thus they have much larger uncertainties. To improve the accuracy of parameterization algorithm, more reliable in-situ observations are needed for different types of clouds over various locations. In contrast, the optimal iteration method seems to have relatively higher accuracies since the retrieval results make the forward model simulations match observations. However, the accuracy of optimal iteration method is highly dependent on the reliability of the forward models. To improve optimal iteration method, better understanding of the cloud processes and more reliable forward models are needed. In summary, more accurate and comprehensive retrieval algorithms are needed in our current cloud remote sensing studies.