Multiscale hypothesis testing theory and methods for aerosol and cloud layer detection of lidar

Abstract

Lidar is a unique instrument for profiling aerosol and cloud layers. Detecting the boundaries of these layers is crucial because a missing layer will not be retrieved later. Many layer detection methods, which can be divided into slope-type and threshold-type methods, have been proposed and applied to a one-dimensional (1D) profile or a two-dimensional (2D) scene using numerous empirical settings. However, many eye-visible missing layers remain, indicating the need for a more effective detector. Herein, we demonstrated that layer detection is essentially a hypothesis testing issue. Theoretically, after providing a possible distribution of the background air signal, the task of layer detection is to test whether a given signal conforms to the provided hypothesis of a background air distribution under a particular significance. Consequently, we propose multiscale hypothesis testing methods (MHTs) for cloud and aerosol layer detection, which can work on 1D profiles and 2D scenes and are named 1D-MHT and 2D-MHT, respectively. The 2D-MHT detected 1.67 times layer area as the CALIOP Version 4.2 Level 2 product totally. Particularly, the results show that the 2D-MHT at 5 km resolution can detect 16.3% and 0.5% more layer area than the CALIOP official product at 5–80 km resolutions at daytime and nighttime, respectively. Evaluation using the extinction retrievals from aircraft lidar and the depolarization ratio of ice clouds shows the reliability of the additional detected layers by 2D-MHT. Thus, the hypothesis testing methods have high potential to effectively improve the accuracy and resolution for retrieving lidar products in the future.