Cloud Detection and Macro-physical Properties Determined from a New Algorithm Using Micropulse Lidar Observation

Description

This study introduces a new algorithm to detect aerosol/cloud layers and separate clouds from aerosols based on micropules lidar (MPL) measurements. A semi-discretization processing (SDP) method is used to prevent the impact of the potentially increasing noises with distance. We then introduce the value distribution equalization (VDE) method for the first time to detect the layer with aerosol/cloud particles. The VDE method can make all signal layers stand out clearly and comparable to each other in signal magnitude, making it more easy and accurate to detect aerosol/cloud layers at far distance. Combined with empirical threshold values, we determine if the signals indicate clouds or not. A synthetic test and a case study indicate that this method can detect useful signal layers and separate clouds and aerosols with high accuracy, while aerosols and clouds could be sometimes misclassified due to seletected threshold values. This algorithm has been applied to 1-year observations at both U.S. SGP site and China Taihu site. At SGP site, cloud frequency shows a clear seasonal variation with maximum values in winter and spring, and also shows a bi-modal vertical distributions with maximum occurrence at around 3-6 km and 8-12 km. The annual averaged cloud frequency is about 50%. Seasonal analysis of cloud base occurrence frequency suggests that the dominant clouds are stratiform in winter and convective in summer. In contrast, at Taihu site, cloud frequency shows no clear seasonal variation and shows a single modal vertical distribution with maximum occurrence at around 1 km. The annual averaged cloud frequency is about 15% higher than that at SGP site. Seasonal analysis of cloud base occurrence frequency suggests that the dominant clouds are likely to be stratiform clouds. Accurate physical understanding of these findings will be done in future.