Publications

Wang, LW; Tian, M; Zheng, YF (2019). Assessment and improvement of the Cloud Emission and Scattering Index (CESI) - an algorithm for cirrus detection. INTERNATIONAL JOURNAL OF REMOTE SENSING, 40(14), 5366-5387.

Abstract
This study evaluates and improves the cirrus detection of the middle and low peak Weighting Function (WF) Cloud Emission and Scattering Index (CESI) by using the Visible Infrared Imaging Radiometer Suite (VIIRS) cloud mask product. Results show that CESI-3 (peak WF about 890 hPa) accurately detect cirrus with the Probability of detection (POD) value reaching 0.69 during both at day and nighttimes. The CESI-1 (peak WF about 550 hPa) falsely detects cirrus in the Sahara Desert, south of the Saudi Arabia, north of Tibetan Plateau and in Australia during the ascending process. More false alarm rates (FARs) are detected in the northern mid-latitude around 60 degrees N because of the surface type and seasonal effects. Furthermore, the thresholds of middle and low peak WF CESIs for cirrus detection are determined and the performance is satisfactory. In addition, Linear Discriminant Analysis (LDA) is employed to train four combination types (combining CESI-1 and CESI-2, combining CESI-1 and CESI-3, combining CESI-2 and CESI-3, and combining CESI-1, CESI-2, and CESI-3). Accordingly, results elucidate that Comb-1-3 (combining CESI-1 and CESI-3) is appropriate to detect cirrus for the ascending process, increasing the POD by 11.2% as compared to CESI-3. However, Comb-1-2-3 (combining CESI-1, CESI-2 with peak WF around 790 hPa, and CESI-3) is most suitable for cirrus detection in the descending process, increasing the PODs by 8.5%. We also find that the POD of cirrus detection over the land increases relatively higher than over the sea. The most significant improvement of cirrus detection occurs over the land in the summer in the daytime with a 35% POD increase from 0.57 to 0.76.

DOI:
10.1080/01431161.2019.1579938

ISSN:
0143-1161