Skip all navigation and jump to content Jump to site navigation
About MODIS News Data Tools /images2 Science Team Science Team Science Team

   + Home
ABOUT MODIS
MODIS Publications Link
MODIS Presentations Link
MODIS Biographies Link
MODIS Science Team Meetings Link
 

 

 

Karlsson, K. -G.; Johansson, E.; Devasthale, A. (2015). Advancing the uncertainty characterisation of cloud masking in passive satellite imagery: Probabilistic formulations for NOAA AVHRR data. REMOTE SENSING OF ENVIRONMENT, 158, 126-139.

Abstract
Two alternative methods for probabilistic cloud masking of images from the Advanced Very High Resolution Radiometer (AVHRR) sensor have been examined. Both methods are based on Bayesian theory and were trained using data from the Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP) lidar onboard the Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO) satellite. Results were evaluated by comparing to independent CALIPSO-CALIOP observations and to a one-year ground-based cloud dataset composed from five different remote sensing systems over the observation site in Cabauw in the Netherlands. In addition, results were compared to two different cloud masks; one derived from the geostationary Spinning Enhanced Visible and Infrared Imager (SEVIRI) sensor and one from the Climate Monitoring Satellite Application Facility Clouds (CMSAF), Albedo and Radiation dataset from AVHRR data (CLARA-A1). It was demonstrated that the probabilistic methods compare well with the referenced satellite datasets and for daytime conditions they provide even better performance than the reference methods. Among the two probabilistic approaches, it was found that the formulation based on a Naive Bayesian formulation (denoted PPS-Prob Naive) performed clearly superior to the formulation based on a linear summation of conditional cloud probabilities (denoted PPS-Prob SPARC) for daytime conditions. For the study based on the observations over the Cabauw site, the overall daytime Kuipers Skill Score for PPS-Prob Naive was 0.84, for PPS-Prob SPARC 0.79, for CLARA-A1 0.74 and for SEVIRI 0.66. Corresponding results for night-time conditions were less favourable for the probabilistic formulations (Kuipers Skill Score 0.74 for PPS-Prob Naive, 0.68 for PPS-Prob SPARC, 0.80 for CLARA-A1 and 0.79 for SEVIRI) but still relatively close to the reference dataset The Cabauw distribution of cloudiness occurrences in different octa categories was reproduced very closely by all methods, including the probabilistic formulations. Results based on Cabauw observations were also largely in good agreement with results deduced from comparisons with the CALIPSO-CALIOP cloud mask.The PPS-Prob Naive approach will be implemented in an upcoming version of the Polar Platform System (PPS) cloud software issued by the EUMETSAT Nowcasting Satellite Application Facility (NWC SAF). It will also be used in the second release of the CMSAF CLARA cloud climate data record based on historic AVHRR GAC data (to be denoted CIARA-A2). (C) 2014 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC-SA license

DOI:
10.1016/j.rse.2014.10.028

ISSN:
0034-4257

NASA Home Page Goddard Space Flight Center Home Page