Cloud discrimination from sky images using a clear-sky index

Masanori Saito, Hironobu Iwabuchi

Research output: Contribution to journalArticlepeer-review

10 Citations (Scopus)


Cloud fraction mainly affects the incoming solar irradiance at the surface and is measured with groundbased sky imagers. Although several methods enable cloud discrimination from digital sky images, cloud discrimination methods are still being developed to improve the classification accuracy. This study presents two methods for effective cloud discrimination from digital sky images using a newly defined clear-sky index (CSI). The CSI represents the degree of similarity of the spectral distribution to that expected for clear sky. In the advanced method (AM), the CSI is obtained from red-green-blue (RGB) signals in RAW format by linear transformation by taking into account the solar spectrum at the top of the atmosphere, the ozone transmittance, and the spectral response of the RGB channels. The simplified method (SM) uses digital signals in a JPEG format assuming prescribed color matching functions and atmospheric states. The AM can correctly classify broken gray clouds as cloud and aureole as clear sky in most cases, and it discriminates clear sky and clouds with a correct classification rate of 93.6% based on a comparison with the zenith-pointing lidar measurements. The SM demonstrates accurate cloud discrimination performance as well as the red-to-blue ratio method does. The use of RAWformat data allows for more accurate image-based cloud discrimination.

Original languageEnglish
Pages (from-to)1583-1595
Number of pages13
JournalJournal of Atmospheric and Oceanic Technology
Issue number8
Publication statusPublished - 2016 Aug 1


  • Algorithms
  • Atm/ocean structure/phenomena
  • Classification
  • Cloud cover
  • Clouds
  • Mathematical and statistical techniques
  • Observational techniques and algorithms

ASJC Scopus subject areas

  • Ocean Engineering
  • Atmospheric Science


Dive into the research topics of 'Cloud discrimination from sky images using a clear-sky index'. Together they form a unique fingerprint.

Cite this