SCANDIA -its accuracy in classifying LOW CLOUD

Typ: Rapport
Serie: Meteorologi 91
Författare: Pia Hultgren, Adam Dybbroe and Karl-Göran Karlsson


Low clouds are of great interest for the airborne users of weather forecasts. Therefore it is important to improve the techniques of forecasting low clouds. One valuable way to detect low clouds is through the information from satellite images. A cloud classfication model (named SCANDIA - described by Karlsson, 1996) is used since many years at SMHI. Cloud classification results are distributed to users at the central forecasting office, at local forecasting offices and at forecasting offices of the Swedish Airforce. Since there are still improvements to make in cloud classification applications, the Swedish Airforce startad this project to join the development and research going on in this area at SMHI.

The study focuses on low clouds. As we know from long term experience and earlier studies, the SCANDIA cloud classification model has problems in specific conditions. These situations are:

  • Low level inversion with no significant cloud signature (due to dawn/dusk illumination or mixed water & ice phases).
  • Sunglint in combination with cold sea.
  • Forward scattering, particularly in moist and hazy atmospheres.

This document reports on the general performance of the SCANDIA cloud classification scheme concerning the treatment of low clouds. Validations and verifications have been made to identify and focus on the specific problems. A database (MSMS = Matching Satellite Model & SYNOP data) was constructed and is continuously being updated and expanded. MSMS is used for the validations and verifications. By studying the information in the database from surface observations, NOAA AVHRR satellite data, and the SCANDIA classification, the problems can be identified, and same ways to improve the classification model might be found and suggested. In a wider scope, it can be seen as a preliminary study for the purpose of improving the analysis of low cloudines inferred from satellite data in the SMHI mesoscale analysis medel MESAN (Häggmark,1997).