Sky Cover: Shining Light on a Gloomy Problem
While human observers record sky cover as part of routine duties, the spatial coverage of such observations in the United States is relatively sparse. There is greater spatial coverage of automated observations, and essentially complete coverage from weather satellites that scan the Americas. A good analysis of sky cover must reconcile differences between manual observations, automated observations, and satellite observations, through an algorithm that accounts for the strengths and weaknesses of each dataset. For example, there has long been a deficiency in ceilometers in high cloud coverage. In contrast, satellites are able to determine the evolution of the highest cloud deck, with clouds in the lower troposphere sometimes obscured. Today's geostationary satellites scan every one-kilometer pixel in the visible band (at nadir) several times per hour, while the surface observer is responsible for the entire celestial dome, which extends to clouds sometimes tens of kilometers from the observation location, no less than once per hour.
This presentation discusses the creation of a blended sky cover analysis, and an optimized sky cover analysis which minimizes absolute error with the one-hour forecast of sky cover from the NWS' National Digital Forecast Database (NDFD). There is additional discussion about how the blended sky cover analysis is compared with the cloud ice, cloud water, rain, snow, and other analysis fields from the High-Resolution Rapid Refresh (HRRR), and then applied to three-, six-, and nine-hour HRRR forecasts. The intent is to suggest a reasonable definition for sky cover and demonstrate a product that can bring consistency to analyses and forecasts of sky cover.
Supplementary URL: http://go.wisc.edu/85a0k4