19th Conference on Probability and Statistics

7.5

GFS-Based MOS Opaque Sky Cover Guidance for the Contiguous United States, Alaska, Hawaii, and Puerto Rico

PAPER WITHDRAWN

Wei Yan, NOAA/NWS, Silver Spring, MD

Sky cover is a very important weather element in public weather forecasting, and it is also of particular interest to the aviation industry. The Meteorological Development Laboratory (MDL) of the National Weather Service (NWS) has developed a statistical guidance package of opaque sky cover, which applies the Model Output Statistics (MOS) technique to National Center for Environmental Predictions (NCEP) Global Forecast System (GFS) model output. Sky cover is defined by the NWS as the average amount in eighths of opaque clouds covering the celestial dome observed from the ground surface. Previous to this development, the NWS produced MOS guidance of total sky cover.

A hybrid approach was adopted so that equations were developed for single stations where the developmental data sample was adequate to support the single-station approach and in regionalized form otherwise. Previous MOS development of total sky cover used only the regionalized approach. The MOS equations were developed to calculate the probability of each of five categories of opaque sky cover, namely clear, few, scattered, broken, and overcast. Categorical forecasts were generated from the probabilities by using probability thresholds. These thresholds were obtained from the developmental sample and were designed to maximize the threat score while keeping the bias within a narrow range around unity.

Since cloud amount is normally observed and reported only up to 12,000 feet in height at Automated Surface Observing System (ASOS) sites, the cloud observations from Satellite Cloud Product (SCP) estimates were used to complement the ASOS observations. The new opaque sky cover MOS guidance provides 3-hourly forecasts from 6 hours to 192 hours after the initial GFS model runs at 0000 and 1200 UTC, and from 6 hours to 84 hours at 0600 and 1800 UTC, for the contiguous United States, Alaska, Hawaii, and Puerto Rico, and for two, 6-month seasons (cool – October to March; warm – April to September). Testing on independent data of one full year shows significant improvement in forecast skill by the new sky cover guidance over the current operational one. We attribute this progress mainly to the use of the hybrid approach in the development of forecast equations.

Session 7, Probability Forecasting
Tuesday, 22 January 2008, 3:30 PM-5:15 PM, 219

Previous paper  Next paper

Browse or search entire meeting

AMS Home Page