Experiments Concerning Categorical Forecasts of Open-Ocean Visibility Using Model Output Statistics.
NAVAL POSTGRADUATE SCHOOL MONTEREY CA
Pagination or Media Count:
This study is an extension of previous statistically oriented research at the Naval Postgraduate School. The method of Model Output Statistics is used to predict open-ocean visibility employing stepwise-selection, multiple linear regression. The visibility predictand is specified categorically with comparisons made to a previous probabilistic approach. Predictors include direct and derived model output parameters provided by the U.S. Navys Fleet Numerical Oceanography Center FNOC, Monterey, California. About 18,000 North Pacific Ocean 30 deg - 60 deg n synoptic ship reports at 0000 GMT from June 1976 and 1977, July 1979, and August 1979 were used as both dependent and independent data sets. Visibility equations for both analysis-time and 24- and 48-hr prognostic times are developed, and are verified using percent correct, Heidke skill score and bias. Levels of skill are less than desirable for operational use. Important predictor parameters are found to be sensible and evaporative heat fluxes, meridional wind component, sea-level pressure, airsea temperature difference, relative humidity, an FNOC fog probability parameter and a visibility parameter derived from a marine aerosol model. Other experiments concerning weighted least squares, predictand transformations and R2 deflation are briefly described. Author
- Statistics and Probability