Accession Number : ADA190803

Title :   Very Short Range Statistical Forecasting of Automated Weather Observations.

Descriptive Note : Final rept. Oct 86-Sep 87,

Corporate Author : NATIONAL WEATHER SERVICE SILVER SPRING MD TECHNIQUES DEVELOPMENT LAB

Personal Author(s) : Miller, Robert G

PDF Url : ADA190803

Report Date : Jan 1988

Pagination or Media Count : 241

Abstract : A procedure is developed for providing weather forecasting guidance over the short range period of 10, 20, 30, ..., 120 minutes. It uses the Automated Weather Observing System (AWOS) elements as predictors and predictands. The model is founded on Markov assumptions and uses multivariate regression as the statistical operator. Details are given on how the Generalized Exponential Markov (GEM) model compares with persistance. Tests are performed on a test sample of almost 400,000 cases. Overall, GEM succeeds in bettering current short range weather forecasting techniques (i.e. persistence) over the twelve projection periods of 10, 20, 30..., 120 minutes. The ability of GEM to successfully predict VFR to IFR, and IFR to LOW IFR changes in both visibility and ceiling is also demonstrated.

Descriptors :   *WEATHER FORECASTING, AUTOMATION, CEILING, METEOROLOGICAL DATA, MULTIVARIATE ANALYSIS, PREDICTIONS, REGRESSION ANALYSIS, SHORT RANGE(DISTANCE), SHORT RANGE(TIME), VISIBILITY, WEATHER, WEATHER STATIONS, AUTOMATIC, MATHEMATICAL MODELS, MATHEMATICAL PREDICTION, STATISTICAL ANALYSIS

Subject Categories : Meteorology

Distribution Statement : APPROVED FOR PUBLIC RELEASE