ESTIMATING CONDITIONAL PROBABILITY AND PERSISTENCE

Abstract : The report describes a statistical model and automated techniques which provide estimates of conditional and persistence probability of meteorological events for periods to 48 and 24 hours, respectively. The products are designed to assist forecasters and planners in instances where conditional and persistence frequencies are not obtainable by directly processing observational data. The model considers the diurnal variability of the event by assuming joint probability according to elliptical distributions defined by known (or estimated) hourly unconditional probabilities and lag correlation coefficients obtained from previously summarized data. A computer program performs the integrations by transforming the elliptical distributions to the circular normal with rotated axes, and then counting the number of mil- frequency units in each of the joint probability zones. Comparison of conditional and persistence probability estimates of categorized cloud cover and ceiling/visibility events with observed frequencies of occurrence reveal root- mean-square differences of the order of 5% to 15%.