Estimating Recurrence Probabilities of Weather Events

Abstract

Hourly observations of precipitation, sky cover, ceiling, visibility, wind speed and temperature, taken over a 13-year period at Washington National Airport, Kennedy International Airport and Raleigh-Durham Airport, were processed to obtain unconditional and recurrence relative frequencies of 10 weather events, in both winter and summer. A model to estimate recurrence probabilities of weather events from unconditional probabilities and a measure of temporal correlation was developed on Washington National Airport data. The model was tested on Kennedy International and Raleigh-Durham data by comparing the probability estimates given by the model with relative frequencies based on observations taken at the two stations. The model estimates for time lags of less than 19 h between observations were always better for all weather events than estimates made on the assumption of statistical independence. They were usually better for all lags of less than 37 h in winter and 63 h in summer.