Abstract
A total of 511 056 hourly observations of precipitation, taken over a 13-year period at nine stations, were studied to obtain a better understanding of the characteristics of persistence, runs and recurrence. Each hourly precipitation observation was categorized as either none, light, moderate or heavy. Probabilities of each category, except heavy, were estimated from relative frequencies determined from this large data sample and were compared with some theoretical models. The heavy category occurred too infrequently at all of the nine stations to obtain statistically stable relative frequencies from a 13-year period of record. The sample provided sufficient information about the other categories to confidently fit models to the data. The models can be applied to estimate probabilities that precipitation will be observed for sequences of x hours, or more; for exactly x hours; or at time t and also at time t + x hours.
