Abstract
Long dry spells (sequence of dry days) are rare events, but they are important because they correlate significantly with the area burned during bad wildfire years. Previous attempts to model the frequency of dry spells have been successful for spells of short duration, but have failed for prolonged dry spells.
In the current study, an empirical method has been developed that yields a realistic estimate of the probability of a spell of any duration. The theoretical framework proposes that the data can be explained partly by the dichotomy of weather into blocked and nonblocked westerly flows. A bimodal distribution of dry consecutive days is a consequence of this dichotomy.
The transitional probability of a dry day following k dry days generally pub at k = 1, declines to a shoulder for small k values, and then rises slowly to an asymptotic value that must be estimated from sparse and highly irregular data.
