To improve understanding and progress towards prediction of waterspout probabilities on weekly to monthly time scales, we analyzed time series of temporally aggregated waterspout counts, and their relationship to atmospheric conditions. As a first step, a dataset of waterspout days, defined as days on which one or more waterspouts have been sighted and reported, was constructed using 2006-2016 Local Storm Reports for the Florida Keys and surrounding waters. A Poisson trend analysis of monthly waterspout days for July and August finds no statistically significant trend; for June it suggests a statistically significant negative trend of roughly 8% per year and for September, a statistically significant positive trend of roughly 10% per year. The total number of reports for the wet season as a whole shows no statistically significant trend.
The distribution of monthly waterspout counts for some months, most noticeably July and August, has a bimodal appearance. This suggests differences in the underlying probability of waterspout days within different months. Prior analysis (Devanas and Stefanova 2017) has demonstrated that the probability of waterspout occurrence on a given day can be successfully modeled with binary logistic regression using a small subset of independent sounding-derived parameters. A similar approach applying Poisson regression to reanalysis-derived parameters can be used to model waterspout day counts on a monthly time scale. Attendant distributions of relevant atmospheric variables are analyzed for insight into determining the characteristics of the monthly mean atmospheric states that are responsible for above or below normal probability of waterspout days within that month. As a next step, the relationship between intraseasonal atmospheric variability and waterspout probability is examined.