Abstract
A multivariate time-series model that uses a factor-analytic approach is shown to provide an effective description of both monthly and annual rainfall in south Florida. In the case of monthly rainfall, the scale model shows that deviations from monthly means are caused primarily by large-scale phenomena that have temporal structure. These sort of phenomena are not accounted for by using conventional contemporaneous ARMA models. In the case of annual rainfall, the majority of variance is associated with random normally distributed large-scale phenomena that do not have temporal structure.
