Abstract
There has always been a good deal of interest in the possibility of cycles being present in weather data. This article presents a class of models which allows cyclical components to be modeled explicitly. The cyclical components are stochastic rather than deterministic and the model may be used for forecasting. The statistical handling of the model is based on the state space form and the application of the Kalman filter.
The model was fitted to data on rainfall in Fortaleza, which is a town in northeast Brazil, an area which often suffers from drought. The model gives not only an excellent description of the properties of the series, but also makes it clear that the gains achieved in forecasting by taking account of the cycle are small.
