Cyclical components of local rainfall data

Abstract

This paper reports on the use of a comparatively simple statistical methodology to study local short time series rainfall data. The objective is to help in agricultural planning, by disminishing the risk associated with some uncertainties affecting this business activity. The analysis starts by assming a model of unobservable components, trend, cycle, seasonal and irregular, that is well known in many areas of application. When series are in the realm of business and economics, the statistical methods popularized by the US Census Bureau and US National Bureau of Economic Research are used for seasonal and cyclical estimation, respectively. The flexibility of these methods makes them good candiates to be applied in the meteorological context, and this is done in this paper for a selection of monthly rainfall time series. Use of the results to help in analysing and forecasting cyclical components is emphasized. The results are interesting. An agricultural entrepreneur, or a group of them located in a single ggeografical region, will profit by systematically collecting information (monthly in our work) about rainfall, and adopting the scheme of analysis described in this paper.