Performance of a climate forecast model applied to seasonal rainfall in the State of Pernambuco - Brazil

In this research the performance of a probabilistic climate forecast model applied to seasonal rainfall of Pernambuco State was assessed. Four rainy seasons (EC) were defined and each one was subdivided in two periods - initial (Xi) and final (Yi) - for three mesorregions of Pernambuco State. The methodology proposed consisted in adjusting proportions of Zi (ratio of accumulated rainfall of the first period (Xi) and the total rainfall of each EC (Xi + Yi) in the corresponding EC, year by year, over a time series of at least 30 years, to the Beta distribution for obtaining the minimum accumulated rainfall expected at 80% probability level for the period Yi of EC. The validation of the model was verified with data collected from 1994 to 2002 for each EC. The Lilliefors goodness-of-fit test was used to evaluate the adjustment of model for each set of Zi values. Results showed a good agreement between the empirical distribution and Beta model. There was an increasing trend of lesser agreement in the prediction of Ymin during dry or rainy years and good agreement in years considered as normal. The performed model showed success in the prognostics of Ymin in more than 80% of the studied localities, over the validation period.