Statistical modelling of 5-day average rainfall in Australia during 1950-2013

Abstract

Many regions across Australia have high rainfall variability, which has various effects on water and food availability. This study examines rainfall patterns over consecutive 5-day periods during 1950-2013 using data from 92 observational stations in Australia. The first model used factor analysis to classify the stations into distinct geographical regions. Gamma generalised linear model (GLM) was then fitted to describe the rainfall amount in each category with season and year factors as the predictors. Factor analysis revealed eight factors, which represent eight geographical regions of Australia. Analysis of the similarities in the seasonal evolutions between regions revealed three seasonal rainfall groupings. The GLM models fitted the data quite well in all the areas and showed that 5-daily rainfall is significantly affected by the period of the year and its annual average in most of the regions. The models could be used to simulate rainfall data for the areas with inadequate rainfall records.

Keywords

• rainfall variability
• classification
• generalised linear models
• factor analysis