Understanding ethiopia rainfall seasonality, trend and cyclic nature from traditional time series analysis

The selection of best stations for this study is made by comparing available data set with number of missing months at each station. Stations with 20% and more than 20% missing data are discarded if they are 1st or 3rd class stations and more than 10% if they are synoptic stations. Further securitization of the data is made by looking at the nature of missing data. Stations with more than eleven months successive missing data are eliminated from the analysis even if the percentage of missing data is within the acceptance range.  For this study station with less than twenty four years data period have not been considered.  A total of 64 stations are selected for the study out of available stations. We have used traditional time series analysis in order to study the time fluctuations of monthly rainfall series of selected stations by decomposing the given monthly rainfall data into its trend, seasonal, cyclic and irregular components. All the sixty four stations data is decomposed into its time series components separately. Trend analysis shows that positive monthly rainfall trend is the case over some parts of central, western south-western and eastern parts while negative trend is the case over the north and south. However the slopes are very small and are not as such very significant. Seasonality is the dominate feature in monthly rainfall records. The maximum zone of seasonal index shifts from one place to the other as the year progresses. In January the maximum zone of seasonal index is located in central and southern Ethiopia where as it shows a slight east ward shift in February. With more additional new higher contours the zone of maximum seasonal index maintains its February’s position in March. During April the zone of maximum values shifts to southeast. The north-western parts begin to see better values as compared to previous four months in May. The zone of maximum seasonal index values moves to Amhara region in July and August. In September the maximum is located over parts of western Ethiopia. In October the situation over the south-eastern and southern parts improved. Though there are some differences in values and location of high seasonal rainfall index values there is generally a resemblance in pattern among January, February and March, between July and August, between November and December. The strength of seasonal index is high in the country during July and August and low during January and December. In rainfall regime C all stations have peak seasonality index in April. In August the index gets more pronounced over rainfall regime A. The case when cyclic values are high and what sort of cyclic repetitions occurs have been examined. We have not found any systematic upward and downward regular and uniform cyclic movements in any of the stations. A definite cycle length cannot be defined.  However some stations have more number of high cyclic indexes than others. The irregular component varies between 0 and 3 and it does not follow any well defined pattern. How we use time series analysis to forecast monthly and seasonal rainfall of a given station is discussed in the next document of this study.