Time Series Modelling of Monthly average Temperature in Gaborone-Botswana

Abstract

The seasonal series on the average maximum temperatures in Gaborone (Botswana) is used to identify the best time series model that can be used for forecasting. The series was found to be highly seasonal. Seasonally adjusting the series prior to applying the Box and Jenkins procedure did not average out the seasonal effects, despite giving a fairly good ARIMA(1,1,1). We also fitted SARIMA(p,d,q)(P,D,Q) with seasonality effects at lags s=12,24,36. Correlograms, Dickey Fuller tests and other model comparison methods led to an ARIMA(1,1,1)(0,1,1)[12]. The seasonal multiplicative SARIMA was found to be parsimonious as compared to an additive seasonal SARIMA.