Presenter Status
Faculty of Engineering and IT
Presentation Type
Oral Presentation
Session
B
Location
Chan Shun 108
Start Date
18-5-2017 1:55 PM
End Date
18-5-2017 2:15 PM
Presentation Abstract
The objective of the study was to use time series forecasting techniques to model Windhoek rainfall based on secondary monthly data from 1891 to 2011. Descriptive summary statistics in the form of measures of centrality and dispersion, time series plots, and autocorrelation functions were generated using R time series statistical software. The Box Jenkin’s ARIMA modelling procedure (model identification, model estimation, model validation) was used to determine the best models for the data. Model diagnostics based on residual analysis were performed to assess the adequacy of the identified models. The final model was then used to forecast monthly rainfall for Windhoek up to year 2047. The forecast values suggest that for instance in the 2046/47 , the winter season monthly rainfall point estimates are around 15mm (June 14.5mm; July 14.5 mm; and August 14.3mm) which can technically be higher than expected. However, the lower 95% confidence limits for the same winter months are zero highlighting the possibility of no rainfall during those periods. Based on the ARIMA modelling of the Windhoek rainfall, despite the seasonal and irregular fluctuations, the mean monthly rainfall levels did not suggest an upward or downward trend over the century. Even though the results indicate constant mean monthly rainfall, the limitation of the results is that the analysis is based on a small spatial area to completely rule out climate change effects. Therefore, more adaptive governance initiatives should be explored on the available secondary sources for water security and the sustainable development of the USB.
Biographical Sketch
Mr Godfrey Tichaona Pazvakawambwa is a Phd Engineering student and Infrastructure Planning Manager at Namibia Water Corporation.
A Time-series Forecasting Model for Windhoek Rainfall, Namibia
Chan Shun 108
The objective of the study was to use time series forecasting techniques to model Windhoek rainfall based on secondary monthly data from 1891 to 2011. Descriptive summary statistics in the form of measures of centrality and dispersion, time series plots, and autocorrelation functions were generated using R time series statistical software. The Box Jenkin’s ARIMA modelling procedure (model identification, model estimation, model validation) was used to determine the best models for the data. Model diagnostics based on residual analysis were performed to assess the adequacy of the identified models. The final model was then used to forecast monthly rainfall for Windhoek up to year 2047. The forecast values suggest that for instance in the 2046/47 , the winter season monthly rainfall point estimates are around 15mm (June 14.5mm; July 14.5 mm; and August 14.3mm) which can technically be higher than expected. However, the lower 95% confidence limits for the same winter months are zero highlighting the possibility of no rainfall during those periods. Based on the ARIMA modelling of the Windhoek rainfall, despite the seasonal and irregular fluctuations, the mean monthly rainfall levels did not suggest an upward or downward trend over the century. Even though the results indicate constant mean monthly rainfall, the limitation of the results is that the analysis is based on a small spatial area to completely rule out climate change effects. Therefore, more adaptive governance initiatives should be explored on the available secondary sources for water security and the sustainable development of the USB.
Acknowledgements
The authors acknowledge Namibia Water Corporation (NamWater) and the Namibia Meteorological Services for providing rainfall data and the University of Namibia for funding this research.