Rainfall is a very important meteorological variable for indicating climate change. Over the years there has been concerns of variability and changes in the climate around the world, including in Botswana, which have increasingly led to undesirable impacts like droughts, floods, destruction of the environment and property as well as loss of human life. The use of changepoint techniques to detect past abrupt shifts in climate time series has gained currency across the different regions of the world. The objective of this paper is to determine trends in the annual maximum average rainfall for Botswana during the years 1901-2012 and detect multiple change points in the series. Trend detection is carried out using the Mann-Kendall trend test. To test for stationarity, the Kwiatkowski, Phillips, Schmidt and Shin (KPSS) test is used. Multiple change points are detected using three widely used search methods, namely, the Segment Neighbourhood, Binary Segmentation and Pruned Exact Linear Time (PELT). The results show that the data are heavy-tailed, show no monotonic trend and are stationary. Both the Segment neighbourhood and Binary Segmentation detected multiple change points that correspond to documented periods of drought in the country. The Pruned Exact Linear Time (PELT) fails to detect a reasonable number of changepoints as it identifies almost all the years as changepoint locations.

Projects

Currently no objects available

Currently no objects available