Bayesian trend analysis in annual rainfall total, duration and maximum in the Kara River basin (West Africa)

Abstract

Study region: The Kara River basin, northern Togo and Benin, West Africa. Study focus: This study investigated long-term trends in annual rainfall (annRAIN), annual rainfall duration (DURATION) and annual maximum rainfall (MAXAN) for seven stations between 1950 and 2010. A Bayesian trend analysis was performed by fitting the Lognormal, Normal and Generalized Extreme Value (GEV) distributions to annRAIN, DURATION and MAXAN, respectively, with a time covariate for both the location and scale parameters. Spatio-temporal variation of the mean decadal rainfall and the seasonality of the mean monthly rainfall were also analyzed. New hydrological insights for the region: The results indicate that the interannual variability of annRAIN is decreasing over time at all stations, and the average annRAIN is also markedly decreasing at several stations. However, DURATION is increasing at most stations suggesting that in those parts of the basin where annRAIN is decreasing, rainfall occurs more frequently but with less intensity. For MAXAN, evidence for decreasing trend is found in two stations, and for increasing trend in one station. It is also shown that the peak of the rainy season shifted from September to August since the 1980s. Furthermore, changes in the spatio-temporal distribution of the mean decadal rainfall are also observed. This study provides valuable new insights into trends affecting rainfall variables in the Kara River basin.

Keywords

• Bayesian trend analysis
• Rainfall variables
• Probability distribution
• Time covariate
• The Kara River basin
• West Africa