In this study, average monthly and annual rainfall
totals recorded for the period 1970 to 2010 from a network of
13 stations across the Lake Kariba catchment area of the
Zambezi river basin were analyzed in order to characterize
the spatial-temporal variability of rainfall across the catchment
area. In the analysis, the data were subjected to intervention
and homogeneity analysis using the Cumulative Summation
(CUSUM) technique and step change analysis using rank-sum
test. Furthermore, rainfall variability was characterized by
trend analysis using the non-parametric Mann-Kendall statistic.
Additionally, the rainfall series were decomposed and the
spectral characteristics derived using Cross Wavelet
Transform (CWT) and Wavelet Coherence (WC) analysis.
The advantage of using the wavelet-based parameters is that
they vary in time and can therefore be used to quantitatively
detect time-scale-dependent correlations and phase shifts between
rainfall time series at various localized time-frequency
scales. The annual and seasonal rainfall series were homogeneous
and demonstrated no apparent significant shifts.
According to the inhomogeneity classification, the rainfall
series recorded across the Lake Kariba catchment area
belonged to category A (useful) and B (doubtful), i.e., there
were zero to one and two absolute tests rejecting the null
hypothesis (at 5 % significance level), respectively. Lastly,
the long-term variability of the rainfall series across the Lake Kariba catchment area exhibited non-significant positive and
negative trends with coherent oscillatory modes that are constantly
locked in phase in the Morlet wavelet space.