Seasonal rainfall predictability over the Lake Kariba catchment area 

Get Complete Project Material File(s) Now! »

Intervention analysis

The rainfall CUSUM values for twelve of the thirteen stations depicted in Figure 4 demonstrates that the monthly rainfall has been above the long-term mean for most of the time in 11 stations. Furthermore, there appears to have been interventions in most of the stations (about 8) from 1982, resulting in a generally downward trend. The most probable dates (months since January, 1970) for the observed interventions were determined and the step change analysis was carried to validate the observed intervention. The intervention dates and the corresponding step change analysis results are given in Table 4. It appears that the test statistic (Z-values) are less than the critical value of 1.96 (5%) in all the 12 stations suggesting that the CUSUM values cannot be confirmed. These results implies that the rainfall time series in Kariba catchment area come from the same climatological region and the area experiences an oscillatory hydro-meteorological signal that has no apparent shifts over the 40 year period.

Homogeneity tests

Annual total rainfall amounts at each of the thirteen stations was tested for homogeneity by the four absolute test methods reported in e.g., Wijngaard et al. (2003) i.e., SNHT, BRT, PT and VNT and the results are given in Table 5. As reported in e.g., Wijngaard et al. (2003), Feng et al. (2004) and Sahin and Ggizoglul (2010), the absolute tests considered here could have different sensitivities to changes in rainfall series. As a result, there are apparent differences in test results across the stations illustrated in Table 5. The VNT scored four in-homogeneities; SNHT scored two in-homogeneities while PT and BRT scored one in-homogeneities each. Additionally, based on the Wijngaard et al. (2003) classification, the present study distinguished the in-homogeneities across the stations by categorizing them depending on the number of absolute tests rejecting the null hypothesis at the 5% significance level i.e., a) class A: zero or one rejection, b) class B: two rejections, and c) class C: three or more rejections.

Seasonal trends

The existence of seasonal trends in rainfall trends was assessed by applying the MK test to the DJF rainfall time series spanning 1970 to 2010. As depicted in Table 8, all the stations did not have any significant trend at the 95% significance level. As shown in Table 8 and Figure 5, there exist both negative (~ 69%) and positive (31%) insignificant trends and this structure is a complete inverse compared to the annual distribution pattern. This means that the Kariba catchment region is becoming subtly drier. This scenario might impact negatively e.g., the agricultural and livestock
production activit es (especially those activities that are time critical) in the area. In addition, decisions related to water management will be impacted due to the associated reduction of water in Lake Kariba. This subtle declining DJF rainfall, high variability and the low statistical significance will often lead to underestimating the importance of climate signals (trends) that could be very catastrophic to society and the economy. The magnitude of the seasonal trends inherent in the DJF series in the Kariba catchment area ranges between (-) 0.022 mm per year Kariba station to (+)
0.013 at Nkayi station.

READ  Tectonic and Geological Setting

Coherence of rainfall variability across stations

The CWT and WC derived from the Morlet continuous wavelet transform have been used to examine the nature of monthly total rainfall patterns across the Kariba catchment area by assessing the presence of common power and the relative phase in the time-frequency space. In particular, the phase relationships between standardized rainfall records are explored given that they have common climatology. The Monte Carlo method (using 1000 ensemble surrogate data set pairs of the red noise based on the lag-one autoregressive (AR1) model coefficients of the standardized rainfall data sets) were used to compute the statistical significance (5%) of CWT and WC. In the wavelet space, the rainfall variability in the northern region of the Kariba catchment area is depicted in Figure 6. As illustrated in Figure 6, the CWT of the standardized rainfall records at Karoi and Chibhero stations have constant in-phase and high common power (at 5% significance level) during 1978-1983 and 1995-2003. There is however a larger area with phase-lock deviation outside the 5% significance level suggesting unreliable phase-lags in rainfall between the two northern stations. Furthermore, the WC depicts larger sections with 5% significance level exhibiting inphase relationship suggesting causality in the rainfall fluctuations. Overall, the modes of rainfall variability in the northern region of the Kariba catchment area are wavelengths varying from about two to four years.

Chapter 1: Introduction 
1.1 Background
1.1.2 Rainfall characteristics in Southern Africa
1.1.3 Predictability of seasonal rainfall
1.1.4 Lake Kariba
1.2 Research problem statement
1.3 Aims and objectives of the study
1.4 Thesis outline
Chapter 2: Seasonal rainfall characteristics over Lake Kariba catchment in the Zambezi river basin, Zimbabwe 
2.1 Introduction
2.2 Study area
2.3 Data and methodology
2.3.1 Data
2.3.2 Methodology Intervention and homogeneity
2.3.3 Trend analysis
2.3.4 Wavelet based coherence analysis
2.4 Results and discussion
2.4.1 Rainfall variability in the Kariba catchment area
2.4.2 Intervention and homogeneity analysis Intervention analysis Homogeneity tests
2.4.3 Trend analysis Annual trends Seasonal trends
2.5 Coherence of rainfall variability across stations
2.6 Conclusions
Chapter 3: Seasonal rainfall predictability over the Lake Kariba catchment area 
Preface 68
Abstract 69
3.1 Introduction 71
3.2 Methods
3.2.1 The archived data of the general circulation model and gridded rainfall data 
3.2.2 Statistical downscaling
3.2.3 Verification
3.2.4 Retroactive forecast skill Relative operating characteristics
3.2.5 Reliability
3.2.6 Deterministic skill assessment
3.3 Economic value of the probabilistic forecasts
3.3.1 Predicting the ‘flooding across southern Africa’
3.4 Discussion and conclusions
Chapter 4: Prediction of inflows into Lake Kariba using a combination of physical and empirical models 
4.1 Introduction
4.2 Study area
4.3 Data
4.3.1 Rainfall data
4.3.2 ECHAM4.5-MOM3 coupled ocean–atmosphere model data
4.4 Methodology
4.4.1 Statistical downscaling
4.4.2 Verification
4.5 Results
4.5.1 Inflow hindcasts
4.5.2 Value of the probabilistic inflow forecasts
4.5.3 Year-by-year hindcasts
4.6 Independent case study: The flood season of 2010/11
4.7 Conclusion and discussion
Chapter 5: Summary, conclusions and future perspectives


Related Posts