Variability analyses for the rainfall over the Nile Basin have been confined mostly to sub-basins and the annual mean of the hydroclimatic variable based on observed short-term data from a few meteorological stations. In this paper, long-term country-wide rainfall over the period 1901–2011 was used to assess variability in the seasonal and annual rainfall volumes in all the River Nile countries in Africa. Temporal variability was determined through temporal aggregation of series rescaled nonparametrically in terms of the difference between the exceedance and non-exceedance counts of data points such that the long-term average (taken as the reference) was zero. The co-occurrence of the variability of rainfall with those of the large-scale ocean–atmosphere interactions was analyzed. Between 2000 and 2012, while the rainfall in the equatorial region was increasing, that for the countries in the northern part of the River Nile was below the reference. Generally, the variability in the rainfall of the countries in the equatorial (northern) part of the River Nile was found to be significantly linked to occurrences in the Indian and Atlantic (Pacific and Atlantic) Oceans. Significant linkages to Niño 4 regarding the variability of both the seasonal and annual rainfall of some countries were also evident.
INTRODUCTION
Overview of River Nile countries and their annual rainfall data
. | . | Area (km2) . | Area (%) . | Rainfall (mm/year) . | ||||||
---|---|---|---|---|---|---|---|---|---|---|
S no. . | Country . | TAC . | ACWB . | PTAB . | PTAC . | MAX . | SD . | Cv . | Cs . | Ku . |
1 | Uganda | 235,880 | 231,366 | 7.40 | 98.10 | 1560.8 | 114.14 | 0.10 | 0.14 | 1.13 |
2 | Kenya | 580,370 | 46,229 | 1.50 | 8.00 | 1067.4 | 120.88 | 0.19 | 0.91 | 1.56 |
3 | Tanzania | 945,090 | 84,200 | 2.70 | 8.90 | 1327.5 | 122.82 | 0.12 | 0.38 | −0.22 |
4 | Rwanda | 26,340 | 19,876 | 0.60 | 75.50 | 1471.6 | 138.11 | 0.12 | 0.02 | 0.03 |
5 | Burundi | 27,834 | 13,260 | 0.40 | 47.60 | 1533.9 | 137.83 | 0.12 | 0.12 | 0.23 |
6 | DRC | 2,344,860 | 22,143 | 0.70 | 0.90 | 1755.3 | 69.19 | 0.05 | −0.29 | 2.31 |
7 | Sudan | 2,505,810 | 1,978,506 | 63.60 | 79.00 | 540.5 | 46.81 | 0.11 | −0.53 | 1.07 |
8 | Ethiopia | 1,100,010 | 365,117 | 11.70 | 33.20 | 1165.6 | 95.44 | 0.11 | 0.59 | 0.70 |
9 | Eritrea | 121,890 | 24,921 | 0.80 | 20.40 | 649.3 | 75.40 | 0.20 | 1.01 | 1.50 |
10 | Egypt | 1,001,450 | 326,751 | 10.50 | 32.60 | 67.3 | 9.76 | 0.23 | 0.36 | −0.42 |
. | . | Area (km2) . | Area (%) . | Rainfall (mm/year) . | ||||||
---|---|---|---|---|---|---|---|---|---|---|
S no. . | Country . | TAC . | ACWB . | PTAB . | PTAC . | MAX . | SD . | Cv . | Cs . | Ku . |
1 | Uganda | 235,880 | 231,366 | 7.40 | 98.10 | 1560.8 | 114.14 | 0.10 | 0.14 | 1.13 |
2 | Kenya | 580,370 | 46,229 | 1.50 | 8.00 | 1067.4 | 120.88 | 0.19 | 0.91 | 1.56 |
3 | Tanzania | 945,090 | 84,200 | 2.70 | 8.90 | 1327.5 | 122.82 | 0.12 | 0.38 | −0.22 |
4 | Rwanda | 26,340 | 19,876 | 0.60 | 75.50 | 1471.6 | 138.11 | 0.12 | 0.02 | 0.03 |
5 | Burundi | 27,834 | 13,260 | 0.40 | 47.60 | 1533.9 | 137.83 | 0.12 | 0.12 | 0.23 |
6 | DRC | 2,344,860 | 22,143 | 0.70 | 0.90 | 1755.3 | 69.19 | 0.05 | −0.29 | 2.31 |
7 | Sudan | 2,505,810 | 1,978,506 | 63.60 | 79.00 | 540.5 | 46.81 | 0.11 | −0.53 | 1.07 |
8 | Ethiopia | 1,100,010 | 365,117 | 11.70 | 33.20 | 1165.6 | 95.44 | 0.11 | 0.59 | 0.70 |
9 | Eritrea | 121,890 | 24,921 | 0.80 | 20.40 | 649.3 | 75.40 | 0.20 | 1.01 | 1.50 |
10 | Egypt | 1,001,450 | 326,751 | 10.50 | 32.60 | 67.3 | 9.76 | 0.23 | 0.36 | −0.42 |
TAC, total area of the country; ACWB, area of the country within the Nile basin; PTAB, as a percentage of total area of Nile basin; PTAC, as a percentage of area of the country; MAX, long-term maximum event; SD, standard deviation; Cv, coefficient of variation; Cs, skewness; Ku, actual excess kurtosis.
The weather system's variability is a crucial issue in water and agricultural management especially in the tropics and its stationarity for any extended period is highly unlikely. According to the International Water Management Institute (IWMI) (2014), over 70% of the people within the Nile Basin depend on subsistence rain-fed agriculture for their livelihoods. This situation is becoming more problematic with increasingly unpredictable rainfall (IWMI 2014). According to Melesse et al. (2011), subsistence and rain-fed agriculture, together with high rainfall variability, is one of the main causes of food insecurity and the most daunting challenge the basin faces. Unpredictability of the rainfall is reflected in terms of the low crop yields, unsustainable water use, etc. Due to the variability in the rainfall, river flows in the Nile Basin will be affected (Melesse et al. 2011). This will lower the reliability of the water available for irrigation, hydropower, etc. in the basin (McCartney & Girma 2012).
The climate variability, through its influences on the occurrences of rainfall extremes, has also greatly shaped the frequency, severity, and locations of floods and droughts in the Nile Basin. In the history of flooding in the Nile Basin, Ethiopia, Tanzania, Kenya, Sudan, and Uganda are the most affected countries in terms of the average number of flooding occurrences (Kibiiy et al. 2010). Heavy rainfall events over Khartoum and Atbara led to severe flooding in Sudan during August–September 1988 (Sutcliffe et al. 1989). The prolonged drought in the Greater Horn of Africa that ended in 2005 was followed by severe flooding in August 2006 (Nawaz et al. 2010). In the Lake Victoria Basin, despite the frequent episodes of floods in the low-lying parts (Gichere et al. 2013), in other areas long periods of drought can occur.
Apart from the questionable data quality, the accuracy in the estimation of quantiles in support of water management applications in the study area is also influenced by climate variability. For data of limited record length, Taye & Willems (2011) showed that the values of the extreme events or quantiles can be biased from those of the long term due to the decadal or multi-decadal climate oscillations. For the hydrometeorological quantiles of the Nile Basin, this has prompted both at-site uncertainty quantification (Kizza et al. 2011; Onyutha & Willems 2013, 2015b) and regionalization procedures (Nyeko-Ogiramoi et al. 2012; Kizza et al. 2013; Onyutha & Willems 2015c). However, the data used in the above studies on the at-site and regional analyses are based on only a few hydrometeorological stations and are of short record lengths.
Understanding variability in rainfall is a critical factor, among other meteorological variables to determine the influence of the climate system on hydrology in support of risk-based water resources management. Possible driving forces of rainfall variability might come from the large scale ocean–atmosphere interactions and/or anthropogenic factors. Assessment of the co-occurrence of the rainfall variability with those of the large ocean–atmosphere interactions can be valuable especially in predicting the upcoming periods of decrease or increase in rainfall for sustainable use and management of water resources. Some of the studies undertaken on rainfall variability in the study area include Ogallo (1989), Semazzi & Indeje (1999), Indeje et al. (2000), Mbungu et al. (2012), Taye & Willems (2012), Nyeko-Ogiramoi et al. (2013), Moges et al. (2014), and Onyutha & Willems (2015a). In some of these studies, the variability analyses were confined mostly to sub-basins and annual mean of the rainfall and, moreover, based on observed short-term data selected over different time periods and from a few meteorological stations. These could altogether affect the spatial coherence of results for the variability analyses. Instead of just annual rainfall, variability in the seasonal volumes is also important for the basin given its dependency on the rain-fed irrigation system. In some of the above studies, e.g., Mbungu et al. (2012), there were no attempts to investigate any possible linkages of the rainfall variability to ocean–atmosphere interactions. In the methods used to analyze variability in the above (cited) studies, rainfall intensity values were applied directly. In such cases, computed variability anomalies may be exaggerated if there are possible outliers in the data. An option would be to use ranks of the data values for the variability analysis as tested in this study. Therefore, this study was aimed at assessing the rainfall variability of both seasonal and annual volumes in all the River Nile countries while exploring any possible linkages to the ocean–atmosphere interactions. This was done while considering the following: the country-wide long-term rainfall data over the same time period; computation of variability anomalies in a nonparametric way; and using a number of climate indices relevant to explaining the variability in the rainfall over the River Nile Basin.
When it comes to the computation of rainfall variability, different methods exist. The key ones are empirical orthogonal functions and autocorrelation spectral analysis. The empirical orthogonal function applies principal component analysis to a group of rainfall time series data to extract coherent variations that are dominant. It entails the computation of eigenvectors and eigenvalues of a covariance or a correlation matrix obtained from a group of original rainfall time series data. This method was applied on rainfall variability in the Nile Basin by, for example, Ogallo (1989), Semazzi & Indeje (1999), and Indeje et al. (2000). However, the analysis of empirical orthogonal functions can be sensitive to the choice of spatial domain and the time period. It is also possible that empirical orthogonal function analysis can create patterns from ‘noise’. Physical interpretation of such patterns from ‘noise’ becomes complicated and unreliable. According to Bretherton et al. (1992), although the method of empirical orthogonal functions is satisfactory over a wide range of data structures, it is certainly not universally optimal. The autocorrelation spectral analysis (Blackman & Tukey 1959; World Meteorological Organization (WMO) 1966) utilizes the connection of autocovariance estimation and spectral analysis by the Fourier transform. This method was applied by Nicholson & Entekhabi (1986) to assess the quasi-periodic behavior of rainfall variability in Africa and its relationship to Southern Oscillation. Recently, Ntegeka & Willems (2008) used a quantile perturbation method to study variability in rainfall extremes at Uccle in Belgium. The quantile perturbation method compares two series to obtain relative perturbation factors that are used to represent the variability. One of the series is taken to be the full time data set, while the other is a block of sample points picked from the entire historical period. This method was applied in a number of studies of the Nile Basin; see Taye & Willems (2012) and Moges et al. (2014) for the Blue Nile Basin, and Mbungu et al. (2012) and Nyeko-Ogiramoi et al. (2013) for Lake Victoria Basin. However, the quantile perturbation method is applicable to quantitative series, and for qualitative data sets such as temperature, Nyeko-Ogiramoi et al. (2013) proposed absolute values to be adopted through conversion from degree Celsius (°C) to Kelvin (K). This conversion from °C to K is a standard procedure to make values of the entire data positive. There are, however, other series with both negative and positive values, e.g., the Atlantic Multidecadal Oscillation (AMO) index, the Southern Oscillation Index (SOI), etc., as used in this study which do not have any standard procedure of conversion to only positive values without losing relevant statistics from the original data. Applying the quantile perturbation method to such series directly may, therefore, yield unrealistic perturbation factors over some time slices. In this study, temporal variability is examined by an approach herein referred to as the nonparametric anomaly indicator method (NAIM). It is mainly based on the rescaling of series nonparametrically in terms of the difference between the exceedance and non-exceedance counts of data points such that the long-term average (taken as the reference) is zero (Onyutha 2016). The use of ranks during the rescaling of series ensures that the NAIM is not affected by non-normally distributed data which frequently occur in hydrometeorology.
DATA
Rainfall
Country-wide monthly series of rainfall over the River Nile countries for the period 1901–2011 were obtained via the British Atmospheric Data Centre (BADC) (2014). The said series were from the Climatic Research Unit (CRU) year-by-year variation of selected climate variables by CountrY (CY) available as CRU CY 3.21 data (Harris et al. 2014). For the procedural details of how the CRU CY 3.21 data were processed and obtained, the reader is referred to BADC (2014). Important to note is that the CRU CY 3.21 data sets prioritized completeness, and had no missing data (BADC 2014). Through pre-analysis checks, the obtained rainfall data for the River Nile countries before their use in this study were confirmed to have no missing data. Although South Sudan became independent from Sudan in 2011, the CRU CY 3.21 data set under the name Sudan comprises averages over the two countries.
From the monthly series, seasonal and annual rainfall totals were obtained for analyses of the temporal variability. Variability of rainfall volumes as considered in this study is important for management of agricultural practices. Table 1 shows for all the countries the data record length, long-term maximum total (MAX), standard deviation (SD), coefficient of variation (Cv), skewness (Cs), and actual excess kurtosis (Ku) of the annual rainfall time series. Since for a normal distribution it is expected that Cs = 0 and Ku = 0, it can be said that the rainfall data for the different countries were, on average, slightly positively skewed (Cs = 0.272) and marginally leptokurtic (Ku = 0.787). Compared with the normal distribution, the central peak of the leptokurtic distribution is higher and sharper, and its tails are longer and fatter. However, the highest absolute values of Cs and Ku for the annual rainfall were 1.01 and 2.31 in Eritrea and DRC, respectively. With respect to the annual rainfall, the Cv ranged from 0.05 (DRC) to 0.23 (Egypt), which showed a moderate variability on a year-to-year basis in the study area. Generally, the latitudinal decrease in the rainfall statistics from upstream to downstream of the River Nile is also reflected in the higher magnitude of the runoff values and its variability in the southern rather than the northern part (Nyeko-Ogiramoi et al. 2012). Ethiopia, Sudan, Eritrea, and Egypt, i.e., countries in the northern half of the River Nile, showed lower values of SD than those of the equatorial region (Uganda, Kenya, Tanzania, Rwanda, Burundi, and DRC). The information presented in columns 2 to 5 of Table 1 was obtained online from Food and Agriculture Organization (FAO) via the link http://www.fao.org/ (last accessed on 18 July 2014).
Series related to the large ocean–atmosphere interactions
An increase or decrease in the sea level atmospheric pressure, i.e., the sea level pressure, can reveal useful information on the atmospheric circulation which brings about the drier and wetter conditions, respectively (Onyutha & Willems 2015a). Changes in the sea surface temperature (SST), on the other hand, can generate an imbalance in the heat-flux field which can also bring about the anomalous atmospheric circulation and rainfall patterns (Horel 1982). To gain an insight into the consequences of rainfall variability due to the pressure changes occurring in different oceans and the anomaly in circulation due to the SST (Onyutha & Willems 2015a), the following readily and freely available climate indices were used:
(1) The SOI is defined as the normalized pressure difference between Tahiti and Darwin. The SOI data based on Ropelewski & Jones (1987) were obtained from the CRU (see Table 2).
(2) The AMO index is defined as the SST averaged over 25 °–60 °N, 7 °–70 °W minus the regression of this SST on global mean temperature (van Oldenborgh et al. 2009). Although Trenberth & Shea (2006) included the tropical region in their definition of the AMO, van Oldenborgh et al. (2009) decided to leave it out since it is also influenced by the El-Niño Southern Oscillation (ENSO). The AMO index was obtained from the Koninklijk Nederlands Meteorologisch Instituut through their climate explorer.
(3) The Pacific Decadal Oscillation (PDO) index is the leading principal component of the North Pacific monthly SST variability (poleward of 20 °N in the Pacific Basin). The digital values of the PDO index of Mantua et al. (1997) were obtained from the database of the Joint Institute for the Study of the Atmosphere and Ocean.
(4) The Indian Ocean Dipole (IOD) is the anomalous SST difference between the western (50 °E to 70 °E and 10 °S to 10 °N) and the south-eastern (90 °E to110 °E and 10 °S to 0 °N) equatorial Indian Ocean. The IOD series was obtained from the Japan Agency for Marine-Earth Science and Technology.
(5) The area averaged Niño SST indices (Trenberth 1997; Rayner et al. 2003) for the tropical Pacific region described by Niño 4 (150 °W to 160 °E and 5 °N to 5 °S) were obtained from the Earth System Research Laboratory of the National Oceanic and Atmospheric Administration.
List of series related to the ocean–atmosphere interactions
Data . | Downloaded from: . | Accessed on: . | Period . |
---|---|---|---|
SOI | http://www.cru.uea.ac.uk/cru/data/soi/ | 29 January 2013 | 1900–2004 |
AMO | http://climexp.knmi.nl/data/iamo_hadsst2.dat | 29 January 2013 | 1900–2012 |
PDO | http://jisao.washington.edu/pdo/PDO.latest | 30 January 2013 | 1900–2011 |
IOD | http://www.jamstec.go.jp/frcgc/research/d1/iod | 20 January 2014 | 1900–2003 |
Niño 4 | http://www.esrl.noaa.gov/psd/gcos_wgsp/Timeseries/Nino4/ | 29 January 2013 | 1900–2012 |
Data . | Downloaded from: . | Accessed on: . | Period . |
---|---|---|---|
SOI | http://www.cru.uea.ac.uk/cru/data/soi/ | 29 January 2013 | 1900–2004 |
AMO | http://climexp.knmi.nl/data/iamo_hadsst2.dat | 29 January 2013 | 1900–2012 |
PDO | http://jisao.washington.edu/pdo/PDO.latest | 30 January 2013 | 1900–2011 |
IOD | http://www.jamstec.go.jp/frcgc/research/d1/iod | 20 January 2014 | 1900–2003 |
Niño 4 | http://www.esrl.noaa.gov/psd/gcos_wgsp/Timeseries/Nino4/ | 29 January 2013 | 1900–2012 |
SOI, Southern Oscillation Index; PDO, Pacific Decadal Oscillation index; AMO, Atlantic Multidecadal Oscillation; IOD, Indian Ocean Dipole.
The above climate indices were considered for this study because they were found to be relevant in explaining the variability patterns of rainfall over the Nile Basin in a number of studies including those of Taye & Willems (2012), Nyeko-Ogiramoi et al. (2013), Moges et al. (2014), and Onyutha & Willems (2015a).
METHODOLOGY
Spatial differences in rainfall statistics
Differences across the study area were assessed in terms of dry and wet seasons. This was done using the long-term mean of the monthly rainfall in the different countries. Differences in the annual rainfall totals were also examined. Strong spatial differences of these rainfall statistics could indicate differences in possible driving forces of the temporal variability in the River Nile countries.
Detection of variability using the NAIM
This nonparametric rescaling (Equation (1)) ensured the original series remained unchanged except for its mean which became zero. This was vital to separate the clusters of rainfall which could appear either below or above the long-term average. The next step was to smoothen the fluctuations in the given rainfall series by passing a moving averaging window of a particular duration (block length Lb) through the full time series.
Therefore, to determine the variability in the given series using the NAIM, the steps included:
(1) selecting the Lb;
(2) rescaling of the series using Equation (1);
where ww = j + (Lb–1) and 1 ≤ j ≤ (n + 1–Lb);
(4) obtaining the absolute maximum value (AMV) from the aggregated series Mv(j) of Step (3);
(5) calculating anomaly, expressed in terms of percentage, as a ratio of the Mv(j) to AMV; and
(6) plotting the anomaly (%) against the mid-points of the time slices.
The time step of the Lb in this study was taken as one year. In Step (6), the average conditions over the long term were represented by zero percent, and the values of the changes below and above were expected to characterize the temporal variation.
To assess the sensitivity of the block length Lb (i.e., a subseries of the full time data set covering the period of interest), preliminary analysis was conducted using block periods of 5, 10, and 15 years. The choice of Lb is subjective and may depend on the objective of the study. The value of Lb = 10 years was found to give a much clearer anomaly pattern in the rainfall data than for 5 and 15 years, and was eventually adopted for this study.
To understand variability in rainfall and/or the driving forces, the NAIM was applied to:
(1) series of annual rainfall totals and mean of climate indices in each year; and
(2) series of seasonal rainfall totals and mean of climate indices in the corresponding season.
Test of significance in the NAIM
The null hypothesis (H0) that the observed variability in the rainfall is caused by only the natural randomness (i.e., there is no persistence in the temporal climate variation) was considered. To verify the said hypothesis at a given level of significance (α%), nonparametric bootstrapping (Davidson & Hinkley 1997) by Monte Carlo simulations was employed. Consider NMC as the number of Monte Carlo simulations; the bounds of variability using (100–α)% confidence interval (CI) were established as follows:
(1) anomaly values were obtained in the original series by applying the NAIM;
(2) the original full time series was randomly shuffled to obtain a new temporal sequence;
(3) the new series was divided to obtain subseries each of length equal to the block length Lb;
(4) the NAIM was applied to the shuffled series to obtain new temporal variation of anomalies;
(5) Steps (2)–(4) were repeated NMC times to obtain NMC anomaly values for each subseries;
(6) the anomaly values were ranked from the highest to the lowest; and
(7) the upper (UL) and lower (LL) limits of the (100–α)% confidence interval (CI) were taken as the [0.005 × α% × NMC]th and [{1–(0.005 × α%)} × NMC]th anomaly values, respectively.
If the anomaly values from Step (1) fell within the CI, H0 (natural randomness) was accepted; otherwise rejected. For the rejection, the UL or LL limit of CI from Step (7) was, respectively, required to be up- or down-crossed by the anomaly values. The values of NMC and α% were set to 1,000 and 5%, respectively.
Results from the NAIM applied to the March to May (MAM) rainfall using a time slice of 10 years over the period 1901–2011; UL (LL) is the upper (lower) limit of the 95% confidence interval (CI).
Results from the NAIM applied to the March to May (MAM) rainfall using a time slice of 10 years over the period 1901–2011; UL (LL) is the upper (lower) limit of the 95% confidence interval (CI).
Correlation analysis
Any possible linkage of the rainfall variability to the large scale ocean–atmosphere interactions was sought using correlation analysis (at both α = 5 and 1%) under the null hypothesis H0 ‘there is no correlation between the variability results of the rainfall and those of the climate indices’. The correlation analysis was carried out at both annual and seasonal time scales.
RESULTS AND DISCUSSION
Spatial differences in rainfall seasons
Spatial differences in the temporal rainfall variability
Results from the NAIM applied to the annual rainfall with time slice of 10 years over the period 1901–2011; UL (LL) is the upper (lower) limit of the 95% confidence interval (CI).
Results from the NAIM applied to the annual rainfall with time slice of 10 years over the period 1901–2011; UL (LL) is the upper (lower) limit of the 95% confidence interval (CI).
In Sudan, rainfall was above the reference in the period 1920s up to around 1960 with a significant increase in the late 1950s. This was followed by a drastic and significant decrease in rainfall from the late 1960s to mid-1980s. In line with this decrease, Hulme (1992) found that the mean annual rainfall in the Sahel region (which spans across Central Sudan) for the 1970s and 1980s declined by 30%. Regarding changes in the wet season structure in Central Sudan, Hulme (1988) maintained that the said rainfall decrease from the 1960s led to much shorter rainy seasons in the 1970s and 1980s, mainly due to an earlier end. In Figure 4(j), it is evident that the periods with a significant increase and decrease in Egypt's rainfall were from the 1900s–1920 and 1930s–1960, respectively. In Eritrea, the periods with an increase (decrease) in rainfall were the 1930s, 1950s, and the late 1970s (1940s, 1960s, and mid-1990s); correspondingly for Ethiopia, the following were noted: the 1930s and 1970s (1940–1960s and 1980–2011). With data of shorter record length (from 1964 to 2004) than those used in this study, Taye & Willems (2011) reported that the upper Blue Nile Basin-wide total annual rainfall was above (below) the reference over the period 1960s to early 1980s (mid-1980s–2004). This is consistent with the results of this study as seen from Figure 4(h). Unlike in the equatorial region, the countries in the northern half of the River Nile over the period between 2000 and 2012 were characterized by their rainfall below the reference (Figure 4(g) and 4(j)). Only results for the temporal variability of the annual rainfall are presented in Figure 4. Explanation of the spatio-temporal differences in the driving forces for the observed variability of the various seasons is also presented, as will now be discussed.
Spatio-temporal differences in the rainfall variability drivers
Correlation between anomaly in full series annual rainfall and annual climate indices
. | . | Annual . | MAM season . | ||||||||
---|---|---|---|---|---|---|---|---|---|---|---|
S no. . | Country . | PDO . | SOI . | IOD . | Niño 4 . | AMO . | PDO . | SOI . | IOD . | Niño 4 . | AMO . |
1 | Uganda | −0.50 | 0.03 | 0.27 | −0.10 | −0.21 | −0.29 | 0.05 | −0.01 | −0.04 | 0.31 |
2 | Kenya | −0.10 | −0.03 | −0.16 | 0.10 | −0.27 | −0.23 | −0.04 | −0.46 | −0.20 | −0.24 |
3 | Tanzania | 0.14 | −0.34 | 0.24 | 0.26 | 0.16 | 0.12 | −0.01 | −0.21 | −0.03 | 0.28 |
4 | Rwanda | −0.29 | −0.15 | 0.50 | 0.14 | 0.23 | −0.16 | −0.14 | 0.46 | 0.17 | 0.33 |
5 | Burundi | −0.19 | −0.27 | 0.59 | 0.19 | −0.06 | −0.20 | −0.08 | 0.20 | −0.09 | 0.06 |
6 | DRC | −0.44 | 0.10 | −0.03 | −0.20 | 0.28 | 0.43 | 0.25 | 0.12 | 0.12 | −0.24 |
7 | Sudan | −0.35 | 0.48 | −0.36 | −0.28 | 0.69 | 0.16 | 0.39 | −0.01 | 0.05 | 0.48 |
8 | Ethiopia | −0.21 | 0.60 | −0.18 | −0.33 | −0.35 | −0.36 | 0.45 | −0.12 | −0.21 | 0.65 |
9 | Eritrea | 0.26 | 0.23 | −0.46 | −0.27 | 0.35 | 0.61 | −0.43 | 0.01 | 0.30 | 0.02 |
10 | Egypt | −0.06 | 0.00 | −0.22 | −0.03 | −0.77 | −0.13 | −0.45 | 0.08 | −0.13 | −0.85 |
. | . | JJAS season . | ONDa (ONDJFb) season . | ||||||||
S no. . | Country . | PDO . | SOI . | IOD . | Niño 4 . | AMO . | PDO . | SOI . | IOD . | Niño 4 . | AMO . |
1 | Uganda | −0.39 | 0.35 | −0.03 | −0.42 | 0.14 | −0.13 | −0.39 | 0.26 | 0.48 | −0.39 |
2 | Kenya | −0.25 | 0.24 | 0.27 | −0.14 | −0.04 | −0.10 | −0.06 | −0.02 | 0.24 | −0.04 |
3 | Tanzania | −0.22 | 0.33 | −0.24 | −0.51 | 0.00 | 0.10 | −0.43 | 0.52 | 0.58 | −0.33 |
4 | Rwanda | −0.08 | 0.04 | 0.38 | −0.02 | −0.10 | −0.38 | −0.16 | 0.36 | 0.28 | 0.03 |
5 | Burundi | 0.06 | 0.04 | 0.19 | −0.09 | −0.20 | −0.21 | −0.48 | 0.59 | 0.50 | 0.00 |
6 | DRC | −0.49 | 0.60 | −0.61 | −0.60 | −0.17 | −0.45 | 0.07 | 0.00 | −0.02 | −0.25 |
7 | Sudan | −0.58 | 0.42 | −0.47 | −0.35 | 0.51 | −0.35 | 0.24 | −0.15 | −0.13 | 0.31 |
8 | Ethiopia | −0.17 | 0.12 | 0.12 | −0.09 | −0.24 | −0.19 | −0.03 | −0.01 | −0.03 | 0.15 |
9 | Eritrea | −0.05 | 0.26 | −0.55 | −0.47 | 0.30 | −0.15 | −0.23 | 0.55 | 0.43 | −0.18 |
10 | Egypt | −0.10 | 0.28 | −0.61 | −0.26 | −0.30 | −0.02 | 0.10 | −0.34 | −0.02 | −0.59 |
. | . | JF season . | . | . | . | . | . | ||||
S no. . | Country . | PDO . | SOI . | IOD . | Niño 4 . | AMO . | . | . | . | . | . |
1 | Uganda | −0.18 | 0.18 | −0.12 | −0.29 | −0.04 | |||||
2 | Kenya | −0.13 | 0.11 | −0.14 | −0.02 | −0.13 | |||||
3 | Tanzania | −0.08 | −0.05 | 0.26 | 0.23 | 0.42 | |||||
4 | Rwanda | −0.21 | 0.13 | 0.29 | −0.15 | 0.48 | |||||
5 | Burundi | −0.05 | −0.11 | 0.46 | 0.20 | 0.24 | |||||
6 | DRC | 0.02 | 0.03 | 0.01 | 0.16 | −0.50 |
. | . | Annual . | MAM season . | ||||||||
---|---|---|---|---|---|---|---|---|---|---|---|
S no. . | Country . | PDO . | SOI . | IOD . | Niño 4 . | AMO . | PDO . | SOI . | IOD . | Niño 4 . | AMO . |
1 | Uganda | −0.50 | 0.03 | 0.27 | −0.10 | −0.21 | −0.29 | 0.05 | −0.01 | −0.04 | 0.31 |
2 | Kenya | −0.10 | −0.03 | −0.16 | 0.10 | −0.27 | −0.23 | −0.04 | −0.46 | −0.20 | −0.24 |
3 | Tanzania | 0.14 | −0.34 | 0.24 | 0.26 | 0.16 | 0.12 | −0.01 | −0.21 | −0.03 | 0.28 |
4 | Rwanda | −0.29 | −0.15 | 0.50 | 0.14 | 0.23 | −0.16 | −0.14 | 0.46 | 0.17 | 0.33 |
5 | Burundi | −0.19 | −0.27 | 0.59 | 0.19 | −0.06 | −0.20 | −0.08 | 0.20 | −0.09 | 0.06 |
6 | DRC | −0.44 | 0.10 | −0.03 | −0.20 | 0.28 | 0.43 | 0.25 | 0.12 | 0.12 | −0.24 |
7 | Sudan | −0.35 | 0.48 | −0.36 | −0.28 | 0.69 | 0.16 | 0.39 | −0.01 | 0.05 | 0.48 |
8 | Ethiopia | −0.21 | 0.60 | −0.18 | −0.33 | −0.35 | −0.36 | 0.45 | −0.12 | −0.21 | 0.65 |
9 | Eritrea | 0.26 | 0.23 | −0.46 | −0.27 | 0.35 | 0.61 | −0.43 | 0.01 | 0.30 | 0.02 |
10 | Egypt | −0.06 | 0.00 | −0.22 | −0.03 | −0.77 | −0.13 | −0.45 | 0.08 | −0.13 | −0.85 |
. | . | JJAS season . | ONDa (ONDJFb) season . | ||||||||
S no. . | Country . | PDO . | SOI . | IOD . | Niño 4 . | AMO . | PDO . | SOI . | IOD . | Niño 4 . | AMO . |
1 | Uganda | −0.39 | 0.35 | −0.03 | −0.42 | 0.14 | −0.13 | −0.39 | 0.26 | 0.48 | −0.39 |
2 | Kenya | −0.25 | 0.24 | 0.27 | −0.14 | −0.04 | −0.10 | −0.06 | −0.02 | 0.24 | −0.04 |
3 | Tanzania | −0.22 | 0.33 | −0.24 | −0.51 | 0.00 | 0.10 | −0.43 | 0.52 | 0.58 | −0.33 |
4 | Rwanda | −0.08 | 0.04 | 0.38 | −0.02 | −0.10 | −0.38 | −0.16 | 0.36 | 0.28 | 0.03 |
5 | Burundi | 0.06 | 0.04 | 0.19 | −0.09 | −0.20 | −0.21 | −0.48 | 0.59 | 0.50 | 0.00 |
6 | DRC | −0.49 | 0.60 | −0.61 | −0.60 | −0.17 | −0.45 | 0.07 | 0.00 | −0.02 | −0.25 |
7 | Sudan | −0.58 | 0.42 | −0.47 | −0.35 | 0.51 | −0.35 | 0.24 | −0.15 | −0.13 | 0.31 |
8 | Ethiopia | −0.17 | 0.12 | 0.12 | −0.09 | −0.24 | −0.19 | −0.03 | −0.01 | −0.03 | 0.15 |
9 | Eritrea | −0.05 | 0.26 | −0.55 | −0.47 | 0.30 | −0.15 | −0.23 | 0.55 | 0.43 | −0.18 |
10 | Egypt | −0.10 | 0.28 | −0.61 | −0.26 | −0.30 | −0.02 | 0.10 | −0.34 | −0.02 | −0.59 |
. | . | JF season . | . | . | . | . | . | ||||
S no. . | Country . | PDO . | SOI . | IOD . | Niño 4 . | AMO . | . | . | . | . | . |
1 | Uganda | −0.18 | 0.18 | −0.12 | −0.29 | −0.04 | |||||
2 | Kenya | −0.13 | 0.11 | −0.14 | −0.02 | −0.13 | |||||
3 | Tanzania | −0.08 | −0.05 | 0.26 | 0.23 | 0.42 | |||||
4 | Rwanda | −0.21 | 0.13 | 0.29 | −0.15 | 0.48 | |||||
5 | Burundi | −0.05 | −0.11 | 0.46 | 0.20 | 0.24 | |||||
6 | DRC | 0.02 | 0.03 | 0.01 | 0.16 | −0.50 |
The bold values are significant at the level of 1%. The critical value of the correlation coefficients at significant level of 5% is 0.20.
aFor stations 1–6; bFor stations 7–10.
PDO, Pacific Decadal Oscillation index; SOI, Southern Oscillation Index; IOD, Indian Ocean Dipole; AMO, Atlantic Multidecadal Oscillation; DRC, Democratic Republic of Congo; JF, January to February; MAM, March to May; JJAS, June to September; OND, October to December; ONDJF, October to February.
Results from the NAIM applied to the March to May (MAM) rainfall and climate indices; the correlation coefficient between the two curves of each chart is put as label in {}.
Results from the NAIM applied to the March to May (MAM) rainfall and climate indices; the correlation coefficient between the two curves of each chart is put as label in {}.
For the equatorial region (stations 1–6 in Table 1), the variability of the October–December (OND) (March-May (MAM)) rainfall was shown to have significant correlations with Niño 4 and IOD (AMO). The variability in the JF rainfall was also noted to be influenced by those from the IOD and AMO. For most countries of the equatorial region, the variability in the annual rainfall was found to be significantly linked with influences from the Indian Ocean. Influences from the Pacific Ocean in the form of PDO were not so conclusive for the annual (except the June to September (JJAS)) rainfall on a regional basis because the signs of the correlation coefficients were not consistent in most, if not all, of the countries. The linkage between the rainfall of the equatorial region and the Indian Ocean was also found by Camberlin (1997), Awange et al. (2013), and Tierney et al. (2013). Significant correlation with Niño 4 is consistent with the results from a number of studies in which the influence of the ENSO on rainfall in the equatorial region was found; see, for example, Nicholson & Entekhabi (1986), Ropelewski & Halpert (1987), Ogallo (1989), Nicholson (1996), Nicholson & Kim (1997), Indeje et al. (2000), Philipps & McIntyre (2000), Schreck & Semazzi (2004), and Awange et al. (2013). It was shown that the signs of the correlation coefficients for the annual and the October–December (OND) rainfall were consistent. This means that drivers of variability in annual rainfall could be those that cause variations in the OND rainfall.
For the countries in the northern half of the River Nile, Table 3 shows that there was significant positive (negative) correlation between variability in the annual rainfall and SOI (Niño 4). Correspondingly for the June–September (JJAS) season, significant linkages were found with IOD (SOI). Significant correlation was found between SOI and the MAM rainfall of the countries in the northern part of the River Nile. These findings, again, agree with results from a number of previous studies. For instance, the influence of variability in the rainfall of the said region by the ENSO was found by Osman & Shamseldin (2002) for Sudan; Grist & Nicholson (2001) for the Sahel belt; and Beltrando & Camberlin (1993), Seleshi & Demarée (1995), Seleshi & Zanke (2004), Segele & Lamb (2005), Korecha & Barnston (2007), Abtew et al. (2009), and Diro et al. (2010) for Ethiopia and Eritrea. The variability of annual and March–May (MAM) seasonal rainfall of Ethiopia was shown to have significant correlation with the occurrences in the Pacific and Atlantic Oceans. This is consistent with the results from a number of studies. Using station-based rainfall data, Taye & Willems (2012) and Moges et al. (2014) concluded that changes in the Pacific and Atlantic Oceans are the major natural causes of the variability in the rainfall over the Blue Nile Basin of Ethiopia. Jury (2010) also found that the rainfall variability of the northern and southern zones of Ethiopia are enhanced by the Atlantic and Pacific SST oscillation pattern in warm and cool phase, respectively. The variability in both annual and seasonal rainfall of Egypt was found to be significantly correlated with the influence from the Atlantic Ocean in terms of the AMO index.
Generally, it was shown that the magnitudes (and sometimes the signs) of the correlation between the NAIM results of rainfall with those of climate indices varied from one country to the next. This might have been due to randomness in the rainfall data or difference in microclimate (micro-scale features). Importantly, this might have also been due to the connection of the rainfall variability with the influence of regional features (topography, water bodies, land cover, etc.) over atmospheric circulation. The highlands which bind the Nile Basin to the east, from Eritrea to Kenya, restrict penetration of the easterlies from the Indian Ocean; an exception is the gap between the Ethiopian and Kenyan highlands (Camberlin 2009). Furthermore, the highlands of the Rift Valley also block the moist, unstable westerly flow of the Congo air from reaching the coastal areas (especially of Ethiopia), leading to complex patterns of rainfall over the Ethiopian highlands (Nicholson 1996). According to Camberlin (2009), the East African Great Lakes, furthermore, tend to develop their circulation in the form of lake breezes which interact with slope circulation. The joint impact of these lake and upslope breezes is the afternoon convection. Quantifying these regional influences due to topography over atmospheric circulation could be important. This, although not undertaken in this study, can be carried out by regional numeric modeling while taking into account local topographical features alongside the three-dimensional lake dynamics to reproduce the region's climate.
CONCLUSION
This paper has assessed, using the NAIM, variability in the seasonal and annual rainfall across the River Nile countries. Possible driving influences of the rainfall variability were sought using a number of climate indices. The NAIM gave results which were consistent with those of previous studies. This showed its credibility for analysis of variability in hydrometeorological variables.
The period of high (low) rainfall was the 1960s to early 1980s (the 1920s) for the equatorial region of the Nile Basin, the 1930s and the 1970s (the 1940s–1960s and the 1980s–2011) for Ethiopia, the 1920s–1960 (the late 1960s to mid-1980s) for Sudan, the 1930s and late 1970s (the 1960s and mid-1990s) for Eritrea, and the 1900s–1920 (the 1930s–1960) for Egypt. The countries in the equatorial region (northern half) of the River Nile, over the period between 2000 and 2012 were characterized by their rainfall above (below) reference. Generally, the variability in the rainfall of the countries in the equatorial (northern) part of the River Nile was found to be correlated with those in the AMO and IOD (SOI and AMO) for the March–May season, and Niño 4 (Niño 4 and IOD) for the June–September season. Correspondingly for the variability of the annual rainfall, links to IOD (Niño 4 and AMO) were found.
Assessment of the possible influences of the large-scale ocean–atmosphere interactions on the variability of seasonal and annual rainfall volumes, as carried out in this study, makes it possible to get an insight into the considerable effect of the climate oscillations on water resources and agricultural management practices. When driving forces of the rainfall variability are known, it can be possible to separate natural variability or oscillations (of decadal cycles) from long-term trends (for instance due to global warming). This may be significantly useful especially in predicting upcoming periods of decrease or increase in the seasonal and annual rainfall volumes. Apart from the large ocean–atmosphere interactions, other possible driving forces of rainfall variability including water bodies, topography, transition in land cover and/or change, etc. should be investigated.
ACKNOWLEDGEMENTS
The author would like to thank the two anonymous reviewers for their insightful comments and suggestions that greatly enhanced the quality of the paper. The author also wishes to acknowledge the British Atmospheric Data Centre (BADC) of the Natural Environment Research Council (NERC) for granting access to the data used in this study.