5. DERIVATION OF NET BASIN SUPPLY TIME SERIES FOR THE EQUATORIAL
5.3 Results from application of the Equatorial Lake Model and segmentation
5.3.1 Lake Victoria
A continuous record of end of month levels at Jinja for the period 1899 – 2008, constructed from published end-month levels at the Jinja gauge and data from the archives of the DWRM in Entebbe, was compiled prior to application of the Equatorial Lake Model. A similar dataset for the period 1899 – 1995 has previously been utilised in studies by the Institute of Hydrology (1993) and Kennedy and Donkin (1996). It is considered to be of very good quality and was found to be consistent with the record published by Mott MacDonald (1998) for the concurrent period. Lake Victoria outflows utilised to run the model were compiled using the method outlined in Section 4.3.2.
The Equatorial Lake model was run to simulate operation according to the Agreed Curve with this data set. The model iterates to compute a mean monthly discharge corresponding with an average monthly water level determined from start and end of month levels. It utilises the elevation – storage characteristics for Lake Victoria (WMO, 1982) in order to compute changes in storage from changes in level and then finally computes the net basin supplies by summing the change in storage and the outflow in the Victoria Nile. Figures 5.2 and 5.3 contain simulated and observed average monthly levels and discharges respectively for Lake Victoria. From Figure 5.2 it is evident that there is very close agreement between the simulated and observed levels, particularly prior to completion of the Owen Falls dam in 1954.
Figure 5.2 Model simulation of Lake Victoria levels.
10 10.5 11 11.5 12 12.5 13 13.5
1899 1908 1917 1926 1935 1944 1953 1962 1971 1980 1989 1998 2007
Jinja Gauge level (m)
Year Observed
Agreed curve
Figure 5.3 Model simulation of Lake Victoria discharges.
As shown in Figure 5.3, the outflows used in this simulation for the period 1889 - 2008, correspond to the actual dam releases from the Nalubaale-Kiira Dams and it is noticeable from Figure 5.3 that releases continued in the period 1954-2004 to follow the Agreed Curve with minor deviations followed by compensatory releases. After 2001 there is a marked difference between the observed and simulated levels and outflows due to sustained departure from the Agreed Curve release policy.
Sutcliffe and Petersen (2007) compared measured lake levels, outflows according to the Agreed Curve and actual dam releases for the period up to 2006 and have derived a corrected natural level series which would have resulted had releases been based on the Agreed Curve.
Comparison of this naturalized series with actual measured levels revealed that the impact of recent over-abstraction on declining levels was of the order of 0.6 m.
The resulting annual net basin supply time series obtained through application of the Equatorial Lake Model were subjected to the segmentation algorithm developed by Hubert (1997, 2000) and the results are illustrated in Figure 5.4.
0 200 400 600 800 1000 1200 1400 1600 1800 2000
1899 1908 1917 1926 1935 1944 1953 1962 1971 1980 1989 1998 2007 Discharge (m3 .s-1 )
Year Observed
Agreed curve
Figure 5.4 Historical annual net basin supply computed by the model (1899-2008).
Figure 5.4 shows that application of the segmentation algorithm to the derived annual net basin supply time series of Lake Victoria for the period 1899 – 2008 produced three segments with different means at a 0.05 significance level of the Scheffe test. The three contiguous periods are 1899 – 1960, 1961 – 1963 and 1964 – 2008. Abrupt shifts occur in 1961 and 1964. The mean net basin supply during the three successive periods rises from 20,040 MCM (Million Cubic metres) to a short lived high of 80,558 MCM during three years and then declines to 33,325 MCM. From this preliminary analysis, the time series of Lake Victoria net basin supply can be construed to be non-stationary from the point of view of significance of occurrence of homogeneous segments with significantly different means.
The year 1918 was particularly dry and characterised by a negative value of net basin supply.
The net basin supply for the year 2005, which is one of the years during which the lake releases were not conforming to the Agreed Curve following the commissioning of Kiira dam, was also negative.
-20000 0 20000 40000 60000 80000 100000
1899 1908 1917 1926 1935 1944 1953 1962 1971 1980 1989 1998 2007
Annual NBS (MCM)
Year Lake Victoria net basin supply 1899 - 1960
1961 - 1963 1964 - 2008
The observed variation of annual net basin supply in Figure 5.4 is similar to the findings in a study for development of a new lake Victoria water release policy (Centre for Ecology and Hydrology, 2008), where the values were derived from 108 years of record (Sept 1899 – August 2008) and the difference is in their definition of a hydrological year i.e. September to August as opposed to the calendar year definition of January to December that is utilised in this study.
|The Institute of Hydrology (1993) evaluated the accuracy of net basin supply time series derived using the indirect approach (Figure 5.4) against those derived using a conventional water balance approach i.e. with lake rainfall and evaporation inputs for the period 1925 to 1990. It was concluded that there is reasonable agreement for both inflow models but the conventional water balance approach particularly provides a better representation of the increase in net basin supply during the 1961-64 period. The errors in net basin supply vary randomly in many periods, and are typically in the range 10,000-20,000 MCM/year. These error magnitudes were established to translate into about 0.15 - 0.30 m per year in terms of depth over the lake surface.
An assessment of the sensitivity of lake levels to variations in lake rainfall or land use change was also undertaken by the study conducted by Institute of Hydrology (1993). The findings demonstrated that Lake Victoria water balance is about 5 times more sensitive to long term changes in rainfall than to changes in basin runoff coefficient.