z

Available online at http://www.journalcra.com

INTERNATIONAL JOURNAL
OF CURRENT RESEARCH
International Journal of Current Research
Vol. 5, Issue, 06, pp.1577-1581, June, 2013

ISSN: 0975-833X
RESEARCH ARTICLE
APPLICATION OF STOCHASTIC ANALYSIS AND MODELING TECHNIQUES IN SIMULATING THE
OUTFLOW SERIES FROM LAKE TANA CATCHMENT, ETHIOPIA
Belete, Mulugeta Azeze1*, Yilma Seleshi3 and Hartmut Eckstädt2
1School
of Water Resources and Civil Engineering, Bahir Dar University, Bahir Dar, Ethiopia,
2Institute of Sanitary Engineering, Rostock University, Rostock, Germany,
3Civil Engineering Department, Addis Ababa Institute of Technology Addis Ababa University, Addis Ababa, Ethiopia

ARTICLE INFO ABSTRACT

Article History:
Lake Tana is the biggest fresh water body in Ethiopia. Recently the Lake and its catchment water
Received 20th March, 2013
resources have been identified as one of the most important sources for hydropower, irrigation, and
Received in revised form
24th April, 2013 fishery development. However, there are huge uncertainties in estimating the water balance terms of
Accepted 10th May, 2013 both the Lake and its catchment. These will have an effect in developing the resources in a
Published online 15th June, 2013 sustainable manner. The purpose of this article is to model and simulate the outflow series from the
catchment measured at the Outlet of the Lake. This will give an opportunity to look the development
Key words: impacts and their sustainability from the net supply perspective. A first order autoregressive model
Lake Tana, Net supply, with model parameters Ø1 = 0.7845 and σa = 0.612 is identified as an adequate model to represent the
Outflow series, Simulation, deseasonalized series of the net supply of the catchment.
and Stochastic Modeling.
Copyright, IJCR, 2013, Academic Journals. All rights reserved.

INTRODUCTION there are still huge uncertainties associated with the hydrology
An important aspect of water resources development is of Lake Tana catchment, the hydrologic study conducted by
planning, designing and operation of reservoirs (Jain, 2003). SMEC International Pvt. Ltd. indicates that the estimated
However, reservoirs are capital intensive and environmentally annual inflow into the lake is about 4.9 Billion cubic meters
sensitive projects that need careful planning and management. (Bm3), direct annual rainfall over the lake 3.8(Bm3),
But in very few places nature gives very indispensable Evaporation from the lake 5(Bm3), and the Lake outflow is
reservoirs such as Lake Tana which is a natural reservoir in about 3.7 (Bm3) (SMEC 2008b). Kebede et al. (2005) have
Ethiopia. In this article the terms reservoir and Lake are used also estimated the water budget component of the lake, and
synonymously. Lake Tana is the largest fresh water body in make a preliminary sensitivity analysis of Lake Level and
Ethiopia with a maximum depth of 14m and mean depth of Lake Outflow to rainfall changes. Wale et al. (2009) has
9m. It is located from 11.62o N to 12.31oN latitude and from estimated the runoff from ungauged catchments using
37.01o E to 37.64oE longitude with average natural altitude of regionalization as 880mm per year with a water balance
1786meter above sea level (m asl). At similar elevation the closure error of 5%. Rientjes et al. (2011) have also dealt with
lake has a surface area of about 3060km2 with length of 74km regionalisation studies for the purpose of simulating Lake Tana
and 68km width and storage capacity of 29km3. The Lake level. But SMEC (2008b) has indicated the tricky in
catchment including the Lake area has an area of 15320 km2 regionalization approach as the catchments generating the base
with an average elevation of 2025m asl. It is the head water of flow and surface run off for some of the major tributaries like
the Blue Nile River which is one of the major tributary to Nile Gilgel Abay are not concurrent. Chebud (2009) has also tried
River. There are four perennial rivers (Gilgil Abbay, Megech, to model the Lake stage and estimated the water balance of
Ribb, and Gumera) and many other seasonal streams in the Lake Tana. The inflow based approaches and attempts in
catchment that fed Lake Tana. Blue Nile, which is named managing the Lake and its catchment water resources are
Abay locally, is the only natural outflow from the Lake. At the associated with huge uncertainties which can be seen clearly in
natural outlet of the Lake a weir has been constructed and the discrepancies of values of the water balance terms
commissioned in 1996. The weir regulates water storage in estimated by various studies.
Lake Tana over a 3 m range from 1,784 m to 1,787 masl. The
active storage of the Lake between these levels is about 9,100 Lake Tana sub-basin has immense natural resources potential
million cubic meters (Mm3), which is approximately 2.5 times for growth. The Government of Ethiopia (GoE)’s in its Plan
the average annual outflow from the Lake. The hydrology of for Accelerated and Sustained Development to End Poverty
Lake Tana catchment is not well documented in the scientific (PASDEP) had identified Tana as the first of the five proposed
literatures except in very few journal articles and reports that growth corridors (zones) in the country (MoFED 2006). Huge
present estimate of the water balance of the Lake. Although investment plans have been prepared to develop the water
resources of the sub-basin. In line with this the constructions
*Correspondence Address: Belete, Mulugeta Azeze, Rostock, Germany of some water conveyance and storage structures for
1578 International Journal of Current Research, Vol. 5, Issue, 6, pp.1577-1581, June, 2013

hydropower and irrigation development have been started. of previous investigators see for example Wurbs and Bergman,
Tana-Beles hydropower and irrigation development project, (1990); Vogel and Bolognese (1995); and Thomas and Dan
that has an installed capacity of 460MW and supplied by Lake Rosbjerg (2004). In general reliability provides a measure of
Tana at an average supply rate of 77m3/s (SMEC, 2008c), is how often a system fails. Resilience provides a measure of
the biggest projects in the sub-basins. The construction of this how long a failure will last, and vulnerability provides a
hydropower project is completed and commissioned in May, measure of how severe failures may become. A failure is
2010. The construction of a storage Dam for irrigation of defined as the inability of reservoir system to deliver a
7000ha of land on Koga River, a tributary to Gilgel Abay, is specified yield or service. The most common definition of
also completed and the constructions of other Dams are still probability of failure is the proportion of time units or periods
going on. In Lake Tana Catchment the availability of hydro- such as months during which the reservoir cannot meet the
meteorological data is very scanty. Relatively longer periods demand under the adopted operating rule.
of reliable records are only available for the Lake Level and
Lake Outflow measured at Bahir Dar, a town located just at the In Lake Tana catchment, the observed Outflow at the outlet of
shore of the Lake and its outlet. Other gauging stations for the catchment is a single realization that has been generated
runoff measurement are located at the upstream locations of from the natural stochastic processes that can be modelled. An
the major tributaries of the lake and many stations suffer from appropriate stochastic model to this series can be fitted and the
sedimentation, bank scouring, and in a few cases from bank fitted model will then be used to generate various statistically
overflow during high river stages (SMEC 2008a). Only 40% of similar realizations. The synthetic generated realizations will
the catchment area is covered by runoff measuring gauge then be used to reassess the characteristics and performance
stations. This indicates not only the inadequacy of the measuring indices of the Lake. Usually in reservoir design and
hydrometric stations spatial coverage but also the problems in analysis synthetic inflows are generated and used to determine
getting reliable data from existing stations. The absences of the characteristics and performance indices of a reservoir.
sufficient and reliable records of inflows to the Lake have Because using historic data alone has a limitation in that the
hindered effective use of conventional approaches to assess the future may not be an exact replica of the past. For this purpose
characteristics and responses of Lake Tana in various looking the problem indirectly from the net supply perspective
development scenarios. However, this need not delay or and attempting to model it is very fundamental. The purpose of
restrict our attempt to find other alternative methods that will this article is therefore to model and simulate the Outflow (net
help us towards the creation of more sustainable water supply) of Lake Tana catchment which can be used as in input
resource systems planning and management scenarios in Lake to indirectly reassess the characteristics and performance
Tana’s catchment. The Outflow measured at the outlet of the measuring indices of Lake Tana and sustainability of other
catchment i.e. at the outlet of the lake is its net supply that can developments endeavours in its catchment.
be used indirectly to evaluate the overall impacts of upstream
developments on the characteristics and performances of the METHODS AND MATERIALS
Lake. If we look the situation from net supply perspective than
inflow we will be able to understand it and make clear The planning and management of water resources systems
conceptualization. involves the collection and analysis of many hydro-
meteorological data which are known with the property of
Everyone involved in water resource systems development and possessing an inherent natural variability. Because
management has an obligation to see that those systems observations measured from natural phenomena posses an
provide sufficient quantities and qualities, at acceptable price inherent probabilistic structure, time series models are
and reliabilities, and at the same time protect the environment employed for modelling water resources systems. Time series
and preserve the biodiversity and health of ecosystems for is defined as a sequence of values arranged in their order of
future generation (Loucks, 2000). Though Lake Tana and its occurrence in time (Box et al., 1994). A set of quantitative
catchment are very valuable resources its management needs observations such as stream flow measured in discrete time
very careful planning and operation that neither withdraw interval and arranged in chronological order is an ideal
significant water from the Lake nor allow the storage of example of time series. The time series to be analyzed may
excessive floods. Therefore, in response to the increased rate then be considered as one realization of the underlying
of development in Lake Tana catchment’s there is an stochastic process. For example, the Outflow series recorded at
intensified need to examine the catchment’s water resources the outlet of Lake Tana catchment is a single realization
system performances and their sustainability. Quantifying among other perhaps many different possible realizations of
performance measuring indices like reliability and resilience of the stochastic process. However, it is obvious that there is not
Lake Tana and reassessing them in light of various just one realization of such a process. In principle there are an
development scenarios and information is one of the classical arbitrary number of realizations which all have the same
problems that have not been solved yet. statistical properties as they all result from the same stochastic
process. One of the major uses of time-series modelling is
Water resources systems performance measures are receiving generation of synthetic values for many other realizations. The
considerably increased attention by the scientific community. generation of synthetic time series often consists of the
Reliability, resilience, and vulnerability indexes have been generation of independent variables in a series defined as a
identified and used as a measure of system performance random process. Salas et al. (1988); Box et al. (1994); and
characteristics. Likewise the performance of a reservoir or Hipel et al. (1994) have detail explanation on time series
reservoir systems is assessed in terms of its reliability, analysis. The usual stochastic model development, calibration,
resilience, and vulnerability. The definitions of reliability, and diagnostic check procedures have been adopted to identify
resilience, and vulnerability have been introduced by a number the most suitable model that can adequately represent the
1579 International Journal of Current Research, Vol. 5, Issue, 6, pp.1577-1581, June, 2013

outflow series from Lake Tana catchment. Monthly data for The quality and characteristics of the naturalized Outflow
Lake Tana water level and its Outflow have been collected for series i.e. the net supply series has been further examined
a period of 36 year since 1973 to 2008. Visual inspection of before running to the modelling process. The presences of
plots of the series against time show clear non-homogeneity outliers, trend, and the stability of variance and mean of the
and seasonality as shown in Figure 1. The reason for the non- data sets have to be examined. Wilcox (2009) has indicated the
homogeneity is because of the regulation of the Outflow since importance of box plot in assessing the existence of outliers.
1996. After detecting the regulation effect the outflow series is Hence, the presences of outliers for the naturalized Outflow
naturalized. This has been done by first naturalizing the Lake series has been assessed using box plot and found that the
Level for the period from 1996 to 2008. Naturalizing the Lake series is free from outlier. Dahmen and Hall (1990) have
level has been made by limiting the maximum and minimum recommended the use of Spearman’s rank-correlation method
levels, based on the characteristics of the unregulated lake to investigate the presences of linear trends. Assessing the
level recorded in the period from 1973 to 1995, and stability of variance and mean is very essential and critical in
interpolating the regulated Lake levels in accordance with the time series analysis especially in the analysis of stationary time
restrictions. After naturalized the Lake level a relationship has series. The stabilities of mean and variance of Lake Tana net
been developed for the unregulated lake level and outflow and supply series have been assessed using the procedures outlined
then the relationship has been applied to estimate the in Dahmen and Hall (1990). The assessment shows the
naturalized outflow in the period from 1996 to 2008. The stabilities of both the mean and variance of the Outflow series.
graphs in Figure 2 show the time series plots of naturalized Some of its statistical characteristic is presented in Table 1.
Lake Level and Outflow.

Figure 1 Time Series Plot of observed Outflow and Lake Level

700 4.5
observ ed OutFlow
Observ ed Lak e Level
600 4.0

500 3.5

Lake Level (m)

OutFlow (m/s)

3.0
400
2.5
300
2.0
200
1.5
100
1.0
0
0.5
Month Jan Jan Jan Jan Jan Jan
Year 1973 1979 1985 1991 1997 2003

Figure 2 Time Se rie s Plot of naturalize d Outflow and Lake le ve l

700 O ut Flow 4.0
Lak e Lev el
600 3.5

500
3.0
Lake Level (m)
Outflow (m/s)

400
2.5
300
2.0
200

1.5
100

0 1.0

Month Jan Jan Jan Jan Jan Jan

Year 1973 1979 1985 1991 1997 2003
1580 International Journal of Current Research, Vol. 5, Issue, 6, pp.1577-1581, June, 2013

Table 1. Statistical characteristics of naturalized Outflow from Lake Tana catchment

Average Std.Dev Coefficient of Minimum Maximum(Mm3 ) Skewness
(Mm3) (Mm3) variation (%) (Mm3)
Jan 71.58 16.24 22.68 35.78 106.18 0.04
Feb 47.45 11.77 24.80 21.20 73.80 0.02
Mar 31.58 8.63 27.33 16.02 50.10 0.28
Apr 19.44 3.99 20.54 12.44 28.84 0.31
May 13.86 2.65 19.13 8.56 19.46 0.09
Jun 13.96 4.75 34.01 6.52 27.24 0.79
Jul 37.25 17.25 46.30 11.55 71.83 0.54
Aug 176.5 84.3 47.74 63.7 330.0 0.51
Sep 334.6 136.4 40.76 160.7 638.1 0.59
Oct 302.9 99.6 32.89 134.0 543.7 0.56
Nov 190.45 57.48 30.18 106.18 307.82 0.41
Dec 120.98 33.17 27.41 69.63 193.53 0.56

Figure 3 Autocorrelation Function for Deseasolized Outflow time series

(with 5% significance limits for the autocorrelations)

1.0
0.8
0.6
0.4
Autocorrelation

0.2
0.0
-0.2
-0.4
-0.6
-0.8
-1.0

1 5 10 15 20 25 30 35 40 45 50 55 60 65
Lag

Figure 4 Partial Autocorrelation Function for Deseasolized Outflow series

(with 5% significance limits for the partial autocorrelations)

1.0
0.8
0.6
Partial Autocorrelation

0.4
0.2
0.0
-0.2
-0.4
-0.6
-0.8
-1.0

1 5 10 15 20 25 30 35 40 45 50 55 60 65
Lag

Figure 5 Time Series Plot of Observed and Simulated Outflows

1800000000 Variable
O bserv ed
1600000000 Simulated1
Simulated2
Simulated3
1400000000

1200000000
Outflow (Mm^3)

1000000000

800000000

600000000

400000000

200000000

Month Jan Jan Jan Jan Jan Jan

Year 1973 1979 1985 1991 1997 2003
1581 International Journal of Current Research, Vol. 5, Issue, 6, pp.1577-1581, June, 2013

Following the data quality and characteristics assessments an Jain, S. K.; Singh, V. P. 2003. Water resources systems
attempt was made to model the Outflow series. The naturalized planning and management. 1st. Amsterdam, Boston:
Outflow series as shown in Figure 2 has a clear seasonality. Elsevier.
Hipel et al. (1994) have recommended the use of Kebede, S.; Travi, Y.; Alemayehu, T.; Marc, V. 2005. Water
deseasonalized models for such kind of time series. The series Balance of Lake Tana and Its sensitivity to Fluctuations
is then deseasonalized by subtracting the seasonal mean from in Rainfall, Blue basinn, Ethiopia. In Journal of
each season and also dividing this by the seasonal standard Hydrology, 316, pp. 233-247.
deviation. Plots of the autocorrelation function and partial Loucks, Daniel P. 2000. Sustainable Water Resources
autocorrelation functions of deseasonalized series, shown in Management. In water International, 25 (1), pp. 3–10.
Figure 3 and 4, are used to identify the suitable candidate Ministry of Finance and Economoc Development (MoFED).
models. Finally, the adequacy of the selected model in 2006. A Plan for Accelerated and Sustained developmnet
describing the time series under consideration has been to End Poverty (PASDEP). Addis Ababa.
assessed and ensured by examining graphs of the residual Rientjes, T. H. M.; Perera, B. U. J.; Haile, A. T.; Reggiani, P.;
autocorrelation function, partial autocorrelation function and Muthuwatta, L. P. (2011): Regionalisation for lake level
normality plots. simulation – the case of Lake Tana in the Upper Blue
Nile, Ethiopia. In Hydrol. Earth Syst. Sci 15 (4), pp.
RESULT AND DISCUSSION 1167–1183.
Salas, J. D. 1980. Applied modeling of hydrologic time series.
The modelling process shows that the deseasonalized volume Littleton, Colo: Water Resources Publications.
metric Outflow series from Lake Tana catchment can be SMEC. 2008a. Hydrological Study of the Tana-Beles Sub-
adequately represent by first order Autoregressive model, Basins. Hydrological monitoring network review and
AR(1). Values of the model parameters, Ø1 and σa2, are recommendations. SMEC international Pty Ltd. Addis
estimated and found to be 0.7845 and 0.612 respectively. It is a Ababa.
good evidence and example for the scientific community in the SMEC. 2008b. Hydrological Study of the Tana-Beles Sub-
area of stochastic hydrology that the presence of storage like Basins. Surface Water Inverstigation. SMEC
Lake Tana has an effect on serial correlations of the outflow international Pty Ltd. Addis Ababa.
series. Synthetic seasonal series of the Outflow can be SMEC. 2008c.: Hydrological Study of the Tana-Beles Sub-
generated by invoking an inverse deseasonalization process. Basins. Ecological Studies with Emphasis on Biological
This can be performed by multiplying the generated Resources. SMEC international Pty Ltd. Addis Ababa.
deseasonalized series by each season standard deviation and Thomas, R.K.; Dan, R. 2004. Choice of reliability, resilience
adding the mean of each season. For this purpose the mean and and vulnerability estimators for risk assessments of water
standard deviation of each season shown in Table 1 can be resources systems. In Hydrological Sciences.
applied. Plots of observed and simulated outflow series are Vogel, R.M.;Bolognese, R.A. 1995. Storage-reliability-
shown in Figure 5. The characteristics and performance resilience- yield relations for over-year water supply
measuring indices of the Lake can then be assessed with systems. Water resources research, pp. 645-654.
current and what if scenario based approaches by applying Wale, A.; Rientjes, T. H. M.; Gieske, A. S. M.; Getachew, H.
certain level of reduction on its net supply capacity. A. 2009. Ungauged catchment contributions to Lake
Tana's water balance. In Hydrol. Process.
Wilcox, Rand R. 2009. Basic Statistics. Understanding
REFERENCES
Conventional Methods and Modern Insights. New York:
Oxford University press.
Box, George E. P.; Jenkins, Gwilym M.; Reinsel, Gregory C. Wurbs, R.A; Bergman, C.A .1990. Evaluation of Factors
1994. Time series analysis. Forecasting and control. 3rd. Affecting Reservoir Yield Estimates. In Journal of
Englewood Cliffs: Prentice Hall. Hydrology, 112, pp. 219–235.
Chebud, Yirgalem A.; Melesse, Assefa M. 2009. Modelling
lake stage and water balance of Lake Tana, Ethiopia. In
Hydrol. Process 23 (25), pp. 3534–3544.
Dahmen, E. R.; Hall, M. J. 1990. Screening of hydrological
data. Tests for stationarity and relative consistency.
Wageningen, the Netherlands: International Institute for
Land Reclamation and Improvement.

*******