Thursday, September 2, 2010

Hydro-climatic trends and water resource management implications based on multi-scale data for the Lake Victoria region, Kenya

Environ. Res. Lett. 5 (July-September 2010) 034005;   doi:10.1088/1748-9326/5/3/034005

Hydro-climatic trends and water resource management implications based on multi-scale data for the Lake Victoria region, Kenya

A. J. Koutsouris, G. Destouni, J. Jarsj√∂ and S. W. Lyon*

Bert Bolin Centre for Climatic Research, Department of Physical Geography and Quaternary Geology, Stockholm University, 106 91 Stockholm, Sweden

*Correspondence e-mail: steve.lyon@natgeo.su.se

Received 4 June 2010, accepted 23 July 2010, published 6 August 2010

Unreliable rainfall may be a main cause of poverty in rural areas, such as the Kisumu district by Lake Victoria in Kenya. Climate change may further increase the negative effects of rainfall uncertainty. These effects could be mitigated to some extent through improved and adaptive water resource management and planning, which relies on our interpretations and projections of the coupled hydro-climatic system behaviour and its development trends. In order to identify and quantify the main differences and consistencies among such hydro-climatic assessments, this study investigates trends and exemplifies their use for important water management decisions for the Lake Victoria drainage basin (LVDB), based on local scale data for the Orongo village in the Kisumu district, and regional scale data for the whole LVDB. Results show low correlation between locally and regionally observed hydro-climatic trends, and large differences, which in turn affects assessments of important water resource management parameters. However, both data scales converge in indicating that observed local and regional hydrological discharge trends are primarily driven by local and regional water use and land use changes.

Keywords:  Lake Victoria, Kenya, hydrology, water resource management, irrigation, climate change, hydro-climatic interaction

1. Introduction
Climatic changes are likely to threaten the Earth's already scarce water supply. Currently, about 40%, or 2.8 billion people, of the world population live in river basins with water scarcity. Improved water resource management is needed to help mitigate the potential influence of climatic changes and better utilize current water supplies. For example, about 1.6 billion people live in areas where the water scarcity has resource, developmental or economic reasons (UN 2008). Economic water scarcity occurs when a lack of financial, human or institutional capital causes incapacity to utilize better otherwise sufficient water resources. Lack of water produces negative effects on food security, health, gender equality and education making it both directly and indirectly connected with many of the United Nations' millennium development goals. As lack of financial capital also limits the capacity to import virtual water, which could otherwise compensate for physical water scarcity, water scarcity is thus primarily a crisis of the poor (UN 2006) that will be exacerbated under climatic changes.

For example, Kenya currently uses only 9% of its available water resources (e.g., water resources that would be exploitable if no financial constraints were present) while approximately 50% of the total population is below the national poverty line (UNDP 2008,Swallow et al. 2007). In addition, physical water scarcity often occurs seasonally in Kenya due to unevenly distributed rainfall throughout the year. This creates local water scarcity in many regions of Kenya due to a combination of economical and physical water scarcity. In the Kisumu district, located by Lake Victoria, 53% of the people live below the poverty line with an unreliable rainfall pattern identified as one of the main causes through its effects on food security (NCAPD 2005). Though the mechanisms behind the poverty levels in rural Kenya are more complex than a cause-and-effect relationship between rainfall and poverty, rainfall patterns may exacerbate existing poverty due to the simple fact that many in rural Kenya are dependent on rain-fed agriculture. In the Kisumu district, approximately 90% of the population is dependent on agriculture for both food and income, causing a large part of the population to be affected directly by droughts and floods. Taken together with the strong seasonality in rainfall, the water resources in Kenya and the Kisumu district are quite sensitive to climatic trends. This is exemplified in Orongo village (figure 1) located in the Kisumu district. It is a typical rural, lowland floodplain area in Kenya and as such it is sensitive to the effects of rainfall variability and water management (Swallow et al. 2007). This makes Orongo village a focal point for the efforts of international assistance agencies (e.g., Engineers Without Borders) whose goals are to provide reliable and sustained water resources to the local population.
Figure 1
Figure 1. Site map showing the location of Lake Victoria and the spatial extent of the Lake Victoria drainage basin in Africa. The location of Orongo village near Kisumu is also indicated.
Climatic change may increase the negative effects of rainfall uncertainty, and both current and future effects of this uncertainty could be mitigated to some extent through improved and adaptive water resource management and planning. Planning for improved and more secure water availability relies on our interpretations and projections of the coupled hydro-climatic system behaviour. Better scientific understanding of hydrological and climatic links, conditions and changes is thus a key issue for effective water resource management and its climate adaptation, in Orongo village as in other parts of the Lake Victoria region (Swallow et al 2008) and other regions of the world.
Our understanding of the coupled hydro-climatic system may be greatly hindered by limitations in data availability and quality. For instance, trends identified on a local scale may differ significantly from trends based on regional scale data (see, e.g., Pielke et al2002 considering temperature in eastern Colorado, USA). This study compares such trend results and exemplifies their water management implications on the basis of local scale data from the Orongo village and regional scale data from the whole Lake Victoria drainage basin (LVDB). The main aims of this analysis are to further investigate the prevalence of hydro-climatic trend differences on the basis of data with differing resolution and on different scales, and the propagation of such differences to important water management parameters, such as water storage requirements. In addition to such differences, this study also aims to investigate potential important consistencies in hydro-climatic system trend assessments, which are robust against the use of differently resolved and quantified data on different scales.

2. The study area and data set descriptions

2.1.  Regional scale: the Lake Victoria drainage basin
Lake Victoria is located in East Africa, southwest of Africa's horn (figure 1), from 31°39 'E to 34°53 'E longitude and 0°20 'N to 3°00 'S latitude. The lake is close to rectangular in shape with an area of around 67 000 km2. The shoreline of the lake is divided between Uganda, Kenya and Tanzania. The water surface level is typically 1140 m above sea level and the lake has a mean depth of 40 m with a maximum depth of around 80–90 m. The LVDB has a land area of about 194 000 km2 (Tate et al 2004) and is inhabited by one of the densest and poorest rural populations of the world (UNESCO 2006), where many are subsistence farmers depending on rain-fed agriculture (Anyah et al 2008).
The climate in LVDB can be classified as equatorial with hot and humid conditions where the main climate drivers are easterly monsoons and the bimodal passing of the inter-tropical convergence zone (Anyah and Semazzi 2007). Mean annual precipitation is 1780 mm and mean annual evapotranspiration is 1537 mm (Nicholson et al 2000). Rainfall occurs mainly during two periods: the long rains in March, April and May; and the short rains in September, October and November. Severe droughts occur approximately every 3–4 years during the short rains, every 7–8 years during the hot dry season (December, January, and February), and every 5–8 years during the long rainy season (Awange et al 2008). Lake Victoria is mainly rain-fed with direct precipitation accounting for approximately 80% of the water inflow to the lake (Sutcliffe and Petersen 2007), and about 10% coming from five main tributaries, with the Kagera River being the main contributor. The remaining 10% comes from various small tributaries. The only significant outlet is the White Nile (Song et al 2004), where the outlet is regulated near Jinja pass (0°25 '21 ' 'N, 33°11 '45 ' 'E). Since 1954, discharge and lake levels have been regulated by the Nalubaale dam (formerly known as the Owen Falls dam).
Data series of temperature and precipitation values at the regional scale of the LVDB were compiled from the spatially distributed CRU TS 2.1 Global Climate Data Set (Mitchell and Jones 2005). Temperature and precipitation data were available for monthly time steps from 1901 to 2002. The spatial extent of LVDB was delineated in ArcGIS 9.3® using the hydrology toolbox and the SRTM 90 digital terrain model (Jarvis et al 2008). Regional scale, basin averages of the annual time series of spatially distributed temperature and precipitation data were calculated for the entire LVDB. The temperature time series was then used to estimate annual actual evapotranspiration from the LVDB area (see the supplementary data available at stacks.iop.org/ERL/5/034005/mmedia). Annual discharge data for Lake Victoria at Jinja pass were obtained from the Global hydro-climatic data network data set (Dettinger and Diaz 2000).

2.2.  Local scale: Orongo village
Orongo village is located near the coast of Lake Victoria within Kenyan territory (figure 1). It is located east of the Winam Gulf and 6 km southeast of Kisumu, the third largest city in Kenya. Orongo village is characterized by pastures, homesteads and subsistence farming. The village has about 3000 inhabitants and an effective population density of 600 people per km2 (Levicki 2005). Two rivers flank Orongo village: Luanda River in the southeast and Nyamasaria River in the northwest. These rivers have their origin in the Nandi Escarpments which serve as the main recharge area for the region, having elevations up to approximately 1900 m. The lower parts of this region make up a part of the Kano Plains. Elevation in this lower section, a characteristically lowland floodplain with flat topography and minimal slope, ranges between 1140 and 1300 m. The principal soil types in the Kano Plains are histosols and vertisols (Onyango et al 2005). The land cover in this area is dominated by marshlands and subsistence agriculture (with maize and millet as the main crops).
Water Resources Management Affairs, Kenya (WRMA), have conducted stream flow observations approximately 30 km northeast from Orongo village. This neighbouring watershed, called the Little Oroba watershed, is the closest reliable stream gauge for Orongo village. The outlet of the 54 km2 Little Oroba watershed is located at 34°58 '15 ' 'E, 0°01 '40 ' 'N. Continuous stream flow data are available for daily intervals from 1932 to 1999. These observations were used to calculate a time series of the local scale average annual stream flow from 1932 to 1999. In addition, daily observations of pan evaporation and precipitation measurements are made near Kisumu by WRMA. Pan evaporation data were used to estimate actual evapotranspiration at this local scale (see the supplementary data available at stacks.iop.org/ERL/5/034005/mmedia). These time series were averaged to obtain average annual time series of evapotranspiration and precipitation over the periods of record. Note that the temperature data are collected by Kenya meteorological department at Kisumu, but that this data set was not available for consideration in this study.

3. Methods

3.1.  Hydro-climatic trend analysis
Simple linear regression was used to analyse the trends in the time series of hydro-climatic data collected at both the local scale and the regional scale. These data include observed precipitation and river discharge, and estimated actual evapotranspiration, which in turn depends on temperature, based on two different methods for the local scale and the regional scale assessments (see the supplementary data available at stacks.iop.org/ERL/5/034005/mmedia).
To allow for direct comparison between the different time series, linear regressions were applied to all the data and time series for the period 1968–1995. During this period, all hydro-climatic data have overlapping observation records at both the local scale and the regional scale. The present analysis thus facilitates a direct trend comparison between the local scale and the regional scale hydro-climatic observations and calculations.

3.2.  Use of trend analysis for water management
In order to exemplify the trend analysis use for concrete water management purposes, we estimated the minimum water storage requirement for an average farmer in the region using both regional scale and local scale data. Minimum storage requirement is defined here as the crop water required under standard climatic conditions. The method used to estimate minimum storage requirement was a sequent peak algorithm (Bouver 1978). The sequent peak algorithm is a graphical method based on the cumulative sum of precipitation surplus (PS) defined as inflow minus outflow and demand. Assuming that inflow is due primarily to direct precipitation, outflow is the water lost due to evaporation, and demand is the water transpired by crops (i.e., losses due to leakage and irrigation inefficiency are assumed relatively small and negligible), the cumulative sum of precipitation surplus PS can be estimated as
Equation (1)
where P is precipitation and ETa is the actual evapotranspiration (evaporation plus transpiration by crops) for each time step t over a record of observation that is n time steps in length. Note that as the demand and outflow may be larger than the inflow, P – ETa may be negative.
The sequent peak algorithm is applied to create a time series of cumulative PS. By plotting such a time series, the first peak and the following sequent peak that is higher than the first peak can be identified. The difference between the first-peak value and the minimum value before the sequent peak in time is the water storage requirement for that particular period. This procedure is carried out for the whole record of data at all peaks, and the largest difference found is then the minimum storage required to ensure sufficient water availability. While fairly basic, the sequent peak algorithm provides at least a first-order estimate of water storage requirements, which is compared here between the different scale data for the example of the average farmer of the Orongo village of the Kisumu district.

4. Results

4.1.  Hydro-climatic trends
Considering the precipitation data at the regional scale (figure 2(A)) and local scale (figure 2(B)), neither time series indicates any significant linear trend for the period 1968–1995. If the entire record of data available at both spatial scales is considered, this result holds and neither time series indicates any significant linear trends. Similar to precipitation records, estimated actual evapotranspiration at the regional scale (figure 2(C)) and that at the local scale (figure 2(D)) show no significant linear trend during the period 1968–1995. Again this lack of significant linear trend holds when considering the entire length of record at both scales.
Figure 2
Figure 2. Time series of annual regional scale precipitation (A), actual evapotranspiration (C), and discharge (E) data, and local scale precipitation (B), actual evapotranspiration (D), and discharge (F) data considered in this study. Trend lines shown are fitted for the period 1968–1995, over which all hydro-climatic data are available at both spatial scales.
With regard to regional discharge, however, there is a significant (p  <  0.05) negative linear trend over the period 1968–1995, following a period of increasing discharge at the outlet of Lake Victoria from 1959 to 1964 (figure 2E). In 1964 the discharge peaks and shifts to the significant negative trend. This shift excludes the application of a meaningful single linear regression analysis over the entire period of discharge observation. These results are consistent with previous studies of Lake Victoria's lake levels, showing a water level increase that peaks in the early 1960s (see e.g. Piper et al 1986Mistry & Conway 2003) and then a decrease through to 2005 (see e.g., Mangeni 2006Awange et al 2008).
Furthermore, while the regional scale discharge data show a significant negative linear trend from 1968 to 1995, the local scale discharge data (figure 2(F)) at the Little Oroba gauging station indicate a significant (p  <  0.05) positive linear trend in discharge during the 1968–1995 period. This trend is also seen when considering the entire length of record.


4.2.  Water storage requirements
The largest water deficit estimate using the sequent peak algorithm based on the regional scale data occurs during the period 1999–2001 (figure 3). From this deficit, the estimated minimum storage requirement for an average farmer in the region is 205 mm. Using local scale data within the sequent peak algorithm, the largest water deficit occurs during the period 1989–1994, with a minimum storage requirement of 592 mm. That is, an estimated water storage requirement based on local scale data nearly three times as large as that based on regional scale data.
Figure 3
Figure 3. Results from the sequent peak algorithm using an annual time step. The difference between the two sets of paired dashed lines shows the difference in estimated storage requirement based on regional scale and local scale data.
These minimum storage assessments can further be used to estimate the number of irrigation ponds needed to meet the storage requirements of the average farmer in the Orongo village. For example, Engineers Without Borders assumes that a typical irrigation pond in this region should have the dimensions of 15 m × 20 m × 2 m or about 600 m3 of storage (Levicki 2009). Using this design recommendation, the regional scale estimate of minimum storage translates into a requirement of about two irrigation ponds per acre of agricultural land. Using the local scale data, however, four irrigation ponds per acre of agricultural land are needed. There is, thus, a large influence of the choice of spatial hydro-climatic data resolution on the practical water resource management and planning for the example of Orongo village in the Kisumu district.

The large difference in estimated storage requirements may be partly due to the use of two different methods for estimating actual evapotranspiration at the local scale and the regional scale (see the supplementary data available atstacks.iop.org/ERL/5/034005/mmedia). The use of different methods, however, is not an independent choice, but a necessity due to differences in data resolution and availability at the two scales. It is often the case that available data dictate which methods can be used to determine unobserved hydro-climatic parameters such as evapotranspiration for use in estimates of water storage requirements.

5. Discussion and conclusion
There is clear disparity between discharge observations at the regional and local scales considered in this study, leading to different hydrological trend assessments based on the data from the different scales. The present results further exemplify and quantify how this disparity leads to large differences in the essential parameter of water storage requirement for provision of reliable and sustained water resources.
The inherent influence of data resolution and scale on the choices of quantification methods for water resource management assessments, as discussed above for the evapotranspiration quantification, is not often considered. This is probably because real-world water resource managers must often make decisions regardless of data availability limitations. Still, it is important to consider the effects of these limitations and associated implicit assumptions and generalizations, when managers are confronted with serious water management and climate-adaptation problems. These effects may be significant for resulting decisions and designs of water resource management options, particularly in hydro-climatically sensitive regions.
In the present results, however, we have also seen one important consistent aspect of regional scale and local scale discharge data implications: during the period of significant trends in inter-annual hydrological discharge (even though the trends are opposite on the different data scales), there are no significant trends evident in the climate variables precipitation and evapotranspiration, with the latter in turn depending on temperature. This aspect indicates that neither the regional nor the local trends in the inter-annual discharge changes are currently climate driven. Rather, the current observed discharge trends are most probably connected to local and regional water use and land use changes. This is consistent with the major land use changes, such as deforestation and agricultural expansion, and population growth observed in Lake Victoria drainage basin (Odada et al 2009Lung and Schaab 2010). This development is not homogeneous within the drainage basin and is mainly seen along rivers and in coastal areas (agricultural expansion and population growth) and in the tropical forest (deforestation). This supports the disconnection between climate trends and discharge trends due to local and regional water use and land use changes.

For example, the recent negative regional discharge trend may be an effect of water regulation and management practices at the Nalubaale dam. Direct quantification of the effects of dams and other local/regional water management practices are outside the scope of the present study. However, such effects have been investigated in detail and led to similar conclusions for other hydro-climatically sensitive parts of the world, such as the Central Asian region of the Aral Sea drainage basin (Shibuo et al 2007). In addition, locally driven changes in discharge, in the absence of observable changes in climatic variables, have also been observed in other hydrological catchments in eastern Africa, for instance in Ethiopia (Collick et al 2009).
While the current climatic trends show little direct influence on discharge trends (regardless of data resolution and scale), there is potential for future climate change to influence water availability in this region. Relevant identification of such large scale climate change effects must then be based on realistic assessments also of the water cycling effects of local/regional water use and land use, in hydrological catchments of scales that are consistent with the specific water management problems and decisions.

Link to rest of paper:  http://iopscience.iop.org/1748-9326/5/3/034005/fulltext

About Me

My photo
If you want to follow new posts on twitter, my user name is TenneyNaumer. Blog on climate science: Climate Change -- The Next Generation, at http://climatechangepsychology.blogspot.com

Followers