Introduction

Groundwater is hydraulically connected to surface waters in many regions of the world and an understanding of this interaction is fundamental to effective water-resource management (Winter et al. 1998; Sophocleous 2002; Brodie et al. 2007). Extensive areas of shallow, open surface water as well as permanent and seasonal wetlands occupy low relief (plateau) surfaces across the Great Lakes Region of Africa (GLRA). This low-relief landscape, known as the “African Surface” (Ollier 1993), has been shaped by alternating cycles of deep weathering and stripping since the Palaeozoic which have produced a regolith possessing a highly variable lithology and depth (McFarlane 1976; Taylor and Howard 1998). Rift-related, tectonic processes since the early Miocene are thought to have predominantly shaped the present landscape of the GLRA through the action of meteoric waters and strongly influenced surface hydrology (Doornkamp and Temple 1966; Doornkamp 1968; de Swardt and Trendall 1969; Taylor and Howard 1998; Trauth et al. 2005). There is, however, no consensus about how their physiographic properties interact with those of the surrounding interfluves (von der Heyden 2004) and thereby influence the interaction between groundwater and surface water.

Surface waters are particularly prominent in the Upper Nile Basin of Uganda where Lakes Victoria and Kyoga feature wetlands (Fig. 1) that occupy approximately 18% of the nation’s land area (NWP 2008). Regionally, these surface waters provide vital developmental (e.g. hydro-electric power generation, water supply), ecological and socio-economic (e.g. fishery) services (Chatterjee et al. 1998; Bugenyi 2001; Mwanja 2004). Water-balance studies in the Upper Nile Basin highlight the dominance of the direct input of rainfall in sustaining water levels in wetlands and surface waters (see electronic supplementary material ESM; Kite 1982, 1984; Shahin 1985; Piper et al. 1986; Sene and Plinston 1994; Yin and Nicholson 1998; Yates 1998; Nicholson et al. 2000; Sene 2000; Nicholson and Yin 2001; Sene et al. 2001; Tate et al. 2004). Climate variability and change are therefore expected to strongly influence the balance between precipitation and evapotranspiration and consequently the depth and areal extent of shallow lakes and wetlands in the Upper Nile Basin.

Fig. 1
figure 1

Map of regional drainage networks for lakes Victoria and Kyoga showing the location of study stations at Bugondo, Jinja and Entebbe and stable isotope sample sites (modified from Temple 1970)

Full size image

The role of groundwater in maintaining lake and wetland water levels during periods of low or absent rainfall on deeply weathered landsurfaces that characterise the African Surface is unknown. Indeed, with the exception of Swenson and Wahr (2009), water-balance studies of Lake Victoria (e.g. Kite 1982; Piper et al. 1986; Sene and Plinston 1994; Nicholson and Yin 2001; LVBC 2006) and Lake Kyoga (e.g. Sutcliffe and Parks 1999; Okonga 2000; ILM 2004) have either ignored the contribution of groundwater or assumed it was negligible. Despite substantial changes (up to 3 m) in the level of Lake Victoria during the early 1960s and more recently (2004–2006; Fig. 2), the interaction between groundwater and surface water in response to these extreme events remains unassessed and unclear. As a result, neither a conceptual nor a numerical representation of the interaction between surface water and groundwater exists in the region (Taylor and Tindimugaya 1996).

Fig. 2
figure 2

Daily measurement of water level of Lake Victoria at Jinja from 1948 to 2008. Variations of up to 3 m were found; the largest lake-level rise occurred from 1961 to 1964 and the largest lake-level decline occurred between 2002 and 2007

Full size image

In this study, the interaction between groundwater and surface waters on the highly weathered, low-relief “African surface” in the Upper Nile Basin has been investigated, using triangulated piezometers installed beside existing lake gauging stations. These in situ observations are supported by evidence from stable isotope ratios of O and H in groundwater and surface waters sampled in this and previous studies in the Upper Nile Basin. Using site evidence of the hydraulic gradient between groundwater and surface water together with the depth and hydraulic conductivity of the interface, Darcy throughflow calculations have been applied to approximate the contribution of groundwater to the water balance of Lake Victoria.

Regional geology and geomorphic evolution

The Lake Victoria and Lake Kyoga Basins are underlain by deeply weathered crystalline rocks of Precambrian age. The western half of the Lake Victoria Basin features Neo-Proterozoic rocks comprising conglomerates, sandstones, quartzites, greywackes, shales, dolomitic limestones and basalts and Palaeo-Proterozoic metasediments (Temple 1966; Schlüter 1997; Schlüter 2006). The northern area of the basin is underlain by the Archaean Granulitic-Gneiss Complex (Schlüter 2006) as well as quartzites, graphitic shales, banded iron formations, lava, tuff and amphibolites. Eastern sections of the basin host undifferentiated granites and Cenozoic sediments and volcanics, whereas the southern regions are mainly covered by granitoid rocks. The Lake Kyoga Basin forms part of largely unmapped regions that are considered to include the Archaean Granulitic-Gneiss Complex and Precambrian mudstones, sandstones, arkoses, conglomerates and silicified rocks (Schlüter 1997). Low-lying areas adjoining Lake Kyoga are covered by lateritised, fluvial-lacustrine sediments.

Warping and back-tilting of E–W flowing rivers (e.g. Kafu, Katonga, Kagera) produced a connected series of major lakes in central and southern Uganda that eventually led to the formation of lakes Victoria and Kyoga which dominate the low relief surface of the Upper Nile Basin (Temple 1970). In contrast to lakes formed by rifting in the GLRA (Lakes Albert, Tanganyika and Malawi), lakes Victoria and Kyoga are the products of these downwarping processes and are linked hydraulically via the Victoria Nile River. Lake Kyoga has both a lake zone (mean depths of 3.5–4.5 m; Temple 1970) and a flow-through conduit (mean depths of 7–9 m; ILM 2004) which channels the main upper Victoria Nile River waters. Natural (e.g. climate) and more recent human activities (e.g. dam operation) have influenced the levels of both lakes. Historically, variations in Lake Kyoga levels are primarily determined by the flow of the Victoria Nile from Lake Victoria (Sutcliffe and Parks 1999).

Study areas

Combined groundwater-level and lake-stage monitoring stations on the shores of Lakes Victoria and Kyoga (Fig. 1) were constructed in areas of contrasting land use and are remote from groundwater abstraction. The monitoring station at Bugondo is located along the southern shore of the main northeastern arm of Lake Kyoga (Owor et al. 2009a; Fig. 3a). At Jinja, the monitoring station is located on the northern shore of Lake Victoria close (~2 km) to its outlet (Fig. 3b). The monitoring station (DWRM) at Entebbe lies within a peninsula that runs approximately NE–SW on the northwestern shore of Lake Victoria (Fig. 3c). At each of Bugondo and Jinja monitoring stations, three piezometers were installed in 2007 in a triangular configuration near the lake to enable determination of direction of groundwater flow (Brassington 1998) adjacent to surface waters. Details of groundwater monitoring well construction parameters as well as hydraulic properties, derived from lithological logs using pedotransfer functions (Cronican and Gribb 2004) and hydraulic tests, are shown in Table 1.

Fig. 3
figure 3

Map showing locations of piezometers near the shores of a Lake Kyoga at Bugondo, b Lake Victoria at Jinja, and c Lake Victoria at Entebbe. The topographic contour interval is 15 m

Full size image
Table 1 Well construction information, depth to water, estimated hydraulic conductivity from textural data, and transmissivity from pumping tests for piezometers drilled in 2007 near Lake Victoria at Jinja (JP01, JP02, and JP03 at Jinja) and Lake Kyoga at Bugondo (BP01, BP02, BP03, and BP04), and for a long-term monitoring well near Lake Victoria at Entebbe (DWRM) (Owor 2010)
Full size table

Preliminary lithologic analyses and geoelectrical resistivity surveys at constructed groundwater/surface-water monitoring stations at Bugondo and Jinja, reported by Owor et al. (2009a), show that the interface between groundwater and surface water features a relatively consistent hydrostratigraphy in which fluvial-lacustrine sands are succeeded with depth by weathered rocks indicative of a sub-aerial in situ origin. Observations from all three stations (Bugondo, Jinja and Entebbe) favour a geomorphic model of landscape evolution by etchplanation (in situ deep weathering), as described by McFarlane (1992), over the pediplanation model (stripping by direct surface runoff) proposed by King (1962).

Hydroclimatic setting and observations

Annual rainfall in the Upper Nile Basin is bimodal featuring two distinct rainfall seasons from March to May (MAM) and September to November (SON); mean annual fluxes are 1,195 mm in Jinja (1903–2007), 1,541 mm in Entebbe (1954–2007) and 1,004 mm (1916–1956; 2001–2008) in Bugondo. The ‘nocturnal lake breeze effect’ enhances rainfall over Lake Victoria by about 30% compared to that over the surrounding land area of the catchment (Yin and Nicholson 1998; Nicholson et al. 2000). Most of the rainfall that falls on the surrounding land area of the lake derives from the south-easterly winds which pick up moisture from the central and western part of the lake that later condenses to form heavy rainfall on the northern and western shores of the lake. Annual potential evapotranspiration (PET) increases from about 1,500 mm on the shore of Lake Victoria to over 2,200 mm in northern Uganda (Hanna 1971). A recent estimate of 1,520 mm for Entebbe derives from monthly pan evaporation data (1987–2008) and an applied pan coefficient of 0.9 (Hanna 1971; Dagg 1972; Taylor et al. 2006).

Hydrogeological setting

Groundwater–lake monitoring sites are all underlain by similar deeply weathered crystalline rock (i.e. saprolite and saprock) but feature different land uses (Fig. 3). The Jinja site is located next to a railway pier; the Entebbe site lies within the premises of the Ministry of Water and Environment on a peninsula of Lake Victoria; and the Bugondo site is within a rural farming district, close to a fish-landing site on Lake Kyoga. At Bugondo, mixed fluvial-lacustrine sands (~1–5 m thick) rest on a lateritic duricrust (ferricrete) and lateritic clays comprising principally kaolinite (~2–4 m thick) whose K values were estimated at <0.02 m d–1 which, in turn, overlies a coarser groundwater-bearing saprolite comprising primarily less weathered, sand-sized clasts (Fig. 4a; Owor et al. 2009a). Indicative hydraulic conductivity (K) estimates at Bugondo vary with depth from fluvial-lacustrine sands (0.4–15 m d–1) to sandy saprolite (0.04–3 m d–1). At Jinja, mixed fluvial-lacustrine sands and infill (~2–7 m thick) overlie a thick sandy saprolite of variable thickness (from ~10 m depth; Fig. 4b). Ferricrete duricrust and lateritic clays are absent and presumed to have been eroded when the lake’s outlet at Jinja was carved out during the late Pleistocene. Indicative K estimates vary with depth from fluvial-lacustrine sands with infill (K of <0.02 m d–1) to sandy saprolite (K of <0.02–3 m d–1). At Entebbe, fluvial-lacustrine sands (~5 m thick), overlie saprolite comprising highly weathered (lateritic) clays (15 m thick) beneath which is the main groundwater-bearing coarse-grained saprolite facies (>17 m thick; Fig. 4c). Bulk transmissivity (T) estimates derived from pumping tests in fluvial-lacustrine and saprolite aquifers on Lake Victoria range from 1.7 to 18 m2 d–1 (Table 1). Pumping tests carried out in two wells (31.75°E, 0.93°S; 31.62°E, 0.47°S) installed within fluvial-lacustrine sediments on the western shore of Lake Victoria, yield bulk T values of 18 and 36 m2 d–1, respectively; these values are equivalent to bulk K estimates of between 0.09 and 1.8 m d–1 for an aquifer thickness of 20 m. The range in K estimates at all sites highlights the variability vertically and horizontally in groundwater pathways that constitute the interface of groundwater and surface water in the Upper Nile Basin. The hydrogeological conditions of the interface between groundwater and surface water were deduced from (1) borehole logs, (2) aquifer responses to changes in barometric pressure, and (3) hydraulic tests.

Fig. 4
figure 4

Cross sections of the lithological interface between groundwater and a Lake Kyoga at Bugondo, b Lake Victoria at Jinja, and c Lake Victoria at Entebbe. Black dots show spatial positions of triangulated piezometers whose static water levels relative to lake stage indicate a general hydraulic gradient towards the lakes

Full size image

Methods

Hydrological observations

Daily rainfall was recorded using standard—British standard 5-inch (12.7 cm) diameter—rain gauges from Entebbe, Jinja and Bugondo stations (Fig. 1). Hydraulic head (h) in each piezometer was used to define the hydraulic gradient (i) between groundwater and the lake levels. Both h and lake stage were measured relative to mean sea level datum. Groundwater levels and temperature in piezometers at the Jinja (except for one) and Bugondo sites were measured hourly from 17 November 2007 to 30 June 2009 using an absolute pressure transducer (Solinst Levelogger Gold) that had a pressure range of 100 m. Atmospheric pressure and temperature near each piezometer also were measured hourly over the same time span using absolute pressure transducers (Solinst Barologger Gold) that had a pressure range of 1.5 m. Daily water levels in piezometer JP02 were measured manually, whereas the other two piezometers (JP01 and JP03) were monitored using pressure transducers. The record at JP02 is consequently more subdued as sub-daily fluctuations are missed. Daily mean groundwater levels from 07 July, 1999 to 08 August, 2009 were determined for the Entebbe well (DWRM in Fig. 3c) using a standard float and chart recorder. Barometric pressure fluctuations were removed from the total pressure head measurements taken during the monitoring phase in order to evaluate hydraulic responses to recharge inputs (Davis and Rasmussen 1993; Rasmussen and Crawford 1997; Kruseman and de Ridder 2000; Spane 2002). Lake levels were recorded by DWRM (Directorate of Water Resources Management) on a daily basis during the course of the study period at the two stations (Entebbe and Jinja) on Lake Victoria and at the Bugondo station on Lake Kyoga. Daily total dam releases from the Nalubaale and later also the Kiira dams have been recorded by the Uganda Electricity Board at the Jinja outlet since 1954.

Stable isotope hydrochemistry

Stable isotope ratios of hydrogen (2H:1H) and oxygen (18O:16O) in groundwater and lake water can be used to identify sources of water as well as processes such as mixing and evaporation (Lloyd and Heathcote 1985). Where direct recharge is significant, δ2H and δ18O in the recharging waters approximate that of precipitation weighted by volume. In the Upper Nile Basin, seasonal variations in δ2H and δ18O of rainfall are observed and reflect the “amount effect”, wherein heavy rainfalls associated with the monsoons yield isotopic signatures that are relatively depleted in heavy isotopes (δ2H and δ18O) (Taylor and Howard 1999). Monsoonal rainfall also originates from the Indian Ocean, a source more depleted in δ2H and δ18O (e.g. δ18O = 0 ± 1‰), than those derived from regional surface waters such as Lake Victoria (e.g. δ18O = +4 ± 2‰) where stable isotope ratios are enriched in heavy isotopes through evaporation (Taylor and Howard 1999). Advantage is taken of these spatio-temporal variations in regional stable isotope ratios to investigate further interactions between groundwater and surface water.

Water samples (25 ml) for stable isotopic analyses were collected in capped PVC bottles and placed in coolers out of sunlight while in transit to laboratories. Stable isotope analyses were carried out at the Bloomsbury Environmental Isotope Facility in University College London (UCL) using a gas-source mass spectrometer (Thermo-Finnigan MAT253, Delta V Advantage and Delta Plus XP). Stable isotope ratios of H and O are reported as deviations (δ2H and δ18O) from a standard, VSMOW (Vienna Standard Mean Ocean Water; Lloyd and Heathcote 1985; Coplen 1993; Hoefs 1997). Uncertainty in the δ18O analyses was ±0.05‰ (1 sigma). Monthly variations in the stable isotopic ratios (δ2H and δ18O) for precipitation at Entebbe (1961–1974) were obtained from the International Atomic Energy Agency (IAEA) Isotope Hydrology Information System-Water Isotope System for Data Analysis, Visualization and Electronic Retrieval (ISOHIS-WISER) website and combined with more recent data (1998–2004) provided by DWRM, analysed by IAEA, Vienna. Additionally, previously published (Taylor and Howard 1999) stable isotope ratios from samples collected from Lake Victoria, groundwater (boreholes, springs, and shallow wells), and rivers flowing into Lake Victoria were included in the analysis of interaction between groundwater and surface water.

Results and discussion

Hydraulic gradients between groundwater and Lake Victoria at Entebbe (1999–2009)

Seasonal variability and a summary of statistics of lake stage and groundwater level monitoring data recorded at Entebbe (Fig. 3c) from January 1999 to July 2009 are listed in Table 2 and shown in Fig. 5. Fluctuations in groundwater levels coincide with those of Lake Victoria levels though the latter are subdued relative to the former. These variations result from the bimodal distribution in annual rainfall from MAM and SON rains (Basalirwa 1995) that produce seasonal recharge. The first rains commonly generate greater recharge. Mean h and lake stage over this period are 1,135.4 masl (metres above sea level) and 1,134.5 masl, respectively (Fig. 5a and b). Groundwater flows into the lake throughout period of monitoring though there is a sustained gap in reliable observations from May 2004 to August 2005 (Fig. 5a). Time-averaged i from 1998 to 2009 is 0.008 (Fig. 5c); temporal variations reflect seasonal changes in the recharge and discharge between monitored groundwater and Lake Victoria. The observed i never exceeds 0.035 which is less than the upper limit of 0.2 commonly measured within relatively permeable lakebed sediments (Hayashi and van der Kamp 2007).

Table 2 Summary statistics of hydraulic heads (h) in masl for selected piezometers near Lake Kyoga and Lake Victoria, lake levels (z) in the two lakes, and hydraulic gradients (i) between piezometers and Lake Kyoga or Lake Victoria over different periods in days between 1998 and 2009
Full size table
Fig. 5
figure 5

a Times series of daily mean lake level (Lake Victoria), groundwater level in DWRM monitoring well, and rainfall at Entebbe from January 1999 to July 2009; the green line represents the elevation of the boundary between fluvial-lacustrine sands and saprolite at monitoring well DWRM; b box plot showing summary statistics of Lake Victoria levels and groundwater levels in well DWRM; and c box plot showing summary statistics of lateral hydraulic gradient between well DWRM and Lake Victoria at Entebbe. Here, and in subsequent figures, bars represent 10 percentile, 25 percentile, median, 75 percentile, and 90 percentile with dots showing outliers

Full size image

From July 2003 to July 2004, the level of the lake fell by ~0.5 m due to anomalously low rainfall in early 2003 and excessive dam releases (Sutcliffe and Peterson 2007; Awange et al. 2007; Swenson and Wahr 2009). Groundwater levels also decline over this period but reliable records cease in May 2004. A brief stabilisation in the lake level in early 2005 results from a sharp reduction in total dam releases (Fig. 6a). The level of Lake Victoria then continues to decline until late 2006; groundwater levels also decline sharply from August 2005 until the MAM rains of 2006 when the hitherto seasonal pattern of recharge resumes. Since early 2007, the level of Lake Victoria has stabilised in response to a more regulated outflow regime (Fig. 6a). Lower dam releases at Jinja led, however, to a sharp decline in the level of Lake Kyoga (Fig. 6b).

Fig. 6
figure 6

Measurements of a 7-day moving mean of daily total dam releases together with agreed curve (rating curve between lake level and total outflow) releases at the outlet of Lake Victoria in Jinja, and b daily levels of Lakes Victoria at Jinja and Kyoga at Bugondo from January 1999 to July 2009

Full size image

Evident from groundwater-level observations in Fig. 5a is that the recession of the borehole hydrograph follows two distinct slopes. A rapid, short-lived response (slope of well hydrograph >0.01 m d–1) is apparent when the groundwater levels lie within the fluvial-lacustrine sands whereas slower response (slope of well hydrograph <0.01 m d–1) occurs when groundwater levels fall below sand horizons into the underlying saprolite comprising primarily kaolinite clays and quartz sand that has a lower K (<0.02 to 3 m d–1 on the Lake Victoria shore). These observations highlight the strong influence exerted by rainfall and lithology on the observed variability in groundwater levels. The time series also reveals the importance of dam releases on groundwater/surface-water interactions.

Hydraulic gradients between groundwater and Lake Kyoga at Bugondo (2007–2009)

Seasonal variability and a summary of statistics of groundwater level and lake stage monitoring at Bugondo from December 2007 to June 2009 are shown in Fig. 7. Groundwater level data (BP01, BP02 and BP04) between February and July 2008 are missing as a result of pressure-transducer malfunctions. Mean total hydraulic heads for piezometers BP01, BP02, BP04 and Lake Kyoga are 1032.9, 1033.4, 1033.4 and 1032.7 masl, respectively (Fig. 7b, Table 2). The time-averaged i from 2007 to 2009 between each piezometer and the lake ranges from 0.002 to 0.004 (Fig. 7c). Additionally, i into the lake from each piezometer is highest during periods of groundwater recharge that predominantly occur during seasonal (MAM, SON) rains (Fig. 7a). There are however brief periods of drought (e.g. late 2007) when the i reverses immediately adjacent to the lake (BP01). These declines in very shallow (< 1 mbgl) groundwater levels adjacent to the lake likely result from sustained evapotranspiration by lakeshore vegetation. Mean monthly potential evapotranspiration, derived from pan evaporation data (1971–1974) from Aduku (32.72°E, 1.98°N) just north of Lake Kyoga, ranges from 120 to 200 mm. Comparatively fast recessions observed in well hydrographs after flood events are also considered to be induced, in part, by evapotranspiration. The well hydrographs for BP02 and BP04 closely follow each other and form broad seasonal (up to 3 months) peaks superimposed on short-term (less than monthly) responses.

Fig. 7
figure 7

a Trends in daily mean Lake Kyoga level at Bugondo in relation to daily mean water level in wells and precipitation at Bugondo from November 2007 to July 2009, b box plot showing summary statistics of Lake Kyoga levels and groundwater levels in wells, and c box plot showing summary statistics of lateral hydraulic gradient between wells and Lake Kyoga at Bugondo. Positive hydraulic gradient means flow is from groundwater to lake and negative hydraulic gradient means flow is from lake to groundwater

Full size image

Asymmetry is observed in the seasonal response of water tables to bimodal precipitation at Bugondo and attributed to lower soil-moisture deficits that develop during the shorter dry season from June to July and greater intensity of the second (SON) rains in the Lake Kyoga catchment (Taylor and Howard 1996; Taylor and Howard 1999). Mean i between groundwater and lake levels during wet seasons range from 0.0034 to 0.0052, whereas in the dry season they vary from 0.0015 to 0.003. Heavy rains induce rapid (few days) and short-lived responses in groundwater levels that are most clearly evident from piezometer (BP01) closest to the lake. The more rapid and intense response observed at BP01, relative to BP02 and BP04, likely results from infiltration of precipitation through a comparatively thinner unsaturated zone at BP01 that has been observed elsewhere (e.g. Hunt et al. 2008). The response of lake levels is more subdued because of greater lake storage per unit change in head. For example, the MAM rainy season of 2008 arrested lake levels from a steady decline observed during the late 2007.

Groundwater levels follow sharp declines in the level of Lake Kyoga in late 2007, early 2008 and again in early 2009. Lake-level declines stem from a reduction in dam releases at Jinja described previously (Fig. 6a and b). During the main dry season of November 2007 to February 2008 (DJF 2007–2008), the i between piezometer BP01 and Lake Kyoga reverses. A year later (DJF 2008–2009), the level of Lake Kyoga is 30–50 cm lower, mainly in response to reduced Lake Victoria outflow over this period, and maintains groundwater flow into Lake Kyoga.

Hydraulic gradients between groundwater and Lake Victoria at Jinja (2007–2009)

Seasonal variability and a summary of statistics of groundwater level and lake stage monitoring at Jinja station from November 2007 are shown in Fig. 8. Daily water levels in piezometer JP02 were measured manually whereas the other two piezometers (JP01 and JP03) were monitored using pressure transducers. Because water levels at JP02 were measured once daily, the hydrograph is more subdued than the hydrographs from hourly measurements at JP01 and JP03. Mean total hydraulic heads for piezometers JP01, JP02, JP03 and Lake Victoria are 1,135.0, 1,136.1, 1,136.7 and 1,134.1 masl, respectively (Fig. 8b). Hydraulic gradients into the lake from all the three groundwater piezometers are more pronounced during wet seasons. Piezometer JP01 has a time-averaged i from 2007 to 2009 of 0.013, JP02 is 0.017, and JP03 is 0.008 (Fig. 8c). At both Bugondo and Jinja (Fig. 4a and b), the hydraulic gradient is highest between the middle-distance well and the lake.

Fig. 8
figure 8

a Trends in daily mean Lake Victoria level at Jinja in relation to daily mean water level in wells and precipitation at Jinja from December 2007 to July 2009, b box plot showing summary statistics of Lake Victoria levels and groundwater levels in wells, and c box plot showing summary statistics of lateral hydraulic gradient between wells and Lake Victoria at Jinja. Positive hydraulic gradient means flow is from groundwater to lake and negative hydraulic gradient means flow is from lake to groundwater

Full size image

At Jinja, greater recharge occurs during the first (MAM) rains which, on the northern shore of Lake Victoria, are typically greater in magnitude than the second (SON) rains. The drainage base (lake) level is stable between November 2007 and June 2009 except for short-term (less than monthly) variations in addition to two broad seasonal peaks. The record of piezometer JP01, adjacent to the lake, is strongly influenced by the lake-level variations. Similar to Bugondo, this hydrograph shows rapid, short-lived responses to heavy rainfall. For piezometer JP03, the static water level is much deeper (about 12 mbgl) and within saprolite. The lag between rainfall and observed recharge pulses is up to 4 days. This furthest piezometer from the lake shows a broader hydrograph rise and recession indicative of a longer resident time of water in the unsaturated zone (Hunt et al. 2008) and additional storage provided by the underlying, thicker saprolite. Comparable ranges in specific yield estimates for fluvial-lacustrine sands (<0.01–0.25) and saprolite (<0.01–0.18) are derived from textural analyses (Kasenow 2008).

Stable isotope hydrochemistry

Summary statistics of δ18O values from a compiled dataset of 599 groundwater samples in the Upper Nile Basin are plotted in Fig. 9 relative to surface water and groundwater samples analysed under this work. δ18O values in Bugondo groundwater (−2.5 ± 0.04‰) are consistent with groundwater isotopic signatures observed throughout the Ugandan Upper Nile Basin (−2.3 ± 0.02‰) which are significantly depleted in 18O relative to Lake Victoria (3.5 ± 0.03‰) and Lake Kyoga (4.8 ± 0.2‰), and derive primarily from direct rainfall-fed recharge during the monsoons (Taylor and Howard 1999; Owor et al. 2009b). However, groundwaters at Jinja (1.4 ± 0.3‰) provide evidence of the influence of surface waters enriched in 18O through evaporation.

Fig. 9
figure 9

Distribution in δ18O values of groundwaters and surface water (‰ VSMOW) sampled in the Upper Nile Basin

Full size image

A comparison of the δ2H and δ18O in rainfall, surface water (Lake Victoria) and groundwater on the northern shore of the lake is presented in Table 3 and Fig. 10. The effect of evaporation on stable isotope signatures in months of lighter rainfall (mean monthly rainfall of <138 mm) at Entebbe for 1961–1974 and 1998–2004 is expressed through a reduced slope of the local meteoric water line (LMWL) relative to the GMWL. At Entebbe, isotopic ratios in shallow groundwater plot close to the both the GMWL and LMWL, which suggest active recharge by local rainfall. The isotopic ratio in groundwater at the DWRM site has a slight but analytically significant difference from the volume-weighted mean rainfall. This difference likely derives from the dominance of heavy rains with isotopic ratios more depleted in the heavy isotopes in generating groundwater recharge (Taylor and Howard 1996; Mazor 2004). Stable isotope ratios in groundwater sampled along the northern shore of Lake Victoria (Fig. 1), not only lie on both the GMWL and Entebbe LMWL but at some sites also plot along a significantly reduced slope (i.e. evaporation line) characteristic of lake waters that are subject to evaporation (Fig. 10). These data which derive from boreholes (mean depth = 20 mbgl) provide evidence of mixing between lake water and groundwater.

Table 3 Summary statistics of δ18O and δ2H stable isotope data (‰ VSMOW) from Upper Victoria Nile Basin waters
Full size table
Fig. 10
figure 10

Plot of δ2H versus δ18O recorded in rainfall at Entebbe showing the local meteoric water line (LMWL) derived from historical (1969–1974) and recent (1998–2004) data for months of light (mean of <138 mm) and heavy (mean of >138 mm) rainfall. For reference the global meteoric water line (GMWL) is plotted along with a local evaporation line that links the Lake Victoria water with the groundwater fringing the lake

Full size image

At Bugondo on the shore of Lake Kyoga (Fig. 3a), groundwater and some wetland samples have low δ2H and δ18O values that regress along the Entebbe LMWL (Fig. 11). A second cluster composed of lake and wetland samples with high δ2H and δ18O values regress along an evaporation line with a slope of 5.2. Figure 12 provides a time series of δ18O observations of lake and wetland samples collected during dry and wet seasons. Groundwater samples are uniformly depleted in 18O (δ18O ranges from −3 to −2‰) and additionally, closely cluster together indicating their similar origin (monsoonal rainfall). δ18O values for lake waters consistently exceed 4‰. δ18O values from two wetland samples were 4.5‰ in November 2007 and January 2008, whereas a third sample was −2.5‰ in November 2008. δ18O values of >4‰ for the wetlands mirror lake water and are consistent with the groundwater heads at BP01, being less than the lake level on those dates. Additionally, the isotopically lighter δ18O value for the wetland in November 2008 is consistent with the groundwater head at BP01, being greater than lake level on that date. The wetland δ18O values indicate contributions from groundwater or lake water at different times that are consistent with lateral gradients between well BP01 and the lake.

Fig. 11
figure 11

Stable isotope ratios (δ2H, δ18O) in groundwaters and surface waters for Lake Kyoga, associated wetlands and groundwaters at Bugondo plotted alongside the GMWL and Entebbe LMWL. Note the local evaporation line that links lake water to groundwater fringing the lake

Full size image
Fig. 12
figure 12

Time series of δ18O for groundwaters and surface waters sampled from Bugondo on Lake Kyoga from November 2007 to December 2008

Full size image

At Jinja on the shore of Lake Victoria (Fig. 3b), there is a distinct and progressive enrichment in 18O in δ18O groundwater signatures (−0.02 to 2.8‰) from the furthest piezometer (JP03) to the lake on three dates between December 2007 and August 2008 (Fig. 13). Similar to Lake Kyoga at Bugondo (Fig. 12), Lake Victoria waters are enriched (3–4‰) in δ18O. Isotopic data show that groundwater evolves along the flow path by mixing with lake waters at Jinja. Mixing is supported by trends in electrical conductivity (EC) data which decrease progressively along the flow path from 3,230 μS cm–1 at JP03 to 1,920 μS cm–1 at JP02 and then to 1,540 μS cm–1 at JP01. In addition, samples from JP01 taken during a hydraulic test show slight enrichment in δ18O (from +2.71 to +2.78‰) after more than 2 h of pumping. This enrichment of 0.07 slightly exceeds the measurement error of ±0.05 (1 sigma), and is thought to reflect contributions from lake water, a source enriched in 18O.

Fig. 13
figure 13

Time series of δ18O for groundwaters and surface waters sampled from Jinja on Lake Victoria on three dates between December 2007 and August 2008

Full size image

Influence of base-level changes on groundwater-lake interactions

Recent dam operations have affected the levels of Lakes Victoria and Kyoga, which in turn, influence i between groundwater and the lakes. To provide pre-dam outflow to catchments downstream of the lake, a relation known as the “agreed curve” was established between the lake level and the total outflow from Nalubaale Dam at the Jinja outlet in 1954. Sene and Plintson (1994) show that dam releases have largely adhered to the agreed curve with compensatory discharges made in order to minimise the effects of departures from expected outflow. This dam release policy was closely followed until about 2002 when a second dam (Kiira) was constructed parallel to the Nalubaale Dam at the lake outlet in Jinja. Agreed curve data are available from January 2004 to September 2009 (Fig. 14). The combined outflows of the Nalubaale and Kiira dams (both regulate outflow of Lake Victoria at Jinja) facilitate exceedance of this dam release policy and are able to induce lake-level reductions (Sutcliffe and Peterson 2007; Awange et al. 2007; Swenson and Wahr 2009).

Fig. 14
figure 14

Daily measurements of the total dam releases (blue line) and the agreed curve (red line) at the outlet of Lake Victoria in Jinja from November 2007 to July 2009

Full size image

From 2004 to the end of 2005, total dam releases rose from 140 to ~240% of the agreed curve (Fig. 15). The level of Lake Victoria fell from May 2004 to February 2006 by ~1.2 m (Fig. 6b). During 2006, total dam releases then reduced to 100% of the agreed curve. Since 2007, dam releases have varied within a mean of 100–140% of the agreed curve (Fig. 14). This reduction in dam releases (back to the agreed curve) is followed from late 2006 by a rise (~0.5 m) in lake level which then stabilises from mid-2007 (Fig. 6b). Dam releases influence lake levels but compensatory releases do not always induce immediate or concomitant responses in lake levels. This fact affirms the hypothesis that other influences such as climate (Sutcliffe and Petersen 2007) also control lake levels. During periods when dam releases are lower, i (between groundwater and the lake) at Jinja reduces as lake levels rise (Fig. 8a). Reduced outflow from the Lake Victoria outlet at Jinja tends to lower the level of Lake Kyoga which amplifies i at Bugondo (Fig. 7a). In contrast, high dam releases at Jinja raise the level of Lake Kyoga that then reduces i.

Fig. 15
figure 15

A general hydrodynamic conceptual model showing the lithological domains and direction of fluxes within the lithologic interface with Lakes Victoria (at Jinja and Entebbe) and Kyoga (at Bugondo)

Full size image

Conceptual models of the hydrodynamic interactions between groundwater and surface waters in the Upper Nile Basin

New observations at Entebbe, Jinja and Bugondo show that fluvial-lacustrine sands are succeeded with depth by weathered rocks indicative of a sub-aerial in situ origin (Fig. 15). Fluvial-lacustrine sands (K = 0.02–15 m d–1) rest on saprolite consisting primarily of highly weathered lateritic clays and occasionally a ferricrete duricrust which, in turn, overlies a thick, coarser saprolite comprising less weathered sands (K = 0.02–3 m d–1). During the wet season, precipitation infiltrates the interfluve profile and provides recharge to the transmissive facies of fluvial-lacustrine sands and sandy saprolite; recharge to the latter necessarily occurs by way of discontinuities (stonelines, factures) in the clayey saprolite and duricrust (where present). The precise pathways remain unclear but the routine observation of seasonal groundwater-level fluctuations within the coarser sandy saprolite in response to heavy rainfall (Taylor and Howard 1996, 1999; Owor et al. 2009a) attests to the presence of sub-vertical pathways enabling direct (rain-fed) recharge. In the dry season, hydraulic data show that groundwater levels are low though i remains predominantly influent to lakes. Evidence from the combined groundwater-lake stage monitoring stations indicate a dual system wherein groundwater discharges rapidly and episodically to the lake via fluvial-lacustrine sands following monsoonal recharge events during the wet season but flows more slowly and primarily from saprolite during the dry season.

Groundwater flow into Lake Victoria: an in situ approximation

A first-order approximation of the flow of groundwater to Lake Victoria is estimated from the application of Darcy’s Law for fluid movement in a porous medium. Groundwater flow to Lake Victoria is from the product of the estimated interface transmissivity (T), perimeter of the lake (P), and observed i between the lake and adjacent groundwater. As discussed previously, in situ T values for the interface derived from pumping tests range from 1.7 to 36 m2 d–1. Published estimates of the perimeter of Lake Victoria vary from 3,116 to 3,647 km (Whitehouse and Hunter 1955; Crul 1995). The time-averaged i between the lake and adjacent groundwater observed at Entebbe and Jinja from 1998 to 2009 ranges from 0.008 to 0.017 m m–1. Estimates of the direct Darcy flux to Lake Victoria based on the lowest (i.e. i = 0.008, T = 1.7 m2 d–1, P = 3,116 km) and highest set of values for Darcy Law (i.e. i = 0.017, T = 36 m2 d–1, P = 3,647 km), range from 0.02 to 0.8 km3 year–1.

Owor (2010) consulted 16 studies of the mean long-term water balance for Lake Victoria carried out between 1900 and 2004 (see electronic supplementary material ESM) and found that rainfall is estimated to contribute 84% of the total lake inflows (113 km3a–1) and riverflow which includes contribution from groundwater (as baseflow), is estimated to be 21 km3 year–1 (16% of the total inflows). The Darcy flux of groundwater directly contributing to the water balance of Lake Victoria (0.02–0.8 km3a–1) is estimated to comprise between 0.01 and 0.6% of the total (rainfall and river) inflows to the lake. The estimate from this study of direct groundwater inflow is two orders of magnitude lower than the estimate of direct and indirect (via baseflow to riverine inflow) contributions of groundwater (~29 km3 year–1, ~21% of total lake inflows) to the lake’s water balance, recently estimated over the period 2003 to 2007 by Swenson and Wahr (2009). Their estimate, derived from data generated under the Gravity Recovery and Climate Experiment (Tapley et al. 2004), is, however, neither reconciled to in situ hydrogeological observations nor the hydrogeological conditions around the lake.

The first-order estimation of the direct contribution of groundwater to the water balance of Lake Victoria does not consider the possibility of regional groundwater flow below the effective base of the bedrock aquifer in Uganda, estimated by Howard and Karundu (1992) and Taylor and Howard (2000) to reside approximately 80 m below ground level. There is currently no evidence of substantial groundwater flow below this depth in deeply weathered crystalline rock aquifers in the Lake Victoria Basin. If deep upwelling from saprock does not occur, groundwater flow represents a tiny (<1%) contribution to the water balance of Lake Victoria. The key implication of the estimate (from this study) of low total groundwater inflow to Lake Victoria and, by extension, Lake Kyoga is that baseflow is insufficient to sustain lake levels during periods of low or absent rainfall.

Conclusions

In situ observations of the interactions between groundwater and surface water on surfaces of low relief in the Upper Nile Basin are reported for the first time. Measured i between triangulated piezometers and lake stage boards indicate groundwater discharges to Lakes Victoria and Kyoga throughout most of the period of observation (1999–2009). Hydraulic gradients are highest during monsoonal rainfall when direct recharge elevates groundwater levels. Stable isotope ratios (2H:1H, 18O:16O) support piezometric evidence of monsoonal recharge and the discharge of groundwater to lakes Victoria and Kyoga; mixing of lake water (Lake Victoria) and groundwater is also suggested to have occurred at Jinja. Layered-aquifer heterogeneity in the geological interface between groundwater and surface water, wherein fluvial-lacustrine sands overlie saprolite, gives rise to both rapid and slow groundwater flow components in the recession of borehole hydrographs following recharge events. During the wet season shallow groundwater discharges rapidly to surface water via fluvial-lacustrine sands overlying saprolite, whereas in the dry season groundwater is primarily transmitted by the underlying saprolite.

Groundwater fluxes into the lakes are strongly influenced by dynamic drainage base (lake) levels that are controlled by regional climate variability and dam operations at Jinja. On Lake Victoria, consistently high dam releases at Jinja decrease the levels of the lake which, in turn, raise i between groundwater and lake. In contrast, low dam releases decrease i as the lake level rises. For Lake Kyoga which receives discharges from the dams at Jinja, the reverse is true. High dam releases enhance lake levels which, in turn, lower i between groundwater and lake at Lake Kyoga, whereas low releases increase i as the lake level declines. Based on the in situ observations, the direct contribution of groundwater to the water balance of Lake Victoria is estimated to be <1% of the total inflows to the lake. The evidence is, however, limited to three locations and more observations are required to develop a fuller understanding of the connectivity of groundwater and surface water regionally. Most critically, expansion of current piezometric networks would enable better representations of the spatial heterogeneity in the groundwater/surface-water interface around each lake.