Internat. Assoc. Great Lakes Res., 2006
The Water Balance and Stable Isotope Hydrology of
Lake Edward, Uganda-Congo
James M. Russell1,*and Thomas C. Johnson2 1Department of Geological Sciences
Brown University Box 1846
Providence, Rhode Island 02912
2Large Lakes Observatory
University of Minnesota Duluth 10 University Drive, RLB Duluth, Minnesota 55812
ABSTRACT. Lake Edward, Uganda-Congo, is one of the least studied of the great lakes of East Africa, and little is known of its physical hydrology. Stable isotope data and modeling and previously published estimates of Lake Edward’s water balance are used to constrain the physical hydrology of the lake, and particularly the relative proportion of surface outflow to evaporative water losses. Stable isotope calcula-tions suggest that Lake Edward loses roughly 50% of its water income by evaporation, while reviews of published hydrologic data together with our calculations suggest that evaporation comprises 54% of water losses. The similarity of these two sets of calculations lends credence to their validity, and provides a new water budget for the lake. Our results have important implications for the chemistry and hydrocli-matic sensitivity of Lake Edward.
INDEX WORDS: Lake Edward, East Africa, rift lake, stable isotope, hydrology.
INTRODUCTION
The Great Lakes of East Africa are rich sources of information about past variations of the African monsoons. The potential for these lakes to record past variations in monsoon intensity is partly due to their hydrologic sensitivity that is driven by hydro-logic budgets for the lakes in which water losses are dominated by evaporation (Spigel and Coulter 1996). Lake Edward, located on the equator at the border between Uganda and the Democratic Repub-lic of the Congo, has received perhaps the least at-tention of the East African Great Lakes, despite paleoclimatic studies that have revealed a rich and varied paleoclimatic history for the lake (e.g., Rus-sell et al. 2003, Laerdal et al. 2002). A thorough understanding of the modern hydrology of Lake Ed-ward is critical to interpreting paleoclimate data using Lake Edward’s sedimentary record; however, estimates of Lake Edward’s hydrologic budget are few and often contradictory (Lehman 2002).
*Corresponding author. E-mail: James_Russell@brown.edu
77
This paper seeks to refine calculations of the modern hydrologic balance of Lake Edward using past measurements and stable isotope data. First, we summarize previous estimates and measure-ments of Lake Edward’s hydrology. We then pre-sent new stable isotopic data for the lake and watershed that help to constrain the water balance of the lake. Our data and literature review suggest that evaporation and outflow each account for roughly 50% of water loss from Lake Edward, esti-mates that are similar to some previous studies and help us to better understand the lake’s physical and chemical structure.
Background:
Regional Geology and Hydrology
Lake Edward (0°–0°40′S, 29° 20′–29° 50′E, 912
m a.s.l.) is situated in a Cenozoic half-graben in the Western Arm of the East African Rift Valley (Fig. 1). The lake is presently open, draining northward to Lake Albert via the Semliki River. Lake Edward
is bounded by the Lubero border fault to the west and the Kigezi highlands to the east, the Ruwenzori mountains to the north, and the Virunga volcanoes to the south (Fig. 2). These four regions, together with the Lake George catchment to the northeast, comprise five major catchment areas that provide runoff to Lake Edward.
Lake George is drained by the Kazinga Channel, which flows sluggishly for 60 km to Lake Edward. The Ruwenzori Mountains to the north of Lake Ed-ward rise from the rift floor to heights of over 5,000 m and are currently glaciated. Principal inflows from the Ruwenzoris to Lake Edward are from the Nyamugasani and Lubilia rivers (Fig. 2), while con-siderable additional inflow from the Ruwenzoris is delivered to Lake Edward via Lake George. Moun-tains to the west of Lake Edward along the Lubero border fault rise steeply from the lake to heights of 2,500 to 3,000 meters within 15 km of the lake shore, and are drained by numerous short, steep rivers. The Kigezi highlands to the east rise more gently to form a low divide between Lakes Edward and Victoria. Principal inflows from the Kigezi highlands to Lake Edward are from the Ishasha, Ntungwe, Nchwera, and Nyamweru rivers. The Virunga volcanoes to the south divide Lakes Ed-ward and Kivu and are very important to the hy-drology and chemistry of Lake Edward (Kilham and Hecky 1973, Lehman 2002). Principal inflows
to Lake Edward from the Virunga region are the Lula, Rwindi, and Ruchuru rivers.
Lake Edward: Bathymetry, Morphology, Limnology, and Climatology
Lake Edward has a surface area of 2,325 km2and
maximum depth of 117 meters, located within 5 km of the western shore (Fig. 3; Lehman 2002). Lake Edward has an oxycline commonly found at about 40 m depth and is thought to be oligomictic (Beadle 1981, Hecky and Degens 1973). Although strati-fied, the temperature difference between surface and deep waters is only about 1°C, with an average annual surface temperature of about 26.5°C (Ver-beke 1957, Beadle 1966). Chemically, Lake Edward is a Na-Mg-HCO3system with a salinity of
approx-imately 0.8 g/L and a pH averaging 8.9 (Talling and Talling 1965).
Rainfall in the Lake Edward region falls in two rainy seasons coinciding with the passing of the In-tertropical Convergence Zone, from October to De-cember and March to May (Viner and Smith 1973, Nicholson 1996). Rainfall in the region, and throughout East Africa, varies strongly with altitude (Viner and Smith 1973, Nicholson 1996). Thus, the highlands surrounding Lake Edward receive consid-erably more rainfall than the lake itself, which ex-periences a more arid climate than much of the region (Hurst 1927, Viner and Smith 1973).
FIG. 1. Map of the East African Great Lakes region. Dashed lines in the left-hand figure indicate the position of the eastern and western arms of the rift. Shaded gray regions indicate major water bodies. A close-up at right of the equatorial lakes region shows political bound-aries (dashed lines) bisecting Lake Edward.
Previous Work and Data Sources
The first estimates of the hydrologic budget of Lake Edward were made by Hurst (1925, 1927) as part of a survey of the Nile River headwaters. Hurst’s work contains single-sample river gauge data and runoff estimates for several rivers in Lake Edward’s catchment as well as Lake Edward’s out-flow. Viner and Smith (1973) provided a hydrologic
budget for Lake George based upon 5 years of de-tailed hydrologic and climatic monitoring. Their data include daily to monthly river gauge data, the only such data available for the Edward basin. Data from these authors, supplemented by other esti-mates of Lake Edward’s hydrologic and limnologic characteristics (Worthington 1932, Damas 1937, Verbeke 1957, Hydromet 1982) form the basis for
FIG. 2. Map of the Lake Edward region showing catchment areas, major rivers, high elevation areas, swamps, and water sampling sites.
our study. Lehman (2002) combined hydroclimatic data with his own hydrologic calculations and an energy balance model into the first physical, hydro-logic, and chemical model of Lake Edward. Lehman’s calculations suggest Lake Edward’s out-flow exceeds annual water losses by evaporation by a factor of nearly five, an estimate that differs con-siderably from previous researchers.
Bathymetric and morphometric measurements for Lake Edward were calculated by Laerdal (2000) and Lehman (2002) (Table 1). Catchment areas for Lakes Edward and George were measured from De-fense Mapping Association maps L-4, L-5, and M-5 (Fig. 2). Our estimate for the catchment area of Lakes Edward and George differs slightly from pre-vious studies (Lehman 2002). Based upon the DMA maps, it appears that Lehman (2002) underesti-mated the catchments of the Ntungwe and Rusangwe rivers by about 90% and 116%, respec-tively. We are uncertain as to why these discrepan-cies exist, but we note that our revised catchment for the Rusangwe River matches that of Viner and Smith (1973), who explored the area extensively. Our revised catchment area increases the proportion of low-elevation areas to the east that drain into Lakes Edward and George, which could affect sur-face runoff into the lake.
The isotopic composition of Lake Edward, in-flowing rivers and springs, and occasional rainfall samples were sampled and analyzed between 1996 and 2003. 20-mL samples from rivers were taken at
road crossings within 15 km of Lake Edward’s shore, and lake waters were sampled from open water at least 5 km from shore. Water samples were collected and stored in high-density polyethylene vials prior to analysis. Analyses were conducted on a Finnegan Delta S mass spectrometer at the Uni-versity of Arizona; results are expressed in delta notation with respect to the SMOW standard. Ana-lytical error was 0.1 ‰ for δ18O and 1.0 ‰ for δD.
The Hydrology and Water Balance of Lake Edward
The fundamental equation for the hydrology of Lake Edward assumes the lake is in a steady state with respect to its volume:
Evaporation + Outflow =
Direct precipitation + Catchment inputs (1)
Previous estimates of the magnitudes of each of these terms will be discussed below.
Direct Precipitation
Hulme (1998) estimates precipitation in the Lake Edward region at 1.217 m/yr, similar to the esti-mates of Lehman (2002) of 1.214 m/yr, as well as the 1.1 m/yr estimated by Hurst (1927). However, the highland regions surrounding Lake Edward re-ceive far more rainfall than the lowlands and the lake itself (Viner and Smith 1973). The estimates above are based upon weighted averages of rainfall stations in all of southwestern Uganda, including
FIG. 3. Bathymetric Map of Lake Edward, with depth contours in meters. The position of crater lakes within the basin are also shown.
TABLE 1. Morphometric and Catchment infor-mation for Lakes Edward and George, Uganda-Congo.
Lake Edward
Surface Elevation 912 m a.s.l.
Surface area 2,325 km2
Volume 767 ×108m3
Max Depth 117 m
Catchment Area (less lake) 15,840 km2 Ruwenzori Catchment Area 1,231 km2 Western Escarpment Area 1,136 km2
Eastern Rivers 5,680 km2 Southern Rivers 7,793 km2 (including Ishasha) Lake George Surface area 250 km2 Catchment 9,976 km2
several stations in the highlands surrounding Lake Edward, and therefore probably overestimate direct precipitation to the lake’s surface (Nicholson 1996). Viner and Smith estimate direct precipitation onto the surface of Lake George averages 0.82 m/yr. Rainfall stations nearest to the surface elevation of Lake Edward, Kasese and Kabale, receive 0.87 and 0.99 m/yr, respectively (National Climate Data Center archive). We have averaged these three val-ues and estimate that direct precipitation to Lake Edward is 0.9 m/yr.
Catchment Inputs
Catchment inputs include river inputs, surface runoff, and groundwater inputs. Groundwater, al-though it may be important to the chemical balance of Lake Edward, is assumed to be negligible in the hydrologic budget (Lehman 2002). Catchment in-puts comprise the largest source of water to Lake Edward (Lehman 2002), yet they are by far the most difficult to estimate due to a nearly complete lack of river gauge data from the Lake Edward basin.
The available catchment-normalized surface runoff data from Lake George demonstrate the het-erogeneity of the region’s hydrology (Table 2; Viner and Smith 1973). All of the rivers draining into Lake George measured by Viner and Smith (1973), except the Mpanga and Kyambura, drain the Ruwenzori Mountains and have very high sur-face runoff rates, ranging from 0.514 to 1.54 m/yr. However, when the less steep, low-elevation
east-ern part of Lake George’s drainage basin is taken into account, the average runoff for the entire George basin is only about 0.2 m/yr. This is likely due to the steeper elevation gradients of the Ruwen-zoris, which yield higher runoff, as well as higher average annual rainfall at higher elevations within the lake’s catchments.
The only river that flows into Lake Edward that has annual gauge data is the Nyamugasani River, which drains the Ruwenzori Mountains (Viner and Smith 1973). Lehman (2002) applied the runoff de-rived from the Nyamugasani catchment, 0.514 m/yr, to the entire Lake Edward basin, and calcu-lated inputs to the lake totaling 8.85 × 109 m3/yr.
Based on the example of Lake George it seems likely that this is an overestimate, given that the slope, climate, and bedrock geology of the Ruwen-zori mountains is prone to high runoff as compared to the Lake Edward catchment as a whole. In point of fact, the slope of the Nyamugasani River is about 6% over the river’s catchment, while the average slope of the rivers draining into Lake Edward from the east is only 1.5%. The average slope of rivers draining into Lake Edward from the south is 3%, while rivers draining from the west have slopes equal to, or higher than, the Nyamugasani River. If we assume that rivers draining from the Ruwen-zoris and the western mountains into Lake Edward have surface-area normalized runoff yields equal to the Nyamugasani River, that rivers draining the eastern slope provide runoff equal to that of the Mpanga River, and that the southern rivers provide runoff intermediate between these two areas, we
TABLE 2. River inputs to Lakes Edward and George from Hurst (1927), Viner and Smith (1973), and Lehman (2002).
River Flow Catchment Runoff Annual Input
River (m3/sec) (km2) (m/yr) (109 m3/yr)
Ruchuru, dry season 40.000
Ishasha, dry season 8.000
Ntungwe, dry season 7.000
Nyamugasani 8.330 507 0.514 0.260
Sebwe (George catchment) 2.040 83 0.777 0.060
Rukoki/Kamulikwezi (George) 4.100 183 0.707 0.129
Mubuku (George) 12.500 256 1.540 0.394
Ruimi (George) 6.000 266 0.711 0.660
Mpanga (George) 11.500 4,670 0.080 0.374
Kyambura (Kazinga Channel inflow) 9.500 660 0.450 0.297
George basin (Viner and Smith, 1973) 61.800 9,976 0.196 1.948
Edward basin, Lehman (2002) 280.000 15,840 0.514 8.850
calculate an average runoff for the Edward catch-ment of 0.25 m/yr, very similar to the value of 0.28 suggested by Hurst (1927). Hurst’s value is inter-mediate between that of Lehman (2002) for Lake Edward and Viner and Smith (1973) for Lake George, and seems reasonable given that the Lake Edward catchment contains a slightly higher pro-portion of steeply sloping terrain than the Lake George catchment. Therefore, we assign a runoff value equivalent of Hurst’s estimate of 0.28 m/yr, or 4.435 × 109 m3/yr, to catchment inputs to Lake
Edward excluding inputs from Lake George. In addition to general catchment inputs, the Kazinga Channel delivers 1.70 ×109m3/yr to Lake
Edward (Viner and Smith 1973), a value deter-mined at its exit from Lake George both by hydro-logic modeling and gauge data. This represents the combined inputs of rivers and precipitation to Lake George, less evaporation from Lake George’s sur-face (Viner and Smith 1973).
Outputs
Surface Evaporation
Published estimates for evaporation from Lake Edward vary widely (Table 3). The most common methods of estimating evaporation from a lake sur-face are energy balance and Penman’s (1948) method. The latter combines a formula for potential evapotranspiration with energy balance and water-mass transfer. Both methods require numerous input variables, including air vapor pressure, lake temperature, cloudiness, and surface radiation.
Input data for evaporation calculations includes surface pressure, dew point, cloud fraction, and wind-speed data from the Kasese weather station (Table 4), which lies between Lakes Edward and George. Lake water temperature is derived from mean monthly measurements reported in Verbeke (1957), which are slightly cooler than more recent values reported from Lehman (2002). Insolation
TABLE 4. Meteorological input data used in evaporation calculations. Top of
Atmosphere Surface Surface Dew Temp
Insolation pressure Air Temp Point Cloud Lake Wind-speed
Month W/m2 mb °C °C Fraction °C m/s Jan 416.2 903.6 23.36 19.01 0.413 25.9 2.41 Feb 431.8 903.5 23.58 17.81 0.305 26.0 2.14 Mar 438.2 903.4 23.63 19.02 0.481 26.1 2.48 Apr 427.1 904.7 23.68 19.73 0.333 26.5 2.14 May 406.2 905.7 23.57 19.66 0.257 27.1 2.00 Jun 392.5 904.9 23.24 19.00 0.292 27.2 1.65 Jul 397.0 905.6 22.81 18.15 0.318 25.8 1.66 Aug 414.7 905.1 22.83 17.44 0.494 25.3 2.22 Sep 429.8 905.1 22.73 18.72 0.353 25.8 2.68 Oct 430.3 904.3 22.88 19.13 0.370 26.8 2.67 Nov 418.2 903.8 23.01 19.40 0.517 27.2 2.33 Dec 409.0 904.3 23.26 19.15 0.420 26.5 2.33
TABLE 3. Evaporation Estimates for Lakes Edward and George, from Hurst (1927), Viner and Smith (1973), Lehman (2002), and Penman and energy balance calculations of this study.
Annual Water Loss,
Author Method Rate, m/yr km3/yr
Lehman (2002) Mass Transfer 1.16 2.59
Hurst (1927) comparison to Lake Victoria 1.20 2.79
Viner and Smith (1973) Penman 1.83 4.24
This Study Energy Balance 1.98 4.60
and surface air temperature were obtained from the National Center for Environmental Prediction (NCEP) Electronic Reanalysis Atlas. Kasese data for windspeed, surface pressure, and average air temperature were checked against NCEP data, and little difference was observed.
Energy Balance
The energy balance method for estimating evapo-ration assumes that heat inputs from net radiation are balanced by latent heat loss and sensible heat transfer. Equations for our energy balance calcula-tions are discussed extensively in Yin and Nichol-son (1998) and will not be repeated here. Briefly, top of atmosphere solar radiation calculated for 0° latitude is modified by cloud cover and lake albedo before entering the lake as incoming radiation. The net longwave flux from the lake is determined as a function of lake temperature, humidity, cloud cover, and water emissivity. The difference between these two terms is the net radiation income to the lake. Calculated radiation income in Lake Edward varies from 140 to 190 W/m2.
The ratio of the energy loss from conduction to that from evaporation is referred to as the Bowen ratio, which compares humidity differences in air with a saturated lake surface:
B = (Ca*(TL– Ta))/(L*(e0– ea)) (2)
where Ca is the specific heat of dry air, TL is the
lake surface temperature, Tais surface air
tempera-ture, L is the latent heat of vaporization, e0 is the
saturation vapor pressure, and eais the measured air
vapor pressure. Monthly Bowen ratio values for Lake Edward vary between 0.1 and 0.16. Solution of the Bowen ratio allows for the calculation of evaporation by converting latent heat loss to evapo-rated water using the latent heat of evaporation at measured lake temperatures.
Values calculated for monthly evaporation vary from 0.131 to 0.191 m/month (Table 5). Our mate of annual evaporation exceeds previous esti-mates for Lake Edward, but is similar to Penman and water-balance-based estimates for Lake George (Viner and Smith 1973).
Penman Evaporation
The Penman (1948) approach has been used in numerous studies (Winter et al. 1995, Turner et al. 1996, Yin and Nicholson 1998). Here we rely on a
form of the equation discussed in Jensen (1974) that has windspeed coefficients modified for use in large lakes:
Evap = {(s/(s + ∆))*(Qn – Qx) + (∆/(∆+ s))
[(15.36*(0.5 + 0.01U))*(e0– ea)]} / L (3)
where s is a parameter determined from the slope of the saturated vapor pressure-temperature curve at the mean air temperature, ∆ is the psychrometric
constant, Qn is net solar radiation, Qx is change in
heat stored in the water body, U is windspeed at 2 m height above the water body, e0 is saturated
vapor pressure, eais the measured vapor pressure at
air temperature and humidity, and L is the latent heat of vaporization.
Both Penman and energy-balance derived evapo-ration estimates exceed previous estimates of evap-oration rates from Lake Edward (Table 5) (Hurst 1927, Lehman 2002). Hurst’s estimate is based upon extrapolation of evaporation estimates from Lake Victoria to Lake Edward; however, subse-quent estimates have shown that evaporation rates for Lake Victoria exceed Hurst’s estimate by at least 30% (Yin and Nicholson 1998). Lehman (2002) estimated evaporation using mass transfer calculations, and used diel temperature variations for Lake Edward calculated from his physical model of the lake. While this approach should yield better estimates of evaporation than our calcula-tions above, the diurnal temperature fluctuacalcula-tions of Lake Edward are not known. Moreover,
evapora-TABLE 5. Monthly evaporation estimates for Lake Edward calculated from using both Penman and energy balance methods.
Energy Penman, Balance,
Month m/month m/month
Jan 0.1695 0.1609 Feb 0.1836 0.1795 Mar 0.1788 0.1564 Apr 0.1718 0.1812 May 0.1886 0.1913 Jun 0.1726 0.1701 Jul 0.1496 0.1689 Aug 0.1642 0.1380 Sep 0.1865 0.1727 Oct 0.2004 0.1770 Nov 0.1921 0.1313 Dec 0.1796 0.1555 Annual 2.1373 1.9828
tion calculations using mass transfer equations are problematic when shore-based climatic data are used (Winter et al. 1995). Penman and energy bal-ance formulations are less problematic in this re-gard due to the prominence of the radiation terms in those equations.
Monthly Penman estimates for evaporation rates exceed most previously published values (Table 5), but are again in rough agreement with both our en-ergy budget calculation and Penman estimates for Lake George (Viner and Smith 1973). Due to the potential problems with Penman-based estimation of evaporation (Winter et al. 1995, Nicholson and Yin 1998) the energy-budget derived estimate is used in water balance calculations below.
Outflow
The final term in our hydrologic budget for Lake Edward is outflow via the Semliki River. Annual-ized discharge estimates for the Semliki vary widely (Table 6), ranging from 3.3 × 109 m3/yr
(Worthington 1932) to 10.8 × 109 m3/yr (Lehman
2002). Comprehensive surveys of the Semliki River were made by the World Meteorological Organiza-tion’s Hydromet survey program at the Semliki’s entrance to Lake Albert (Said 1993, Hydromet 1982). Between the two lakes, the Semliki drains roughly an area of about 7,000 km2, including the
extremely wet western side of the Ruwenzori Mountains. Thus, although Hydromet measure-ments cannot directly tell us of the Semliki’s
dis-charge from Lake Edward, they do provide upper limits for the amount of water that exits Lake Ed-ward assuming that water losses by evaporation from the Semliki River are at least balanced by water inputs from the catchment between the two lakes.
Assuming that our hydrologic estimates for direct precipitation, river inflows, and evaporation are correct, solution of equation 1 provides an estimate of Semliki River discharge of 3.9 ×109 m3/yr,
simi-lar to those of Hurst (1927) and Worthington (1932). If the additional drainage received from the Semliki catchment (assuming inputs of 0.3 m/m2/yr,
similar to the Edward catchment, from 7,000 km2),
is subtracted from Hydromet (1982) gauge mea-surements, the Semliki discharge from Lake Ed-ward is about 3.7 × 109 m3/yr, very similar to our
estimate of 3.9 × 109 m3/yr based upon Lake
Ed-ward’s water balance.
Stable Isotope Hydrology of Lake Edward
Numerous authors have used stable isotopic and hydrologic measurements of lakes to constrain less-easily measured components of lake’s hydrologic budgets (see Gat 1995). Although a lack of compre-hensive data for the Edward catchment precludes a detailed discussion of the lake’s isotope hydrology, stable isotope data nevertheless provide important constraints on Lake Edward’s water budget.
Assuming groundwater is a negligible hydrologic
TABLE 6. Estimates of the annual rate of Semliki outflow. Sources are given at left.
Flow Rate Annualized flow
Author Site and Date (m3/sec) (109m3/yr)
William Garstain, reported in
Hurst, 1927 L. Edward, dry season, 1903 97 NA
Hurst, 1925 L. Albert, Mar 1924 175 NA
Hurst, 1925 L. Albert, Apr 1923 90 NA
Hurst, 1927 L. Edward, estimated NA 5.0
Worthington, 1932 L. Edward, dry season, 1930 104 3.3
Damas, 1937 L. Edward 65 NA
Hydromet, 1982,
reported in Said, 1993 L. Albert, measured 1956–60 NA 3.8 Hydromet, 1982
reported in Said, 1993 L. Albert, measured 1962–70 NA 5.9
input and output, the isotopic mass balance of a lake can be described with the following equation:
dVδ
lake/dt = Qrainδrain + Qinflowsδinflows –
Qoutflowδlake – Qevapδevap (4)
where V is the volume of the lake, dt is the time pe-riod of interest, Q represents hydrologic fluxes, δ
represents the isotopic composition of a given vari-able, and the isotopic composition of a lake’s out-flow is assumed to be identical to that of the lake water. Applying this equation to Lake Edward, and assuming steady state conditions (current dV/dt equals zero), this equation can be expressed as:
Qrainδrain + QKazingaδKazinga+ Qother inflowsδother inflows = QSemlikiδlake + Qevapδevap. (5)
Assuming that the inflow fluxes are relatively well constrained, this equation can be rearranged to
solve for the ratio of water losses by outflow rela-tive to evaporation:
Qrainδrain + QKazingaδKazinga+ Qother inflowsδother inflows = QSemlikiδlake + (1 – QSemliki)δevap. (6)
The isotopic composition of Lake Edward was mea-sured in 1996, 2001, and 2002, and displays little variation, with an average of 4.3 ‰ for δ18O and 30
‰ for δD (Table 7). Wet and dry season
measure-ments of Lake George in 2002 and 2003 also show little variation, while the Kazinga Channel varied slightly and averaged about 0.5 ‰ for δ18O and 10
‰ for δD. River samples include wet and dry
sea-son measurements in 2002 and 2003 from all the major tributaries from the eastern side of Lake Ed-ward, several rivers draining the Ruwenzoris, and springwater samples from near the eastern border fault. Together, these samples cover 65% of Lake Edward’s catchment area, and average –2.2 ‰ for
δ18O and –2.8 ‰ for δD. It should be noted that TABLE 7. Results of stable isotopic analysis (δδ18O, δδD) of lakes, rivers, and springs
from the Lake Edward catchment sampled in 1996–2003.
Sample Date δ18O, SMOW δD,SMOW
Lake Edward surface (5 m depth) May-96 4.3 29
Lake Edward hypolimnion (45 m depth) May-96 4.5 31
Lake Edward surface (1 m depth) Jan-01 4.2 29
Lake Edward surface (1 m depth) Jan-02 4.2 30
Lake George surface (0.5 m depth) Jan-02 1 14
Lake George surface (0.5 m depth) May-03 1.0 10
Kazinga Channel Jan-01 0.6 11
Kazinga Channel May-03 0.3 8
Nyamugasani River, East tributary Jan-02 –2.7 –4
Nyamugasani River, West Tributary Jan-02 –2.2 –1
Lubilia River Jan-02 –2 0
Bwera River (Lubilia Tributary) Jan-02 –1.4 2
Mubuku River (Lake George inflow) May-03 –4.4 –16
Nyamweru River Jan-02 –1 5
Nyamweru River May-03 –2.8 –7
Ishasha River Jan-02 –1.7 1
Ishasha River May-03 –2.9 –9
Ntungwe River Jan-02 –1.1 3
Ntungwe River May-03 –2.7 –8
Nchwera River Jan-02 –1.1 3
Nchwera River May-03 –2.4 –6
Maramagambo forest spring, “Blue Pool” Jun-03 –2.1 –2
Maramagambo forest unnamed spring Jul-03 –2.2 –3
Rain, Ft. Portal 4 Jan 2002 2.2 33
Rain, Ft. Portal 1 Jan 2002 2.1 28
Rain, Ft. Portal 29 Dec 2001 –1.6 3
this average does not include the Kazinga Channel, which is strongly affected by evaporation of waters impounded within Lake George.
The isotopic composition of rainfall in the region is poorly constrained. Four rainfall measurements taken in the Lake Edward basin in 2001–2003 aver-age –0.4 ‰ for δ18O and 17.25 ‰ for δD but
ex-hibit considerable scatter. The nearest rainfall station to Lake Edward, at Entebbe, Uganda, has a mean weighted composition of –2.91 ‰ for δ18O
and –11.2 ‰ for δD (Rozanski et al. 1996), but is
likely influenced by water evaporated from Lake Victoria. Moreover, evapotranspired moisture within the Congo River basin may bring isotopi-cally heavy rainfall from the west into the Edward region (Rozanski et al. 1993), thereby further dis-tancing the isotopic composition of rainfall near Lake Edward from Entebbe.
Lake Edward, Lake George, river, and spring samples are plotted in δ18O vs. δD space together
with the global meteoric water line (GMWL, δD =
8 * δ18O + 10) of Craig (1961) and the African
me-teoric water line (AMWL, δD = 7.4 * δ18O + 10.1)
(Fig. 4). The latter was adopted by Cohen et al. (1997), who showed that stations in the interior of East and Central Africa define a δD vs. δ18O trend
that differs from the GMWL due to the extreme continentality of rainfall in interior Africa. The va-lidity of the AMWL for Lake Edward is confirmed by the fact that rivers from the Edward basin plot on or closer to the AMWL than the GMWL (Fig. 4). Following the reasoning of Craig (1961), the tersection of the line linking Lake Edward to
in-flowing rivers yields the mean isotopic composition of Lake Edward’s source waters. Solution of these equations gives –0.91 ‰ for δ18O and 3.36 ‰ for δD.
These values are somewhat heavier than the aver-age values of rivers draining the northern and east-ern catchments of Lake Edward. Moreover, if we assume that the mean isotopic composition of rivers sampled within the Edward basin equals that of rainfall, we can estimate the weighted isotopic com-position of inputs to Lake Edward (the Kazinga Channel, river inputs, and rainfall) to be –1.56 ‰ for δ18O and 0.1 ‰ for δD. The assumption that the
isotopic composition of rivers is not strongly al-tered by evaporation, and therefore can be used to estimate the composition of rainfall, is supported by the position of those rivers on or near the AMWL and GMWL in Figure 4. Were the rivers strongly affected by evaporation, we would expect them to plot off the meteoric water lines along the regional evaporative trend defined by Lake Edward (Craig 1961). The differences between these compositional estimates of the source waters for Lake Edward suggest an unmeasured heavy isotopic source-water to Lake Edward, likely related to moisture from the Congo basin from the unsampled catchments to the south and west of Lake Edward (Rozanski et al. 1993). At present there is no objective method for determining the precise isotopic composition of source waters to Lake Edward, so we assume this composition is intermediate between our weighted composition and the composition calculated using the AMWL: –1.24 ‰ for δ18O and 1.73 ‰ for δD.
We note that this estimate is conservative in that it is isotopically heavy relative to our measured val-ues. Isotopically lighter input values will result in higher estimates of the importance of evaporation, calculated below.
The isotopic composition of evaporated water vapor from Lake Edward, δevap, has not been
mea-sured. However, it can be calculated using the fol-lowing equation from Benson and White (1994) that describes the isotopic equilibration of lake-de-rived evaporated water with regional humidity across a turbulent mixed layer:
δevap/1000 = {[(Rlake/ev) – hfRair] /
[((1 – h)/k) + h(1 – f)]} – 1 (7)
where Rlake = 1 + δlake/1000 and Rair = 1 + δair/
1000. In this equation, evis the equilibrium enrich-ment factor that depends on lake temperature (ev = FIG. 4. δD vs. δ18O for rivers, springs, and lakes
exp(1137TL–2 – 0.4156TL–1 – 2.0667 × 10–3,
Ma-joube 1971), h is relative humidity of the region, f is the fraction of humidity that has been advected into the basin, δair is the isotopic composition of
moisture advected into the basin, and k is the ki-netic fractionation factor that depends on wind speed and equals 0.994 for wind speeds less than 6.8 m/s (Merlivat and Jouzel 1979). We used an erage relative humidity of 74%, and the annual av-erage lake temperature data of Verbeke (1957) for Lake Edward to calculate ev (Table 4). δair is
as-sumed to be in isotopic equilibrium with regional rainfall at surface air temperatures, and regional rainfall is assumed to have the same average iso-topic composition as rivers in the region (Friedman
et al. 1962, Benson and White 1994).
The calculation of δevapis very sensitive to
com-binations of f and h (Benson and White 1994). De-creasing humidity causes isotopically lighter values of δevap due to faster exchange across the mixed
layer near the lake surface. The value of f for Lake Edward is unknown, and will depend on factors such as regional climate, humidity, and winds as well as basin morphology. f can vary between 0 and 1, but is likely low in large lakes such as Lake Ed-ward (e.g., Ricketts and Johnson 1996, Benson and White 1994). By substituting equation 6 into equa-tion 5 and varying f, we can calculate a range of
possible values for the percentage of the water in-come to Lake Edward lost by evaporation (Fig. 5). Using the same suite of regional input variables, humidity and wind speed data from Kasese, and hy-drologic and isotope variables from Viner and Smith (1973) and measured in the present study, we performed the same calculation for the oxygen iso-tope balance of Lake George. The latter calculation allows us to estimate the validity of our results for Lake Edward, as the hydrological fluxes for Lake George are reasonably well-known (Viner and Smith 1973).
Viner and Smith (1973) show that Lake George loses 21% of its water income by evaporation, while solution of equations 5 and 6 for Lake George estimate evaporative losses of 22 to 25% of water income as f varies from 0 to 0.7. Our esti-mates are thus remarkably similar to measured val-ues given the uncertainty in our estimates of the isotopic composition of rainfall in the region. Ap-plying these equations to Lake Edward, calculations of the percentage of the net water income that is lost from Lake Edward by evaporation differ for
δ18O and δD by an average of 12%. It seems likely
that this is due to errors in calculating the composi-tion of source water to the lake. Regardless, it is ap-parent that, at a minimum, evaporation represents 40% of the net water output from Lake Edward. Unfortunately, the value of f cannot be known with certainty for Lake Edward. However, at values of f < 0.4, which seem likely for a lake the size of Ed-ward, and with δ18O calculations using values set at
mean variables listed in Table 4, the most likely evaporative loss is between 50 and 60% of the water income.
DISCUSSION AND RECOMMENDATIONS
The East African Great Lakes comprise an im-portant economic resource for riparian countries. Despite their importance, considerable uncertainty remains with regards to the Great Lakes’ physical hydrologies, including that of Lake Edward. Within the present study, surface runoff, outflow, evapora-tion, and the isotopic composition of water income to Lake Edward remain poorly constrained. More-over, it should be noted that we have averaged hy-droclimatic data from the Lake Edward region across several decades, introducing potential errors into our estimates that we cannot quantify. Never-theless, some preliminary conclusions may be drawn, and we hope that this work will spur future
FIG. 5. Isotopic simulations of water loss by evaporation as a function of f (fraction of advected moisture over the lake) calculated for δδD and δ18O. Error bars represent the range of
varia-tion when the net source composivaria-tion is allowed to vary between the values calculated by mean weighting of hydrologic inputs and by the inter-section of the AMWL with the evaporative trend defined by Lake Edward.
research into the physical hydrology of this impor-tant lake.
Our hydrologic estimates for the water budget of Lake Edward based upon literature review suggests that evaporation comprises about 54% of the water losses from Edward (Table 8). This seems reason-able in light of the results of our isotopic analyses that constrain the ratio of evaporation/total water losses to between 0.5 and 0.6. Our revised hydro-logic estimates for Lake Edward suggest that evap-oration is much more important to water losses than previous researchers have indicated (e.g., Hurst 1927, Lehman 2002). Based upon our analysis, it appears that previous analysts may have overesti-mated the magnitude of river inputs to Lake Ed-ward and thereby annual discharge from the Semliki River. Indeed, comparing the hydrologic estimate of this study to Lehman (2002) highlights the importance of obtaining accurate runoff esti-mate from the Lake Edward basin: The higher sur-face runoff values used by Lehman (2002) yield 85% higher water inputs to the lake than the present study.
Our results have important implications for the modern-day chemistry of Lake Edward, and the po-tential for developing paleohydrologic records from Lake Edward. Lake Edward waters are slightly brackish (0.7 ppt TDS) with a chemistry dominated by Na+, Mg2+, K+, and HCO
3-. Kilham and Hecky
(1973) attributed this chemistry to the influence of the alkaline, ultramafic rocks of the Virunga volca-noes. Unfortunately, there are almost no chemical data from the rivers draining the Virunga region into Lake Edward, severely limiting our ability to develop hydrochemical mass balance models of Lake Edward. Lehman (2002) produced the first chemical model for Lake Edward, and balanced the lake’s bicarbonate budget using Hurst’s (1927) measurement of the alkalinity of the Ruchuru River of 17.2 meq. This alkalinity is more than twice that of Lake Edward’s; however, Hurst (1927) also
states that reagents for measuring chemical analyses were made from local natural waters, potentially corrupting the alkalinity data.
Marlier (1951) measured the conductivity of the Ruchuru River at 408.7 µS/cm, a value much too small to allow an alkalinity of 17.2 meq/l. Other rivers and lakes in the region with conductivities ranging from 300–600 µS/cm have alkalinities be-tween 2.1 and 6.8 meq/l, while Lake Edward has a conductivity of ~880 µS/cm and an alkalinity of ~9 meq/l (e.g., Damas 1954, Talling and Talling 1965). In sum, our hydrologic estimates imply that Lake Edward’s salinity is significantly concentrated rela-tive to its inputs; we estimate a concentration factor of ~2 for conservative solutes. Furthermore, our es-timates imply that Lake Edward’s salinity and chemistry should be particularly sensitive to changes in the hydrologic balance and concomitant changes in salinity concentration factors, particu-larly changes in rainfall as suggested by Lehman (2002).
Considerable ambiguities about the hydrology of Lake Edward remain and will not be resolved with-out additional measurements of the lake’s physical properties. While some of these, such as rainfall, lake temperature, and cloudiness, may be most ef-fectively monitored using remote sensing tech-niques (e.g., Lehman 2002), others, such as surface runoff and evaporation, will require additional field measurements using river gauges and lake-based meteorological buoys. Both Lehman (2002) and this study highlight cloudiness, humidity, diurnal temperature, and unmeasured runoff from the southern rivers as key variables needed to clarify our understanding of Lake Edward. These variables remain unmeasured, and must be quantified to fur-ther our knowledge of both the modern and paleo-limnology of this Great Lake.
CONCLUSIONS
Calculations based on stable isotopic and hy-drometeorological data provide similar estimates for Lake Edward’s water budget. These data indi-cate that Lake Edward loses between 50 and 60% of its water income by evaporation from the lake sur-face. Hydrologic inputs to the lake are dominated by river inputs from the catchment. Thus, although Lake Edward loses significantly more of its water income to outflow than other East African Rift lakes, the large evaporative flux from Lake Edward should make the lake’s water level and chemistry highly sensitive to hydroclimatic variations.
TABLE 8. Our calculated summary water budget for Lake Edward based upon previous surveys and stable isotope mass balance calculations.
Direct Precipitation 2.04 ×109m3/yr Kazinga Channel Discharge 1.7 ×109m3/yr Other catchment inputs 4.75 ×109m3/yr Evaporation 4.61 ×109m3/yr Semliki River Outflow 3.88 ×109m3/yr Water Residence Time 20 years
ACKNOWLEDGMENTS
We wish to thank the Government of Uganda, and in particular the Ugandan National Council of Science and Technology and Ugandan Wildlife Au-thority for permission to conduct field work. Dirk Verschuren, Hilde Eggermont, Kristina R. M. Beun-ing, and the International Decade for East African Lakes program are also acknowledged for assis-tance with field work. Sharon Nicholson and John T. Lehman provided very helpful reviews of an ear-lier version of this manuscript. This research was supported by NSF Earth System History program grant # 0314832. Any opinions, findings and con-clusions expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation.
REFERENCES
Beadle L.C. 1966. Prolonged stratification and deoxy-genation in tropical lakes I. Crater Lake Nkugute, Uganda, compared with Lakes Bunyoni and Edward.
Limnol. Oceanogr. 2:152–163.
———. 1981. The Inland Waters of Tropical Africa: An
Introduction to Tropical Limnology. London:
Long-man.
Benson L.V., and White J.W.C. 1994. Stable isotopes of oxygen and hydrogen in the Truckee River-Pyramid Lake surface-water system. Limnol. Oceanogr. 39: 1945–1958.
Cohen, A.S., Talbot, M.R., Awramik, S.M., Dettman D.L., and Abell, P. 1997. Lake level and paleoenvi-ronmental history of Lake Tanganyika, Africa, as inferred from late Holocene and modern stromatolites.
GSA Bulletin 109:444–460.
Craig H. 1961. Isotopic variations in meteoric waters.
Science 133:1702–1703.
Damas, Z.H. 1937. Recherches dans les Lacs Kivu,
Édouard, et Ndalaga. Exploration du Parc National
Albert. Mission H. Damas (1935–1936). Institut Royal Colonial Belge, Brussels, Belgium II (1). ——— , 1954. Étude limnologique de quelques lacs
rundais, I le cadregéographique. Institut Royal
Colo-nial Belge, Mémoires Collection Tome XXIV, fasc. 2. Brussels, Belgium.
Friedman, I., Machta L., and Soller, R. 1962. Water-vapor exchange between a water droplet and its envi-ronment. J. Geophys. Res. 67:2761–2766.
Gat, J.R. 1995. Stable isotopes of fresh and saline lakes. In Physics and Chemistry of Lakes, 2nd Ed., pp. 139–162. A. Lerman, D. Imboden, and J.R. Gat (eds.), Springer-Verlag, Berlin.
Hecky, R.E., and Degens, E.T. 1973. Late
Pleistoncene-Holocene chemical stratigraphy and paleolimnology of the Rift Valley Lakes of central Africa. Woods Hole
Oceanographic Institution Technical Report 73, Wood’s Hole, MA.
Hulme M. 1998. Global precipitation dataset, http:// www.cru.uea.ac.uk/~mikeh/datasets/global
Hurst, H.E. 1925. The lake plateau basin of the Nile. Egypt Ministry of Public Works, Physical Department Paper No. 2.
———. 1927. The lake plateau basin of the Nile. Egypt Ministry of Public Works, Physical Department Paper No. 23.
Hydromet. 1982. Hydrometerological survey of the
catchments of Lakes Victoria, Kyoga, and Mobutu Sese Seko. Project findings and Recommendations.
World Meteorological Organization (WMO) and United Nations Development Program (UNDP), Geneva. Report no. RAF/73/001.
Jensen, M.E. 1974. Consumptive use of water and
irriga-tion water requirements: a report. The American
Society for Civil Engineers Technical Committee on Irrigation Water, New York.
Kilham, P., and Hecky, R.E. 1973. Fluoride: geochemi-cal and ecologigeochemi-cal significance in East African waters and sediments. Limnol. Oceanogr. 18:932–945. Laerdal, T. 2000. Lakes Edward, George, and Victoria
(Uganda): A study of late Quaternary rift tectonics, sedimentation, and paleoclimate. PhD thesis, Univer-sity of Bergen.
———, Talbot, M.R., and Russell, J.M. 2002. Late Qua-ternary sedimentation and climate in the Lakes Edward and George area, Uganda-Congo. In The East
African Great Lakes: Limnology, Paleolimnology, Biodiversity, Eds. EO Odada and D. Olago, pp.
429–470. Kluwer. Adademic Publishers, Dordrecht, The Netherlands.
Lehman, J. 2002. Application of AVHRR to water bal-ance, mixing dynamics, and water chemistry of Lake Edward, East Africa. In The East African Great
Lakes: Limnology, Paleolimnology, Biodiversity,
Eds. EO Odada and D. Olago, pp. 236–260. Kluwer. Adademic Publishers, Dordrecht, The Netherlands. Majoube, M. 1971. Fractionnement en oxygene-18 et en
deuterium entre l’eau et sa vapor. J. Chem. Phys. 197: 1423–1436.
Marlier, G. 1951. Recherches hydrobiologiques dans les riviéres due Congo Oriental. Composition des eaux. La conductibilité électrique. Hydrobiologia 3: 217–227.
Merlivat, L., and Jouzel, J. 1979. Global climatic inter-pretation of the deuterium-oxygen 18 relationship for precipitation. J. Geophys. Res. 84:5029–5033.
Nicholson, S.E. 1996. A review of climate dynamics and climate variability in Eastern Africa. In The
Limnol-ogy, climatolLimnol-ogy, and paleoclimatology of the East African Lakes. Eds. T.C. Johnson and E.O. Odada, pp.
Penman, H.L. 1948. Natural evaporation from open water, soil, and grass. Proceedings of the Royal
Soci-ety of London Series A 76372–383.
Ricketts, R.D., and Johnson, T.C. 1996. Climate change in the Turkana basin as deduced from a 4,000 year long δ18O record. Earth Planet. Sci. Lett. 142:7–17. Rozanski, K., Araguás-Araguás, L., and Gonfiantini, R.
1993. Isotope patters in modern global precipitation. In Climate Change in Continental Isotopic Records. P.K. Swart, K.C. Lohmann, J. McKenzie, and S. Savin (eds.), pp. 1–36. American Geophysical Union Monograph 78, Berkeley, CA.
———, Araguás-Araguás, L., and Gonfiantini, R. 1996. Isotope patterns of precipitation in the East African region. In The Limnology, climatology, and
paleocli-matology of the East African Lakes, Eds. T.C.
John-son, and E.O. Odada,, pp. 79–94. Gordon and Breach Publishers, Newark, NJ.
Russell, J.M., Johnson, T.C., Kelts, K.R., Laerdal, T., and Talbot, M.R. 2003. An 11,000-year liithostrati-graphic and paleohydrologic record from equatorial Africa: Lake Edward, Uganda-Congo.
Palaeo-geograhy Palaeoclimatology Palaeoecology 193:
25–49.
Said, R. 1993. The River Nile: Geology, Hydrology, and
Utilization. Oxford: Pergamon Press.
Spigel, R.H., and Coulter, G.W. 1996. Comparison of Hydrology and Physical Limnology of the East African Great Lakes: Tanganyika, Malawi, Victoria, Kivu, and Turkana (with reference to some North American Great Lakes In The Limnology,
climatol-ogy, and paleoclimatology of the East African Lakes,
Eds. T.C. Johnson and E.O. Odada, pp. 103–140. Gor-don and Breach Publishers, Newark, NJ.
Talling, J.F., and Talling, I.B. 1965. The chemical com-position of African lake waters. Internationale Revue
gesamten Hydrobiologie 50:421–463.
Turner, B.F., Gardner, L.R., and Sharp, W.E. 1996. The hydrology of Lake Bosumtwi, a climate-sensitive lake in Ghana, West Africa. J. Hydrol. 183:243–261. Verbeke, J. 1957. Recherches écologiques sur la faune
des grands lacs de l’est du Congo Belge. Resultats scientifiques de l’exploration hydrobiologique (1952–1954) des lacs Kivu, Edouard et Albert. Institut
Royal Colonial Belge, Brussels, Belgium 3 (1). Viner, A.B. and Smith, I.B. 1973. Geographical,
histori-cal, and physical aspects of Lake George. Proc. Royal
Society of London Series B 184:235–270.
Winter, T.C., Rosenberry, D.O., and Sturrock, A.M. 1995. Evaluation of 11 equations for determining evaporation for a small lake in the north central United States. Wat. Res. Res. 31:983–993.
Worthington, E.B. 1932. A report on the fisheries of
Uganda Investigated by the Cambridge Expedition to the East African lakes, 1930–31. Part I. Lakes Edward and George and the Kazinga Channel.
Zoo-logical Laboratory, Cambridge, June 1932.
Yin, X., and Nicholson, S.E. 1998. The water balance of Lake Victoria. Journal of Hydrological Sciences 43:789–811.
Submitted: 7 March 2005 Accepted: 24 November 2005
