The underwater light conditions in the African Great Lakes depend on the complex dynamics of ecological and hydrological forces, and are strongly influenced by local environmental conditions and global biogeochemical cycles. Changes in the optical conditions in these lakes have direct impacts on ecosystem productivity, carbon dynamics and nutrient availability. A central role in the underwater light climate is played by dissolved organic matter which is present in all aquatic ecosystems. The chromophoric fraction of these compounds can mediate ecosystem change through its influence on the attenuation of ultraviolet and PAR radiation, microbial carbon cycling and radiative transfer. In the African Great Lakes, little information is available regarding the dynamics of dissolved organic matter and those sources and sinks which control its presence in the water column. We present an extensive spatial analysis of three major bays on the Kenyan and Ugandan shores of Lake Victoria. We use these data to examine the dynamics of chromophoric dissolved organic matter in different bays and we develop a model to estimate its flow from these bays to the Lake, considering both conservative mixing and photodegradation processes. While some bays release chromophoric dissolved organic matter practically unmodified into the Lake, increased residence time and exposure to solar ultraviolet radiation create conditions where chromophores are lost before entering the open lake.
Variations in the penetration of solar radiation can strongly modify the chemical and biological environment of aquatic ecosystems. A key element which controls the spectral conditions in tropical ecosystems is dissolved organic matter. Whether produced by coastal vegetation, phytoplankton biomass or catchment activities, the spectral properties and concentration of dissolved organic matter can influence primary productivity, secondary production and radiative transfer. Phototransformations of dissolved organic matter have recently received much study (Helbling and Zagarese, 2003). Photodegradation and photobleaching (defined here as the loss of absorption in the ultraviolet and visible wavelengths) of chromophoric dissolved organic matter (CDOM) will depend on the spectral intensity of the incident irradiance in the water column, photolability of the CDOM present and exposure time (Bertilsson and Tranvik, 2000). Photochemical transformation of CDOM can result in the production of low molecular weight organic compounds with lower absorbivity in the wavelengths absorbed (Carrillo et al., 2002). The resultant consequences on the aquatic ecosystem can be multiple, including: the formation of toxic photoproducts (Osburn and Morris, 2003; Zepp, 2003), the increase in the pool of bioavailable carbon substrates (Bertilsson and Tranvik, 1998) and the increase of secondary productivity, as well as inducing changes in the spectral attenuation of the water column with implications on productivity and radiative transfer (Neale, Helbling and Day, 2007).
In the present study, we examine the spectral conditions of the inshore waters of Lake Victoria, concentrating on three large bay areas (Katonga Bay, Nsonga Bay and Nyanza Gulf), each with different optical and hydrological conditions, and each with different influences on the open lake (Figure 1). In these areas, catchment and littoral zone conditions play an important role in controlling the variability of the chemical and physical conditions of the water column (Lung'Ayia et al., 2000; Cózar et al., 2004, Bracchini et al., 2007). Recent increases in nutrient loads, changes in the stability and extension of the vertical stratification and major changes in the trophic web have brought these waters to eutrophic conditions (Mugidde, 1993; Talling and Lemoalle, 1998).
The dominant wetland vegetation in Lake Victoria is a highly productive emergent C4 sedge (Cyperus papyrus L.) (Saunders et al., 2007) which can generate large quantities of organic carbon in favorable conditions of high solar insolation and abundant water. The anaerobic accumulation of organic matter and its eventual release into the water column is strongly linked to wetland extension and management. Likewise, intensive land use within the catchment has been shown to lead to major increases in the transport of organic and nutrient rich matter (Njiru et al, 2008).
To better understand the importance of CDOM on the optical conditions of these important areas, we characterise and compare spatial conditions of conservative and non-conservative components of the water column. We develop a model of photobleaching to help explain differences in the spatial variability of optical conditions.
We examined the optical and chemical conditions of coastal areas throughout the Ugandan and Kenyan areas of Lake Victoria, in May 2003 and June 2004 (Loiselle et al., 2008). Measurement stations (277) represent all major coastal types.
Vertical profiles of temperature (accuracy ± 0.1°C) and specific conductance (accuracy ± 2.0 μ S cm− 1) (Hydrolab Datasonde 4, Hach Company, Loveland, CO) and solar irradiance (PAR waveband, 313 nm, 340 nm with a PUV 514, Biospherical Instruments, San Diego, CA) were measured at each sampling site, beginning just below the water surface and measuring to near the lake bottom. The profiling radiometer was manually lowered from the sunny side of the boat, obtaining diffuse irradiance measurements every 0.3 s together with depth. On average, 400 measurements (irradiances and water depth) were obtained for each profile. Profiles in which changes in surface solar irradiance occurred (clouds) were discarded and the measurements repeated. The resulting radiation profile was corrected by removing the dark current signal (< 0.00001 W m− 2 nm− 1) and plotted against the water depth. The diffuse attenuation coefficient Kλ was calculated by fitting the best exponential curve to the profile data (Iz = Ioe(− Kλ (z − z0))) where the fitting of the exponential decay was superior to R2 > 0.98. Specific conductance was determined from conductivity normalized to a temperature of 25°C. The conductivimeter was calibrated every day using a standard KCl solution (Crison, Spain). Incident solar photosynthetically available radiation (PAR) was measured 3 m above the water surface (Skye Instruments LTD, UK) during all measurements.
Mixing depth was defined graphically by the intersection of trend lines fitted to the thermal profiles, one through the upper part of the epilimnion and the second through the metalimnion(Wetzel, 2001). The determination of chlorophyll a and dissolved organic matter was obtained from samples obtained at a depth of 0.5 m. Chlorophyll a concentrations were estimated using a calibrated portable fluorometer (Turner designs: wavelength excitation = 485 nm, measured wavelength = 680 nm) which was calibrated daily using a solid standard (Turner Designs). Fluorometer calibration was made using measurements of chlorophyll-a concentrations according to Strickland and Parsons (1972) from different lakes including Lake Victoria. Despite of the different sources of variability in the chlorophyll-a fluorescence (e.g. quenching, temperature), the measured concentrations and fluorescence signals showed a good linear correlation. CDOM samples were filtered using single use 0.22 μ m polyethersulfone filters (Millex GP Filter Unit, Millipore S.A., Malsheim, France) following Grzybowski (2000). Samples were stored in acid washed glass bottles, after rinsing each bottle with filtered lake water. Filled bottles were covered and stored in the dark at 4°C for 1 to 5 days before absorbance measurements. Absorbance scans were performed using a Lambda 25 spectrophotometer (PES, Perkin Elmer, speed scan of 4500 nm min− 1). A single quartz cuvette (pathlength 0.01 m) and a Milli-Q blank were used. After acclimation in a temperature controlled room (18°C), three scans were performed for each sample and average values computed for all samples where the differences between scans were less than 10− 3. Absorbance (aλ.) was measured over a wavelength range of 200 to 700 nm with a spectral resolution of 1 nm. The absorbance measurement at 700 nm was subtracted from each measurement to remove variations due to baseline changes. Absorption coefficients (Aλ) were calculated using Aλ = 2.303 l− 1* aλ. where l is the pathlength in meters.
A small number of filtered samples (4) were tested for photo-lability using quartz bottles with a high UV transparency (average transmittance 94% in UVB and UVA). These samples had been previously filtered in-situ (0.22 μ m) after sampling in Nyanza Gulf (Figure 3c) and stored in the dark at 4°C for 10 days prior to the photobleaching experiment. The quartz bottles were placed in a controlled temperature bath (20 ± 1.8°C), together with MilliQ blanks and exposed for three days to natural solar radiation in the Siena laboratories. Prior to initiating the experiment, samples were stored for 8 hours in the dark at room temperature to examine changes in optical properties not related to photodegradation. Surface irradiance measurements were obtained every minute throughout the experiment using single channel sensors (SKR 420, 430, Skye Instruments LTD, UK) for ultraviolet A (UVA, 315–380 nm) and ultraviolet B (UVB, 280–315 nm). The sensors are cosine corrected with cosine error of 3% to 80 degrees, linearity error < 0.2%, absolute calibration error of about 3%. Total dose (kJ m− 2) was calculated by considering the average solar irradiance in the UV wavelengths, the total time of exposure and spectral transmission of the quartz bottle, weighted for solar spectral irradiance in the UV wavelengths. Samples within the bottles were mixed twice daily during sampling. Samples (15 ml) were obtained twice daily (pre and post exposure) from each bottle after rinsing the external parts with sterilized water. Samples were housed in dark glass containers, covered and stored at 4°C until spectrophotometric analysis at the conclusion of the experiment.
The optical variables showed significant spatial variability in all study areas. Diffuse vertical attenuation coefficients were constantly higher near to the littoral zone (Figure 2a–c). Nearshore KPAR values averaged well over 2 m− 1 while KPAR values in the open lake averaged below 1 m− 1. In all cases, a horizontal gradient was present, which is more pronounced for the measurements of diffuse vertical attenuation of solar UV radiation, determined at 313 nm (K313) (Figure 3a–c) and 340 nm. Absorption of solar UV radiation was related to CDOM absorption, with a significant and linear correlation between K313 and A365 (df = 95, R2 = 0.75).
Measurements of absorption of the dissolved fraction (A365) were highest nearest to the littoral zone, especially when wetland lined rivers flowed into the bay (Katonga River into Katonga Bay, Sondu and Nyando rivers into Nyanza Gulf). Absorption decreased as the CDOM rich nearshore waters moved into the open lake. This reduction in CDOM absorption is dominated by two principle processes, degradation of chromophores and the conservative mixing of CDOM rich bay waters with the open lake water. The two processes are not exclusive and contribute in different degrees to changes in CDOM absorption over time and space.
In ocean estuaries, several authors (e.g. Stedmon and Markager, 2003) have compared changes in salinity and changes in CDOM adsorption to differentiate between conservative mixing and photodegradation processes. As riverine waters enter into the sea, there is a gradual increase in salinity. If salinity can be considered as a conservative property (ignoring evaporation), relative changes in salinity can be compared to changes in CDOM absorption over space. Such a comparison can be used to determine whether losses in CDOM absorption are due to conservative mixing with low concentration CDOM waters or due to the non-conservative sinks (loss of chromophores) such as photodegration.
In Lake Victoria, an inverse situation occurs. About 84% of water input into Lake Victoria comes as rain (Talling and Lemoalle, 1998), resulting in a lake with a very limited ionic content with respect to the ion rich waters from the catchment where anthropogenic activities are present. In such conditions, differences in specific conductance have been used to trace water inflows from the catchment to the lake (Cozar et al., 2004), treating specific conductance as a near conservative tracer of water movement. Differently from estuaries, a decrease in specific conductance is expected to occur from riverine waters to open lake waters. Where decreases in CDOM absorption and the specific conductance follow similar patterns, conservative mixing is likely to dominate the loss of CDOM absorption. Where changes in CDOM concentrations and specific conductance are unlinked, sources or sinks of CDOM should be present. In Katonga Bay and Nyanza Gulf, major changes in both conductance and CDOM absorption occurred (Figure 4a, b). In both cases, a reduction in the specific conductance for inner bays (> 100 μ S cm− 1 in Katonga Bay and > 140 μ S cm− 1 in Nyanza Gulf) to open water (< 80 μ S cm− 1) occurs. Nsonga Bay, without a major river source presents less significant differences in specific conductance.
In Katonga Bay, a comparison between specific conductance and CDOM absorption (Figure 4a) shows that both measurements behave similarly, while in Nyanza Gulf CDOM absorption and specific conductance show different trends (Figure 4b). This is also confirmed by examining the correlation between specific conductance and CDOM absorption. In the case of Katonga Bay, a significant correlation was observed (df = 19, R2 = 0.77) while the correlations for Nyanza Gulf (df = 18, R2 = 0.27) and Nsonga Bay (df = 22, R2 = 0.34) were not significant (p > 0.001).
Measurements of photodegradation, performed using quartz bottles on 4 samples from Nyanza Gulf show the gradual reduction in UV absorption over increasing dose (Figure 4c). The resultant photodegradation rate (change in absorption over increasing UV dose) ranged from 0.001 – 0.0001 m kJ− 1 m− 2. Higher rates characterised the inner Gulf (K1) and lower rates characterised the outer Gulf (K15, K20). Blank measurements of MilliQ water showed no absorption and samples not exposed to solar radiation did not exhibit a reduction in absorption over time.
A reduction in CDOM absorption, K313 and KPAR in relation to distance from the coast occurred in both Katonga Bay and Nyanza Gulf. This reduction in CDOM absorption is related to both dilution and photodegradation of CDOM from most probably terrestrial sources (Kowalczuk et al., 2003). While the degradation of phytoplankton biomass may also contribute to CDOM absorption (DeGrandpre et al.,1996; Nelson et al. 2004), no significant correlation between chlorophyll a concentrations and CDOM absorption was observed in the study bays. Undoubtedly, the importance of phytoplankton biomass in the production of CDOM increases in the open lake waters (Loiselle et al., 2008).
Conservative mixing between the open lake waters and the CDOM rich inner bay waters will depend on river flowrate, overall rainfall and the horizontal mixing regime between the lake and the bay. While the rainfall and river flow, normalised for bay area are a similar magnitude for both Nyanza and Katonga, the exchange rate between the bay and the lake are significantly different. Katonga Bay is an open bay (with the Bay aperture making up 28% of the total Bay perimeter) which allows for a significant exchange with open lake waters. Nyanza Gulf is closed by Rusinga Island and the Mbita causeway and presents a reduced channel (Rusinga Channel) with respect to its overall area (< 5% of total perimeter). The high exchange (and mixing) rate between Katonga Bay and the lake is clearly seen in the quick reduction of specific conductance over distance. This differs significantly from the spatial distribution of specific conductance in Nyanza Gulf.
As mixing between the river borne CDOM rich waters and the open lake waters is favoured in Katonga Bay, the short residence time does not favour photodegradation within the bay. This is further inhibited by the high attenuation of solar UV radiation (K313) in the inner Bay (due to high concentrations of dissolved and particulate organic matter) further reducing overall UV dose of the CDOM rich water. The result is that conservative mixing, rather than photodegradation appears to be the dominating mechanism in modifying the optical properties (absorption) of CDOM in Katonga Bay. In such conditions, the organic matter released into the bay from the Katonga River and surrounding wetlands does not undergo major modification before it enters into the open lake. Using the average flowrate of the Katonga River (5.1 m3 sec− 1, COWI 2002) and roughly estimating dissolved organic carbon concentrations from CDOM absorption (Mazzuoli et al., 2003), it is possible to estimate that approximately 7 × 103 kg of dissolved organic carbon enter the bay each day, with a wide variation in relation to season floods. This CDOM is then released, relatively unmodified, into Lake Victoria. It should be noted that such a rough calculation will underestimate the overall flowrate of organic carbon as we do not consider the release of organic carbon from papyrus wetlands (through diffusion and wave action) located in other parts of the Bay.
Nyanza Gulf presents different conditions, both in the behaviour of CDOM and specific conductance. Due to its morphology and extension, there is a reduced exchange between the Gulf and the Lake, in comparison to Katonga Bay. This reduced exchange (increased residence time) and higher penetration depth of solar UV radiation create conditions in which photodegradation may play a relatively important role in modifying the optical properties of CDOM. To better understand such modifications, we modelled the combined effects of conservative mixing and photobleaching of CDOM in relation to distance from the riverine sources, the Sondu and Nyando rivers, using the measured photobleaching rates. For the photo-induced loss of absorption, we calculated a hypothetical reduction in CDOM absorption (A365) in relation to distance from the riverine sources (residence time),Figure 5) it was possible to compare the roles of conservative mixing and photodegradation at different distances from the river sources (Figure 5). Differences between the model and the measurements can be associated with the shortcomings of our modelling approach, in particular that CDOM sources are limited to river/wetlands in the inner Gulf and CDOM sinks are limited to photodegradation. Further information on CDOM sources from photoplankton/macrophyte production and secondary sinks (microbial degradation, flocculation) within the Gulf need to be considered to refine the estimate of the fate of organic carbon.
Using the average water flowrate of the Sondu and Nyando rivers and estimating dissolved organic carbon concentrations from CDOM absorption (Mazzuoli et al., 2003), it is possible to estimate that approximately 17 × 103 kg of dissolved organic carbon are released into the Gulf each day. However, due to the relatively limited mixing between the Gulf and the lake, photodegradation processes are allowed to modify CDOM absorption (loss of chromophores) before Gulf waters enter the lake. Of the 17 × 103 kg of chromophoric dissolved organic carbon released, approximately one-third (7 × 103 kg) reach the open Lake. The remaining two-thirds are lost due to sinks.
The comparison between Nyanza Gulf and Katonga Bay points to major differences in the roles that bays play on Lake Victoria. Semiclosed bays such as Nyanza Gulf (and to a lesser degree Murchison Bay) provide conditions where riverine and wetland released dissolved organic matter have sufficient time to undergo degradation mechanisms before they enter the open lake. Open bays such as Katonga, Nsonga or Entebbe Bay, Karungu Bay, Muhuru Bay release largely unmodified CDOM into the Lake. Mixing, residence time and UV penetration depths needs to be investigated in major bays such as Speke Gulf and Murchison Bay to better understand the dynamics of sources and sinks of dissolved carbon in Lake Victoria.
This research was supported by the European Commission RTD INCO programme (ICA4CT2001-10036) and the Italian Interuniversity Consortium CSGI. We thank our colleagues at the Kenyan Marine Fisheries Research Institute and the Ugandan Department of Water Resources Management for their collaboration in data gathering and analysis. We thank R.T. Heath for his constructive comments to the manuscript.