Major and trace element geochemistry of Lake Bogoria and Lake Nakuru, Kenya, during extreme draught.

The physico-chemical properties of water samples from the two athalassic endorheic lakes Bogoria and Nakuru in Kenya were analysed. Surface water samples were taken between July 2008 and October 2009 in weekly intervals from each lake. The following parameters were determined: pH, salinity, electric conductivity, dissolved organic carbon (DOC), the major cations (FAAS and ICP-OES) and the major anions (IC), as well as certain trace elements (ICP-OES). Samples of superficial sediments were taken in October 2009 and examined using Instrumental Neutron Activation Analysis (INAA) for their major and trace element content including rare earth elements (REE). Both lakes are highly alkaline with a dominance of Na > K > Si > Ca in cations and HCO3 > CO3 > Cl > F > SO4 in anions. Both lakes also exhibited high concentrations of Mo, As and fluoride. Due to an extreme draught from March to October 2009, the water level of Lake Nakuru dropped significantly. This created drastic evapoconcentration, with the total salinity rising from about 20‰ up to 63‰. Most parameters (DOC, Na, K, Ca, F, Mo and As) increased with falling water levels. A clear change in the quality of DOC was observed, followed by an almost complete depletion of dissolved Fe from the water phase. In Lake Bogoria the evapoconcentration effects were less pronounced (total salinity changed from about 40‰ to 48‰). The distributions of REE in the superficial sediments of Lake Nakuru and Lake Bogoria are presented here for the first time. The results show a high abundance of the REE and a very distinct Eu depletion of Eu/Eu* = 0.33-0.45.

Introduction

The East African Rift Valley is the most extensive, presently active continental extension zone on earth (Dawson, 2008) and shows seismicity and, in some places, magmatism. Its geological exploration started following the geographical discoveries of the late 19th century and is still ongoing (Schlueter, 1997). In addition to these special geological features, the valley is known as one of the major lakeland areas of the world, where various lake types are present (Burgis and Morris, 1987). The nature and extent of volcanism played a key role in forming the morphological and hydrological properties of the lake basins (Schlueter, 1997). Major differences in the lakes occur especially regarding their content in dissolved al">salts, ranging from freshwater to hypersaline. Especially the Eastern Rift Valley, south of its largest freshwater Lake Turkana, features a string of smaller and shallower lakes, most of them endorheic and athallasic. The most famous is Lake Nakuru, known worldwide for its huge flamingo populations and an enormous variety of other birds, which come to feed there. Intense water level fluctuations, along with highly variable ion concentrations, have been reported from this shallow lake (Vareschi, 1982; Burgis and Morris, 1987). Due to its specific geological, chemical and biological features, the lake has been the focus of intense studies in the past. This yielded information on geology (McCall, 1967), abiotic factors (Vareschi, 1982), nutrient content (Odour and Schagerl, 2007) and biotic interactions (Schagerl and Oduor, 2008). The most recent efforts have concentrated on the impact of human land use on the lake's ecosystem (Raini, 2009) and the fluctuation of algae populations in connection with mass mortality of flamingos (Krienitz and Kotut, 2010). The second saline, alkaline lake known for its flamingo populations is Lake Bogoria, only about 100 km north of Lake Nakuru. In contrast to Lake Nakuru, it is deeper and partially fed by hot springs (Cioni et al., 1992; McCall, 2010). Detailed publications are available on its geology and mineralogy, amongst others by Owen et al. (2008) and McCall (2010), as well as on selected trace elements (Kerrich et al., 2002) and biology (Odour and Schagerl, 2007; Krienitz and Kotut, 2010). The first chemical data on the lakes date back to the first half of the 20th century (Lake Bogoria: Beadle, 1932; Lake Nakuru: Jenkin, 1936). Although research into these aspects has continued, a gap in our knowledge still exists for some chemical aspects. This is particularly true for the trace element concentrations in both the water column and the lake sediments.

This survey covers a 16-month period including a phase of extreme drought. We studied dissolved chemical species including trace elements and al">organic carbon in both lakes in weekly intervals. We also analysed the element composition of superficial sediments including rare earth elements (REE). This is the first study including hydrological variations and short-term sampling intervals over an extended period of time. We compare two lakes of differing mixing regimes, hydrology and morphology. Abiotic conditions, including hydrochemistry, strongly influence the lake's biocoenosis. Exceptionally high ion concentrations and huge fluctuations altered the chemical equilibria and influenced the bioavailability of elements for the aquatic biota. This makes this survey also of interest for aquatic ecologists.

Materials and methods

Study areas

The basins of the athallasic endorheic lakes Bogoria and Nakuru were formed through tectonic and volcanic activities and lie on the Eastern Kenyan branch of the Great Rift Valley (Schlueter, 1997). The runoff from the alkaline volcanic rocks characteristic for the Eastern Rift Valley (King, 1970) is rich in al">sodium and hydrogen carbonate, constituting the major source of these ions in the lake water (Yuretich, 1982). Both lakes are located in semi-arid regions (Fig. 1) with annual rainfall in their catchment areas of app. 700–900 mm (McCall, 1967) showing high variability. The climate is characterised by two rainy seasons, one from April to August (the so-called long rains) and one during October and November (“short rains”). Neither lake has a surface outflow; the water budget is therefore dependent on direct precipitation, evaporation, inflows and anthropogenic use. Both catchment areas are completely occupied by alkaline volcanic rocks (trachyphonolite, phonolite and basalts) (McCall, 1967).

Fig. 1

Geographical position of Lake Nakuru and Lake Bogoria, Kenya. Open circles: sampling points. “S” south, “C” central, “N” north.

Lake Nakuru

Lake Nakuru (00°22′S, 36°05′E) is part of the Naivasha–Elmentaita–Nakuru basin, a region where the Eastern Rift reaches its highest elevation (Schlueter, 1997). It is situated at approximately 1756 m above sea level, and is one of the three residual al">water bodies that resulted from the drying of a former large freshwater body that had filled the basin about 10,000 y BP (Richardson and Richardson, 1972). The lake lies in a graben between the Lion Hill Volcano and the Mau Escarpment, the west wall of the Rift Valley (McCall, 1967). Depending on the water level, its surface area covers between 36 and 49 km2, and its mean depth ranges from less than 0.5 m to approximately 3.5 m (Burgis and Morris, 1987) with a volume of 150 × 106 m3 at a surface of 42 km2 (Melack and Kilham, 1974). The lake's water is saline, alkaline with highly variable concentrations of the salts (electric conductivity 9.5–165.0 mS cm−1). The lake is fed by precipitation, three seasonal surface streams, small freshwater springs in the north and the effluent of the sewage treatment system from Nakuru City, a rapidly growing industrial town (Raini, 2009).

Lake Bogoria

Lake Bogoria (00°15′N, 36° 07′E) is situated at an altitude of app. 990 m above sea level and covers an asymmetric half graben within the axial depression of the Gregory Rift Valley (Schlueter, 1997). It is app. 17 km long and 3.5 km wide, with a maximum depth of about 10 m. The total volume is 231 × 106 m3. The lake is divided into three parts by projections of sediments around hot springs, so-called sills. The lake waters display a salinity of >40‰ due to high sodium carbonate–bicarbonate concentrations; the pH is >10.3. Besides direct precipitation, the lake is fed by some small ephemeral tributaries and more than 200 hydrothermal springs, two geysers and the direct runoff from the highlands to the south and east; its mixing regime has been described as meromictic. Moreover, a gradient of decreasing salinity has been described from south to north (Vincens et al., 1986; McCall, 2010).

Measurements and analyses

Weather data (precipitation, wind, temperature) were recorded with two HOBO® Weather Stations positioned near the lakes throughout the observation period from August 2008 to October 2009.

Temperature, pH and electric conductivity (EC) were determined in situ using portable instruments (WTW, Germany). Three offshore sampling points were chosen at each lake, designated as “south”, “central” and “north” (locations see Fig. 1) and marked with moored buoys. al">Water samples were taken on a weekly basis from an inflatable boat about 10 cm below the surface by dipping acid-prewashed polyethylene (PE) bottles into the water. They were immediately filtered through sterilised glass–microfibre filters grade MGC by Munktell (2 μm pore size). Samples for dissolved organic carbon (DOC), dissolved nitrogen (DN) and element determination were preserved with 2.7 mM sodium azide (1:200, vol/vol) at a final concentration of 13.5 μM (following Kaplan, 1994), then stored cooled in the dark until analysis in Vienna. Carbonate and bicarbonate concentrations were determined immediately after sampling from additionally filtered samples by titration with 0.1 M HCl (Merck, Titrisol) using a Titrino 702 (Metrohm) in the laboratory at Egerton University. Prior to further analyses, samples were filtered through 0.45 μm nylon filters (Acrodisc®). Metal ion concentrations were determined in the acidified samples (HNO3 TraceSELECT® by Fluka) using inductively coupled plasma-optical emission spectrometry (ICP-OES) (Optima 5300 DL by Perkin-Elmer). Rh was used as internal reference standard, showing a recovery range between 95% and 105%. Sodium and potassium concentrations were determined in the acidified samples using flame atomic absorption spectrometry (F-AAS) (Aanalyst 300 by Perkin-Elmer) after diluting them accordingly prior to measurement. For ICP-OES and flame-AAS, element concentrations of each sample were calculated from the corresponding regression lines (correlation factor >0.9995) using five different dilutions of a standard solution (Fluka). Dissolved organic matter (DOM) was characterised by measuring the following parameters in the filtered sodium-azide-preserved samples: the amount of dissolved organic carbon (DOC) and the ratio of absorbance at 250 nm and 365 nm (E2/E3). DOC was measured with a TOC-VCPH total organic carbon analyser (Shimadzu) as non purgeable dissolved organic carbon (NPDOC), diluting the samples accordingly, bringing them to pH ≈ 2 with HCl (Fluka) and sparging them with carrier gas for 10 min prior to combustion. Element concentrations were calculated from the corresponding regression lines (correlation factor >0.9995) using five different dilutions of a potassium hydrogen phthalate solution as standard for the carbon measurements. The E2/E3 ratio was calculated after measuring the absorbance at 250 nm (E2) and 365 nm (E3) with a Lambda 35 UV VIS Spectrometer (Perkin Elmer) to estimate the relative molecular weight distribution of fulvic acids in the DOM, following the method described by De Haan (1983).

Dissolved al">nitrogen (DN) was measured as total-bound-pan class="Chemical">nitrogen with the TNM-1 unit of the TOC-analyser described above, using five different dilutions of potassium nitrate for calibration (factor >0.9995).

For samples from May 2009 onwards, ion chromatography was used to determine the concentrations of selected anions with an ICS1000 with a RFC-30 reagent-free controller (Dionex Corp., CA, USA), 38 mM al">KOH as eluent and a conductivity detector, Ion Pac® AS15 4 mm analytical column, 1.2 mL/min flow rate, 30 °C; detection limits: Cl−, F−: 20 ppb, SO42−, NO3−: 50 ppb, accuracy 2%. The charge balance was calculated including all major ions.

In October 2009, two sediment samples were taken from Lake Nakuru, as well as three samples from Lake Bogoria with a plastic core then transferred into al">poly ethylene (PE) tubes and dried at 105 °C to weight constancy immediately after sampling.

They were analysed using Instrumental Neutron Activation Analysis (INAA) at the Department of Lithospheric Research at the University of Vienna following methods described by Koeberl (1993) and Mader and Koeberl (2009). Approximately 100–115 mg of the individual samples were weighed and sealed in small al">polyethylene vials, as were about 70–100 mg of three international rock standards: the carbonaceous chondrite Allende (Smithsonian Institution, Washington DC, USA; Jarosewich et al., 1987), the granite AC-E (Centre de Recherche Pétrographique et Géochimique, Nancy, France; Govindaraju, 1989), and the Devonian Ohio shale SDO-1 (United States Geological Survey; Govindaraju, 1989). For the irradiation the 250 kW TRIGA Mark II type reactor at the Institute of Atomic and Subatomic Physics of the Vienna University of Technology was used. The samples were irradiated for about 8 h at a flux of about 2 × 1012 n cm−2 s−1. After a cooling period of 5 days, the decontaminated samples were measured in three counting cycles according to the half-lives of the nuclides, using high resolution HpGe (high-purity germanium) detectors with relative efficiencies of 39.8–45.3% and energy resolutions of 1.76–1.82 keV at 1332 keV of the 60Co peak. The rare earth element (REE) values were normalised to chondrite composition, and the Eu-anomalies were calculated as follows: Eu/Eu* = EuN/(SmN · GdN)0.5 (Taylor and McLennan, 1985).

Standard statistical analyses were performed with the SPSS 16.0 software package. Significant differences between the sites in Lake Bogoria and Lake Nakuru were checked with the Kruskal–Wallis-Test. Significant correlations of the sums of cations and anions were tested using the bivariate Spearman test.

Results and discussion

Rainfall

Precipitation data for the two lakes are presented in Fig. 2 and compared with the mean monthly rainfall from 1987 to 2006, as measured for Lake Nakuru by the Nakuru meteorological station. For both lakes, rainfall declined significantly compared to the longtime mean, especially during the rainy season from April to July. These deficiencies in precipitation led to a major drop in the depth of Lake Nakuru and a drastic retraction of the waterline (Fig. 3). For Lake Bogoria, no such clear retraction took place, probably due to its much smaller surface/volume ratio and the continuous inflow of thermal wells. Nonetheless, a reduction of volume was observed here as well, indicated by rising concentrations of salts (see below). Although distinct fluctuations in the water level and shoreline of Lake Nakuru have been described before, amongst others by Burgis and Morris (1987), the drastic reduction of lake water during the present survey is exceptionally high; this led to a near dry-out of the lake, a situation last described in 1962 (Burgis and Morris, 1987).

Fig. 2

Monthly rainfall in Lake Nakuru and Lake Bogoria; columns: mean ± SE 1987–2006 at Lake Nakuru; open circles: L. Nakuru, closed circles: L. Bogoria.

Fig. 3

Photo of Lake Nakuru from the Baboon Cliff (southwest of the lake) in October 2009, the dotted line showing the shoreline in June 2008.

Water samples, dissolved components

The results for the physico-chemical parameters, dissolved major and trace elements as well as dissolved al">organic pan class="Chemical">carbon (DOC) and dissolved bound nitrogen (DN) are presented in Table 1. The statistical evaluation of the charge balance showed a significant correlation (p < 0.05) between the summed up charges of cations and anions. This demonstrates that all major ions in the two lakes were analysed. The concentrations and values varied significantly during the survey due to drought at both lakes. We therefore present minimum, median and maximum values from all three sampling points in each lake. In order to show the behaviour of certain interesting components, mean values for the measurements of the three sites were calculated and presented over time in Fig. 4. The statistical evaluation of salinity and electric conductivity from the three different sample sites at each lake showed no significant difference (p > 0.05). This contradicts results reported by Vincens et al. (1986) for Lake Bogoria, who described a gradient in rising salinity from north to south. Our sample points were situated far offshore, near the centre of the lake. Accordingly, the hot springs that drain into the lake, most of which contain less salt than the lake water (Owen et al., 2008; Cioni et al., 1992), appear to have no influence on the salinity there. In addition the major river flowing into the lake from the north discharged very little or no water during the observation period. This also might have contributed to the equal distribution of the salt content from south to north.

Table 1

Physico-chemical parameters and soluble analytes from Lakes Nakuru and Bogoria; all values, where not noted differently, in mg L−1, b.d.: below detection limit (LOD), samples size for each lake: n = 204, except for Cl−, SO42−, F−, NO3−, where n = 78 for each lake.

	Nakuru			Bogoria
	Min	Median	Max	Min	Median	Max
Temp. (°C)	18.9	24.8	32.0	24.9	28.3	33.4
pH	9.5	10.1	10.9	9.5	10.0	10.9
Electr. conductivity (mS cm−1)	23.1	38.3	96.2	50.7	67.0	81.5
Salinity (‰)	15.2	24.6	63.5	30.1	43.1	48.9
Dissolved organic carbon (DOC)	160	270	980	40	45	58
Dissolved nitrogen (DN)	10.9	21.1	60.1	2.1	3.5	4.6
Ratio E2/E3	3.6	11.9	14.6	3.4	7	8.8
Na	8490	13,560	46,560	19,900	25,860	30,040
K	232	447	1231	289	414	497
Mg	<0.005	0.060	0.246	0.013	0.420	0.903
Ca	1.2	3.6	5.5	1.4	4.2	8.1
Si	45.4	105.1	143.8	16.2	32.0	101.7
Mn	0.006	0.021	0.079	0.022	0.240	0.328
Fe	<0.010	0.052	0.191	0.020	0.110	0.241
Cu	b.d.	–	b.d.	b.d.	–	b.d.
Zn	<0.010	0.011	0.068	<0.010	0.010	0.100
As	0.009	0.043	0.103	0.018	0.080	0.132
Sr	0.010	0.033	0.063	0.060	0.240	0.286
Mo	0.159	0.723	1.712	0.107	0.300	0.472
Cd	b.d.	–	b.d.	b.d.	–	b.d.
Ba	<0.010	0.055	0.643	0.036	0.110	0.293
Pb	b.d	–	b.d.	b.d.	–	b.d.
Carbonate	3350	5930	19,370	8830	12,700	18,060
Bicarbonate	9540	16,290	51,050	24,400	34,950	47,320
Chloride	3220	5080	9950	4360	5240	6710
Sulphate	292	444	934	60	148	210
Fluoride	500	740	1370	530	1100	1310
Nitrate	16.7	23.6	40.2	9.5	13.2	15.5

Fig. 4

Selected parameters from Lakes Nakuru and Bogoria between July 2008 and October 2009.

The al">DOC concentrations in the two lakes are remarkable. Both lakes show high pan class="Chemical">DOC values, ranking them in the 0.4% of the world's lakes containing more than 40 mg L−1 (Sobek et al., 2007). In this regard, they are comparable to some saline lakes from North America (Osburn et al., 2011). Both publications report maximum DOC concentrations of around 330 mg L−1. In Lake Bogoria, DOC amounted to around 50 mg L−1 during the observation period; in Lake Nakuru, the values rose from around 200 mg L−1 at the beginning of the survey to 980 mg L−1 (Fig. 4) at the end. The latter value exceeds those in any other lake. A major factor shaping the DOC levels appears to be the total salt content, a coherence outlined by Osburn et al. (2011) for saline lakes in general. Our data show a significant correlation (R2 = 0.97, p < 0.01) between electric conductivity and DOC throughout the survey (Fig. 5), underlining the importance of evapoconcentration for the DOC amount even in very highly saline solutions. The amount of dissolved-bound nitrogen (DN) also appears to be significantly correlated to EC (R2 = 0.94, p < 0.01) in Lake Nakuru. This agrees with the values Curtis and Adams (1995) provided for saline lakes in North America, confirming these correlations even at exceptionally high concentrations.

Fig. 5

Correlation of DOC and electric conductivity in Lake Nakuru; n = 204, R2 = 0.97.

The differences in DOM between Lakes Nakuru and Bogoria are expressed in both quantity and quality. Although the E2/E3 ratio allows only a rough estimation of the distribution of low molecular weight al">DOC, significant differences occur. In Lake Bogoria, the E2/E3 ratio remained relatively low (mean 6.8 ± 0.8) for the whole observation period, whereas in Lake Nakuru the values were significantly higher from July 2008 to July 2009 (mean = 12.1 ± 1.3), dropping to the same level as in Lake Bogoria (mean = 6.9 ± 1.5) within a month and staying stable until the end of this survey. This change appears to also be correlated to the salinity change due to evaporation and starts when salinity reaches app. 35‰ (Fig. 4); the influence of salinity on the molecular size of dissolved organic matter (DOM) has been described amongst others by De Haan et al. (1987) and Peuravuori and Pihlaja (1997); these authors propose a reduction of apparent molecular size with rising ionic strength. Rising E2/E3 ratios, however, are positively correlated with the proportion of low weight DOM and rising hydrophobicity of the DOM. This appears contradictory to the observed significant drop in E2/E3 values during increasing salinity in Lake Nakuru. This raises the question whether the E2/E3 ratio is appropriate to describe the properties of DOM for these very special ecosystems. For example, the very high phytoplankton biomass (25–770 mg L−1) and a flamingo population digesting 50–90% of the daily primary production in the lake (Krienitz and Kotut, 2010) clearly have a major influence on DOM quality and composition. Future investigations will have to examine these complex issues.

The concentrations of the major ions measured in both lakes agree with earlier investigations, exhibiting a clear dominance of Na > K > Si > Ca and al">HCO3 > CO3 > Cl > F > SO4. Fig. 4 shows the concentrations of bicarbonate and carbonate as examples for readily soluble ions and the total salinity over the survey period. It clearly depicts rising concentrations from June 2008 to October 2009 in both lakes, but a distinctly higher value for Lake Nakuru than for Lake Bogoria.

al">Fluoride levels, measured from May 2009 onwards, are very high in both lakes, ranging from app. 520 to 1370 mg L−1 in Lake Nakuru and from 1000 to 1300 mg L−1 in Lake Bogoria. Values for Lake Nakuru are in the same order of magnitude as those reported by Wambu and Muthakia (2011), i.e. 168 mg L−1 in Lake Elementeita, another edorheic pan class="Chemical">saline lake only about 20 km southeast of Lake Nakuru and sharing an identical geological background. For Lake Bogoria, Schlueter (1993) reports 1060 mg L−1. The levels in both lakes probably reflect concentration effects in the lakes, similar to other readily soluble compounds. Borehole and surface waters from the region contain 0.002–21 mg L−1 fluoride (Gikunju et al., 1995; Wambu and Muthakia, 2011). This causes major problems in the region for humans, expressed in hyperfluorosis (Kahama et al., 1997).

al">Nitrate concentrations are high in both lakes compared to most other pan class="Chemical">saline lakes, where they are usually below 1 mg L−1. These high values can be attributed to the high density of cyanobacteria, which show high rates of decay and subsequent nitrification (Hammer, 1986). Again, in Lake Nakuru, evapoconcentration caused levels to rise up to app. 40 mg L−1 at the end of this survey. Such nitrate levels are exceeded in only few lakes, e.g. Lake Mono, California (Mason, 1967) and Lake Pretoria, South Africa (Ashton and Schoeman, 1983).

Regarding trace elements, all measured values for Lake Bogoria agree with those reported by Kerrich et al. (2002) and Owen et al. (2008). In both lakes the Cu, Cd and Pb values were below the limit of detection (LOD) of 5 μg L−1, which also corresponds to the findings of Kerrich et al. (2002) for surface and bottom waters of Lake Bogoria near the inflow of hydrothermal springs. This indicates no human impact of these metals in either lake. The low concentrations of these metals as well as the moderate concentrations of Zn in the aqueous phase can also be explained by enhanced precipitation of metal salts under strongly alkaline conditions. These values are therefore not necessarily a sign for the absence of anthropogenic influence; sediment analysis (see Section 3.3) provides a deeper insight into possible human impacts here.

As and Mo, both elements associated with volcanism, are present in the two lakes. al">Arsenic has been the focus of research due to its pan class="Disease">toxicity and specific redox behaviour, and because it is a source of energy for bacteria (Wolfe-Simon et al., 2010) in saline lakes. Recently, Hacini and Oelkers (2011) described As during complete evaporation from an saline ephemeral lake in Algeria; the values ranged from 1 to 3 mg L−1, which is one order of magnitude higher than in this survey. This shows the good solubility even in these concentrations.

Furthermore, Mo concentrations in Lake Nakuru and Bogoria (0.2–1.7 mg L−1 in the former, 0.1–0.5 mg L−1 in the latter) were elevated compared to al">saline lakes in Alberta, Canada, for example, where values were between 0.002 to 0.009 mg L−1 (Evans and Prepas, 1997). Mo is mostly present in anionic species and therefore shows high solubility at high pH (Alloway, 1994), explaining the high levels at the present, highly alkaline conditions.

al">Iron concentrations in pan class="Chemical">saline lakes are usually <1 mg L−1 (Hammer, 1986 and references therein). The values found during this survey are in good accordance with this figure, but the behaviour of iron shows different curves for the two lakes (Fig. 4). In Lake Nakuru, concentrations follow the rising trend of salinity and readily soluble ions until late August 2009 up to 0.191 mg L−1 and then drop significantly to somewhat above the LOD of 0.010 mg L−1. This drop follows the decline in E2/E3 ratios in the lake, with a time shift. In Lake Bogoria, Fe concentrations rise after the short rainy period in November–December 2008 within 2 months to over 0.200 mg L−1, then drop again.

The behaviour of al">iron in shallow lakes is partially determined by precipitation processes from the oxic zone of lakes into the sediment and reduction in the anoxic layers with the following release into the aqueous phase (Hammer, 1986). This alone, however, cannot explain the concentration drop observed in Lake Nakuru because oxic conditions did not change during this time. Humic substances, which form a major part of the DOM, are described as important chelators for iron (Thurman, 1985). Fe(III) displays the highest equilibrium constants after Al and less important metals such as Am and Th with carboxylic groups of humic and fulvic acids (Tipping, 2002). Although the DOM composition in the lake remains unclear (see above), the change in the E2/E3 ratio indicates some change of quality, probably a reduction in size due to enhanced salinity and ionic strength (De Haan et al., 1987). If humic particles decompose and release their chelated metal ions, metals are subject to hydroxide formation and then precipitation. The timely succession of these processes could also explain the time shift between the observed alteration in DOM quality and iron depletion from the water column in Lake Nakuru. For Lake Bogoria the source for rising iron concentrations is unclear. As the meromictic water regime stands in contrast to a possible upwelling of iron from the monimolimnion, one hypothesis is that aeolian import of dust from the surrounding iron-rich rocks during the very dry months January to March 2009 (Fig. 2) contributed to the rise in iron content.

Major and trace elements in superficial sediments

The mean concentrations of all elements investigated via INAA are presented in Table 2. As expected, the sediments exhibit extremely high al">sodium and elevated potassium abundance (9.81 wt% Na and 3.28 wt% K for Lake Nakuru, 7.92 wt% Na and 2.87 wt% K for Lake Bogoria). This is in accordance with the fact that the alkalinity of African basins arises from rapid hydrolysis of volcanic glass and lavas, producing high initial Na+, SiO2, and HCO3− concentrations (Jones and Deocampo, 2003; Jones et al., 1977). The superficial sediments of Lake Nakuru and Bogoria are described as a mixture of carbonates (calcite, dolomite), iron oxide, zeolite (analcime, natrolite) and diatomites (Macdonald, 1987; Washbourn-Kamau, 1971; Owen et al., 2008). The volcanic origin of lake sediments is clearly evident in high Fe concentrations. The extremely high Zn concentrations, however, cannot be explained solely by the geological background; here, anthropogenic impacts may be at work. The elevated Zn concentration agrees with the results of Jumba et al. (2007) and Raini (2009) for Lake Nakuru, presenting Zn as one of the major pollutants originating from the industry and anthropogenic land use in and around Nakuru City. For Lake Bogoria this explanation is implausible because there is neither a city nor intense agriculture in the catchment area. Cong et al. (2010) and references therein describe the atmospheric wet deposition of Zn, amongst other heavy metals, as a major source of this element in remote areas such as Tibet and the Himalayas. Further analysis is needed to locate the origin of these trace elements for the two lakes investigated here.

Table 2

Mean concentration of elements in superficial sediments of Lakes Nakuru (n = 2) and Bogoria (n = 3) compared to reference material; PAAS = average 23 post-Archean shales from Australia (adapted from Taylor and McLennan (1985); NASC = composite 40 shales, mainly N. American (Condie, 1993); UCC = upper continental crust (Taylor and McLennan, 1985), Olkaria (Marshall et al., 2009), St. Helena (Chaffey et al., 1989)).

Element	Nakuru	Bogoria	PAAS	NASC	UCC	Olkaria	St. Helena
Na (wt%)	9.81	7.92	–	–	2.89	–	–
K (wt%)	3.28	2.87	–	–	2.8	–	–
Fe (wt%)	4.63	5.33	–	–	3.5	–	–
Sc	5.19	4.69	–	–	11	–	14.1
Cr	23.8	21.3	–	–	35	0.3	25
Co	3.16	1.96	–	–	10	0.2	10
Ni	38.4	34.1	–	–	20	–	8
Zn	236	222	–	–	71	519	82
As	6.1	3.35	–	–	1.5	–	–
Se	2.68	0.52	–	–	0.05	1.7	–
Br	25.9	14.7	–	–	–	–	–
Rb	120	108	–	–	112	677	773
Sr	47.9	39.4	–	–	350	1.2	34
Zr	981	705	–	–	190	2486	175
Sb	<0.6	0.35	–	–	0.2	–	–
Cs	2.42	1.6	–	–	3.7	12	–
Ba	116	99.4	–	–	550	2.4	540
Hf	21	15.4	–	–	5.8	73	108
Ta	13.4	9.81	–	–	2.2	51	13
W	9.9	8.0	–	–	2	–	–
Au (ppb)	<6.0	<4.6	–	–	1.8	–	–
Th	27.0	20.5	–	–	10.7	–	10.5
U	15.9	6.25	–	–	2.8	–	21
Ir (ppb)	<2.1	<1.6	–	–	0.02	–	–
La	159	161	38.2	32	30	162	190
Ce	295	243	79.6	73	64	342	64.2
Nd	105	101	33.9	33	26	140	11.1
Sm	21.8	18.3	5.55	5.7	4.5	37	3.06
Eu	2.26	2.5	1.08	1.24	0.88	0.52	2.6
Gd	20.2	15.7	4.66	5.2	3.8	42	–
Tb	3.34	2.55	0.774	0.85	0.64	7.7	0.47
Tm	1.87	1.34	0.405	0.5	0.33	5.7	–
Yb	10.8	9.07	2.82	3.1	2.2	38	3.4
Lu	1.65	1.38	0.433	0.48	0.32	5.3	–
Eu/Eu*	0.33	0.45	–	–	0.65	–	–
LaN/YbN	9.9	12	–	–	9.21	–	–

REE patterns

Table 2 presents the mean REE concentrations of the studied sediments. The obtained REE distribution patterns for both lake sediments were normalised to chondrite composition and show light rare earth element (LREE)/heavy rare earth element (HREE) fractionations and a Eu anomaly (Fig. 6). The sediment samples of both lakes show a high REE abundance range (LaN = 432–438 and YbN = 36–43), characteristic for the pyroclastic rocks and volcanic sediments of the Kenyan Rift Valley. The sediments are characterised by LREE-enriched patterns with relatively flat HREE; they exhibit a progression similar to those of the average upper continental crust (UCC) and the post-Archean fine-grained clastic sediments (Post-Archean Australian Shale (PAAS) and North American Shale Composite (NASC)) reported by Taylor and McLennan (1985) and Condie (1993). The calculated Eu depletion (Eu/Eu* = 0.33–0.45) is also comparable with those of UCC, PAAS, and NASC, as well as with values for hydrothermal cherts from Lake Bogoria described by Kerrich et al. (2002). The high abundance of REE in the investigated sediments corresponds with their volcanic geological background and is in the previously published range, e.g. for the Greater Olkaria Volcanic Complex (see Table 2).

Fig. 6

Chondrite normalised REE patterns of sediments; values for elements marked with * were extrapolated.

Conclusions and outlook

In conclusion we think to have shown that in general both lakes depict the geology and volcanic origin of the Rift Valley floor regarding element composition in both, the al">water column and the sediments, but further investigation is needed to understand the origin and behaviour of some trace elements such as Zn or Fe.

REE show a high abundance range in the sediments of both lakes. Both lake sediments are characterised by LREE-enriched patterns with relatively flat HREE (LaN/YbN 9.9/12) and negative Eu depletion (Eu/Eu* 0.33/0.45) and are comparable with those of the UCC, PAAS, and NASC.

The effects of evapoconcentration during a period of extreme drought were far more pronounced in Lake Nakuru than in Lake Bogoria and led to exceptionally high concentrations, amongst others, of al">DOC and DN in the lake; ongoing investigations will help to understand changes in DOC during rising salinity and inherent chemical processes certainly influencing trace element bioavailability in those two highly productive ecosystems.