A 14,000-Year Oxygen Isotope Record from Diatom Silica in Two Alpine Lakes on Mt. Kenya

See allHide authors and affiliations

Science  22 Jun 2001:
Vol. 292, Issue 5525, pp. 2307-2310
DOI: 10.1126/science.1059612


Oxygen isotopes are sensitive tracers of climate change in tropical regions. Abrupt shifts of up to 18 per mil in the oxygen isotope ratio of diatom silica have been found in a 14,000-year record from two alpine lakes on Mt. Kenya. Interpretation of tropical-montane isotope records is controversial, especially concerning the relative roles of precipitation and temperature. Here, we argue that Holocene variations in δ18O are better explained by lake moisture balance than by temperature-induced fractionation. Episodes of heavy convective precipitation dated ∼11,100 to 8600, 6700 to 5600, 2900 to 1900, and <1300 years before the present were linked to enhanced soil erosion, neoglacial ice advances, and forest expansion on Mt. Kenya.

Paleoclimate records from tropical regions are essential to understanding past changes in Earth's climate system, equator-pole linkages (1), and the sensitivity of tropical regions to future climate change (2). High-resolution oxygen isotope records from Andean ice cores reveal the response of tropical mountain climates to global forcing during the Last Glacial Stage and Holocene (2). On Mt. Kenya, however, ice-core studies of the rapidly retreating mountain glaciers have proved unsuccessful because of meltwater percolation (3,4). As an alternative, 18O measurements can be made on diatom silica (δ18Odiatom) from the sediments of high-altitude lakes. A 3000-year δ18Odiatom record obtained from Hausberg Tarn [4370 m above sea level (asl)], a glacier-fed lake on the northwest flank of Mt. Kenya (5), contained short-term minima in δ18Odiatom attributed to glacier melting and longer term minima attributed to increased water temperatures (5,6), in contrast to the positive relation assumed between δ18O and air temperature in ice-core studies (7). Here, we present a 14,000-year δ18Odiatom record compiled from two adjacent but hydrologically independent tarns located on the climatically sensitive northeast flank of Mt. Kenya. Their catchments have not been glaciated during the last 14,000 years (6).

Small Hall Tarn (SHT; 0°9′S, 37°21′E, 4289 m asl) is a topographically closed lake (0.5 m deep in January 1996) in the Gorges Valley. Simba Tarn (ST; 0°9′S, 37°19′E, 4595 m asl) is an open lake near the head of the Gorges Valley; in January 1996 it had a maximum depth of 5 m and an electrical conductance of just 18.9 μS cm−1. Precipitation in the summit zone of Mt. Kenya falls mainly from March to May and in October and November, although some occurs in every month. Total precipitation declines with altitude above the tree line (∼3000 m asl) and is <900 mm a−1above 4500 m asl, where it mainly falls as snow (3). The main moisture source is the southwest Indian Ocean, because of southeasterly airflow in the lower troposphere during the rainy seasons (3, 8). Mean annual temperatures are close to freezing at both sites (+1.3°C at SHT, –0.7°C at ST), with a monthly variability of ±2°C but a large diurnal range (10° to 20°C) (3). Terrestrial plant cover is moderate (SHT) to very sparse (ST) at present.

Piston cores raised in 1996 from SHT and ST were dated by accelerator mass spectrometry, yielding 12 14C dates for SHT and 1014C dates for ST (9). They were analyzed for volume magnetic susceptibility, diatoms, pollen, and green algae (SHT only) (10). Oxygen isotopes were measured on cleaned diatom silica from 72 levels in ST and 50 levels in SHT (11) (Figs. 1 and2). Benthic, circumneutral, freshwater taxa, notably small Fragilaria spp., dominated the diatom assemblages. Changes in the diatom floras are coincident with isotope changes, but we do not observe systematic isotope-diatom species relations.

Figure 1

Small Hall Tarn: summary of volume magnetic susceptibility, diatom analysis, pollen influx, and δ18Odiatom data. Pollen taxa are grouped according to their modern distribution: local (Gramineae, Cyperaceae,Alchemilla, and Compositae), ericaceous belt, and montane forest (including dry, moist, and tree-line species). The14C date in brackets was excluded from the age-depth model.

Figure 2

Simba Tarn: summary of volume magnetic susceptibility, diatom analysis, and δ18Odiatom data. The 14C date in brackets was excluded from the age-depth model.

Strong agreement between the multiproxy data from SHT and ST enables us to identify two contrasting paleoenvironmental states. The first, characterized by stable, enriched δ18Odiatomvalues (+27 to 37 per mil in SHT and +33 to 34 per mil in ST), corresponds to zones 1, 3, 5, and 7. Diatom-species richness was generally greatest in these zones, particularly in the late-glacial sediments of SHT (zone 7). Here Brachysira brebissonii, a cold-water, oligotrophic species (12) found in the highest tarns on Mt. Kenya today, was codominant alongside the ubiquitousFragilaria spp. Diversity was also high in zone 1 of Simba Tarn, suggesting an expansion of epiphytic niches available to the diatoms. Green algae, including Pediastrum andScenedesmus, were most abundant during the periods of strongest isotopic enrichment in SHT.

The second paleoenvironmental state, characterized by more variable, strongly depleted δ18Odiatomvalues, corresponds to zones 2, 4, and 6. In SHT, δ18Odiatom values of <20 per mil occurred in all three zones; similar negative excursions are also found in ST, but only the changes in zone 4 are of comparable amplitude. High magnetic susceptibility values signifying allochthonous sediment influx are strongly correlated with isotopic depletion and low-diversity,Fragilaria-dominated diatom assemblages. Increased erosion of partially vegetated slopes may have created the turbid, disturbed habitats that Fragilaria readily occupy.

Influxes of pollen from terrestrial plants reflect changes in three major environmental variables: precipitation, temperature, and (at least during the late glacial period) atmospheric CO2pressure (13). Even-numbered isotope zones (depleted δ18Odiatom values) correspond to modest increases in the influxes of local pollen (Gramineae, Cyperaceae,Alchemilla, and Compositae) together with montane-forest taxa transported from lower altitudes. Enhanced graminoid, shrub, and forest cover is consistent with wetter and/or warmer conditions.

During the last 14,000 years, four major negative shifts in δ18Odiatom can be traced from basin to basin (Fig. 3). First, from 11.1 to 8.6 thousand years ago (ka), a decrease of 13 per mil was observed in SHT and 3 to 4 per mil in ST. Second, two abrupt δ18O minima occurred between 6.7 and 5.6 ka. The amplitude of these isotopic excursions was 18 per mil in SHT and 14.2 per mil in ST. Third, a decrease in δ18Odiatom occurred between 2.9 and 1.9 ka in all three isotope records from Mt. Kenya (Fig. 3). In SHT, δ18Odiatom declined sharply by 16.6 per mil after 3.1 ka, whereas in ST this isotopic shift took place in two steps, ∼3.4 ka and ∼2.8 ka, giving a composite decrease of 2.5 per mil. In Hausberg Tarn the isotope signal was noisier, but the largest depletion of 2.3 per mil occurred at ∼2.5 to 2.4 ka (5). Finally, the uppermost samples in SHT and ST show a substantial decrease of 3 per mil and 8 per mil, respectively, within the last 1.3 ka.

Figure 3

Isotope data from Simba Tarn, Small Hall Tarn, and Hausberg Tarn (5). These data are compared to lake levels in the Ziway-Shala Basin, Ethiopia, derived from 14C-dated shorelines (32).

The similarity of the ST, SHT, and Hausberg Tarn isotope curves indicates underlying climate forcing modified by local hydrological factors. At tropical stations, the δ18O values of monthly rainfall (δ18Oprecip) exhibit a far stronger correlation with total precipitation (the “amount effect”) than with air temperature (14, 15). By analogy with the Rayleigh-distillation model developed to explain seasonal variations in the isotopic values of surface snow on tropical ice caps (16), the main factors governing the isotopic composition of tropical alpine lakes are (i) the δ18O value of the water vapor over the source region; (ii) the progressive rainout of the heavy isotope (18O) along the air-mass trajectory from the source region, which is partially offset by recycling from the surface and evaporation of falling raindrops; (iii) orographic uplift; (iv) uplift in convective shower clouds; and (v) surface processes. In the case of lakes, the latter are dominated by evaporative enrichment of 18O in lake water. Formation of diatom silica results in fractionation of the oxygen isotopes by around –0.2 to –0.5 per mil °C−1 (17). Moreover, an increase in 18Odiatom of 3 to 10 per mil has been reported during early diagenesis in ocean sediments (18).

Although insufficient data are available to quantify all these effects, the most important causes of isotopic variation can nevertheless be identified. General circulation model simulations for the Last Glacial Maximum suggest that the δ18O values of rainfall (δ18Oprecip) at low altitudes in the tropics were probably 1 to 3 per mil higher than today (19,20): The net effect of a 1.6 per mil increase in the18O content of mean ocean water (21) is the lowering of tropical sea surface temperatures (SSTs), resulting in a smaller temperature gradient between the vapor source and the site of condensation, and hence the weakening of the African and Asian monsoons (reduced amount effect). The same factors would have contributed to the high values of δ18Odiatom (maximum +36.7 per mil) recorded on Mt. Kenya during the late glacial period (14 to 11.2 ka).

Unfortunately, the modern isotopic composition of precipitation on Mt. Kenya is unknown, apart from spot dry-season measurements of ∼ +0 to +2 per mil at 4200 m asl (5). However, δ18Oprecip values at Kericho (2130 m asl) and Muguga (2070 m asl) in the Kenya Highlands are ∼ –4.4 and –3.8 per mil, respectively (14), the most negative isotopic values occurring during the wet seasons. The monthly δ18Oprecip values for these stations (22) lie close to the global meteoric line, indicating that precipitation formation occurs under isotopic-equilibrium conditions, with minimal evaporation of rainfall (23). For δ18O, the impact of orographic uplift is ∼ −2.7 to −2.9 per mil km−1 in East Africa (24). This suggests that δ18Oprecip at our two study sites averages around −10 to −11.5 per mil. These estimates are compatible with spot measurements of –9 per mil on river runoff from 4000 m asl (24), –8 per mil on surface snow from the Lewis Glacier (4), –6 to –7 per mil on Hausberg and Oblong Tarns (open lakes fed by glacial meltwater), and +1 per mil on Naro Moru Tarn (a closed lake) (5).

During wetter intervals in the past, the mean isotopic values of lake water would have been lowered by increases in precipitation (14,16, 25) and cloud height [giving lower cloud-top temperatures (23, 26)], strongly reinforced by decreased evaporation and possibly by lake overflow (in the case of SHT). Note, however, that any reduction in water temperatures resulting from increased cloud would have partially counteracted these effects through its impact on isotopic fractionation by diatoms.

A comparison of the Holocene variations in δ18Odiatom (Fig. 3) with alkenone SST estimates for the southwest tropical Indian Ocean, the source of much of the precipitation on Mt. Kenya, shows that the isotopic minima around 9 ka and at 6.9 to 5.8 ka corresponded to high SSTs (27). At present, heavy seasonal rainfall totals over the eastern Kenya Highlands are strongly linked to positive SST anomalies over the tropical South Atlantic and Indian Oceans, as well as to El Niño–Southern Oscillation events (28). The heavy rains of 1961–1962 provide a modern analog for the abrupt, high-amplitude δ18Odiatomminima. In November 1961, for example, the precipitation on the footslopes of Mt. Kenya exceeded 275% of normal (28) as a result of onshore flow from a large area of anomalously warm SSTs in the western Indian Ocean. Evaporation was also greatly reduced by dense cloud (29). A 145-year tree-ring study from Narok Mau, Kenya, confirms a 30 per mil decrease in mean δD, which is highly correlated with δ18O (24) from 1953–1958 to 1959–1963 (30).

The importance of moisture balance in causing the δ18Odiatom minima is supported by the pollen evidence for wetter and/or warmer conditions (Fig. 3). Notwithstanding a modest increase in plant cover, unusually heavy precipitation may have led to severe erosion of exposed volcanic soils on Mt. Kenya (31), resulting in the magnetic-susceptibility peaks. Our data suggest that anomalously heavy snowfall on the peaks of Mt. Kenya may contribute to the neoglacial ice advances dated >5.7 ka, 3.2 to 2.3 ka, and 1.3 to 1.2 ka (6). Lake-level curves from the East African–South Asian monsoon region (32, 33) support our climatic interpretation of the δ18Odiatom data (Fig. 3), as does pollen evidence for generally wetter and warmer conditions in Kenya at ∼6.8 ka (34). Environmental changes on Mt. Kenya are therefore symptomatic of the same climatic-forcing mechanisms that affected low-altitude tropical areas.

We conclude that centennial- to millennial-scale fluctuations in the18O content of diatom silica from alpine lakes on Mt. Kenya primarily reflect variations in moisture balance and cloud height, driven by SST anomalies. Hence, they provide a valuable new data source to supplement the sparse and rapidly deteriorating (7) isotopic archives in tropical glaciers.

  • * Present address: Scottish Agricultural College, Crichton Royal Farm, Dumfries DG1 4SZ, UK.

  • Present address: Department of Geography, University of Newcastle, Newcastle-upon-Tyne NE1 7RU, UK.


Stay Connected to Science

Navigate This Article