Content uploaded by Alfred Wuest
Author content
All content in this area was uploaded by Alfred Wuest on Feb 15, 2022
Content may be subject to copyright.
778
Limnol. Oceanogr., 49(3), 2004, 778–783
q
2004, by the American Society of Limnology and Oceanography, Inc.
Response of Lake Kivu stratification to lava inflow and climate warming
Andreas Lorke
1
Applied Aquatic Ecology, Limnological Research Center (EAWAG), CH-6047 Kastanienbaum, Switzerland
Klaus Tietze
PDT GmbH, Physik-Design-Technik, Sensorik und Consulting, D-29227 Celle, Germany
Michel Halbwachs
Universite´ de Savoie et Coordination de la Recherche Volcanologique, CNRS-INSU, F-73376
Le Bourget du Lac Cedex, France
Alfred Wu¨est
Applied Aquatic Ecology, Limnological Research Center (EAWAG), CH-6047 Kastanienbaum, Switzerland
Abstract
During the eruption of Nyiragongo Volcano in January 2002 about 10
6
m
3
of lava entered Lake Kivu. The high
concentrations of CO
2
and CH
4
dissolved in the deep waters of Lake Kivu raised serious concerns about a potential
gas outburst with catastrophic consequences for the population in the Kivu-Tanganyika region. Therefore, 3 weeks
after the volcanic eruption, we performed an ad hoc lake survey of the stability of the water column stratification.
Vertical profiles of temperature and turbidity revealed signatures of the lava, which had penetrated to 100 m depth;
however, there was no substantial warming or destratification of the gas-containing deep layers below. The deep
double-diffusive structures also remained unaltered. Based on these observations, we conclude that a thermally
driven gas outburst in Lake Kivu is not to be expected from future eruptions of comparable dimensions. In addition,
the recent measurements allowed for an update and gave new insight into the stratification and double-diffusive
mixing phenomena in Lake Kivu. A comparison with former measurements revealed a warming of the upper part
of the lake of up to 0.5
8
C within the last 30 yr, which could be attributed to climate variability.
The Nyiragongo Volcano (Democratic Republic of Congo,
East-Central Africa) erupted on 17 January 2002. Volumi-
nous lava flows, originating from different eruptive vents,
destroyed the central part of the city of Goma and about one
million m
3
of lava entered Lake Kivu, which is located about
18 km south of the Nyiragongo Volcano in the western
branch of the East African Rift Zone (Fig. 1). The lake cov-
ers a surface area of 2,400 km
2
and has a maximum depth
of 485 m. There were serious concerns about a thermally
driven limnic gas outburst due to the high concentrations of
dissolved CO
2
and CH
4
at great depth in the permanently
stratified water column (Deuser et al. 1973; Tietze et al.
1980). It was feared that the hot lava could warm up and
locally destabilize the density stratification in those deep lay-
ers and, thus, trigger a massive gas outburst. Such a gas
outburst in Lake Nyos (Cameroon) in 1986 (Kling et al.
1
To whom correspondence should be addressed. Present address:
Limnological Institute, University of Constance, Mainaustrasse 252,
D-78464 Konstanz, Germany (Andreas.Lorke@uni-konstanz.de).
Acknowledgments
We thank Cle´ment Mudaheranwa (Kigali, Rwanda) for the logis-
tic support and for courtesy of the digital map (Fig. 1). We are
grateful to J. J. F. de Vos (Bralirwa Heinecken, Gisenyi) forhis kind
assistance in the field. We also acknowledge Martin Schmid’s com-
ments on an earlier draft of this manuscript. The Humanitarian Aid
Office of the European Community (ECHO) funded the study. A.L.
was supported by EAWAG.
1987; Tietze 1992), most probably triggered by rock fall,
released a dense cloud of carbon dioxide that flowed through
the surrounding valleys and asphyxiated approximately
1,800 people (Freeth et al. 1990). Taking into account that
the total volume of dissolved gas in Lake Kivu is about three
orders of magnitude higher than in Lake Nyos (3
3
10
9
m
3
compared to 3
3
10
6
m
3
, respectively) and given the con-
siderably larger population density at Lake Kivu, a cata-
strophic disaster could not be excluded a priori. Paleolim-
nological evidence for such catastrophic events in the past
5,000 years was found in sediment cores from Lake Kivu
by Haberyan and Hecky (1987).
Critical for a potential gas outburst is the relative satura-
tion of the dissolved gases, which reaches a maximum at
about 270 m depth with 8% carbon dioxide and 43% meth-
ane saturation relative to the hydrostatic pressure (Deuser et
al. 1973; Tietze et al. 1980). The gases remain in the deep
waters of the lake because the seasonal convective mixing
of the surface boundary layer extends only to about 50 m
depth (Tietze et al. 1980). Below, the water is anoxic and
concentrations of dissolved carbon dioxide and methane in-
crease continuously with depth. Although the carbon diox-
ide, which is thought to be mainly of mantle origin, enters
the lake (in dissolved form) by groundwater inflow, the
methane is thought to be produced within the lake by an-
aerobic bacteria, which can use both acetate from decom-
posing organic matter and magmatic CO
2
as a carbon source
(Deuser et al. 1973; Tietze et al. 1980; Schoell et al. 1988).
779Lake Kivu stratification
Fig. 1. Map of Lake Kivu with sampling stations and the lo-
cation of the Nyiragongo volcano (18 km north of the city of
Goma). The two bold and numbered CTD stations in the center of
the northern basin are referenced within the text.
Fig. 2. Profiles of potential temperature, electrical conductivity
corrected to 20
8
C, and light transmissivity, LT, from a CTD cast in
the center of Lake Kivu northern basin (CTD Sta. 1 in Fig. 1).
Measurements
A short-term investigation was carried out on 7–10 Feb-
ruary 2002 to assess the potential of a gas outburst and to
investigate the impact of the lava inflow on the density strat-
ification of Lake Kivu. A SeaBird-19 conductivity–temper-
ature–depth (CTD) probe was used to measure vertical pro-
files of electrical conductivity, temperature, and light
transmissivity (at a wavelength of 660 nm along a light path
of 0.1 m). Profiles were measured off the city of Goma in
the vicinity of the lava inflow, as well as along a 26-km
transect from Goma to the island of Idjwi in the southern
part of the lake (Fig. 1).
Water samples were collected for calibration of the con-
ductivity measurements at 10, 25, 150, and 450 m depth,
using a 5-liter Niskin bottle; a remotely operated underwater
camera (ROV) was used to visually inspect the lava intru-
sions at different locations and depths.
Results
The general features of the CTD profiles are illustrated in
Fig. 2, showing a cast from the center of the northern basin
(Fig. 1). The stratification is characterized by a seasonal ther-
mocline within the uppermost 50 m depth and a permanently
stratified hypolimnion with increasing temperature stabilized
by increasing salinity with depth. Below 200 m, the density
stratification is further stabilized by increasing concentra-
tions of dissolved CO
2
and weakly destabilized by increasing
concentrations of dissolved CH
4
(Degens et al. 1973). The
salinity was estimated to vary between 1.15 g kg
21
at the
surface and 4.47 g kg
21
at 450 m depth, based on the ionic
composition of the water samples and following the proce-
dure described by Wu¨est et al. (1996). An interesting feature
of the overall stratification is the step-like structure of the
profiles within the permanently stratified part of the water
column, showing three distinctive mixed layers (200–250 m,
275–300 m, and 320–355 m) separated by strong gradients.
At least the lowest two mixed layers must have been actively
mixed very recently and do not exhibit any diffusion-in-
duced gradients of the measured parameters.
The light transmission has a constant bulk value of about
80% transmissivity (at 660 nm over 10 cm path length) and
strong peaks of higher turbidity (i.e., lower light transmis-
sivity) at each density gradient. However, based on our data
it is not clear whether these turbidity peaks are related to the
volcano eruption and subsequent lava inflow or whether they
are a rather typical feature of Lake Kivu, since no detailed
turbidity profiles were measured in the lake before. Lake
Kivu has no significant river inflows (only one outflow, the
Ruzizi River, which discharges into Lake Tanganyika); riv-
erine intrusions can be excluded as alternative sources of the
observed turbidity layers.
A signature of the lava inflow, however, is the thin layer
of slightly increased temperature at about 80 m depth (Fig.
2). The layer thickness is approximately 5 m, and the tem-
perature increase is less than 0.05
8
C. Temperature and light
transmissivity profiles along the transect are shown in Fig.
3a and 3b, respectively. The temperature profiles close to the
lava inflow show large fluctuations between 50 and about
120 m depth that resulted from the local heating by the lava
and subsequent advective isopycnal transport. The observed
780 Lorke et al.
Fig. 3. Successive profiles of (a) temperature and (b) light transmissivity following a north-to-south transect from the lava inflow in
Goma to the island of Idjwi (Fig. 1). Successive profiles are offset by 0.21
8
C and 2.2%, respectively. The numbers above the profiles
indicate the distance from the lava inflow in kilometers. The inset plot in Fig. 3b shows the changes of the bulk values of the light
transmissivity as a function of distance from the lava inflow. The bulk values were estimated as the mean background light transmissivity
below 50 m depth (disregarding the fluctuations). All profiles were measured within 4 h, and temporal changes of the water column
characteristics can hence be neglected.
temperature inversions do not cause local convection, since
the strong salinity gradient within that depth range stabilizes
the density stratification of the water column. Farther away
from the lava inflow, the temperature fluctuations are in-
creasingly smoothed out, and only one distinct peak of high-
er temperature remains. This peak is most pronounced in the
profile taken 14.1 km away from the lava inflow with a tem-
perature anomaly of
;
0.14
8
C and disappears completely in
the profiles measured farther away. This strongest peak can
be related to the earliest and major inflow events. Taking
into account the time between the Nyiragongo eruption and
the profile measurements (23 d) and setting 14 km and 20
km as the lower and upper limits for the horizontal disper-
sion results in a horizontal advection velocity between 0.7
and1cms
21
, respectively. This is in good agreement with
tracer measurements in several lakes of comparable and
smaller size (Peeters et al. 1996). Neither the main structure
of the vertical stratification nor the heat content changed
significantly as a result of the lava inflow.
The heat input of the 10
6
m
3
of hot lava flowing into the
lake can be estimated to about 2.4
3
10
15
J, assuming typical
values (Dragoni et al. 2002) for temperature (1,000
8
C), heat
capacity (850 J kg
21
K
21
), and density (2,800 kg m
23
). The
assumption that this heat was solely distributed within this
thin, 5-m thick layer of water over an area spanned by a
segment of a circle centered in Goma with an opening angle
of 130
8
(following the shorelines) and a radius of 17 km
results in a temperature increase of 0.3
8
C. This estimate is
in reasonable agreement with the observed maximum tem-
perature increase of 0.14
8
C, if we take into account that a
significant amount of heat was probably lost to the atmo-
sphere due to evaporation or was mixed into the rest of the
water column. Since a local destabilization of a 5-m thick
layer within the respective depth range would require a tem-
perature increase of 0.5
8
C, even the total amount of heat
introduced by the lava was not sufficient to force convective
mixing of this layer over the observed spreading area.
The transmissivity profiles along the CTD transect are
shown in Fig. 3b. Again, the profiles close to the lava inflow
show strong fluctuations, which become smoother with in-
creasing distance. A direct correlation between the strength
of the well-defined temperature peaks and the peaks in the
transmissivity could not be observed. However, in contrast
to the temperature, the bulk or background transmissivity, as
estimated from the profile sections, which show constant
transmissivity values, increases with distance from the city
of Goma (inset plot in Fig. 3b). Since this bulk transmissiv-
ity shows no vertical structure, contrary to what would be
expected for settling particles, the changes can be attributed
to the dilution of the particles only, with settling not being
a dominant factor.
Discussion
Several peculiarities have made Lake Kivu the subject of
comprehensive investigations in the past. These include the
high concentrations of dissolved gases (Tietze et al. 1980)
and their origin (Deuser et al. 1973; Tietze et al. 1980;
Schoell et al. 1988), the resulting density stratification (New-
man 1976; Tietze 1981), the special underground geological
situation (Degens et al. 1973), and the potential exploitation
of the dissolved methane (Tietze 2000). In addition, double-
diffusive mixing, caused by the geothermal heating at the
bottom of the lake, was observed and analyzed (Newman
1976; Tietze 1981). All of those investigations were made
more than 25 yr ago and provide the opportunity for a long-
term comparison with the recent profiles.
A comparison of the recent temperature profile with the
one measured by Tietze in 1975 (Tietze 1981) in Fig. 4 re-
veals a distinct warming of up to 0.5
8
C within the perma-
781Lake Kivu stratification
Fig. 4. Comparison of one recent (February 2002) temperature
profile and one measured by Tietze in 1975 (Tietze 1981). Both
profiles were measured with a vertical resolution of a few decime-
ters and a proven temperature calibration. Fig. 5. Detailed structure of the temperature stratification
around the mixed layers. The two profiles were measured 11 (CTD
1) and 17 km (CTD 5) off Goma (see Fig. 1) with 3 days time
difference in between. The inset plot shows the double-diffusive
staircase in CTD 1.
nently stratified part of the water column between 50 and
250 m depth. Both profiles were measured with a proven
accuracy better than 0.01
8
C and a high vertical resolution of
;
10 cm. A similar increase in temperature was already ob-
served by Deuser et al. (1973) by comparing their profiles
to earlier profiles of Damas (1937); however, their accuracy,
which was stated to be ‘‘within a few tenths
8
C’’ in the
original publication (Degens et al. 1973), is not sufficient to
allow for a direct comparison. In contrast to Fig. 4, the for-
merly observed warming affected the entire water column,
and Deuser et al. (1973) proposed the energy released by
the formation of methane to be responsible. However, since
the methane concentration increases significantly only below
a depth of 250 m, where no recent temperature increase
could be observed from Fig. 4, and the amount of warming
increases toward the surface, this recent increase in temper-
ature below the depth of seasonal mixing can most probably
be attributed to climate variability or change. Comparable
climatic warming was observed in other African lakes (Half-
man 1993; Hecky 1993; Plisnier 2000; O’Reilly et al. 2003;
Verburg et al. 2003), as well as in other lakes around the
world (e.g., Scheffer et al. 2001).
Interestingly, the near-bottom temperature of the lake has
not changed at all in the past 25 yr, indicating a stable equi-
librium between the geothermal heat flux into the lake and
the vertical transport within the water column. However, the
stratification did change slightly. The most obvious change
occurred between 250 and 300 m depth, where a well-mixed
layer has been established in a former gradient region. The
two other well-mixed layers, at 200–250 m depth and 320–
350 m depth, became thicker compared to 25 yr ago. At
least these two layers do not exhibit any gradients of the
measured physicochemical parameters, as one would expect
due to the smoothing effect of diffusion. Hence they can
only be sustained by active mixing. The only possible mech-
anism for the local production of the necessary mechanical
energy is convection, which may be driven by double dif-
fusion or by layer merging of thin, 1- to 2-m thick double-
diffusively mixed layers as shown in the inset plot in Fig.
5. The spikes in the recent profiles at the edges of those
mixed layers (at 170 and 250 m depth in Fig. 5) also indicate
convective activity, although they are stabilized by salinity.
An unusual peculiarity is that the double-diffusive mixing in
Lake Kivu is not solely characterized by the classical coun-
tergradients of temperature- and salinity-related density pro-
files, but also the increasing concentrations of dissolved gas-
es must also be taken into account: whereas the carbon
dioxide increases the density, the dissolved methane decreas-
es the density with increasing gas concentration.
The density ratio R
r
, an indicator for the susceptibility of
the stratified water column to double diffusion, is defined as
the ratio of the stabilizing forces and the destabilizing forces
and can be expressed in this special case as
]
S
]
[H CO ]
23
b
1
b
SCO2
]
z
]
z
R
5
(1)
r
]
T
]
[CH ]
4
a
1G 2
b
CH4
1212
]
z
]
z
where Sand Tdenote salinity and temperature, zis the depth,
a
and
b
S
are the coefficients of thermal and haline expan-
sion, [H
2
CO
3
] and [CH
4
] are the dissolved gas concentra-
tions, and
b
and
b
the respective expansion coefficients
CO CH
24
(note that
b
is negative), whereas
G
denotes the adiabatic
CH4
lapse rate. The ionic carbon species (HCO and CO ) are
222
33
included in S, and [H
2
CO
3
] was estimated following Schmid
et al. (2003), whereas [CH
4
] was taken from Tietze et al.
(1980). Double-diffusive staircases have been observed in
systems with 1
,
R
r
#
10, but the susceptibility to double
diffusion increases with decreasing R
r
until the density ratio
approaches 1, where the stratification becomes unstable
(Turner 1973).
Similar to that found by Schmid et al. (in press) in Lake
Nyos, the density ratio is below 10 throughout the entire
water column, but it falls below 3 within two depth ranges,
at 130–170 m and 380–400 m depth. Double-diffusive stair-
782 Lorke et al.
cases could only be observed within the upper range (see
Fig. 5) and were best resolved in the profile shown in Fig.
2 (CTD Sta. 1 on the map in Fig. 1). This profile shows a
typical double-diffusive staircase with layer heights between
2.4 and 4.3 m, increasing with depth. Applying the semi-
empirical flux law from Kelley (1990) results in an average
double-diffusive heat flux of 0.03 W m
22
. Although using
the original flux law from Turner (1973) results in a higher
heat flux of 0.05 W m
22
, these values are more than one
order of magnitude smaller than those found by Newman
(1976). Within the same depth range he estimated an average
heat flux of 1.6 W m
22
, which could be attributed to the
much smaller layer thicknesses of 0.85 m. Although in dis-
crepancy with former measurements, the observed upward
heat flux within the double-diffusive staircase is within the
wide range of bottom heat flux estimates of 0.017–0.17 W
m
22
determined by Degens et al. (1973).
However, relating the bottom heat flux to the double-dif-
fusive heat flux between 130 and 160 m depth raises ques-
tions about the heat transport within the intermediate depth
range, where no small-scale staircases could be observed.
The apparent diffusivity of heat K
T
within the double-dif-
fusive regime can be estimated by applying Fick’s law to
the observed heat flux and the local temperature gradient and
results in K
T
ø
2
3
10
26
m
2
s
21
, about one order of mag-
nitude greater than the molecular diffusivity. Since it can be
assumed that turbulent mixing in the interior of a strongly
stratified and deep lake is comparably weak (Schmid et al.
in press; Wu¨est and Lorke 2003), with vertical diffusivities
close to the molecular ones, the enhanced vertical transport
within the double-diffusive depth ranges could become
transport-limited from below.
The continuing double-diffusive transport with limited
heat flux at the lowest interface could lead to progressive
layer merging until the entire depth range becomes mixed.
This process could be responsible for the formation and
maintenance of the special pattern in the Lake Kivu strati-
fication, showing these three mixed layers separated by
strong gradients (Fig. 2). In fact, the three mixed layers ob-
served in the recent measurements correspond to the three
distinct depth ranges with strong double-diffusive staircase
formation in the measurements of Newman (1976). Further
indication of active convective mixing across these layers is
given by the observation of spikes (which are stabilized by
salinity) in the temperature profiles adjacent to the layers
(Fig. 5). The oppositional (bimodal) occurrence of these
spikes in the two profiles shown in Fig. 5 reveals the either
temporal or spatial dynamics of the fine structure of the strat-
ification in Lake Kivu.
Beyond all the unique findings, it is important to point
out that the overall stability of Lake Kivu is relatively high
and that a much stronger impact than that of the January
2002 lava inflow would be necessary to trigger a gas out-
burst from the deep waters. Clear signatures of the lava in-
flow were found down to 100 m depth and as far as 14 km
away from the location of the lava inflow, but the stability
and the heat content of the gas-rich layers below were not
affected. The observed general increase in temperature with-
in the upper part of the permanently stratified part of the
water column is related to long-term changes, which are
most probably caused by climate variability. Even the dou-
ble-diffusive structures, which are based on a small-scale
equilibrium of molecular processes, were well preserved.
Therefore, besides the acquisition of new concentration pro-
files of dissolved gases, future research should focus on the
potential impact of lava outflows from submersed vents as
well as on the potential impact of baroclinic dislocation in
the water column triggered by the enhanced seismic activity
at the lake bottom.
References
D
AMAS
, H. 1937. La stratification thermique et chimique des lacs
Kivu, Edourd et Ndalaga (Congo Belge). Verh. Int. Ver. Lim-
nol. 8: 51–68.
D
EGENS
,E.T.,R.P.
VON
H
ERZEN
, H.-K. W
ONG
,W.G.D
EUSER
,
AND
H. W. J
ANNASCH
. 1973. Lake Kivu: Structure, chemistry
and biology of an East African Rift Lake. Geol. Rundsch. 62:
245–277.
D
EUSER
, W. G., E. T. D
EGENS
,
AND
G. R. H
ARVEY
. 1973. Methane
in Lake Kivu: New data bearing its origin. Science 181: 51–54.
D
RAGONI
, M., F. D’O
NZA
,
AND
A. T
ALLARICO
. 2002. Temperature
distribution inside and around a lava tube. J. Volcanol. Geo-
therm. Res. 115: 43–51.
F
REETH
, S. J., G. W. K
LING
,M.K
USAKABE
,J.M
ALEY
,F.M.
T
CHOUA
,
AND
K. T
IETZE
. 1990. Conclusions from Lake Nyos
disaster. Nature 348: 201.
H
ABERYAN
,K.A.,
AND
R. E. H
ECKY
. 1987. The late Pleistocene and
Holocene stratigraphy and paleolimnology of Lakes Kivu and
Tanganyika. Paleogeogr. Paleoclimatol. Paleoecol. 61: 169–197.
H
ALFMAN
, J. D. 1993. Water column characteristics from modern
CTD data, Lake Malawi, Africa. J. Gt. Lakes Res. 19: 512–
520.
H
ECKY
, R. E. 1993. The eutrophication of Lake Victoria. Verh. Int.
Ver. Limnol. 25: 39–48.
K
ELLEY
, D. E. 1990. Fluxes through diffusive staircases: A new
formulation. J. Geophys. Res. 95: 3365–3371.
K
LING
,G.W.,
AND OTHERS
. 1987. The 1986 Lake Nyos gas disaster
in Cameroon, West Africa. Science 236: 169–170.
N
EWMAN
, F. C. 1976. Temperature steps in Lake Kivu: A bottom
heated saline lake. J. Phys. Oceanogr. 6: 157–163.
O’R
EILLY
, C. M., S. R. A
LIN
, P.-D. P
LISNIER
,A.S.C
OHEN
,
AND
B.
A. M
C
K
EE
. 2003. Climate change decreases aquatic ecosystem
productivity of Lake Tanganyika, Africa. Nature424: 766–768.
P
EETERS
, F., A. W
U
¨EST
,G.P
IEPKE
,
AND
D. M. I
MBODEN
. 1996.
Horizontal mixing in lakes. J. Geophys. Res. 101: 18361–
18375.
P
LISNIER
, P.-D. 2000. Recent climate and limnology changes in
Lake Tanganyika. Verh. Int. Ver. Limnol. 27: 2670–2673.
S
CHEFFER
, M., D. S
TRAILE
,E.H.N
ES
,
AND
H. H
OSPER
. 2001. Cli-
matic warming causes regime shifts in lake food webs. Limnol.
Oceanogr. 46: 1780–1783.
S
CHMID
, M., A. L
ORKE
,C.D
INKEL
,A.W
U
¨EST
,
AND
G. T
ANYILEKE
.
In press. Double-diffusive convection in Lake Nyos, Came-
roon. Deep-Sea Res. I.
,,A.W
U
¨EST
,M.H
ALBWACHS
,
AND
G. Z. T
ANYI
-
LEKE
. 2003. Development and sensitivity analysis of a model
for assessing stratification and safety of Lake Nyos during ar-
tificial degassing. Ocean Dynamics 53: 288–301.
S
CHOELL
, M., K. T
IETZE
,
AND
S. M. S
CHOBERTH
. 1988. Origin of
methane in Lake Kivu (East-Central Africa). Chem. Geol. 71:
257–265.
T
IETZE
, K. 1981. Direct measurements of the in-situ density of lake
and sea water with a new underwater probe system. Geophys-
ica 17: 33–45.
783Lake Kivu stratification
. 1992. Cyclic gas bursts: Are they a ‘‘usual’’ feature of Lake
Nyos and other gas-bearing Lakes? p. 97–107. In S. J. Freeth,
K. M. Onuoha, and C. O. Ofoegbu [eds.], Natural hazards in
West and Central Africa. Vieweg-Verlag.
. 2000. Lake Kivu gas development and promotion-related
issues: Safe and environmentally sound exploitation, Report
PDT GmbH No. 520002, Republic of Rwanda, Ministry of
Energy, Water and Natural Resources, Kigali (Rwanda).
,M.G
EYH
,H.M
U
¨LLER
,L.S
CHRO
¨DER
,W.S
TAHL
,
AND
H.
W
EHNER
. 1980. The genesis of methane in lake Kivu (Central
Africa). Geol. Rundsch. 69: 452–472.
T
URNER
, J. S. 1973. Buoyancy effects in fluids. In G. K. Batchelor
and J. W. Miles [eds.], Cambridge monographs on mechanics
and applied mathematics. Cambridge Univ. Press.
V
ERBURG
, P., R. E. H
ECKY
,
AND
H. K
LING
. 2003. Ecological con-
sequences of a century of warming in Lake Tanganyika. Sci-
ence 301: 505–507.
W
U
¨EST
, A.,
AND
A. L
ORKE
. 2003. Small-scale hydrodynamics in
lakes. Annu. Rev. Fluid Mech. 35: 373–412.
,G.P
IEPKE
,
AND
J. D. H
ALFMAN
. 1996. Combined effects
of dissolved solids and temperature on the density stratification
of Lake Malawi, p. 183–202. In T. C. Johnson and E. O. Oda-
da, [eds.], The limnology, climatology and paleoclimatology of
the East African Lakes. Gordon and Breach.
Received: 5 March 2003
Amended: 10 December 2003
Accepted: 11 December 2003
