The Evolution of Thermal Structure and Water Chemistry in Lake Nyos

Journal of Volcanology and Geothermal Research, 39 (1989) 151-165 151 Elsevier Science Publishers B.V., Amsterdam -- Printed in The Netherlands

The evolution of thermal structure and water chemistry in Lake Nyos

GEORGE W. KLING 1, MICHELE L. TUTTLE 2 and WILLIAM C. EVANS 3 1The Ecosystems Center, Marine Biological Laboratory, Woods Hole, MA 02543, U.S.A. 2U.S. Geological Survey, Federal Center, Denver, CO 80225, U.S.A. 3U.S. Geological Survey, Menlo Park, CA 94025, U.S.A. (Received November 10, 1988; revised and accepted April 10, 1989 )

Abstract

Kling, G.W., Tuttle, M.L. and Evans, W.C., 1989. The evolution of thermal structure and water chemistry in Lake Nyos. In: F. Le Guern and G. Sigvaldason (Editors), The Lake Nyos Event and Natural C02 Degassing, I. J. Volcanol. Geotherm. Res., 39: 151-165.

We collected a time series of physical and chemical data to gain a better understanding of the dynamics of Lake Nyos. Measurements of water and gas chemistry, and temperature made during January, March, and May 1987 are compared to data taken in September 1986 just after the initial C02 gas release. There is no pattern of change in overall heat content of the lake, although heat input to bottom waters ( 185-208 m) has occurred at a rate of 1600 mW m -2. This increase in heat content translates to a change from 23.38 to 24.12°C at 200 m and can be explained by geothermal heat flow and addition of thermal spring water. Concentrations of Ca 2+, Mg2+, Na +, K +, Fe2+ and alkalinity have increased only in bottom waters. In situ lake processes such as sulfate and iron reduction are unable to account for the changes in alkalinity. Observed chemical changes are consistent with a scenario where slightly thermal soda water is being input to the bottom of the lake. Measurements of pC02 at depth ranged from 18 to 28% of saturation and exhibited horizontal variability. Overall recharge of C02 in bottom waters is negligible. Mainly because of increasing ion concentrations in bottom water, total stability of the water column increased 33 % from 48,800 J m- 2 in September 1986 to 64,700 J m -2 in May 1987. As long as C02 concentrations remain the same, this level of stability is higher than could be disrupted by common limnologic or meteorologic processes. There is thermal and chemical evidence that a buildup of dissolved iron and C02 in bottom waters must have preceded the August 1986 gas release. In addition, a survey of all crater lakes in Cameroon indicates that only Lakes Nyos and Monoun contain high concentrations of dissolved iron and C02. Thus there is a low probability of any other Cameroonian lake releasing a substantial volume of CO2.

Introduction study. By characterizing basic properties and processes in this system, we hope to resolve On August 26, 1986 an enormous volume of questions about the state of the pre-event lake, CO2 was released from Lake Nyos that killed the mechanism of gas release, and any future about 1700 people (see papers in this volume). dangers. Two possible explanations for this event center As a form of introduction, we summarize the around a limnologic hypothesis (Kling et al., analyses and conclusions made by the Ameri- 1987 ) and a volcanic eruption hypothesis (Ta- can Scientific Team studying the event (Kling zieff et al., 1987). This rare natural disaster is et al., 1987; Tuttle et al., 1987). On the basis of complex, little understood, and difficult to 14C, He and $1sC-C02 data from dissolved gas

0377-0273/89/$03.50 © 1989 Elsevier Science Publishers B.V. 152 G.W. KLING ET AL. in the lake it is apparent that most of the C02 tion; and (4) sores that predated the event. Ob- released during the event was magmatic in ori- served skin blisters were associated with ex- gin. However, undisturbed sediment and clear tended unconsciousness, similar to symptoms water at depth, low water temperatures, ab- found in comatose drug overdose patients sence of acid gases, and the low sulfur nature of (Baxter and Kapila, 1989, this issue). the system indicate that no major volcanic One intriguing aspect of the Lake Nyos gas eruption occurred. Much if not all of the CO2 release and a similar but smaller release of CO2 released was stored in the lake prior to the event. from Lake Monoun (Sigurdsson et al., 1987) is Accumulation of CO2 in the lake starts when that both events occurred in August and were CO2-rich gas of magmatic origin rises to the only two years apart. There are recent trends of Earth's surface and contacts groundwater. Ox- decreasing air temperature and insolation rel- ygen and hydrogen isotope data plus the rela- ative to long-term means in this region of Ca- tive proportions of solutes in surrounding meroon. These trends suggest that weakening springs suggest a common origin for ground and of lake stratification, coupled with a predict- lake water. This common origin is consistent able yearly interval of reduced lake stability with the hypothesis of groundwater transfer of during August, may be responsible for the tim- dissolved CO2 into the lake. In addition, mea- ing of these events (Kling, 1987a). sured values of 1sO and 2H indicate that evap- The purpose of this paper is to describe the orative concentration is not responsible for the evolution of thermal structure and water chem- greater concentrations of major ions in Lake istry in Lake Nyos. From this we hope to better Nyos relative to other Cameroonian crater define the location of gas before the release, the lakes. Post-event depth profiles of major ions possible triggers and the mechanism of gas re- imply that stable stratification existed before lease, and the potential for future releases of the release, and that only partial mixing of lake CO2. Site descriptions of Lake Nyos and the waters occurred during the release. surrounding area are given in Kling et al. (1987) The trigger mechanism responsible for the gas and papers in this volume. release from the lake is unknown but has received much speculation. We know of no data Methods that require as explanation a single, specific trigger mechanism. In fact, ifpCO2 levels were Temperature was measured with a recently near saturation prior to the event, almost any calibrated thermistor ohm-meter combination physical process common in lakes could have interconnected by 4-conductor cable (Sass et moved water vertically enough to cause local al., 1981; accuracy= 0.1 ° C, precision = oversaturation and thus initiate the release. 0.001 ° C). Water samples were collected with a Pathological studies indicated that victims PVC Niskin or van Dorn bottle and filtered rapidly lost consciousness and died of C02 as- through 0.1 Hm filters in the field. Conductiv- phyxiation. There was no evidence for chemical ity, pH, and alkalinity by potentiometric titra- burns on victims or survivors as would be ex- tion were also measured in the field. Non-car- pected from volcanic sulfur gases, and Kusak- bonate alkalinity was measured on selected abe et al. (1989, this issue) provide further samples and was always < 2% of total alkalin- chemical evidence against the possibility of such ity. The pH values of deep water samples were burns. The skin lesions were attributable to affected strongly by C02 degassing. In these some combination of the following: (1) expo- samples pH was calculated using measured sure to a direct heat source such as a cooking HCO~- and CO2 concentrations and was always fire; (2) pressure sores from prolonged lying in higher than 5.10. Water for oxygen analysis was a fixed position; (3) post mortem decomposi- fixed immediately with Winkler reagents and THE EVOLUTION OF THERMAL STRUCTURE AND WATER CHEMISTRY IN LAKENYOS 153

Areo x I0e mI tionship presented in Figure 1. Water density, 0 0.5 l.O 1.5 lake stability, and C02 fugacity were determined using methods detailed in Appendix 1.

E 4O Results and discussion .oI Temperature and heat content , ol/ /.eo All temperature profiles taken since the event are similar in overall character to the May 1987 profile presented in Figure 2. Temperatures decrease with depth until they reach a subsurface minimum, and then they steadily increase with depth to the bottom. In September 1986 the 0 5 I0 15 thermocline was weak and poorly defined. Volume x I0 "r m3 Water temperature began to increase slightly at

Fig. 1. Hypsographic curve and depth-volume relationship a depth of ~ 5 m and the water was anoxic be- for Lake Nyos. low 10 m. By January 1987 a thermocline had developed at 7 m and below 11 m water temper- titrated with phenylarsine oxide within six ature began to increase. In May 1987 a thermal hours of collection. Dissolved gases were col- structure more typical of an undisturbed lake lected in situ using evacuated stainless steel existed. The thermocline had deepened to 11 m cylinders. The cylinders were depressurized in and a distinct upper mixed layer (epilimnion) the lab and gases collected and analyzed using had developed. At this time increases in water procedures similar to those used by Evans et al. temperature began at 22 m and dissolved oxy- (1988). Collection of dissolved gases in pres- gen was present to 15 m. surized cylinders, as opposed to collection of Y bubbles from samplers brought to the surface, ff eliminates errors caused by differing gas solu- bilities. Major anions were analyzed by ion 4C LAKE NYOS chromatography. Water samples for major cat- 18 Moy 1987 1030 ions were acidified in the field using HC1 and analyzed by atomic absorption or induction- 80 coupled plasma spectrophotometry. Major ions E =- in water collected in the pressurized cylinders T= ¢0 were analyzed similarly. Ammonium was deter- r~ mined by specific ion electrode in the field. Val- ues for 613C-XC02 and 613C-CH4 were determined on a Finigan MAT 251 mass 160 spectrometer. Ion and solid phase chemical equilibria were calculated using the computer program 200

PHREEQE (Parkhust et al., 1980). Calculations 25 25 27 involving water masses were made using the Temperature, °C hypsographic curve and depth-volume rela- Fig. 2. Temperature versus depth for Lake Nyos. 154 G.W. KLINGET AL.

The phenomenon of increasing water tem- TABLE 1 perature with depth in Lake Nyos exists due to Lake Nyos water temperature at selected depths in Sep- controls on water density by chemical stratifi- tember 1986, January 1987 and May 1987 cation. Heat accumulates in bottom waters (hypolimnion) because geothermal heat flux to Depth Temperature (°C) the surface is slowed by strong density gra- (m) September N= 2 January N= 3 May N= 4 dients and low eddy diffusion (see Turner, 1973). At present in Lake Nyos, calculated 55 22.875 22.796(0.003) 22.770(0.058) 65 22.875 22.817(0.004) 22.793(0.057) Brunt- V~iis~il~ifrequencies (a measure of strat- 75 22.900 22.850(0.001) 22.830(0.060) ification strength) approach -~ 10- ~ s- 1 across 85 22.925 22.880(0.004} 22.870(0.056) the density boundary between depths of ~ 5-15 95 22.950 22.921(0.005) 22.915(0.053) m. These frequencies correspond to low eddy 105 23.125 22.960(0.004) 22.959(0.053) 115 23.400 23.020{0.010) 23.021(0.057) diffusion rates compared to rates in other strat- 125 23.113 23.070(0.009) 23.078{0.054) ified lakes (Quay et al., 1980). Before the event, 135 23.163 23.122(0.003) 23.133(0.053) density gradients would have been even stronger 145 23.188 23.182(0.005) 23.172(0.054) between dilute surface water (conductivity= 155 23.225 23.197(0.003) 23.184(0.056) 40.5/~S cm-1; Kling 1987b) overlying a hypo- 165 23.225 23.202(0.001) 23.190{0.053) 175 23.275 23.210(0.001) 23.210(0.055) limnion close to saturation with CO2. In Lake 185 23.275 23.270(0.021) 23.358(0.060) Nyos dissolved CO2 accounts for most of the 190 23.325 23.312{0.029) 23.468(0.039) increase in water density with depth. In such a 195 23.338 23.486(0.042) 23.698(0.055) situation of stable stratification, a geothermal 200 23.375 23.689(0.149) 24.121(0.042) 204 23.438 23.936(0.077) 24.158(0.052) heat flow similar to a global average of about 50 mW m -2 (Cerm~ik et al., 1984) would take 100 Standard deviation of the mean given in parentheses. years to accumulate the excess heat found below 55 m after the event, compared to the heat The corresponding mean increase in heat con- content of an isothermal temperature profile. tent of the 185-208 m layer was calculated us- This time value assumes no heat loss by vertical ing temperature profiles taken in various parts diffusivity, but also ignores any heat input from of the lake. Heat content increased from 236.7 thermal springs. It shows that it is quite possi- kJ cm-2 (mean of two measurements) in Sep- ble for the observed temperature profile to de- tember to 238.2 kJ cm -2 (N=3, S.D.=0.45) in velop once strong chemical stratification was January and finally to 240.3 kJ cm -2 (N=4, established. Thus the inverse temperature S.D. = 0.52 ) in May. A regression equation de- stratification in Lake Nyos argues for stable scribing this change over 256 days is significant stratification and accumulation of CO2 well be- at the P<0.0001 level and is given by HEAT fore the August 1986 gas release. (kJ cm-2)=236.5+(14.14*DAY), r2=0.91. Changes in thermal structure from Septem- Thus the heat flow into the lake required to ac- ber 1986 to May 1987 have been small except count for these significant changes in heat con- near the surface and close to the lake bottom. tent must be about 1600 mW m -2. The mag- Surface waters followed the general pattern of nitude of geothermal heat flow below Lake Nyos seasonal change in temperatures in this region is unknown; this value of 1600 mW m -2 is (Kling, 1987a) by cooling from September to higher than the range of 20-130 mW m -2 found January and then warming from January to in various settings worldwide (Cerm~ik et al., May. Mean temperature values at the middle 1984; Beck, 1985). depths (65-175 m ) deviated little between dates A second potential source of this heat is from (Table 1 ). However, temperature below _~ 185 advection of thermal groundwater. Thermal m has increased steadily since September 1986. soda springs in Cameroon can reach tempera- THE EVOLUTION OF THERMAL STRUCTURE AND WATER CHEMISTRY IN LAKE NYOS 155 tures of 74°C (Marechal, 1976). In May 1987 temperatures measured in Lake Nyos just after we sampled the Ahoy Ekenzu spring (number the August 1986 gas release (Barberi et al., 1989, 94 in Marechal, 1976), which had a chemistry this issue) prove that significant heat was as- similar to that of Lake Nyos bottom water sociated with the gas release and that it was vol- (spring conductivity= 1530 ItS cm -1, HCO~- canic in nature (see Sigvaldason, 1987). An = 19.0 mM). It discharged free plus dissolved equally viable explanation for these tempera- CO2 at an estimated 500 L min -1, and was ture measurements is solar heating of the sur- slightly thermal at 27 ° C. Assuming an input of face water. For example, enhanced rates of water similar to this spring at 27°C into the heating are commonly observed in lakes when bottom of Lake Nyos, the flow required to ac- particulate concentrations are high (Schreiner, count for the heat changes observed in the 185- 1984). Such heating rates can be modeled in any 208 m layer since September is only 7 L s-1. water by a modified version of the equation pre- This flow rate over 256 days amounts to about sented in Zaneveld et al. (1981) as follows: 1.3% of the volume of water below 185 m. An OT analogous system is that of Lake Kivu in East Ot = [ (Qs/pCpD) (1--e(-kD)) ] -- (B/pCpD) Africa, which is heated from below by thermal springs (Degens et al., 1973; Tietze, 1978). In where T is temperature in ° C, t is time, Qs is Lake Kivu the upward heat flux of 710-1600 short wave radiation incident at the lake sur- mW m -2 (Newman, 1976) is similar to the 1600 face, p is water density, Cp is specific heat of mW m -2 value calculated for Lake Nyos. On water, D is depth of mixed layer, k is extinction the basis of temperature measurements alone, coefficient, and B is surface heat loss {long wave however, the relative heat contributions of geo- radiation + evaporation + sensible heat trans- thermal flux versus groundwater advection in fer). This equation was applied to Lake Nyos Lake Nyos are difficult to separate. for September 1986, just after the event when Rates of temperature increase in bottom particulates (mainly iron minerals) were high waters are constrained by buoyancy effects. On in surface waters. Values of Qs = 880 W m- 2 and the basis of measured water density at 204 m in B = 225 W m -2 were estimated from values in May 1987 (Appendix 1), water temperatures the Smithsonian Tables and values for nearby would need to reach 35 °C in order to ascend lakes (Kling, 1987b) assuming cloudless con- buoyantly. This upward mixing would redis- ditions and winds < 1 m s -1. An extinction tribute heat in the water column and thus set a coefficient of 1.7 m- 1 was calculated from mea- maximum attainable temperature of bottom sured secchi depth and an empirical relation waters. In the last nine months, however, the between k and secchi depth (log[k]= mean increase in water density at 204 m caused 0.1952- (0.9959) (log[secchi]), r2=0.84; raw by solute input ( + 0.00034 g cm -3 ) has roughly data in Kling, 1988). Taking a 1-m mixed layer matched the decrease in density caused by ris- depth, the theoretical rate of heating in Lake ing temperatures (-0.00027 g cm-3). Such a Nyos on 3 September is 0.43 °C h-1. This is a stable system inhibits the turbulent movement mean temperature increase integrated over the of water and limits, to the process of eddy dif- chosen mixed layer, and the heating rate near fusion, exchange of mass and heat from the bot- the surface or in specific subsurface particulate tom to upper water levels. It will be interesting layers could be higher. The calculated rate is to follow these density changes through time, similar to differential rates of heating observed although at present the short run of data places by Schindler et al. (1981) that were up to 0.5 ° C projected temperature increases in the realm of h-' in particulate layers at a depth of 2 m. speculation. Heating at such rates accumulated over several It has been argued that the high surface water days could account for much of the anoma- 156 G.W. KLING ET AL. lously high surface temperatures recorded by to the bottom (Figs. 3 and 4; Table 2). In this Barberi (Sigvaldason, 1987). bottom layer, concentrations of all ions have As a test of this theoretical model of heating increased. The largest change has been HCO~- rates an actual balance of surface energy was concentrations; alkalinity increased by about calculated for Lake Nyos on May 17, 1987 and 3.92 meq L -1 from September 1986 to May compared to model predictions. From 08.10- 1987. 13.40 the upper 6 m of the water column gained This large increase in HCO~- is most likely 1650 × 101° J of heat, which translates to a heat- due to input of concentrated groundwater rather ing rate of 0.077°C h -1. Using k=0.4 m -1, than to dissolution of carbonates or other in situ Qs= 880 W m -2 and B=225 W m -2, the theo- processes that generate alkalinity such as re- retical mean heating rate predicted for the same duction of SO~-, Fe(OH)3 and MnO4 or de- upper 6 m of water is 0.082 °C h-1. This com- nitrification. For example, concentrations of Ne pares very well to the measured rate. Kusakabe and Mn 2+ would be expected to increase if et al. (1989, this issue) present an analysis of MnO~- reduction or denitrification were impor- potential temperature increases resulting from tant sources of alkalinity. However, their con- iron oxidation in surface waters, although mea- centrations in Lake Nyos are very low and have sured iron concentrations in Lake Nyos seem not changed significantly from September to too low for this mechanism to be as important May (Tables 2 and 3). Second, almost 2 mmol as solar heating. L- 1 of sulfate would need to have been reduced over this time period to account for the ob- Water chemistry served increase in alkalinity. But SO~- concentrations have remained at less than 5 #mol L- 1 Observed changes over time in profiles of ion and reduced sulfur burial in the sediments is concentrations in Lake Nyos are presented in very small (total sediment S= 0.11%; Kling et Figures 3 and 4. There has been a slight fresh- al., 1987). Finally, reduction of Fe (OH) 3 formed ening of the surface water as the upper mixed in oxygenated surface waters and transported layer deepened since September. Below this, to anoxic bottom waters also would produce al- major ion concentrations have remained fairly kalinity. As a result, Fe 2+ would accumulate at constant except in the depth layer from 185 m lower depths. Table 2 shows, however, that increases in Fe 2+ concentrations in bottom waters TABLE2 are insufficient to account for the observed alkalinity change. Our equilibria calculations in- Chemical changes in Lake Nyos bottom waters over time dicate that siderite (FeCO3) was oversaturated (in mg L -1) at 200 m in May 1987. These calculations used Date Sep. 86 Jan. 87 Mar. 87 May 87 a pH of 5.18 and resulted in a saturation index Depth (m) 200 198 198 198 (SI-FeCO3) of 0.51, i.e., the product of Fe e+ and CO~- concentrations was 3.2 times ( 10°'51 ) that HCOa- 725 836, 846 903, 1020 964 Ca 2+ 46 47, 50 57, 66 66 required (theoretically) for precipitation of Mg e+ 57 62, 77 75, 85 78 FeCO3. However, a loss of iron in bottom waters Na + 19 17, 22 20, 23 27 from precipitation of FeCOa cannot be invoked, K + 7 6,6 7,7 8 Fe ~+ 71" 88, 87 97, 115 95 because precipitation of FeCO3 and its burial in Mn 2+ 1.6 1.85, 1.86 2.18, 2.21 1.75 the sediments would consume the alkalinity Si 26 31, 33 36, 39 39 produced by reduction of Fe (OH)3. NH4 + 3.4 Another possible explanation for the ob- *An underestimate because iron hydroxide precipitated be- served alkalinity increase is the dissolution at fore filtering. depth of carbonates formed in surface waters. THE EVOLUTION OF THERMAL STRUCTURE AND WATER CHEMISTRY IN LAKE NYOS 157

0 O" A Sep. 1986 & Sep. 1986 I Jan. 1987 • Jan. 1987 -25 -25- • May 1987

-50 -50

-75, -75 o~ ¢n o~ -100 -100 E E ~_ -125 :~ -125

a - 150 -150

-175 -175

-200 .200

-225 -225 .2 .4 .6 .8 1 1.2 1.4 1.6 1.8 .5 1 1.5 2 2.5 3 3.5 Calcium, mM Magnesium, mM

0, A Sep. 1986 ~1~ • Sep. 1986 • Jan. 1987 -25. • Jan. 1987 .25. X • May 1987

-50 -50.

-75 -75, u~ if) -100 -100.

.c X ~ -125" ¢Z -125 Q -150 - 150

-175 -175.

-200, -200.

-225 -225 .04 .06 .oa .1 .12 .14 .16 .18 .2 .22 .2 .3 .4 .5 .6 .7 .8 .9 1 1.1 1.2 Potassium, mM Sodium, mM

Fig. 3. Depth profiles of major ions (mM) in Lake Nyos.

Chemical equilibria calculations show that planation for observed changes in bottom water CaCO3 and MgCO3-3H20 are undersaturated chemistry. by at least 1.7 orders of magnitude at all depths A ternary diagram plotting the proportional and dates in Lake Nyos (SI-CaCO3 < - 1.7; SI- cationic composition of Lake Nyos waters com- MgCOa-3H20 < -4.7). Thus precipitation of pared to the composition of various spring these minerals in surface waters followed by waters supports this assertion (Fig. 5). Hypo- downward transport is unimportant to bottom limnetic chemistries from all dates plot very water chemistry. Given the concurrent in- close to values for water issuing from ultra- creases in all major ions at depth in Lake Nyos, mafic rocks (Hem, 1970; Siever and Woodford, the input of groundwater is the most viable ex- 1979), which are similar in composition to pyro- 158 G.W. KLING ET AL,

0 ~ • Sap. 1986 O~-~L • Sep. 1986 • Jan. 1987 II~l-~_..~.,~=.~_ • Jan. 1987 ~25 "25t ~,~ . May 1987

-50

-75 -75

-loo ~ -10

.,1=0. -125

-150

-175 175. ~ ~xlBX~

-200 -200.

-225 -225 4 6 8 10 12 14 16 18 0 .2 .4 .6 .8 1 1.2 1.4 1.6 1.8 Bicarbonate, mM Iron, rnM

Fig. 4. Depth profiles of major ions (mM) in Lake Nyos.

TABLE 3 masses, consistent with a hypothesis of partial lake turnover caused by the August 1986 gas re- Concentrations of dissolved gases (#mol kg - ] ) in Lake Nyos lease. There may also be some in situ weather- Depth N2 02 Ar CO CH4 CO2 H2S ing of exposed subsurface rocks contributing to (m) the changes in Lake Nyos bottom water chem- 23Jan. 1987 istry, although it is unknown what surface area 17 628 <0.5 10.9 <0.5 0.3 22500 <0.05 35 205 <0.5 4.64 < 0.5 78.7 48600 <0,05 of rock is exposed to weathering, how much in- 76 198 < 1.0 3.6 <1.0 111 70000 <0,1 creased pCO2 affects weathering rates, and to 107 225 33.1 4.1 < 1.0 186 110000 <0,1 what extent this mechanism of ion input has 131 196 81,6 3.1 < 3 118 108000 <0,3 198 146 2.6 4.5 < 2 308 159000 <0.2 altered the trajectory shown in Figure 5. 23Mar. 1987 Changes in Lake Nyos bottom water chem- 99 86.7 1.1 2.1 < 0.1 81.1 74900 <0.1 198 159 1.1 2.2 <0.4 205 177000 <0.4 istry since the gas release can be measured, but 198 58 0.72 1.4 <0.3 170 146000 <0.3 chemical evolution of the lake prior to the event 18MayI987 must be inferred. The chemical character of 61 226 1.8 4.7 121 62800 bottom water in the two lakes that have re- 99 75.7 2.0 0.71 79.1 78750 99 3110 330 37.2 162 92260 leased CO2, Nyos and Monoun, is unique among 143 172 6.0 3.2 255 129500 all crater lakes in Cameroon. The most distin- 171 166 <0.8 2.9 272 131800 guishing features are the high concentrations of 198 55.4 2.3 < 1.2 154 124400 198 95.5 3.9 < 2.3 255 184200 reduced iron and dissolved CO2 (Sigurdsson et 198 1560 21.8 16.8 413 193200 al., 1987; this paper). If the CO2 released during the 1986 disaster was moved through the lake from below by volcanic injection, as suggested clastic flow deposits around Lake Nyos (Tuttle by several authors (Tazieff et al., 1987; papers et al., 1987; Kusakabe et al., 1989, this issue). in this issue), then it becomes difficult to ex- Surface waters in Lake Nyos lie on the trajec- plain the high concentrations of reduced iron tory from ultramafic derived waters to lake in- and alkalinity now present. For example, how flow, nearby springs issuing from granite, and did the iron accumulate and what prevented its pre-event surface waters. This indicates simple loss by iron carbonate precipitation? Subsur- but incomplete mixing of upper and lower water face weathering in the high pCO2 environment THE EVOLUTIONOF THERMALSTRUCTURE AND WATERCHEMISTRY IN LAKENYOS 159

Na&K

.~ LAKENYOS PREEVENT SURFACE WATER A • VAN DORN SAMPLES • PRESSURIZEDBOMB SAMPLES S SPRINGSIN GRANITE I STREAMINFLOW U~ SPRINGIN ULTRAMAFICROCK \ B BASALTLEAC.ATE

L NYOS SURFACE ~]

L NYOSBOTTOM~ Ca/ \ Mg&Fe

Fig. 5. Ternary diagram showing the proportion of major cations (molar basis ) in surface (van Dorn ) and bottom (pressurized bomb) water samples from Lake Nyos; stream inflow (Kusakabe et al., 1989, this issue); springs issuing from granitic rocks above Lake Nyos; rain in Cameroon and surface water before the event (Kling, 1987b); spring water associated with ultramafic rocks (Hem, 1970); and leachates from experimental weathering of basalts (Siever and Woodford, 1979).

below Lakes Nyos and Monoun may supply low can account for this phenomenon. For ex- more iron to these lakes relative to supply rates ample, a common mechanism of deep water en- in other Cameroonian lakes. However, sub- trainment in lakes and oceans is Langmuir stantial accumulation of these ions only begins circulation. This circulation forms a set of con- when dissolved CO2 added to bottom waters in- vectional helicies resulting in upwelling and creases stratification strength and water col- windrows of particulate material such as plank- umn stability. In addition, high pC02 stabilizes ton on the surface (Faller, 1978). The win- the reduced iron by lowering pH and CO32- drows are oriented roughly parallel to the wind concentrations. Thus the current chemistry of but often remain unseen for lack of surface Lakes Nyos and Monoun supports the idea that marker. The iron hydroxide streaks observed in pCO2 was high in the hypolimnion prior to these Lake Nyos are, as far as we know, a novel form gas releases. of surface marker and indicate entrainment of Observations of reddish patches and streaks iron-rich anoxic water to the surface. In addi- on the surface of Lake Nyos have been used to tion, these streaks were not accompanied by argue for renewed volcanic activity beneath the noticeable gas bubbles (S.G. Tebor, pers. com- lake (Tazieff, 1987). These patches are caused mun., 1989). This is consistent with the hy- by the formation of iron hydroxide as dissolved pothesis that shallow, anoxic waters were en- Fe 2+ is brought from depth into oxygenated trained to the surface by Langmuir circulation, surface waters. However, entrainment of deep rather than deep, CO2-1aden waters being dri- water from above as well as injection from be- ven upward by volcanic events. 160 G.W. KLINGET AL,

Gas chemistry March, and May of 1987 {Table 3) are similar for each depth to values reported by Tietze Chemical composition of dissolved gases in (1987) for November 1986 and by Kusakabe et Lake Nyos is presented in Table 3. Comparison al. (1989, this issue) for October 1986. There of these values with values in Kling et al. (1987) is, however, distinct horizontal variation in show that the general character of dissolved Pco2 at any depth. For example, two samples gases has changed little since September 1986. from May 1987 at 198 m have Pco2 values of Carbon dioxide is still the dominant dissolved 3.57 and 5.55 atm (at 24°C ), which are the low- gas and makes up greater than 97% of all gases est and highest Pco2 values measured on any even in samples taken as shallow as 17 m. The date at this depth. The cause of this heteroge- nonvolcanic nature of the dissolved gases has neity is unclear; it could be due to localized in- not changed from samples taken directly after puts of C02 to bottom waters, internal waves, the event, and is most apparent in the low con- or chemical reactions in the sample cylinders centrations of CO, H2, and H2S. after collection. Concentrations of dissolved C02 in Lake The relative contribution of chemical reac- Nyos ranged from 18 to 28% of saturation (Fig. tions in sample cylinders to Pco2 variability can 6). We calculate that if water at all depths was be estimated from measured concentrations of saturated the lake could hold ~ 1.45 km 3 CO2 N2, 02, and At, because excesses of these gases {at 20 oC; or -~ 5.3 × 10 l° mol C02). In May 1987 result from leakage of evacuated cylinders to the there remained ~0.38 km 3 C02 in the lake atmosphere prior to filling. For example, the (1.41×101° mol CO2). Thus a maximum of worst case of contamination is in the 99 m sam- about 1 km 3 of CO2 could have been released ple taken in May 1987. Assuming that all 3110 during the August 1986 event. This volume is /~mol kg- 1 of N2 is from atmospheric contami- sufficient to account for the distribution of fa- nation, a corresponding concentration of 02 talities around the lake (Kling et al., 1987). equal to about 830/Jmol kg-1 would have en- There has been little change in average CO2 tered the sampler, which contained only 330 concentrations over time; values in January, /Jmol kg -1 when analysed. According to the reactions: ...... ,..:.....:.:.. Fe 2+ + ~ 02 +H + =Fe 3+ + ½ H20 "'"'""-...,.,...... ""...... Fe 3+ + 3H20=Fe(OH)3 +3H + 50 saturation distance each mole of 02 consumed can produce 8H +, E converting 8HCO~- to 8CO2¢aq)+ 8H20. Thus a 1- I00 I-- maximum of 4000/1tool CO2 kg-1 could have Q,. I¢J87 hi Q 0 Jan been added to this sample, resulting in an ov- • Mor 150 Moy erestimate of in situ CO2 concentration by 4%. In all other samples the potential for oxygen LAKE NYOS contamination is much less.

20q 0 • 0 'A" Cr Progressive temperature changes character- , , , , , , , , , , , , i , , I , i b , 0 50 I00 150 200 istic of internal waves were not seen in the me- C02 (aq)' mmol kg "~ talimnion or hypolimnion in May. A primary mode seiche in the metalimnion of Lake Nyos Fig. 6. Depth profile of C02 concentrations in Lake Nyos. Dotted line gives depth of saturation for values of CO2 shown in May would have a period of 2-3 hr. Because on x-axis; for example, water with 200 mmol kg -1 of C02 lengths of temperature observation were on the would be saturated at a depth of 45 m. order of 30 rain, any strong seiche probably THE EVOLUTION OF THERMAL STRUCTURE AND WATER CHEMISTRY IN LAKE NYOS 161 would have been detected. In addition, the ver- event. For samples below 20 m, ~13C-C02 val- tical density structure of Lake Nyos constrains ues for XC02 are - 3.0 to - 3.5 %0 PDB. Meth- the amplitude of an internal seiche. As a rough ane from samples below 90 m yielded ~1~C val- estimate, and using equations in Hutchinson ues of -45 to -52%o PDB and JD values of ( 1957 ), wind speeds of 20 m s- ~would produce -240 to -270%o SMOW. None of the three seiche amplitudes of no more than 5 m in the isotopes show regular changes with depth over metalimnion, and no more than 15 m in the hy- the time interval from September 1986 to May polimnion. Thus the observed C02 distribution 1987. It appears from this isotopic evidence that in Lake Nyos may be affected to some degree the major sources of these dissolved gases are by internal waves, as well as by localized inputs unchanged; the C02 is magmatic in origin and of CO2. Such localized inputs have yet to be the methane is probably from biological decom- identified. position (see Kling et al., 1987). Given the horizontal variability O~co2, a nine month period of observation is marginal to es- Lake stability and degassing timate a true recharge rate of C02 into the lake. This is especially true if recharge events are ep- Overall water column stability (S) is a fun- isodic or seasonal. However, our data indicate damental measure of the strength of thermal that no major recharge has occurred to date. plus chemical stratification in a lake. The Concentrations of CH4 and N2 are important strength of stratification constrains the initia- in that they contribute to the total gas pressure tion of degassing and can be used in predicting and thus to the potential for degassing. For ex- the likelihood of future gas releases. More for- ample, in the May 1987 samples the relative mally, S is the amount of energy required to mix contribution of CH4 + N2 amounts to 6% of the a lake to uniform density without the addition total gas pressure at 198 m. Their contribution or subtraction of heat. In Lake Nyos the con- increases to 24% of the total pressure at 61 m. cept of stability is complicated by the high pres- In May, PCH4 ranged from 0.03 to 0.29 atm (at sure of dissolved gas. As a parcel of water is lifted 24°C) at 24 m and 198 m, respectively. How- to the surface, at some point the total ambient ever, CH4 pressures before the event could have pressure decreases sufficiently to cause over- been much higher. For example, hypolimnetic saturation and bubble formation. The efferves- methane pressures in other Cameroonian cra- cence of gas that occurs produces an increase in ter lakes sampled in May 1987 are up to 7 times bulk buoyancy of the fluid and also an increase that found in Nyos: Lake Manengouba, 91 m, in turbulence. In such an event S will be re- PCH4=0.95 atm; Lake Benakuma, 130 m, duced independent of external energy inputs. It PCH4 = 2.08 atm; Barombi Mbo, 98 mPCH4 = 0.85 is important to note that in a gas-charged lake, atm. If pre-event pressures in Lake Nyos were higher than present, they would have facili- when gas concentrations approach saturation tated initial exsolution because CH4 is about 25 the conventional models of water column sta- times less soluble than COe. In any case, con- bility break down. In the prediction or assess- centrations of CH4 and N2 should be monitored ment of potential gas releases then, stability in Lake Nyos in order to accurately predict to- values should be interpreted in conjunction with tal gas pressure and thus the stability of the an estimate of the minimum vertical distance a system. water mass must move in order to create oversaturated conditions. Isotopes Water column stability increased from 48,800 J m -2 in September of 1986 (assuming a CO2 The isotopic composition of carbon species profile identical to the averaged profile mea- in the lake is relatively unchanged since the sured in January 1987), to 50,700 J m -2 in Jan- 162 G.W. KLING ET AL. uary, and to 64,700 J m -2 in May. Water den- events) could destabilize Lake Nyos and pre- sity increases due to the input of solutes to cipitate large-scale degassing of C02. bottom waters over this time period are mainly Two possible models of large-scale degassing responsible for the higher S values in January are isenthalpic (see Henley et al., 1984) or is- and May. In addition, surface waters cooled entropic (see Kieffer, 1982). In either process slightly from September to January and warmed there is a rapid transition from low velocity slightly from January to May. This explains the buoyant gas flow to a high velocity flow regime smaller increase in S during this first time pe- driven by the expansion force produced during riod compared to the increase in S from Janu- exsolution. For example, water at 200 m in Lake ary to May. These stability values are ten times Nyos contains about 0.7 mol L -1 of CO2 at sat- greater than those for Barombi Mbo, the most uration. Isenthalpic decompression of this C02 stable low-CO2 lake in Cameroon (Kling, 1988). exsolved at the bottom of Lake Nyos (21 atm) In addition, the Lake Nyos values are more than and brought to the surface (1 atm ) would pro- twice that for Lake Atitlan, which is the most duce 928 J of energy through expansion. This stable low-CO2 lake known (Kling, 1988). By is sufficient to drive 1 kg of water to a height of comparison, using data for gas-charged Lake 95 m above the lake. Isentropic decompression Kivu in East Africa taken from Tietze (1978) would produce even more energy. Regardless of and Degens et al. (1973), we calculate Kivu's which model is chosen to best represent the stability to be ~340,000 J m -2. Lake Kivu is thermodynamics of degassing at Lake Nyos, the extremely stable because of its great depth (485 energy gained from this expansion force is suf- m) and because solutes in its bottom water are ficient to generate the tremendous water surges five times more concentrated than those found caused by the August event (described by Kling in Lake Nyos. These comparisons among gas- et al., 1987). charged lakes are for perspective only, because localized gas saturation can quickly reduce stability. Regional survey The vertical movement of water required to achieve localized gas saturation is equal to sample depth minus saturation depth. In May this We completed our survey of all Cameroon 'saturation distance' ranged from 170 m for the crater lakes in May 1987. Morphometric de- 198 m samples, to 49 m for the 61 m sample scriptions, thermal profiles, and some chemical (Fig. 6). In other words, if water at 61 m was characteristics of these lakes are presented in raised to 12 m, exsolution of gas would occur Kling (1988). Bottom water temperatures in all given suitable nucleation sites. In this specific lakes were no higher than those in other tropi- case, degassing of a small volume of water cal lakes at similar elevation and latitude; we brought from 61 to 12 m would not trigger a found no abnormal input of heat attributable massive gas release, because CO2 concentra- to volcanic activity. Only Lakes Nyos and tions in epilimnetic water are small. The com- Monoun had elevated concentrations of dis- mon limnological processes that transport water solved gas or ions in the hypolimnion. Given vertically, such as an internal seiche or currents that the buildup of CO2 and solutes must have driven by wind or convection, are insufficient preceded the gas bursts in Lakes Nyos and in magnitude to move water 50 m or more Monoun (as discussed above), in the absence through the strong density gradients in this of premonitory chemical changes in bottom lake. If saturation levels remain unchanged, waters there is a low probability of any other only large scale external inputs of energy Cameroonian lake releasing a substantial vol- (rockslides, subsurface injection of gas, seismic ume of C02. THE EVOLUTION OF THERMAL STRUCTURE AND WATER CHEMISTRY IN LAKE NYOS 163

Summary creased recently due to solar warming of surface waters and input of solutes to the There is now good evidence suggesting that hypolimnion from groundwater. At present the the Lake Nyos disaster was non-volcanic in na- level of stability and the level of C02 saturation ture and that the released CO2 was stored in the combine to create conditions that could be dis- hypolimnion rather than in or below the lake rupted only by strong external inputs of energy sediments. For example, several observations such as rockslides or seismic events. Of more require high CO2 concentrations in the lake immediate concern appears to be the potential prior to the August 1986 event. Increasing sol- for catastrophic flooding resulting from loss of ute concentrations with depth argue for gas the weak natural dam holding back the upper buildup before the event in order to create 40 m of Lake Nyos (Lockwood et al., 1988). strong density stratification. This density Such an event could trigger the release of much stratification is needed also to allow for the ac- of the gas remaining in the lake. This danger cumulation of heat and to maintain the inverse seems to warrant continued evacuation of low- temperature profile of Lake Nyos. Without high lying areas downstream of Lake Nyos. pC02 values before the event, Fe 2+ would pre- cipitate as FeCO3 and thus dissolved iron con- Acknowledgements centrations would remain low. Most likely the effervesence of gas due to lo- We thank the Cameroon Ministry of Higher cally oversaturated conditions was the mecha- Education and Scientific Research, especially nism of release during the event. The trigger the Institute for Geology and Mines Research, creating these conditions is unknown. If the re- for their cooperation and support in this work. lease of gas was initiated in near surface waters, This research was supported also by NSF grants as described by the model of Tietze (1987), sev- 8400532 and 8544206 to D.A. Livingstone and eral of the more common water movements such 8709978 to G.W.K., by the U.S. Geological Sur- as internal seiches or strong penetrative con- vey, and by the U.S. Embassy in Cameroon. vection may have triggered the release. A model Greg Tanyileke, Gamnje Ngala, Eyike Emman- of near surface release does not, however, elim- uel, Peter Enyong, Ndoni Paul, Jack Lock- inate potential triggers originating in deep wood, Haroldur Sigurdsson, and Steve Tebor water, such as a thermal plume or subsurface provided field assistance. Paul Briggs and Doug landslide. For example, the only possible loca- White helped with chemical analyses. We also tion of an injection of CO2, accumulated in dia- remember Helimission pilot Steve Bartalsky, treme voids, is from cliffs below the water sur- who died in a helicopter crash only 3 months face that are clear of sediment. Such an injection after helping us reach the most remote lakes in also could act as a trigger for the initial gas re- Cameroon. lease as suggested by Kling et al. (1987). It must be emphasized that at present there are no data Appendix 1 which demand a unique interpretation of the processes involved. (1) Lake stability (S) was calculated follow- There is distinct horizontal variation in the ing Idso (1973) and expressed in J m-2: concentration of CO2 in Lake Nyos and overall its distribution is still poorly characterized. It S=g/Aof(Z-Z. ) (p~ -

)Az dz is apparent that no major recharge of CO2 has where g is acceleration due to gravity, Ao and Az occurred since the 1986 event, although it would are surface area and area at depth z, z. is the be premature to assign a recharge rate with sta- depth of mean density, pz is the density at depth tistical bounds. Water column stability has in- z, and

is the mean density calculated as: 164 G.W. KLINGET AL.

mol- 1 at 295 ° K (Weiss, 1974 ). CO2 concentra-

= (i/Vo)fAzp~ dz tions were then calculated as: (2) Fluid density at each depth was calcu- [CO2] =Kofexp[ (1-P)v*/RT] lated as: where Ko is the virial equation of state derived p= [1000+ (m M) l/[ (m~v) + (1000/p0) ] from the van't Hoff equation and v* is the par- where m is the mean molality of the lake water, tial molal volume of CO2 (Weiss, 1974). M is the mean molecular weight of solutes in the water, po is the density of pure water at tem- References perature T (°C) and pressure P (bar), and ~v is the mean apparent molal volume of the lake Barberi, F., Chelini, W., Marinelli, G. and Martini, M., 1989. water (Millero, 1972). Values for the density of The gas cloud of Lake Nyos (Cameroon, 1986): Results pure water were calculated using the relations of the Italian Technical Mission. In: F. Le Guern and in Chen and Millero (1977). In mixed electro- G. Sigvaldason {Editors), The Lake Nyos Event and Natural C02 Degassing. J. Volcanol. Geotherm. Res., 39: lyte solutions the apparent molal volume (0v) 125-134 (this issue). of an ion is a function of the molar concentra- Baxter, P.J. and Kapila, M., 1989. Acute health impact of tion {c) of the medium as shown by the Red- the gas release at Lake Nyos, Cameroon, 1986. In: F. Le lich-Meyer equation (Redlich and Meyer, 1964) Guern and G. Sigvaldason (Editors), The Lake Nyos where: Event and Natural CO2 Degassing. J. Volcanol. Geoth- erm. Res., 39:265-275 (this issue). Ov =f)° + Sv(pom) ½+ Bv(Pom ) Beck, A.E. (Editor), 1985. Terrestrial heat flow and thermal regimes. Tectonophysics, 121: 1-97. is the ionic partial molal volume. Conven- Cerm~k, V., Rybach, L. and Chapman, D.S. (Editors), 1984. tional partial molal volumes reported in Mil- Terrestrial heat flow studies and the structure of the lithosphere. Tectonophysics, 103: 1-356. lero (1972) were corrected to ionic partial mo- Chen, C.-T. and Millero, F.J., 1977. The use and misuse of lal volumes using Vo(H + ) =-5.0 cm 3 tool -I. pure water PVT properties for lake waters. Nature, 266: The ¢~¢ of C02 was taken from Weiss (1974) 707-708. and that of CH4 was taken from Tiepel and Degens, E.T., Von Herzen, R.P., Wong, H.-K., Deuser, W.G. and Jannasch, H.W., 1973. Lake Kivu: Structure, chem- Gubbins (1972). S~ is a theoretical limiting istry and biology of an East African rift lake. Geol. slope given in mixed solutions by kz2I ½where k Rundsch., 62: 245-277. is determined by temperature, z is ionic charge, Evans, W.C., White, L.D. and Rapp, J.B., 1988. Geochem- and I is ~Xcizi2. Bv is an empirical deviation istry of some gases in hydrothermal fluids from the southern Juan de Fuca ridge. J. Geophys. Res., 93: 15305- constant determined for each ion. These devia- 15313. tion constants were calculated assuming addi- Failer, A.J., 1978. Experiments with controlled Langmuir tivity and using Bv for NaC1, KC1, CaC12, MgCI2, circulations. Science, 201: 618-620. KF, NaF, and K2S04 from Millero (1971), and Hem, J.D., 1970. Study and interpretation of the chemical characteristics of natural water. U.S. Geol. Surv., Water- B. for NaHCO3 from Perron et al. (1975). Mean Supply Pap. 1473. apparent molal volume was then calculated us- Henley, R.W., Truesdell, A.H., Barton, P.B., Jr. and Whit- ing the mixture rule of Young and Smith (1954) ney, J.A., 1984. Reviews in Economic Geology, Vol. I. where: Fluid-mineral Equilibria in Hydrothermal Systems. Economic Geology Publishing Co., E1 Paso, 267 pp. (I)v =X mi~)vi/ X mi Hutchinson, G.E., 1957. Treatise on Limnology, Vol. I. Wiley, New York, N.Y., 1015 pp. (3) Fugacity of CO~ was determined by the Idso, S.B., 1973. On the concept of lake stability. Limnol. use of the fugacity-pressure relationship In Oceanogr., 18: 681-683. (f/P) = BtP/RT, where [ is the fugacity at pres- Kieffer, S.W., 1982. Dynamics and thermodynamics of volcanic eruptions: Implications for the plumes of Io. In: sure P and absolute temperature T, R is the D. Morrison (Editor), Satellites of Jupiter. Univ. of Ar- universal gas constant, and Bt=-0.1233 L izona Press, Tucson, Ariz., pp. 647-723. THE EVOLUTIONOF THERMALSTRUCTURE AND WATERCHEMISTRY IN LAKENYOS 165

Kling, G.W., 1987a. Seasonal mixing and catastrophic de- erty, D.L., 1981. Importance of particulate layer for- gassing of tropical lakes, Cameroon, West Africa. Sci- mation and internal radiation absorption in the devel- ence, 237: 1022-1024. opment of the thermocline. 44th Annu. Meet., Am. Soc. Kling, G.W., 1987b. The comparative limnology of lakes in Limnol. Oceanogr., Milwaukee, WI. Cameroon, West Africa. Ph.D. thesis, Duke University. Schreiner, S.P., 1984. Particulate matter distribution and Kling, G.W., 1988. Comparative transparency, depth of heating patterns in a small pond. Verh. Int. Verein. mixing, and stability of stratification in lakes of Came- Limnol., 22: 119-124. roon, West Africa. Limnol. Oceanogr., 33: 27-40. Siegel, D.I. and Pfannkuch, H.O., 1984. Silicate mineral Kling, G.W., Clark, M.A., Compton, H.R., Devine, J.D., dissolution and pH 4 and near standard temperature and Evans, W.C., Humphrey, A.M., Koenigsberg, E.J., pressure. Geochim. Cosmochim. Acta, 48: 197-201. Lockwood, J.P., Tuttle, M.L. and Wagner, G.N., 1987. Siever, R. and Woodford, N., 1979. Dissolution kinetics and The 1986 Lake Nyos gas disaster in Cameroon, West the weathering of mafic minerals. Geochim. Cosmo- Africa. Science, 236: 169-175. chim. Acta, 43: 717-724. Kusakabe, M., Ohsumi, T. and Aramaki, S., 1989. The Lake Sigurdsson, H., Devince, J.D., Tchoua, F.M., Presser, T.S., Nyos gas disaster: chemical and isotopic evidence in Pringle, M.K.W. and Evans, W.C., 1987. Origin of the waters and dissolved gases from three Cameroonian cra- lethal gas burst from Lake Monoun, Cameroon. J. Vol- ter lakes, Nyos, Monoun and Wum. In: F. Le Guern and canol. Geotherm. Res., 31: 1-16. G. Sigvaldason (Editors), The Lake Nyos Event and Sigvaldason, G.E. (Editor), 1987. Conclusions and rec- Natural C02 Degassing, I. J. Volcanol. Geotherm. Res., ommendations of the international scientific confer- 39:167-185 (this issue). ence on the Lake Nyos Disaster. UNESCO, Yaounde, Lockwood, J.P., Costa, J.E., Tuttle, M.L., Nni, J. and Te- Cameroun, March 1987, 21 pp. bor, S.G., 1988. The potential for catastrophic dam fail- Tazieff, M.H., 1987. Lake Nyos new phreatic eruptions. A ure at Lake Nyos maar, Cameroon. Bull. Volcanol., 50: note distributed at the Hawaii Symposium on How Vol- 34O-349. canoes Work, Hilo HI, January 1987. Marechal, A., 1976. G~ologie et G~ochimie des Sources Tazieff, M.H., Faivre-Pierret, R.X., LeGuern, F. and Chev- Thermomin~rales du Cameroun. ORSTOM, Paris, 169 rier, R.M., 1987. La catastrophe de Nyos, Republique pp. du Cameroun. Centre des Faibles Radioactivit~ CNRS Millero, F.J., 1969. The partial molal volumes of ions in Contrib. No. 811, 76 pp. seawater. Limnol. Oceanogr., 14: 376-385. Tiepel, E.W. and Gubbins, K.E., 1972. Partial molal vol- Millero, F.J., 1971. The molal volumes of electrolytes. Chem. umes of gases dissolved in electrolyte solutions. J. Phys. Rev., 71: 147-176. Chem., 76: 3044-3049. Millero, F.J., 1972. The partial molal volumes of electro- Tietze, K., 1978. Geophysikalische Untersuchung des Ki- lytes in aqueous solutions. In: R.A. Horne (Editor), vusees und seiner ungewShnlichen Methangaslager- Water and Aqueous Solutions. Wiley-Interscience, New st~itte-Schichtung, Dynamik un Gasgehalt des Seewas- York, N.Y., pp. 519-595. sers. Thesis, Christian-Albrechts Univ., Kiel. Newman, F.C., 1976. Temperature steps in Lake Kivu: A Tietze, K., 1987. The Lake Nyos gas catastrophe in Ca- bottom heated saline lake. J. Phys. Oceanogr., 6: 157- meroon: cause, sequence of events, consequences. In: 163. W.H. Graf (Editor), Topics in Lake and Reservoir Hy- Parkhurst, D.L., Thorstenson, D.C. and Plummer, L.N., draulics. Proc. 22nd Congress of IAHR, Lausanne, 31 1980. PHREEQE-a computer program for geochemical August-4 September 1987, pp. 93-98. calculations. U.S. Geol. Surv., Water-Resour. Invest., 80 Turner, J.S., 1973. Buoyancy Effects in Fluids. Cambridge (96): 1-209. Univ. Press, New York. Perron, G., Desnoyers, J.E. and Millero, F.J., 1975. Appar- Tuttle, M.L., Clark, M.A., Compton, H.R., Devine, J.D., ent molal volumes and heat capacities of some sulfates Evans, W.C., Humphrey, A.M., Kling, G.W., Koenigs- and carbonates in water at 25°C. Can. J. Chem., 53: berg, E.J., Lockwood, J.P. and Wagner, G.N., 1987. The 1134-1138. 21 August 1986 Lake Nyos gas disaster, Cameroon. U.S. Quay, P.D., Broecker, W.S., Hesslein, R.H. and Schindler, Geol. Surv., Open-File Rep. 87-97, 58 pp. D.W., 1980. Vertical diffusion rates determined by tri- Weiss, R.F., 1974. Carbon dioxide in water and seawater: tium tracer experiments in the thermocline and hypo- the solubility of a non-ideal gas. Mar. Chem., 2: 203- limnion of two lakes. Limnol. Oceanogr., 25: 201-218. 215. Redlicb, O. and Meyer, D.M., 1964. The molal volumes of Young, T.F. and Smith, M.B., 1954. Thermodynamic prop- electrolytes. Chem. Rev., 64: 221-227. erties of mixtures of electrolytes in aqueous solutions. J. Sass, J.H., Kennelly, J.P., Jr., Wendt, W.E., Moses, T.H., Phys. Chem., 58: 716-724. Jr. and Ziagos, J.P., 1981. In-situ determination of heat Zaneveld, J.R.V., Kitchen, J.C. and Pak, H., 1981. The inflow in unconsolidated sediments. Geophysics, 46: 76- fluence of optical water type on the heating rate of a 83. constant depth mixed layer. J. Geophys. Res., 86: 6426- Schindler, J.E., Waldron, M.C., Eubanks, J.O. and Daugh- 6428.