In 1986 the small, roughly circular Lake Nyos in the Cameroon highlands silently released a massive cloud of carbon dioxide. Being a dense gas it hugged the ground and flowed down valleys for up to 25 km. 1700 local people perished by suffocation, together with their livestock (See Geohazards 2000). Having a recent volcanic origin, the lake is fed by springs in its bed that contain dissolved CO2 emitted from the residual magma chamber below. At 200 m deep the bottom water is sufficiently pressurised to retain the dissolved gas so that signs of the potential hazard remain hidden until such a limnic eruption occurs. Far larger, with a surface area of 2700 km2, Lake Kivu bordered by Rwanda and The Democratic Republic of Congo, is even deeper (up to 470 m). It too lies within a volcanically active zone, in this case the western arm of the East African Rift System. Being one of the most nutrient-rich bodies of fresh water on Earth, its biological productivity is extremely high, so as well as bottom water enriched in dissolved CO2 – a staggering 256 km3 – methane (CH4) is also present in very large amounts (~65 km3). This comes partly from anaerobic decay of dead organisms and from microbial reduction of the magmatic CO2 passing through its bottom sediments. Sulfate-reducing bacteria also generate toxic hydrogen sulfide (H2S) in the anoxic bottom waters – Lake Nyos contains less dissolved salts and did not emit H2S.
So Kivu presents a far greater hazard than the volcanic lakes of Cameroon and an emission of a dense gas mixture might fill the rift valley in the area to a depth of about a hundred metres. Being highly fertile the valley around the lake has a high population (2 to 3 million), so the death toll from a limnic eruption could be huge. A further hazard stems from tsunamis generated by such gas bursts. Once bubbles form at depth the bulk density of water drops, so large masses of water surge to the surface rather than the gas itself; a phenomenon known to happen in the periodic eruptions of Lake Nyos. What might trigger such an event in Lake Kivu? The East African Rift System is seismically active, but recent earthquakes did not result in limnic eruptions. Subaqueous volcanic eruption is the most likely to set one off. A surface lava flow from the nearby Mount Nyiragongo entered the lake at the town of Goma in 2002 but, fortunately, did not reach the threatening deeper part of Kivu. Sediment samples from the lake reveal periodic transport of land vegetation to its deeper parts, roughly every thousand years. The sediments with plant fossils also contain abundant remains of aquatic animals, suggesting both tsunamis accompanied by toxic emissions.
Mitigating the hazard of limnic eruptions at Lake Nyos was made possible in 2002 by linking its bottom waters to the surface by plastic piping. After initial pumping, the release of bubbles at shallower depths and the resulting fall in bulk water density set off something akin to a large soda siphon, slowly relieving the deeper layers of their load of dissolved CO2. This resulted in 50 m high fountains of what was effectively soda ‘pop’. In 2009 this was repeated on a far larger scale on Lake Kivu, the operation being paid for by separation and sale of methane. Yet even this attempt at mitigation has its risks: first of destabilising what may be a fragile equilibrium to trigger a limnic eruption; second by lifting nutrient-rich bottom water that would encourage algal blooms at the lake surface and potential deoxygenation. The current issue of the Journal of African Earth Sciences includes a detailed review of the issues surrounding such dual-purpose hazard mitigation (Hirslund, F. & Morkel, P. 2020. Managing the dangers in Lake Kivu – How and why. Journal of African Earth Sciences, v. 161, Article 103672; DOI: 10.1016/j.jafrearsci.2019.103672). By 2015 the Rwandan KivuWatt Methane Project had a capacity for 25 MW of electrical power generation.
Running at full capacity, degassing the depths of Lake Kivu would provide the economic benefit of low-cost electricity for Rwanda and the DRC, at a maximum generating capacity of 300 mW using the most efficient power plant, as well as removing the risk of a catastrophic gas release. Yet the release of CO2 from the lake and from methane burning would increase atmospheric greenhouse warming significantly, albeit less than if the methane was simply released, for CH4 has 25 times the potential for trapping outgoing heat. Hence the dilemma. Either way, there remains the risk of turning Kivu’s surface water into an anoxic algal ‘broth’ with devastating effects on its fishery potential. Burial of the dead phytoplankton, however, might generate more methane by bacterial decay; a possible source of renewable biofuel that ‘recycles’ the atmospheric CO2 consumed by algal photosynthesis. The geohazards, according to Hirslund and Morkel, are really the ultimate driver for development of Lake Kivu’s fossil fuel potential, now that they are better understood as a real and present danger to millions of people. The authors calculate that a catastrophic gas release may be on the cards in the late 21st century. Yet there are other resource issues bound up with the health of the lake’s surface waters. Preserving the layered structure of the lake water to some extent is also important. Until the rates of natural infiltration of volcanic CO2 and biogenic production of methane are known, a minimum rate of gas extraction to make the lake safe is impossible to calculate. Perhaps matching those rates with gas removal should govern future operation. The total methane content of Lake Kivu is just 1.5 times the annual production from the UK sector of the North Sea. It is sufficient for power generation at 300 MW, at most, for 50 years, which would roughly double Rwanda’s current installed generation capacity – mainly from hydropower. Although Kivu is shared equally between Rwanda and the DRC even half of the short term power potential would be a significant benefit to Rwanda’s ~11 million people, though considerably less to the ~81 million living in the DRC; if access was equitable.