Nyiragongo and Nyamuragira: a review of volcanic activity in the Kivu rift, western branch of the East African Rift System

Abstract

Nyiragongo and Nyamuragira are two active volcanoes of the Western branch of the East African Rift in the Virunga area. They were built at the Kivu rift axis ca. 12,000 years ago and set above two tectonic steps separated by the Kameronze Fault. Both volcanoes have displayed a succession of intra-crater and flank eruptions that have been observed and documented since the end of the nineteenth century. Here, we have collated and reviewed these publications and reports. Nyiragongo is famous for its semi-permanently active lava lake, which at the time of writing (2020) was the largest in the world. During the construction of the main stratovolcano, which ended a few centuries ago with a caldera collapse, the lava composition changed from melilitite to leucite and then to melilite-bearing nephelinite. The historically active lava lake is believed to be directly fed from an upper intra-volcano reservoir, a shallow reservoir situated a few kilometres below the volcano in the granite basement, and a deeper intra-crustal magma chamber. Historic activity has been documented since 1894 and can be divided into eight stages, on the basis of sudden changes between lava filling and draining, with cycles of rising lava lake activity and overflows, followed by sinking and complete or partial drainage. Twice in recent history (in 1977 and 2002), major flank eruptions were accompanied by complete drainage of the lava lake and the upper plumbing system. The lava that filled the crater since 1948, and then again after the 1977 and 2002 drainage events have been calculated at a cumulative volume of around 324 × 10⁶ m³. In comparison, the 1977 and 2002 flank eruptions involved 47 × 10⁶ m³ of lava. The average annual output rate associated with crater filling is thus estimated at between 4 and 13 × 10⁶ m³. Lava lake behaviour changes from equilibrium, with alternation between gas pistoning and spattering regimes through disequilibrium with intermittent activity, to complete disappearance of the lava lake. These changes can be related to the conditions of the descent of dense degassed magma from the upper conduit into the shallow reservoir. However, since 1959, the chemical composition of the leucite and melilite-bearing nephelinite lavas has not significantly changed, which implies a magma supply from the same magma batch. Nyamuragira was characterised by a shield building phase of activity until a caldera collapse, ca. 300 to 500 years ago. The post-caldera phase has involved effusive activity in the caldera and at numerous flank fissures. The plumbing system consists of an upper reservoir roughly at the basement-volcano interface and averaging a volume of 400 × 10⁶ m³, a shallow upper-crust stratified reservoir, and a middle-crust mafic magma chamber. Volcanic activity has involved a succession of filling and emptying events at the upper reservoir. Lava volumes of historic eruptions reveal annual output rates averaging 14 × 10⁶ m³ between 1901 and 1976, and 40 × 10⁶ m³ between 1976 and 2012. A drastic increase in activity occurred in December 1976. This event coincided with the January 1977 flank eruption of Nyiragongo and resulted from a main tectonic event in the rift basement that improved the efficiency of magma ascent at both volcanoes. The historic lava composition can be related to six cycles of magma accumulation in, and withdrawal of, the upper reservoir from the shallow stratified reservoir. Similar magma storage and transport systems are known in many effusive systems: the Nyiragongo lava lake shares behavioural characteristics similar to those observed at Kilauea, Erebus, and Erta’Ale. At Nyiragongo and Nyamuragira, magma supply and persistent activity with sudden changes of the magma output rates in relation to tectonic events are also comparable with those of Kilauea, Piton de la Fournaise of Réunion island, and Mount Etna.

ANNEX. Nyiragongo historic activity

Nyiragongo’s historic activity is described in eight stages.

• 1. Quietness to moderately active stage. From 1894 to 1927 no eruptions were reported. Evolution was limited to enlargement of the sink in the upper platform, from two small circular holes in 1894 to three holes in 1911, which coalesced and widened into a single hole in 1918. From time to time (1894, 1905, 1911–1914, 1918–1924) large clouds of steam rose from the vents. They revealed activity at depth may be linked to the formation of the second inner platform.

• 2. Sub-continuous lava lake activity (1927–1966), a deep stage. At the beginning of 1927 steam clouds rapidly increased in volume. From 1928 a permanent red glow at night suggested the existence of an active lava lake which was discovered in 1930 by R. D. Burtt (Record in the Virunga National Park’s Guest Book and Hoier 1950). The lava lake level was roughly estimated at 3070 m a.s.l. Only scattered observations are available from 1930 to 1948. Accurate reports of the activity were made in 1948, 1953, 1954 to 1959, 1965, and 1966. The crater exhibited three nested sinks, with, from top to bottom, the upper wall (160–200 m in height), the first platform at 3270 m a.s.l., the second wall (190 m in height), the second platform at 3080 m a.s.l., the third wall and the third platform where the lava lake was located.

  A large block that detached and fell from the first or the second platform was floating in the lava lake and was named “the crag” (to remember the crag of the Halemaumau crater of the Kilauea volcano of Hawaii). It was a 220-m-wide trapezoidal-shaped boulder tilted 20° towards the SE and emerged from the lake at a height of 50 m. The lake level was variable. It was at the second platform level in 1948 (ca. 3080 m a.s.l.), 20 m below the second platform in 1953, 40 m in February 1956, 30 m in July 1956 and 1958, 40 m in 1959, and 10 m in March 1965. It again reached the second platform in December 1965. These annual variations are averages of weekly or daily oscillatory ascents and drops of more or less 10 m. The “crag” has followed these oscillations and gone up and down with the lava lake level. Eruptive activity alternated between violent explosions with abundant spattering and lava overflowing or with only convective motion. The duration of both highly active and calm periods ranged from a few days to a few weeks. Variations of the intensity of the lava lake activity have been detected by the brightness of the nightglows.

  In 1966, the lava temporarily filled the second inner pit and the activity changed to intermittent eruptions with inactive periods. This change was due to a widening of the crater from 360 to 650 m in diameter, when the lava covered the second platform.

  Interesting indications of the lava lake working have been given by the setting of the “crag”. From 1948 to 1976, the buoyant behaviour of the crag has resulted in a vertical displacement of at least 190 m. It is clear that the crag was supported by molten lava. The density of solid vesicular lava (ca. 2.7) is slightly lower than that of the molten lava (ca. 2.8), so the crag floated like an iceberg. The underlying magma has filled a deep cylindrical pipe averaging 350 m in diameter. The surface lava seeped all around the crag and appeared in a large open and crescent-shaped area about 100 m wide to the west of the floating boulder. The semi-permanent existence of a lava lake means that deep heat supply equalled the heat loss that was limited by the shield effect of the crag. Variations of the intensity of the eruptive events were due to imperfect equilibrium. During periods of high heat loss, the magma head is cooled, becoming more viscous, which limited the outgassing. In turn, fountaining decreases and the lava lake is covered with a thick black skin. However, heat and gaseous transfer continued and trapped beneath the solid skin, which was rapidly destroyed by the gas flux overpressure to begin a new eruptive phase.

• 3. Episodic intra-caldera eruptions (1966–1972). In February 1966, the active lava level was close to 3100 m a.s.l., approximately 170 m below the first platform, with limited activity at the west side. Cessation of the activity was followed by a 10-m collapse in the middle part of the pit in December 1966. After a three-year inactive period, the eruptive regime resumed in 1970 with an alternation of eruptive events and inactive phases. Eruptive events occurred every 16 to 21 days (Pouclet 1973b). These events began with fountaining and the formation of five to eight spatter cones in a sub-circular fissure 300 to 400 m in diameter, directly above the former margin of the third pit of the second stage. Thin lava flows spread out and covered the lower platform as a temporary enlarged lava lake. After 5 to 6 days, the activity waned and ended. Upon cessation of activity, a collapse of a few metres of the inner part of the active circle occurred with withdrawal of the lavas. However, at each eruptive event, the floor has risen of 1 to 2 m by overflowing, and its levels progressively reached 3130 m a.s.l. in June 1970, 3170 m a.s.l. in September 1970, 3210 m a.s.l. in March 1971, 3235 m a.s.l. in July 1971, 3260 m a.s.l. in January 1972, and 3270 m a.s.l. in April 1972. The crag ascended with the magma, but in July 1971 it disappeared below its surrounding flows. In early April 1972, lava flows covered the upper platform for the first time. This was followed by a collapse of 20 to 25 m in the central pit, which created a sink 300 to 400 m in diameter. The lava withdrawal to the new pit cleared the last part of the second wall and revealed a 2-m circular cliff in the first platform.

• 4. Semi-continuous lava lake activity (1972–1977). After the lava withdrawal to the new inner pit, the lava lake remained active (Durieux 2002a; Pottier 1978). Its level fell in 1972 and rose again in 1973, when lava flows again overflowed the first platform. At the end of 1973, a collapse occurred in the same area of the former third pit, as a circular sink 300 m in diameter. In this restricted area, the lava lake was still active. The lake level ascended in the middle of 1974, then subsided until July 1976, after which it rose again to reach the first platform by the end of 1976. From 1975 to 77, semi-continuous activity of the lava lake resulted in multiple overflows above the first platform.

  It is noteworthy that, at each ascent event, the lava never climbed much higher than the first platform level (3270 m a.s.l.). This height would appear to be the barometric equilibrium for the magma column.

• 5. Flank eruption, 1977. On 10 January 1977, the first historic flank eruption occurred on the northern and southern flanks along N–S fractures (Fig. 5). To the north, three fractures, around some hundred metres in length, opened in the Baruta crater and along the northern and north-western flanks. To the south, a fissure cut the Nyiragongo and Shaheru craters and a 3 km-long fracture extended from the upper slope of Shaheru to the lava plain. Highly fluid lavas spread out from these fractures mainly to the southern lava plain within a period of less than one hour.

  The magma column sink was immediately followed by a collapse of the whole terrace system that created an 800 m-deep new pit below the rim of the crater. The volume of this new pit from the upper platform level is calculated at 210 × 10⁶ m³. This volume of rock, which is greater than the volume of the erupted lava, disappeared into and beneath the volcano. Seismic activity registered before and after the 1977 eruption showed volcanic earthquakes at the northern part of Nyiragongo, which indicated a normal faulting with the tension axis nearly perpendicular to the strike of the rift axis. At the same time, tectonic earthquakes were located at the southern flank of Nyiragongo and in Lake Kivu with epicentral distribution along the rift axis (Hamaguchi et al. 1992). These data confirm the field observation of a major tectonic motion along the continuation of the N–S Kisenyi Fault (Figs. 3 and 4).

  Some authors have thought that the simultaneous eruptions of Nyamuragira (December 23, 1976) and Nyiragongo (January 10, 1977) may indicate a connection between the two magmatic reservoirs (Tazieff 1977). This cannot be true, as petrological and chemical differences between the magma compositions of the two volcanoes preclude any communications between their volcanic crustal reservoirs (Pouclet et al. 2016) but it is true that both eruptions were triggered by the same major tectonic event in the Kivu rifting system.

• 6. The 1982 renewal and 1994–1995 episodic activity. Following four years and five months of quiescence, new lava erupted in the sink floor, at 2670 m a.s.l., on 21 June 1982, and rapidly filled the crater (Krafft and Krafft 1983; Tazieff 1984). The molten lava ascended to 3090 m a.s.l. in October-November 1982 and crusted. After a quiescence of 11 years and 7 months, activity was renewed in June 1994 and the lava lake level reached 3175 m a.s.l. in December 1995 (Tedesco 2002) and finally 3190 m a.s.l. Activity ceased from 1996 to January 2002.

• 7. Second flank eruption, 2002. On 17 January 2002, a series of fractures opened from the upper southern flank to the lava plain along a 17-km-long N–S network (Fig. 6). Lava flows erupted from eight vents for 24 to 48 hours (Komorowski et al. 2002). A 2-m-thick flow reached Lake Kivu and destroyed the eastern part of the town of Goma. The lava volume is estimated at 25 × 10⁶ m³. This eruption was immediately preceded and followed by numerous large tectonic earthquakes. Epicentres have been localised from Nyiragongo’s upper flanks to Lake Kivu northern basin (Kavotha et al. 2002). At least 18 large shocks (M ≥ 3) occurred in January followed by many aftershocks during 2002. The volcanic activity clearly resulted from reactivation of the N–S rift-related fault system and, as in 1977, from newly formed fractures. However, the seismic activity was much more intense in 2002. A NW–SE radial fracture at the upper NW upper slope propagated from the north, at 2800 m a.s.l., to the south, at 1580 m a.s.l., close to the outskirts of Goma. Along most of its course it formed a 5 to 20 m-wide graben structure intruded with dyke magma. Only one short fracture, trending WNW–ESE (vent 6 of Komorowski et al. 2002), was controlled by the transverse fault between the Virunga shoal and the northern basin of Lake Kivu (Fig. 5).

  The eruption was followed by crater collapse in the night of 22–23 January. This collapse was limited to the middle part of the 1995 lava floor, averaging 670 m in diameter and 540 m deep below the 1995 platform (bottom at 2635 m a.s.l.). The volume of this new funnel-shaped crater is calculated at 63 × 10⁶ m³. This is smaller than for the 1977 sink, which involved the whole upper platform area. It is worth noting that both collapses reached about the same depth. This level could be the transition zone from the upper wide pipe to the lower narrow conduit. As in 1977, the volume of rock that disappeared is greater than that of the spilled-out lava. This disappearance could be caused by a ring-faulting mechanism as suggested by Shuler and Ekström (2009). These authors have studied five unusual earthquakes with anomalous frequency contents that occurred near Nyiragongo in 2002–2005. They interpret these events “as being generated by slip on inward-dipping conical ring faults under the volcano”. Evacuation of the magma column by the flank eruption causes an underpressure in the shallow magma chamber provoking collapse of the roof of the magma chamber along cone-shaped ring faults.

  Unlike the 1977 collapse, the 2002 one did not happen during lava lake activity. The lava lake was solidified at 3165 m a.s.l. An increase of volcanic seismicity was recorded in December 2000 and during 2001, but at that time, Nyamuragira was very active with lava filling of the caldera basin and it may be the site of the recorded seismic activity. However, in the Nyiragongo crater, there was a significant increase of heat flux and vapour degassing through the lava lake crust. The initial stage of the eruption of 17 January 2002 at the upper vents was only effusive and devoid of explosive features (no scoria or spatter products). It is likely that this degassed lava drained from residual stored molten lava below the encrusted carapace of the 1995 platform (Komorowski et al. 2002). The volume of 2002-stored lava was not sufficient to provide all the erupted lavas. Indeed, protracted flank eruptive activity was explosive with lava fountains building scoria and spatter cones along propagating dykes. This eruptive style involves the supply of new magma from a deep source. Komorowski et al. (2002) have suggested that this new magma bypassed the upper edifice conduit that was depressurised after the lava lake drainage. Firstly, during 2001, evidence of significant heat transfer indicated an ascent of magma in the volcano conduits and fractures that were created by tectonic shocks. Secondly, drainage of the stored lava was initiated by the flank fracturing, which subsequently allowed the eruption of fresh lava. The 2002 eruption thus resulted from the addition of two factors: magma overpressure and tectonic fracturing. Modelling of the remote sensing InSAR data suggests the opening of two dykes in the upper crust and the volcano shield according to Wauthier et al. (2012).

• 8. Last renewal. Four months after the 2002 eruption, a small lava fountain appeared on the floor of the crater in May 2002. Sporadic fountaining led to the development of a new lava lake in 2003. Compared with the 1977 eruption this activity came into sight very soon after the crater collapse. But it was weaker than during the 1982 renewal. First, in 1977, the collapse involved a much more important rock volume than in 2002. Second, the time span between 1977 and 1982 may explain the high level of magma pressure that was not reached in May 2002. Intense activity with sporadic ascent of the lava lake level was observed from 2003 to 2007 but few measurements are available due to civil war unrest. Observations were done from helicopter. A noticeable ascent seems to have occurred between August and December 2005. The lake level could have been around 2875 m a.s.l. It reached 2975 m a.s.l. in June 2009 and 3060 m a.s.l. in June 2010. The lava lake was still active in a crater of 250 m in diameter. Periodically, lava flows poured out from this crater and covered the whole floor of the lower platform. However, this apparent increase in the ascent of the lava level sporadically changed to rapid drops of tens of metres, particularly in June 2011. In December 2011, an ascent of 90 m was registered. In the crater, level variations of 10 m are commonly observed over periods of minutes to days (Burgi et al. 2014; Bobrowski et al. 2017; Barrière et al. 2018, 2019). In July 2014, the lake level was 70 m below the lower platform but again ascended rapidly and reached the third platform.

  From time to time, overflows have inundated the sink bottom with a succession of ascents and falls. During periods of high level of lake activity, a suite of spatter cones was built around the crater at the platform rim and constituted a ring levee. Sometimes, lava flows passed above the ring and spread out over the platform. On 29 February 2016, a fracture opened at the eastern edge of the lower platform to the foot of the 2002 sink wall with the building of a spatter cone above the new vent (Global Volcanism Program 2016; Burgi et al. 2018). This activity spanned 2.5 months. New lava covered the lower platform and cascaded into the lava lake. With this supply, the lower platform rose up as well as the lake level. However, the lake level again dropped by the end of 2016, then it rapidly ascended in 2017 and 2018. A new drop occurred in July 2019 (Global Volcanism Program 2019). However, step by step, the lower platform level became higher, with overflowing either from the lake or from the circular vents. By the end of 2019, the eastern vent reactivated and contributed to the rapid ascent of the lower platform level. In 2020, the lava level is only some tens of metres below the 1995–2002 platform height of 3175 m a.s.l.