Thermo-rheological magma control on the impact of highly fluid lava flows at Mt. Nyiragongo



[1] In January 2002 Mount Nyiragongo erupted foiditic lavas that covered the Southern volcano flank devastating vast urban areas. Lava flows originated from vents at different heights on the eruptive fissure displayed different velocities, from tens of km/h at the highest vents to slow-advance (0.1–1 km/h) in Goma town several km away from the volcano. To understand the different behavior of lava flows and their threat to the local population, we undertook a multidisciplinary study involving textural and rheological measurements and numerical simulations of heat transfer during magma ascent. We demonstrate that pre-eruptive cooling and syn-eruptive undercooling of magma determined the different rheological behavior of lava flows erupted from vents at diverse heights. Venting at lower altitudes is expected to produce viscous, slowly advancing lavas, although development of fluid, faster flows should be included among possible future eruptive scenarios.

1. Introduction

[2] On 17 January 2002 several lava flows originated along a 10 km fissure system from the summit area of Mount Nyragongo volcano, (Democratic Republic of Congo), 3649 m a.s.l., to the suburbs of Goma town (ca. 500,000 inhabitants) [Allard et al., 2002; Komorowski et al., 2004; Tedesco et al., 2002; D. Tedesco et al., The January 2002 volcano-tectonic eruption of Nyiragongo volcano, Democratic Republic of Congo, submitted to Journal of Geophysical Research, 2006, hereinafter referred to as Tedesco et al., submitted manuscript, 2006], resulting in the most outstanding example of lava flow impact on a large town. At 08:30 local time the pre-existing North Shaheru (NS) fissure (Figure 1), 2600–2800 m a.s.l., generated a “curtain of fire”, and, in the following hours, the fissure system propagated southward, reaching the outskirts of Goma, at 1550 m a.s.l., at about 16:00. Two large lava flows erupted from different vents entered the town, one of them reaching Lake Kivu at night. At that time, approximately 15% of the town was covered by lava, one third of the airport runway was destroyed, and about 120,000 people found themselves homeless [Komorowski et al., 2004; Baxter et al., 2004]. Eye-witness reports and field evidence indicate that lava flows from the highest vents were extremely fluid and fast-moving (tens of km/h, by analogy with estimates for the similar 1977 lava flows [Tazieff, 1977]), instead those in town were more viscous and slowly-advancing (0.1–1 km/h). Accordingly, the threat posed to people was dramatically different. Those living in villages on the flanks of the volcano had little chance to escape, whereas hundreds of thousands of people from Goma could flee in a rapid exodus, mostly towards nearby Rwanda [Baxter et al., 2004].

Figure 1.

Location of collected samples and lava flow paths. Landsat 7 ETM + image of the Nyiragongo Volcano, January 2003. Lava flow paths of the 17 January 2002 eruption also reported. Dotted lines indicate the eruptive fractures. Chronology of eruptive activity reported in boxes. Ellipses refer to location of samples [after Favalli et al., 2006].

[3] Here we investigate the texture and rheology of the 2002 Nyiragongo products with the aim of exploring the described different lava flow emplacement and dynamics. Our results point to a central role of pre-eruptive cooling in determining different lava rheology and flow behavior, and, as a consequence, the smaller threat posed to Goma inhabitants by the advancing flows.

2. Field Observations

[4] Field investigation shows that solidified lava from NS consists of a very smooth 5–15 cm thick blanket mantling the topography, although abundant marks on trees suggest that active flows were about 1 m thick close to the fissure. Only scattered scoria related to the waning phase of the eruption is present. Hundreds to thousands of meters downflow the thin lava blanket is covered by a clinkery aa lava surface concentrating within depressions. Three kilometers from the fissure the flow front consists of a 1 m thick, individual, aa lava flow.

[5] The Munigi (MN) fissure, which fed the most devastating flow in Goma (Figure 1), is surmounted by a 1 km chain of scoria and spatter cones. The solidified lava has a blocky aa surface and is >2 m thick a few meters away from the fissure. Lava flows from fissure vents located at intermediate altitudes display morphological and structural features transitional between those from NS and MN fissures.

3. Textural and Rheological Measurements

[6] Nine samples consisting of lava (4), scoria (4) and spatter (1) erupted from different vents were thin-sectioned and investigated texturally. Sample location is reported in Figure 1, and details of the procedure can be found in the work of Polacci et al. [2006]. Major and trace elements in bulk samples, and the mineralogical assemblage (mostly melilite and nepheline), are the same within analytical uncertainty all along the fissure system [Landi et al., 2004; Tedesco et al., submitted manuscript, 2006]. NS and MN lavas have similar vesicularity (∼30–50 vol%), vesicle geometry and size distributions. Microphenocrysts (≥100 μm) with equilibrium textures represent 5–18 vol% of the vesicle-free MN samples investigated here, and up to 50 vol% for distal samples investigated elsewhere [Landi et al., 2004; Santo et al., 2004], while they are absent in NS samples. Lavas from both NS and MN have holocrystalline groundmass textures, the former having groundmass crystals (<100 μm) one order of magnitude smaller than the latter. MN scoria shows a higher vesicularity (55–68 vol.%) with rounded vesicles to >1 cm. Microphenocryst content and distribution are similar to those in lavas, but the groundmass consists of microlite-free, clear glass.

[7] Newtonian viscosity was measured, after homogenization, via concentric cylinder and micropenetration rheometry [Dingwell et al., 1996; Dingwell, 1989]. In dynamic cooling experiments, a sample of lava was melted at 1770 K for several hours, then progressively cooled at controlled rates within the concentric cylinder. In separate experiments, samples were equilibrated isothermally, quenched, and microscopically inspected.

[8] Dry viscosities of Nyiragongo liquid are among the lowest measured in natural magmas (Figure 2). Viscosity measurements during dynamic cooling reveal a clear departure from the pure liquid trend at temperature <1350 K. In a range of only 15–30 K a minimum viscosity increase of one order of magnitude, termed “rheological cut-off”, occurs for cooling rates of 1–3 K/min. For a cooling rate of 5 K/min the dynamic viscosity deviates only slightly from the inferred pure liquid trend.

Figure 2.

Viscosity of investigated melts. (a) Isothermal (pure liquid) and dynamic cooling (liquid + crystal) viscosity of Nyiragongo melts. (b) Enlargement at the rheological cut off. Lines represent best Arrhenian fits for the pure liquid at high and low temperatures.

[9] Samples quenched to about 1100 K after cooling at a rate of 5 K/min, as well as those equilibrated at 1370 K, show no microphenocrysts, and clear glass with 0–10 vol% groundmass crystals. Conversely, samples cooled at 1 K/min display holocrystalline textures with <100 μm-1 mm melilite crystals. Equilibration at 1320 K results in up to 50 vol% crystals.

[10] The above results suggest superliquidus conditions for the NS lava, implying T ≥ 1370 K, while the microphenocryst content of MN lava suggests an equilibrium temperature in the range 1320–1370 K. The experiments also show that cooling rates ≥5 K/min do not allow crystal nucleation and growth.

4. Numerical Simulations

[11] To evaluate the eruptive temperature at MN we performed finite element numerical simulations of cooling during dike flow, by employing the GALES code developed by Longo et al. [2006]. The physico-mathematical model adopted solves the equation of conservation of energy (Table 1) for magma flowing in the dike and for country rocks in the shallowest portion beneath the earth's surface. The 2D simulations calculate the balance between heat advection due to magma flow and diffusive heat loss from magma toward and through the country rocks. The algorithm adopts a double space-time discretization as in Shakib et al. [1991]. The rectangular space elements are 10 m in the vertical direction, while the horizontal dimension is 0.01 m inside the dike, and it progressively increases to as much as 500 m far away from the dike walls. The different horizontal and vertical size of space elements inside the dike reflects the general dike geometry, and the faster heat transfer by convection along the vertical direction with respect to heat transfer due to diffusion along the cross-flow direction. The time step is 1 s.

Table 1. Energy Conservation Equation, Input Parameters, Initial, and Boundary Conditions of Simulationsa
Input ParameterValues of Parameters
  • a

    Energy conservation equation is equation image , where T is temperature, t is time, v is velocity, κ is thermal diffusivity, and x and y denote spatial directions. Simulation labels in Figure 3 corresponding to each adopted value are reported within square brackets (s8 parameters are fixed after 3600 s of s5).

Dike horizontal length1.4 km
Dike vertical length0.5 km [4–8, 10–12, 14–15], 2 km [9]
Dike width10 cm [4–6, 8–12, 14–15], 20 cm [7]
Flow rate80 m3/s [4, 10, 12, 16], 225 m3/s [15], 450 m3/s [5–9, 11, 14]
Average velocity in parabolic profile0.6 m/s [4, 10, 12, 16], 1.6 m/s [15], 3.2 m/s [5–9, 11, 14]
Magma temperature at dike base1320 K [5, 7, 9, 12, 14, 15, 16], 1370 K [4, 6, 8, 10, 11]
Initial rock temperature300 K
Magma gas content0 vol% [4–9], 40 vol% [10–12, 14–15]
Rock diffusivity10−6 m2/s
Magma diffusivity2.5·10−7 m2/s

[12] A constant Newtonian velocity profile was assumed by minimum and maximum estimates of the average mass flow-rate at MN fissure. The horizontal component of velocity was assumed to be zero, due to much larger vertical dike dimension. The physical properties of magma and rocks, the initial and boundary conditions are reported in Table 1. The presence of 40 vol% gas bubbles was also considered in some simulations. Degassing-induced cooling and latent heat of crystallization were neglected. The simulations took into account dike lengths of 500 and 2000 m, and two different dike widths of 10 and 20 cm. The rising lava was assumed to ascend in a pre-existing dike, which was progressively filled at a constant velocity throughout the simulation. The heat transfer equation was solved by defining the temperature at the dike inlet as a boundary condition. The simulations start with a dike totally filled with magma at constant (inlet) temperature, and country rocks at constant ambient temperature. The results are however inspected starting from a time (defined as time zero in Figure 3) when the entire magma in the conduit has been replaced by new magma coming from below, in order to approximate magma cooling and rock heating during magma rise before the start of the eruption. The numerical results are therefore more appropriate for the thermal state of magma at an advanced eruption stage, when the effects of the assumed initial conditions are minima. Two magma temperatures, 1320 and 1370 K, are taken as representative of the pre-eruptive magmatic temperature range as inferred from crystallization experiments. Each simulation takes about 3 days of 3 GHz CPU.

Figure 3.

Average magma temperature at fissure exit of erupted Nyiragongo lavas: (a) average exit magma temperature and (b) average cooling rate of rising magma. Magma temperature at dike base: 1320 K for solid lines, 1370 K for dashed lines. Simulations corresponding to each labelled line are described in Table 1.

[13] Figure 3a shows the calculated exit magma temperatures. The initial temperature is low and progressively increases to an almost constant value after 1–4 hours of extrusion. This trend is due to initial diffusion-dominated heat transfer determined by the low temperature of country rocks, followed by an increase of the role of thermal advection as long as country rocks are heated up by the rising magma. The asymptotic exit temperatures are from 10 to 200 K less than the magma temperature at the dike base.

[14] Figure 3b shows the average cooling rate along the dike, determined as the overall temperature change from the dike base to its exit divided by the dike transit time. With one exception related to the case of large dike width (20 cm), the calculated cooling rates exceed 5 K/min. Comparison with dynamic cooling experiments suggests that magma cooling due to rise through country rocks may not involve crystal nucleation.

5. Discussion and Conclusions

[15] A consistent picture emerges from the present investigation. Crack formation and dike propagation toward the MN vents likely initiated months before the 2002 eruption [Tedesco et al., 2002], resulted in magma cooling to a temperature close to 1320 K and in the growth of a few vol% melilite crystals. Syn-eruptive ascent of this magma produced additional cooling at a rate sufficiently large to prevent crystal nucleation and growth. The eruption discharged therefore microlite-free, crystal-bearing, supercooled magma with an initial viscosity of about 103 − 105 Pa s (T = 1250 − 1150 K, Figures 2 and 3) that produced abundant scoria. During the eruption the lava temperature increased and viscosity decreased to values in the range 102 − 103 Pa s (T = 1310 − 1230 K, Figures 2 and 3), allowing gas to escape from the liquid and resulting in a more passive emission of lava. Once discharged, the undercooled magma rapidly nucleated abundant microlites increasing further the viscosity of lava flows.

[16] In contrast, the NS magma had an initial superliquidus temperature ≥1370 K, implying a maximum viscosity of 60 Pa s (Figure 2). Although the simulations in Figure 3 with T0 = 1370 K adopt mass flow-rates and initial country rock temperatures typical of the Munigi fissure, the calculated T values produce conservative estimates of eruptive viscosities about one order of magnitude less than those for the MN magma. From the results in Figure 2, cooling rates in the range 5–1 K/min would imply a minimum of 20–40 minutes to increase the viscosity of NS magma to values similar to those for the MN magma. During this time the lava could flow at high velocities, leaving a thin solidified pahoehoe sheet at the contact with the cold ground surface.

[17] The above picture is consistent with the U-Th disequilibria which show that the fissure system tapped magma from progressively larger depth, therefore magma which had to flow a longer distance to emerge at the surface, with progressively increasing distance from the central volcano crater (Tedesco et al., submitted manuscript, 2006).

[18] Although we cannot exclude that a different volatile content characterized magma discharged by NS and MN vents, the equal major, minor and trace element distribution in the two magmas suggests that the original volatile content was also the same, implying a substantially similar volatile content in the liquid at 1 bar pressure. The different dynamics of lava flows from NS and MN vents are likely due to different eruptive temperatures rather than different volatile contents. This work points to slow magma cooling during dike propagation in the weeks to months preceding the eruption as the trigger for the different lava flow rheological behavior and emplacement dynamics at Nyiragongo. Only a few tens of people died as a direct consequence of the eruption [Baxter et al., 2004], but this number could have been orders of magnitude higher if the lava flows which impacted Goma were as fluid as those originated at high altitude from NS vents. If in the future magma erupted from vents close to Goma were to undergo less efficient pre-eruptive cooling (e.g. as a consequence of more rapid rock fracturing and magma intrusion or of repeated intrusion and heating of country rocks), then the town might be inundated by fluid lava flows. This scenario would pose a much more severe threat to the large population of Goma, since people would have little or no time to escape from the extremely fast-moving lava flows.


[19] This work was funded by the Italian DPC-INGV Nyiragongo project in the frame of the 2001–2003 GNV Program, and by the EU Volcano Dynamics RTN project. P. Papale benefited from two UN-OCHA contracts in January and November 2002. We are grateful to F. Colarieti for thin section preparation and SEM assistance, P. Pantani for graphic assistance, G. Chirico for providing Figure 1, and M. Potuzàk for experimental assistance.