The initiation of the 1996 jôkulhlaup from Lake Grimsvôtn, Vatnajôkull, Iceland

Transcription

1 Tilt: Extremes of the Extremes: Extraordinary Floods (Proceedings of a symposium held at Reykjavik. Iceland. July 2000). I AI IS Publ. no The initiation of the 1996 jôkulhlaup from Lake Grimsvôtn, Vatnajôkull, Iceland TÔMAS JÔHANNESSON Icelandic Meteorological Office, Bt'isladavegitr 9, IS-150 Reykjavik, Iceland ti@vedur.is Abstract The jôkulhlaup from Lake Grimsvôtn in Vatnajôkull Ice Cap in 1996 was initiated by a different physical mechanism to that assumed in traditional theories of jôkulhlaups. The maximum discharge was reached only 16 h after the start of the flood at the terminus. Water outbursts, through a >300 m thick glacier near the terminus, indicate water pressures several bars in excess of ice overburden at the beginning of the flood. A deep ice canyon was formed in the ice cap near Lake Grimsvôtn extending along about 10% of the subglacial floodpath. Frozen sediments formed in crevasses and frazil ice on the surface of the flood waters indicate the flow of supercooled water in the terminus region. These observations can be interpreted such that the jôkulhlaup was initiated by the movement of a localized pressure wave that travelled 50 km in 10 h from Lake Grimsvôtn to the terminus, forming a subglacial pathway along the glacier bed. Shortly after this wave reached the terminus, the jôkulhlaup was flowing at a high discharge through a tunnel which would have needed much a longer time to form by ice melting as assumed in existing theories of jôkulhlaups. The observations also indicate that in current theories the rate of heat transfer from subglacial flood water to the overlying ice is greatly underestimated. Key words jôkulhlaup; Grimsvôtn, Iceland; heat transfer; supercooling; frozen sediments INTRODUCTION Jôkulhlaups from the subglacial Lake Grimsvôtn in Vatnajôkull Ice Cap, southeastern Iceland, are an interesting subject for scientific investigation. These jôkulhlaups have often caused substantial damage to farmlands, roads and communication lines and disrupted travel between the inhabited areas to the east and west of the Skeiôarârsandur outwash plain. Data about floods from Lake Grimsvôtn have been used in some of the most important studies of the physics of jôkulhlaups and they are probably more studied than jôkulhlaups from any other location in the world. A general description of jôkulhlaups from Lake Grimsvôtn is given by Mrarinsson (1974) and Bjôrnsson (1974, 1988). The jôkulhlaup from Lake Grimsvôtn in 1996 (Fig. 1) was caused by a subglacial volcanic eruption north of Lake Grimsvôtn (Guômundsson et al, 1997). The flood started more abruptly and reached higher discharges than jôkulhlaups from Lake Grimsvôtn have done in recent decades. The maximum discharge out of the reservoir was estimated to be somewhat less than 40 x 10 3 m 3 s" 1 (Bjôrnsson, 1997) and the maximum flood at the terminus slightly above 50 x 10 J m J s" 1 (Snorrason et aï., 1997). In spite of the many previous studies of the physics of jôkulhlaups from Lake Grimsvôtn (Nye, 1976; Bjôrnsson, 1992), several aspects of the jôkulhlaup in 1996 were unexpected and highlight a need to improve existing theories (Bjôrnsson et aï., 2001 ).

2 Tômas Jôhannesson Gjà^ xwnajok/ullx ^ h ) ICECAP 1 v - ^SGrimsyotfi If J0, Grîfhsfjair O 3 <0 2, <5 15 c 10 km Fig. 1 The outlet glacier Skeiôarârjôkull Glacier from the Vatnajôkull Ice Cap. The location of the Grimsvôtn geothermal area and subglacial lake and the Gjâlp Volcano are shown. The extent of the 1996 jôkulhlaup on the Skeiôarârsandur outwash plain at approximately 24:00 h on 5 November 1996 near the time of maximum discharge is also shown (based on Snorrason et al, 1997).

3 The initiation of the 1996jôkulhlaup from Lake Grimsvôtn, Vatnajôkull, Iceland 59 Although the discharge of most jôkulhlaups from Lake Grimsvôtn rises almost exponentially over a period of a week or more, several jôkulhlaups from Lake Grimsvôtn are known to have started abruptly in a similar way, although perhaps not quite as rapidly, as the 1996 jôkulhlaup. Such jôkulhlaups did for example occur in 1861, 1892 and 1938 (Mrarinsson, 1974). The abrupt rise of the discharge in these jôkulhlaups, which was sometimes preceded by a period of much more slowly increasing discharge, indicates that two different physical processes may be important during the start of jôkulhlaups. Slowly rising jôkulhlaups then correspond to a balance between melting of ice and release of potential energy described by traditional jôkulhlaup theories whereas rapidly rising jôkulhlaups would be governed by a fundamentally different physical mechanism (Snorrason et al, 1997; Bjôrnsson et al, 2001). Rapidly rising jôkulhlaups do not only occur in Lake Grimsvôtn. Jôkulhlaups from the ice cauldrons Skaftârkatlar in western Vatnajôkull Ice Cap (Bjôrnsson, 1977, 1992; Zôphôniasson & Pâlsson, 1996) often rise rapidly, that is to m 3 s" 1 in 1-2 days and jôkulhlaups from several other locations in Iceland are reported to have risen equally or even more rapidly, in particular very large jôkulhlaups caused by volcanic eruptions in the Katla Volcano in Myrdalsjôkull Ice Cap. The jôkulhlaup from Katla Volcano in 1918 reached an estimated maximum discharge of about 3 x 10 5 m 5 s" 1 in only a few hours and lasted less than a day (Tomasson, 1996). All these jôkulhlaups are, like the 1996 jôkulhlaup from Lake Grimsvôtn, difficult to explain with traditional theories on jôkulhlaups. The traditional theories have in fact never been used in an analysis of the well-known jôkulhlaups from the Myrdalsjôkull Ice Cap because it was clear that the sudden, catastrophic floods from the Katla Volcano are outside the scope of the theories. The rapidly rising jôkulhlaups are large and small and come from both subglacial and marginal lakes at many different locations. Some of them were induced by volcanic eruptions, but others do not appear to have been directly caused by eruptions. The extensive available observations from the 1996 jôkulhlaup from Lake Grimsvôtn (Haraldsson, 1997) give valuable indications of the improvements that are needed in existing jôkulhlaup theories. These observations are discussed in this paper and a new mechanism for a rapid initiation of jôkulhlaups is suggested. Many rapidly rising jôkulhlaups, where traditional theories do not seem to apply as mentioned above, indicate that these ideas could apply to jôkulhlaups at other locations. This paper summarizes a more detailed paper on the same subject (Jôhannesson, 2002) where the observations from the 1996 jôkulhlaup are described more thoroughly and the interpretation suggested here is supplemented with more quantitative arguments. THE START OF THE 1996 JÔKULHLAUP The discharge of the 1996 jôkulhlaup from Lake Grimsvôtn rose very rapidly. The outflow from the reservoir started about 10 h before the flood reached the terminus and during this time about 0.6 km"' of water accumulated below the glacier (Bjôrnsson, 1997). The discharge reached its maximum of about 50 x 10 J m J s" 1 only about 16 h after the flood wave burst out at the terminus region (Snorrason et al, 1997). This

4 60 Tômas Jôhannesson rapid rise cannot be explained by the positive feedback between water flow and tunnel enlargement assumed in traditional jôkulhlaup theories as is most easily seen by the fact that the potential energy corresponding to the 0.6 km J of water that accumulated below the glacier before the flood reached the terminus could only have melted km J of ice downglacier from the ice canyon near Lake Grimsvôtn, as described below (assuming for simplicity the water was uniformly distributed with a uniform potential gradient over the 50-km-long floodpath). The jôkulhlaup burst through several hundred metres thick ice at many locations near the terminus (Fig. 1) (Snorrason et al, 1997; Roberts et al, 2000a; Russell et al, 1999) indicating widespread areas of water pressure 5-10 bars (0.5-1 MPa) above ice overburden. The progressive forming and the short duration of the outbursts indicates the propagation of a subglacial pressure wave at the front of the jôkulhlaup with a higher subglacial water pressure than after the initial stage of the jôkulhlaup was over. The propagation of a wave of high pressure has no counterpart in traditional theories of jôkulhlaups where a spatially uniform potential gradient is assumed to drive the flood through a tunnel with a comparatively uniform cross-sectional area along its axis. In addition to the observations of the rapid rise and the high subglacial pressure the 1996 jôkulhlaup from Lake Grimsvôtn provided insight of a more thermodynamic nature into the physics of jôkulhlaups. The jôkulhlaup melted a 100-m-deep, 1-kmwide and 6-km-long canyon in the ice cap along the uppermost part of the floodpath (Bjôrnsson, 1997). The dimensions of the canyon indicate that flood water with an initial temperature of about 8 C released essentially all its thermal energy within the 6 km of the 50-km-long floodpath where the canyon was observed and thereby melted the 0.3 km J of ice required to form the canyon. Supercooling will lead to the formation of ice particles when the flood water flows out of the subglacial channel, since flowing water with suspended sediments cannot be maintained at a temperature below the melting point at atmospheric pressure. The particles quickly float to the surface after the flood water exits from the glacier and form a mushy layer of floating ice particles or frazil ice on the surface of the flood. Some of the particles will float to the sides away from the most rapidly flowing water near the middle of the flood channel and accumulate in calmer parts of the flood and in ponded flood water near the banks. Such a layer covering several square kilometres with an estimated thickness of 3-10 cm was in fact observed east of Skeiôarâ River during the start of the 1996 jôkulhlaup (Snorrason et al, 1997; Sigurôsson & Snorrason, personal communication). Frazil ice was also observed floating on the surface of the main flood. It is very difficult to estimate the total volume of the frazil ice on the surface of the main flood, but its average thickness could have been 1-10 mm. An order of magnitude analysis of these admittedly limited observations indicates that the heat released by the freezing of the frazil ice is of the same order of magnitude as the heat required to raise the temperature of the flood water from near the pressure melting point of about -0.2 C below the glacier to approximately 0 C at atmospheric pressure at the outlet (Jôhannesson, 2002). A widespread occurrence of frozen sediments in crevasses in the terminus region of Skeiôarârjôkull Glacier was discovered after the 1996 jôkulhlaup (Roberts et al, 2000a,b, 2002; Tweed et al, 2000). The sediments consist of sandur material

5 The initiation of the 1996jôkulhlaup from Lake Grimsvôtn, Vatnajôkull, Iceland Distance (km) Fig. 2 A longitudinal section along the path of jôkulhlaups from Lake Grimsvôtn to the terminus of Skeiôarârjôkull Glacier (the geometry is simplified from a figure in Bjôrnsson, 1997). The shaded areas at the bedrock and ice surface show a qualitative picture of the subglacial flood wave propagating from Lake Grimsvôtn and the associated lifting of the ice surface at the initiation of the 1996 jôkulhlaup. The vertical dimensions of the flood wave and the lifted area are chosen for illustrative purposes and are not intended to show their proper size relative to the ice thickness. The inserted smaller figure shows a closer view of the propagating bulge at the front of the jôkulhlaup. The variation of the ice overburden (short dashes, the curve indicates the water level above bedrock corresponding to the ice overburden) and of the potential fj) = p. g z h + p. g h. + p. (long dashes, drawn as the equivalent water level) is also shown in the area of the bulge. The pressure difference Ap = p. - p, between the flood water and the ice overburden is positive where the curve representing the potential is higher than the curve representing the ice overburden (see Jôhannesson, 2002, for further explanation). embedded in a matrix of ice and have a well developed stratification parallel to the crevasse walls. They seem to have been formed by freezing of suspended sediments in supercooled flood water onto crevasse walls. Since the Skeiôarârjôkull Glacier is temperate, there is no other plausible mechanism for the creation of the frozen sediments than supercooled flood water. The appearance of supercooled water during the flood requires that the water first reaches temperatures below 0 C at a high pressure below the glacier and then flows rapidly towards lower pressure so that potential energy is not released fast enough in the flow to maintain the temperature at or above the local pressure melting point of ice. The 6-km-long canyon in the ice cap near Lake Grimsvôtn is relatively short compared with the 50-km-long floodpath. This, together with the discovery of the frozen sediments in crevasses and the frazil ice on the surface of the flood water, implies that the rate of heat transfer from the flood water to the overlying glacier ice is between one and two orders of magnitude more rapid than expected according to traditional jôkuhlaup theories.

6 62 Tômcis Jôhannesson A SUBGLACIAL PRESSURE WAVE An interesting observation from the Lake Grimsvôtn 1996 jôkulhlaup is the uniform, almost linearly increasing discharge out of the reservoir during the first 16 h of the flood, before the discharge reached its maximum of about 40 x 10' 1 nr' s"' (Bjornsson, 1997, Fig. 9; Bjornsson et al, 2001). There does not seem to be any effect of the outburst of the flood from the terminus on the measured outflow from Lake Grimsvôtn. There is also no indication that the outflow from Lake Grimsvôtn is affected by the passage of the front of the flood wave through areas of different ice thickness and widely varying ice surface or bedrock slopes along the floodpath. Thus, it appears that variations in the geometry of the floodpath far away from Lake Grimsvôtn had little effect on the discharge out of the lake and there was no significant step in the discharge when the resistance provided by the glacier to the propagation of the subglacial flood front disappeared and the flood burst out at the terminus. The measurements of outflow from the lake imply that there was little direct connection between the subglacial lake at Lake Grimsvôtn and the propagating flood front further downglacier. The flood front was therefore not directly pushed forward by a pressure head extending all the way from the lake. Rather the pressure wave at the flood front must have been propagating downglacier as an independent dynamic feature in the subglacial hydraulic system. The flood front was of course fed from behind by the water flow from the lake, but this flow must have been essentially determined by a local potential gradient and by the local tunnel dimensions. A qualitative picture of the initiation of the 1996 jôkulhlaup from Lake Grimsvôtn is shown in Fig. 2. This picture is based on observations of the 1996 jôkulhlaup and conclusions about the dynamics of the flood wave which may be drawn from them. Firstly, the subglacial volume occupied by the flood water further than about 6 km from Lake Grimsvôtn is almost entirely created by the lifting of the glacier along the floodpath. The width of the floodpath is of the order of a kilometre, although this is not well known except close to Lake Grimsvôtn, and the thickness of the subglacial water layer is of the order of 10 m (Bjornsson, 1997). Secondly, the subglacial water pressure must be comparatively close to ice overburden except close to the propagating flood front. The total pressure head corresponding to the difference between the water level in Lake Grimsvôtn and bedrock and ice surface elevations along the floodpath is of the order of 100 bars (10 MPa), whereas the observations indicate that the difference between the subglacial water pressure and the ice overburden is of the order of 10 bars (1 MPa) or less near the front. This means that the potential gradient in the subglacial water flow upglacier from the flood front must be such that the difference between the water pressure and the ice overburden is of the order of 10% or less of the total pressure head from the lake level to the propagating front. At the same time the water pressure at the front must be maintained sufficiently in excess of the local ice overburden to drive the propagation of the front. A central problem for the understanding of the dynamics of the propagation is how such a delicate balance of the water pressure and the ice overburden at the flood front is maintained. Assuming that the geometry of the subglacial pathway at the front of the jôkulhlaup is a subglacial bulge as shown in Fig. 2, with a thinner tail extending back to Lake Grimsvôtn, the potential driving water flow will decrease relatively uniformly

7 The initiation of the 1996jôkulhlaup from Lake Grimsvôtn, Vatnajôkull, Iceland 63 in the downglacier direction, except in the bulge near the flood front where the potential gradient will be smaller and the variation in the pressure will tend towards a hydrostatic pressure distribution. This has interesting consequences for the difference between the water pressure and the ice overburden. If the bulge, where the potential gradient is comparatively small, is several kilometres long, then this pressure difference is predicted to be of the order of 10 bars higher at the front of the bulge compared with the tail of the bulge. This applies to both downsloping and upsloping parts of the subglacial geometry because the distribution of the pressure difference in the bulge arises mainly from the slope of the ice surface rather than of the bedrock. A more detailed analysis given in Jôhannesson (2002), indicates that pressure differences that arise naturally in a bulge shown schematically in Fig. 2 are consistent with the speed of propagation of the front of the 1996 jôkulhlaup and with obseivations of propagating areas of high subglacial water pressures near the terminus of Skeiôarârjôkull Glacier when the jôkulhlaup burst out of the glacier. NEED FOR A NEW THEORY Traditional theories of jôkulhlaups cannot explain several important aspects of the 1996 jôkulhlaup from Lake Grimsvôtn. Some of these aspects were known before from other jôkulhlaups, but others, such as the frozen englacial sediments and the frazil ice on the surface of the flood waters indicating flow of supercooled water through the glacier, had not been observed before. The 1996 Lake Grimsvôtn jôkulhlaup thus highlights a need for a new physical theory of jôkulhlaups capable of explaining these new aspects, but maintaining of course the part of the older theory that successfully explains the hydrographs of many, comparatively slowly rising jôkulhlaups. Two main aspects need to be addressed by an improved theory. Firstly, the subglacial water pressure must be able to become locally higher than the ice overburden, producing a subglacial pressure wave that can travel downglacier and push out a subglacial pathway through which the flood subsequently flows. The theory needs to explain why this seems to happen only in some jôkulhlaups and not in others. In order to truly incorporate such propagating pressure waves, the theory must abandon an assumed simple geometry of the subglacial channel and allow both the height and the width of the channel to be dynamically determined by the theoretical physical description based on the geometry of the bedrock and ice surface of the glacier. Secondly, the heat transfer from the flood water to the tunnel walls must be reformulated so that it is in agreement with the existing observations. A simple semiquantitative analysis of the start of the 1996 jôkulhlaup from Lake Grimsvôtn is described briefly in the previous section and presented in more detail by Jôhannesson (2002). This theory has not been developed in quantitative detail so far but it highlights the most important physics of a propagating subglacial pressure wave that needs to be addressed by a new theory. Acknowledgements Data about the extent of the 1996 jôkulhlaup on Skeiôarârsandur on the maps in Fig. 1 were supplied by the Hydrological Survey of the National

8 64 Tômas Jôhannesson Energy Authority and Skûli Vikingsson drew the map. I thank Oddur Sigurôsson and Magnus Tumi Guômundsson for describing to me many observations of the jôkulhlaup, which I used in this paper, and Oskar Knudsen for bringing the frozen sediments in crevasses on Skeiôarârjôkull to my attention. A part of the ideas described in this paper were developed while the I worked at the Science Institute of the University of Iceland and at the Hydrological Survey of the National Energy Authority. I thank Helgi Bjornsson, Ami Snorrason, Magnûs Mâr Magnûsson and Helgi Jôhannesson for useful comments. REFERENCES Bjornsson, H. ( 1974) Explanation of jôkulhlaups from Grimsvôtn, Vatnajôkull, Iceland. Jôkull 24, Bjornsson, II. (1977) The cause of jôkulhlaups in the Skaftâ River, Vatnajôkull. Jôkull 27, Bjornsson, 14. (1988) Hydrology of Ice Caps in Volcanic Regions. Societas Scienliarum Islandica, University of Iceland, Reykjavik, Iceland. Bjornsson, H. (1992) Jôkulhlaups in Iceland: prediction, characteristics and simulation. Ann. Glaciol. 16, Bjornsson, H. (1997) Grimsvatnahlaup fyrr og nil. In: Vatnajôkull. Gos og hlaup 1996 (ed. by II. Haraldsson), Iceland Public Roads Administration, Reykjavik, Iceland. Bjornsson, If, Pâlsson, F., Flowers, G. E. & Magnûsson, M. T. (2001) The extraordinary 1996 jôkulhlaup from Grimsvôtn, Vatnajôkull, Iceland. EOS Trans. Am. Geophys. Un. 82(47), 2001 Fall Meeting Suppl. Abstract IP21A Guômundsson, M. T., Sigmundsson, F. & Bjornsson, 11. (1997) Ice-volcano interaction of the 1996 Gjâlp subglacial eruption, Vatnajôkull, Iceland. Nature 389, Haraldsson, H. (ed.) (1997) Vatnajôkull. Gos og hlaup Iceland Public Roads Administration, Reykjavik, Iceland. Jôhannesson, T. (2002) Propagation of a subglacial flood wave during the initiation of a jôkulhlaup. Accepted for publication in Hydrological Sciences Journal. Nye, J. F. (1976) Water flow in glaciers: jôkulhlaups, tunnels and veins. J. Glaciol. 17(76), Roberts, M. J., Russell, A. J., Tweed, F. S. & Knudsen, Ô. (2000a) Ice fracturing during jôkulhlaups: implications for englacial floodwater routing and outlet development. Earth Surf. Processes and Landforms 25, Roberts, M. J., Lawson, D. E., Knudsen, Ô., Russell, A. J. & Tweed, F. S. (2000b) Glaciohydraulic supercooling at two Icelandic glaciers: implications for englacial sediment accretion and basal ice formation. EOS Trans. Am. Geophys. Un. 81(48), 2000 Fall Meeting Suppl. Abstract II6IG-10. Roberts, M. J., Russell, A. J., Tweed, F. S. & Knudsen, Ô. (2002) Controls on the development of supraglacial flood-water outlets during jôkulhlaups. In: The Extremes of the Extremes: Extraordinary Floods (ed. by Â. Snorrason, 11. P. Finnsdôttir & M. Moss) (Proc. Reykjavik Symp., July 2000). IAHS Publ. no. 271 (this volume). Russell, A. J., Knudsen, Ô., Maizels, J. K. & Marren, P. M. (1999) Channel cross-sectional area changes and peak discharge calculations in the Gigukvisl River during the November 1996 jôkulhlaup, Skeiôarârsandur, Iceland. Jôkull 47, Snorrason, Â., Jônsson, P., Pâlsson, S., Ârnason, S., Sigurôsson, O., Vikingsson, S., Sigurôsson, Â. & Zôphôniasson, S. ( 1997) I llaupiô â Skeiôarârsandi haustiô Ûtbreiôsla, rennsli og aurburôur. In: Vatnajôkull. Gos og hlaup 1996 (ed. by II. Haraldsson), Iceland Public Roads Administration, Reykjavik, Iceland. Tômasson, H. (1996) The jôkulhlaup from Katla in Ann. Glaciol. 22, Tweed, F. S., Roberts, M. J., Finnegan, D. C, Russell, A..1., Knudsen, Ô. & Gomez, B. (2000) Englacial flood routes during the November 1996 jôkulhlaup, Skeiôarârjôkull, as revealed bv aerial photography EOS Trans. Am. Geophys. Un. 81(48), 2000 Fall Meeting Suppl. Abstract H12B-04. Zôphôniasson, S. & Pâlsson, S. (1996) Rennsli i Skaftârhlaupum og aur- og efnastyrkur i hlaupum 1994, 1995 og Report OS-96066/VOD-07, National Energy Authority. Reykjavik. Iceland. Pôrarinsson, S. (1974) Vôtnin stria. Saga Skeiôarûrhlaupa og Grimsvatnagosa. Bôkaûtgâfa Menningarsjôôs, Reykjavik, Iceland.