Warnings for hazards in practice

The experience gained in warning service is maybe the most significant in extending the alert system and for developing real-time research. During the PRENLAB-2 period there were three eruptive crisis in Iceland, two of which have already ended with eruptions and two earthquake sequences occurred in S-Iceland where warnings and information was significant.

• Short-term warning and information before the eruption in volcano Hekla, February 2000:

  The eruption that started in the volcano Hekla in S-Iceland at approximately 18:17-18:19 GMT on February 26, 2000, was preceded by precursory signals on seismographs and volumetric strainmeters during the last 79 minutes prior to the outbreak.

  The National Civil Defence of Iceland was warned of a probable imminent eruption about an hour before the ash plume was observed. By time the precursory signals became more prominent and at 17:53 a prediction was issued to the Civil Defence claiming that an eruption was certainly to be expected within 15-20 minutes, with the recommendation that a warning should be issued and broadcast to the public. The main immediate hazard caused by recent Hekla eruptions is the 10 km high ash plume during the first minutes of the eruption that endangers flight traffic. Therefore, well before the final prediction was made, probable ash trajectories were calculated and the Civil Aviation Administration was notified that an ash plume from Hekla could be expected to rise to cruising altitudes of air traffic within short. The first signs of the coming eruption were seen at 17:00 from a seismometer situated within 2 km of the top of the volcano. These were earthquakes of magnitude below 1. The earthquake swarm became more intense in the next half hour and events were located by the automatic location system and detected by the alert system at 17:29. Earthquake sequences of similar size are unknown at Hekla except as a prelude to eruptions. A volumetric strainmeter at a distance of 15 km from Hekla showed rapidly increasing contraction starting 30 minutes before the eruption. Based on modelling of the 1991 Hekla eruption this was a clear sign of an opening of a feeding dyke for an eruption. This observation was used to refine the timing of the final prediction. The rapid response to the premonitory signals of the eruption was possible because of a combination of improved instrumentation in the region of Hekla, improved facilities for real-time analysis and alerting, and raised general alert caused by recent unrest in the neighbouring volcanoes, mainly Katla and Eyjafjallajökull (Stefánsson et al. 2000b; Ágústsson et al. 2000).

• Stress forecast before the magnitude 5 earthquake in November 13, 1998, in the Ölfus area:

  The Hengill-Ölfus area in SW-Iceland (Figure 1) was a very significant research and test area for the PRENLAB-2 project as it was for the PRENLAB project for development of warning algorithms as well as for general understanding of earthquake related crustal processes (Rögnvaldsson et al. 1998a) (Figure 2).
  The reason is basically the very high seismic activity in this area since 1994. It has been possible to carry out deformation measurements of various kinds to keep track of the deformation, both by GPS and SAR in addition to very detailed observations of frequent seismic swarms and individual earthquakes up to 5.1 in magnitude. Stress modifications related to the largest earthquakes have been observed. Thus an earthquake cycle has been observed from the start time of build-up of stress on June 4, 1998, in a large area towards concentration of stress in a focal region and foreshocks of an earthquake that occurred on November 13, 1998. After the earthquake of November 13, it was then observed how an E-W fault zone served as a stress guide, and how a sequence of earthquakes was observed related to that guide.

  
  On basis of experience in studying shear-wave splitting time patterns in this area and on basis of the general understanding achieved of crusal processes there, stress forecast was issued by Stuart Crampin at UEDIN.DGG, on November 10, 1998, to IMOR.DG. This forecast said that an earthquake of magnitude 5-6 could occur anytime between the issuing of the forecast (M=5) and the end of February 1999 (M=6) if stress kept increasing. An earthquake of magnitude 5 occurred near the center of the region included in the forecast on November 13. Although this kind of forecast is far from being a complete earthquake prediction this is a step forward for short-term warnings. It does not in itself specify the epicenter of the earthquake. In this case the most likely epicenter and size had been guessed based on former activity, i.e. to complete an ongoing seismic cycle, as had been described by IMOR.DG scientists. In hindsight it was observed that the earthquake of November 13 had foreshock activity, which in fact defined the most likely epicenter for the earthquake, and also indicated that it was impending within short. Of course it is always a question if a sequence of microearthquakes is a foreshock activity or not. However, the pattern and characteristics of the foreshock activity in this case and methods for automatic evaluations of observations, give hopes that procedures can be developed to complete such a stress forecast by observations which aim at finding the place and the time of the earthquake nucleation before it ruptures (see further Subproject 3).

• Hazard assessments and short-term warning related to the two M=6.6 earthquakes in the South Iceland seismic zone in June 2000:

  In June 2000 two earthquakes with magnitude 6.6 (Ms) struck in the central part of the South Iceland seismic zone (SISZ), immediately followed by seismic activity along zones of 100 km length (Stefánsson et al. 2000c). This occurred after 88 years of relative quiescence in the 70 km long EW transform zone in SW-Iceland (Figure 4).

  
  Earthquakes in this region have, according to history, frequently caused almost total destruction in areas encompassing 1000 km²and have been a threat for inhabitants of this area, which is a relatively densely populated farming area. In spite of open surface faults and measured accelerations higher than 0.8 g no serious injuries were reported and no homes collapsed. In light of the fear which the expected SISZ earthquakes have caused among Icelanders, the relatively minor destruction has lead to some optimism regarding the safety of living in the area. Many of the ideas about the nature of strain release in the area have been confirmed. As far as the epicenter of the first earthquake is concerned, hazard assessments or long-term predictions were confirmed, and in hindsight precursors have been observed. Useful short-term warnings regarding the epicenter, size and time of the second earthquake could be issued beforehand. Significant observations were made of the earthquakes as well as of their premonitory and following processes which will lead to better models of the SISZ earthquakes as well as to better hazard assessments and warnings in the future. Among significant observational systems are the SIL system, which is especially aimed at retrieving information from microearthquakes, strong motion instruments with a good coverage in the area, continuous GPS measurements, borehole strainmeters and hydrological observations in boreholes. Earlier GPS net measurements were repeated after the earthquakes, and detailed analysis of extensive surface fissures was carried out. It is expected that no more than 1/4 of the moment build-up in the zone has been released in these two earthquakes, and that even larger earthquakes may occur in the zone during the next decades.
  
  The short-term warning, hazard assessment and premonitory activity are further described in Stefánsson et al. (2001).