The Slungawarning algorithm involves to watch in real-time seismicity of microearthquakes as well as of their source dimensions, and to detect anomalies based on a physically established source model and experience of observations of the SIL microearthquake system for 9 years.
The physical basis for developing and testing such an algorithm is a recently evolving rate-and-state dependent model of friction of fault planes. This model suggests that for any given earthquake generating fracture and loading system, there exists a minimum fault radius for earthquakes, again suggesting that different parts of the faults with different minimum fault radius of earthquakes may be active at different stages of the premonitory process leading to large earthquakes. Thus by evaluating microearthquakes it may be possible to resolve a process that sometimes has been assumed and treated as totally chaotic.
Based on such physically based ideas and observations a number of earthquake warning parameters have been defined and an earthquake warning algorithm has been developed which has been tested retrospectively on SW-Iceland (Slunga et al. 2000).
A study of the microseismicity during the 9 years within SW-Iceland where the large earthquakes occurred has been made (Slunga et al. 2000). A detailed description of that work is given in Slunga (2001). Here some main results are summarized.
The most obvious result is the close relation between the microearthquake fault radii and the geodynamic processes (coinciding with large earthquakes, and the large eruption below Vatnajökull glacier in 1996). The 3 largest earthquakes are preceded by large fault radius microearthquakes 3-4 months before the events. In addition the larger earthquakes in the Hengill area are preceded by an increase in microearthquake fault radii one or a few days before the events.
Most larger earthquakes are also preceded by an increased microearthquake activity culminating with foreshocks within the last day or days before the events. For the largest earthquake of the period the foreshock activity is lacking ML>1 events. For the quiet part of the SIL area with large earthquakes on N-S faults a "sandwich" seismicity ratio method was proposed. It compares the seismicity rate within a N-S strip to the seismicity of the surroundings. It was found that the last month before the large June 17, 2000, earthquake the surroundings became silent while the N-S strip became more active (the microearthquake sizes typically between ML= -0.2 to 0.4). All these aspects were tested within an EQWA (EartQuake Warning Algorithm) and it was found that during times of warnings (24 hour after each warning and within a distance of 6 km) the probability to have a large earthquake within the warned area was about 500 times larger than the probability based on total randomness.
Finally it was also found that most false warnings were at places of coming earthquakes. Such warnings during the last years before major earthquakes may increase the awareness of the risks.
In general the results of the study (Slunga et al. 2000) are in agreement with other precursory studies of shallow strike-slip earthquakes in other places of the world.
This algorithm is now being installed to be applied routinely for short-term warnings and for testing at IMOR.DG.
A thorough description of the method and of results of tests is found in Slunga (2001).