ESSG meeting on Geological Hazards

Wednesday February 19, 2003, 2pm
The University of Reading, Environmental Systems Science Centre

Tea 3:30, Meeting ends 5:00pm

Ian Main (The University of Edinburgh)
Earthquake Statistics and Predictability
The mitigation of earthquake hazard depends on a quantification of the point process of earthquake recurrence in space and time. Combined with wave propagation information, and the seismic response of local building sites, this information can be used to produce seismic hazard maps that can be used for land use planning and building design codes. Current practice centres on time-independent (Poisson) hazard calculations that are stationary in time, although non-stationarly processes such as earthquake foreshocks and aftershocks are known to occur. Here I consider the population dynamics of earthquakes, and test hypotheses of earthquake occurrence against natural data. First I briefly describe the statistical mechanics of earthquakes, and show cellular automaton models that predict (a) quasi-stationary Boltmann fluctuations in total system energy about a steady-state and (b) a gamma form for the probability density function for the earthquake energy. The results map to those expected at the critical branching model in statistics, the percolation threshold in stochastic flow problems, or the critical point in second-order phase transitions. These predictions are compared to data from the global frequency-energy catalogue, and confirmed using a Bayesian Information Criterion which tests alternative hypotheses with varying numbers of model parameters. In order to test the time-dependent component, we examine the global catalogue for triggered event pairs, compared to a Poisson background process. the results show a trigerring correlation length of 10-20 km or so, similar to the brittle thiockness of the crust, and 'anomalous' diffusion <x>~t**H, with an exponent H<<0.5. Some 3% of these potentially damaging earthquakes (magnitude>5.5) are triggered, showing that the time-dependent component (for large events) is significant compared to the background, but small in an absolute sense. Prospects for tuning seismic hazard with a conditional probability to reflect this amplification would be straightforward to implement. Such methods apply to time-dependent hazard based on a population. The prospects for predicting individual events (place, magnitude and time, above chance) remain beyond reach with current understanding.

Adrian Matthews (University of East Anglia)
Predicting Volcanic Activity with Rainfall
Dome-forming volcanic eruptions cyclically extrude bodies of lava over several months, which then become gravitationally unstable and collapse, generating pyroclastic flows. On 29 July 2001 extreme rainfall over Montserrat coincided with a major collapse of the Soufriere Hills lava dome. We present rainfall and seismic records that demonstrate, for the first time, a relationship between intense rainfall and lava dome collapse, with associated pyroclastic flow generation. After seven months of little rain and a period of sustained dome growth, the onset of intense rain was followed within hours by dome collapse and pyroclastic flows. The large-scale weather system responsible for the rain was identifiable in satellite images and prUse of uninitialized value in concatenation (.) or string at E:\listplex\SYSTEM\SCRIPTS\filearea.cgi line 455, line 102. edicted by meteorological forecasts issued 60 hours prior to the volcanic activity. It is suggested that weather prediction of intense rainfall be incorporated with existing geophysical and geochemical measurements to improve warnings of these hazardous events.

Louise Burt (RUWPA, University of St Andrews)
Mapping volcanic flow deposits from stratigraphic records
The stratigraphic record of deposits from volcanic eruptions is incomplete, due to, among other things, erosion. Thus, maps based on this record alone, will underestimate the likelihood that an area may be affected during future eruptions. A statistical algorithm will be presented which can be used to compensate for the incompleteness of the stratigraphic record. Cartographic functions can then be applied to produce various types of hazard map. The method is illustrated using the eruption history of Montagne Pelée.

Willy Aspinall (Aspinall and Associates)
Statistics and subjective probabilities: inputs to real-time management of the Montserrat volcanic eruption crisis

Howard Grubb:
envstat list page:
Updated: January 27, 2003