The decision by an Italian court to prosecute 6 Italian scientists for failure to ‘predict’ the L’Aquila earthquake in 2009 is a gross example of the misunderstanding of the processes governing earthquake occurrence. Contrary to this prosecution’s apparent logic, earthquakes are fundamentally, or inherently unpredictable. All available geological evidence indicates that their prediction in time is just not possible. The logic used by the scientists in stating that they considered the situation before the magnitude 6.3 event in 2009 ‘normal’ was in keeping with best available knowledge of earthquake generation processes. Italy is a region characterized by frequent seismic activity, as first mapped by Robert Mallet, an Irish engineer and enlightened geoscientist of the 19th century.
Detailed catalogs of events in well-monitored seismically active areas (e.g. fault-riddled mines) show that ‘big’ events do not have reliable fore-warnings such as increases in or sets of smaller earthquake events and/or changes in other earth processes e.g. changes in radon gas levels, water pressure etc. The processes seem to be governed at a micro-scale by the shape of the fault surfaces (best represented by a fractal distribution of unevenness) or more directly, the friction between the opposing fault surfaces. Opposing sides of a fault are being pressed together by tectonic forces that are continuous. However stress release is not continuous, and occurs as a quick jolt that we feel as an ‘earthquake’ (essentially a slippage or ‘rupture’ in the temporary join across the fault).
‘Where’ a slippage or break in the contact between the fault surfaces might occur may be possible to forecast (not ‘predict’ however) as McCloskey et al (2005) demonstrated by successfully modelling stresses on the Sumatran fault zone following the tragic magnitude 9.2 Sumatra–Andaman earthquake on the Sunda Trench megathrust (a type of deep-earth fault). The Sumatra–Andaman earthquake gave rise to the St Stephen’s Day tsunami of 2004. However, the same advanced modelling of seismic stress fields that leads to such advances uses at its core an underlying model that also rules out the possibility of allowing earthquake prediction in time (i.e. the ‘when’), in addition to space (the ‘where’).
The way that the physics captured by the model works to faithfully reproduce earthquake catalogs and regional stress fields, and predict the location of fault stress release, says something fundamental about earthquake generation processes. Essentially, very small changes in the tectonic stress being experienced by a fault can lead to very large earthquakes, as a critical threshold may be passed within the system. As McCloskey et al (2005) put it when describing their modelling of the 2nd rupture along the Sunda megathrust: ‘The size of this triggering stress [~0.1 bar] illustrates the extreme non-linearity of the earthquake nucleation process’.
The ability of humans to predict ‘big’ earthquakes is controlled by the same non-linear, inherently unpredictable, geological processes. Failure to predict the timing of major seismic events should not therefore be considered a criminal offense, and sets a grossly unfair, ill-informed and dangerous precedent in the areas of hazards research and management.