The announcement that six Italian scientists and one government official have been charged with manslaughter following the L'Aquila earthquake in 2009 in Italy may have grave implications for scientists who try to provide advice on low probability events such as earthquakes in the context of high uncertainties but also high consequences. The charges alleged that members of the National Commission for Forecasting and Predicting Great Risks who held a meeting in L'Aquila the week before the earthquake provided "incomplete, imprecise and contradictory information" and "failed to adequately evaluate and communicate the risk to the local population". Here we present some of the scientific background to the case.
A magnitude 6.3 earthquake at 03:32 local time on 6 April 2009 struck the city of L'Aquila in the Abruzzo region of central Italy. The earthquake killed over 300 people, injured over 1,500, destroyed around 20,000 buildings and displaced over 65,000 people. Seismic activity in the L'Aquila area had increased above normal background levels in January 2009. A number of these earthquakes were felt, causing some alarm. The largest was a magnitude 4.1 earthquake on 30 March. Two earthquakes with magnitudes of 3.5 and 3.9 occurred a few hours before the mainshock.
Central Italy has long history of destructive earthquakes including events in 1349, 1461, 1703 and 1762. The 1461 earthquake caused the almost complete destruction of Onna, Poggio Picenze, Castelnuovo and Sant'Eusanio Foronese. An earthquake near Avezzano in 1915 (about 40 km south of L'Aquila) killed around 30,000 people and was one of the deadliest European earthquakes of the 20th Century. More recently, a magnitude 5.9 Mw earthquake in October 2002 struck the town of San Giuliano di Puglia, 140 km south of L'Aquila, destroying a local school building. In September 1997 two earthquakes with magnitudes of 5.8 and 6.0 Mw struck the Umbria region close to the city of Assisi, killing eleven people and seriously damaging many buildings. In addition, earthquake swarms, sequences of earthquakes clustered in time and space without a clear distinction of main shock and aftershocks, are relatively common in this part of Italy.
Since we generally know the places where earthquakes can occur, over longer periods of time, such as many years, we are in a position to make reasonably reliable estimates about earthquake hazard in different places. The hazard is usually expressed as the probability of a given level of ground motion within a certain period of time. This can help us to plan for future earthquakes. In this sense, the hazard at L'Aquila was well-known and maps of the seismic hazard were readily available at INGV.
Buildings that have been designed appropriately generally sustain much lower levels of damage in earthquakes. However, many of the historic masonry buildings in the region were vulnerable to earthquakes. Buildings that had been retrospectively strengthened fared much better. This highlights the critical importance of fully implementing building codes and retrospective strengthening measures.
Earthquakes occur in sequences that are clustered in space and time. The largest earthquake in the sequence is known as the mainshock. The mainshock is followed by aftershocks, the number of which increases with the magnitude of the mainshock. Earthquakes that occur before the mainshock are known as foreshocks. These can only be identified after the mainshock. Analysis of data worldwide (Reasenberg, Pure and Applied Geophysics, 155, 355-379 (1999)) shows that only about 15% of earthquakes are preceded by a foreshock that is within one unit of magnitude of the mainshock and within 10 days and 75 km of the origin time and epicentre.
This is an area of ongoing research and the issue was discussed in a report by the International Commission on Earthquake Forecasting for Civil Protection following L'Aquila. A number of statistical models have been used to examine changes in the probability of an earthquake with time for given sequences of seismicity, e.g. Marzocchi and Lombardi (2009) examine this for the L'Aquila sequence. These models indicate that the probability of a large earthquake can increase in the aftermath of small ones, but nonetheless remains low. Furthermore, the uncertainties in these models are substantial and do not provide a basis for reliable forecasting. Attempts to forecast with these models would lead to large false alarm rates.
There will be an appeal, but whatever happens, the L'Aquila case has important implications for the way that scientists communicate uncertainty to both decision makers and the general public. Scientists need to be completely open about the data that we collect and the models that we use, as well as the uncertainties associated with both. Unfortunately, this case may make scientists much more cautious about any interpretation of high consequence events.