Speaker
Mrs
Ludmila Provost
(IRSN)
Description
Magnitude estimates of earthquakes occurred before the instrumental period are a key issue in seismic hazard assessment. For such earthquakes, the only information available is provided by historical sources. These are first translated into macroseismic intensity by means of intensity scales and then intensities are used to estimate earthquake parameters such as epicentral intensity, magnitude and sometimes depth. Methods for computing earthquake parameters from intensity data differ within and between countries, leading to the long-lasting discussions about the reference earthquake catalogue that should be used for seismic hazard assessment. Within the SHARE project (http://www.share-eu.org/), the SHEEC 1000-1899 catalogue (https://www.emidius.eu/SHEEC/; Stucchi et al., 2013) first attempted at homogenizing data and procedures for the assessment of the parameters of historical earthquakes. However some issues remained open, including the way depth is taken into account in the methodology and the way uncertainty is quantified.
We compare two methods, Boxer (Gasperini et al 2010) and Quake-MD (Provost and Scotti, 2017) applied to a set of 62 events located along the French-Italian border using the same macroseismic data (SisFrance database, http://www.sisfrance.net) and epicentral location.
In Boxer parameter estimates result from the application of a single intensity prediction equation (IPE) to each isoseismal. The final estimate is the mean of the magnitudes associated to each isoseismal. The standard deviation of the mean depends on the weight given to each isoseismal, which in turn, depends on the number of data for the given event as well as the number of data used in the IPE calibration procedure for the given intensity class and its associated standard deviation. In the Quake-MD method, on the other hand, several IPEs are calibrated in order to fit the intensity decrease with distance and applied through a least-squares procedure that results in a solution for each IPE with an associated standard deviation. A space of acceptable solutions is then defined for each IPE by rejecting magnitude-depth solutions that are incompatible with the allowed epicentral intensity values and the given IPE. The sum of the acceptable solutions quantifies the epistemic uncertainty of the earthquake parameters in the Magnitude-Depth-Epicentral Intensity space.
We show that the treatment of depth has a major impact on magnitude estimates. We compared the barycenter of the Quake-MD space of solutions and Boxer magnitude estimates. Boxer does not assess depth; whereas in the Quake-MD method depth is inverted. In order to quantify differences between the two methods we run a first test assuming that depth is fixed at 10 km depth. In this case, application of the two methods to the 62 macroseismic data sets lead to differences between magnitude estimates characterized by a mean of 0.02 and a standard deviation of 0.26 with only 6% of the events presenting a difference in magnitude estimates greater than 0.5. We then ran a second test for the Quake-MD methodology, where depth can be inverted for and is allowed to range between 1 and 25 km. In this case differences in magnitude estimates are characterized by a mean of 0.05 and a standard deviation of 0.4.
In order to understand the statistical significance of such differences, we compared quantified uncertainties using both methodologies. The space of acceptable solutions resulting from the Quake-MD method can be quite large compared to the uncertainty estimates proposed by Boxer. For the first test case we show that 92% Boxer mean magnitude estimates fall within the space of solutions proposed by the Quake-MD methodology.
This work underlines the importance of quantifying uncertainties in parametric earthquake catalogues and the need to propagate such uncertainties in seismic hazard assessments, be it probabilistic or deterministic.
Primary author
Mrs
Ludmila Provost
(IRSN)
