Evaluation of earthquake prediction rates allowing for magnitude uncertanties

• Autorzy: Lasocki S.
• Języki publikacji: EN

Abstrakty

EN

One of the possible ways of evaluation of the earthquake prediction efficiency is a posterior count of successful predictions compared with the number of all issued predictions and the number of all events which should be predicted. Two efficiency parameters are in general use: the success rate defined as a percentage of the successful predictions in the total number of issued predictions and the alarm rate defined as a percentage of the total number of issued predictions and the alarm rate defined as a percentage to be successful predictions in the number of events which should be predicted. These parameters cannot be easily assessed when the identification of targets of prediction is incertain. When the prediction system aims at the events of magnitudes not lower than a given threshold value, the uncertainly can be generated by errors in magnitude evaluations. An estimation of the succes and alarm rates under encertain identification of prediction targets is formulates as a probabilistic transition from the actual to observed magnitudes. In general, the problem can be solved only when the probability of incidental correlation of issued predictions with the 'other' events that are not targets of prediction, but that are included a test set, is taken into account. A parameter controlling these incidental correlations, called the incidental correlation rate, is defined as the ratio of the number of incidential correlations to the total number of 'other' events. A complete solution for the two-class and multiclass cases is given. The two-class case, in which an event both actually falls into and ois observed as falling into one of two categories, is solved on the basis of Bayes theorem. In the multiclass case, coresponding to the actual situation in the evaluation of earthquake prediction efficiency, the transition from the actual to observed magnitude classes of events that originally follow the magnitude-frequency relation, is considered. This case is solved by an analysis of the prediction and transition processes, given the prediction and incidental correlation rates and the transition probabilities. Since the final system of equations is strongly ill-conditioned, an approximate way of estimation of the prediction efficiency is proposed. Numerical tests, carried out for various combinations of initial conditions and Monte Carlo simulations of the prediction/transition processes, have proved that the solution is correct and its approximation is reasonable. The same tests show taht the estimators inferred from previous suggestions of several authors can lead to unacceptable results.

Słowa kluczowe

EN

earthquake prediction; prediction efficiency; prediction rates

• Wydawca: Instytut Geofizyki PAN
• Czasopismo: Acta Geophysica Polonica
• Rocznik: 1998
• Tom: Vol. 46, nr 3
• Strony: 233--259
• Opis fizyczny: bibliogr. 6 poz. rys., tab.

Twórcy

autor

Lasocki S.

• Akademia Górniczo-Hutnicza 30-059 Kraków Al. Mickiewicza 30,