Routine Estimation of the Prospect of Individual Forecasts and Their Efficiency

Abstract

One of the main tasks facing the Kamchatka Branch of the Russian Expert Council on Earthquake Prediction and Seismic Risk Assessment (KB REC) is to evaluate submitted forecasts of large earthquakes. The experience at the KB REC showed that forecasts predicting the future evolution of a seismicity situation submitted by researchers very frequently contain merely some individual elements necessary for prediction, which is clearly insufficient. Examination of such forecasts is difficult, and occasionally outright impossible, which complicates the use of such data in the analysis of seismicity situations and the development of a comprehensive forecast evaluation. While the evaluation of the success of a forecast causes no difficulty to a first approximation, a preliminary (at the time the forecast is submitted) evaluation of the quality of a forecast in any form is frequently difficult. This especially concerns novel or recently developed prediction procedures, that is, procedures without a “history” at the KB REC. It is such forecasts which constitute the object of this study. The tasks we set ourselves to resolve are (1) developing a procedure applicable to the evaluation of a forecast at the time it is submitted, including the estimation of the probability of random occurrence, (2) estimation of the success of a forecast when any statistics based on the use of this particular prediction procedure is absent. This paper presents the approaches used at the KB REC for routine evaluation of the success of a forecast and/or precursors. All parameters used to evaluate prediction procedures and their practical application are formalized and possess clear physical meaning.