Modeling of magnitude distributions by the generalized truncated exponential distribution

TL;DR

This paper introduces the generalized truncated exponential distribution (GTED) to improve modeling of seismic magnitudes, addressing limitations of existing exponential models and demonstrating its effectiveness with empirical data.

Contribution

The paper proposes the GTED, a flexible model that generalizes existing exponential distributions, allowing better seismic magnitude distribution modeling and parameter estimation.

Findings

GTED outperforms TED based on Akaike information criterion

GTED can be applied to geographic seismic data

The model effectively captures the distribution near the upper bound

Abstract

The probability distribution of the magnitude can be modeled by an exponential distribution according to the Gutenberg-Richter relation. Two alternatives are the truncated exponential distribution (TED) and the cut-off exponential distribution (CED). The TED is frequently used in seismic hazard analysis although it has a weak point: When two TEDs with equal parameters except the upper bound magnitude are mixed, then the resulting distribution is not a TED. Inversely, it is also not possible to split a TED of a seismic region into TEDs of sub-regions with equal parameters except the upper bound magnitude. This weakness is a principal problem as seismic regions are constructed scientific objects and not natural units. We overcome it by the generalization of the above-mentioned exponential distributions: the generalized truncated exponential distribution (GTED). Therein, identical exponential distributions are mixed by the probability distribution of the correct cut-off points. This distribution model is flexible in the vicinity of the upper bound magnitude and is equal to the exponential distribution for smaller magnitudes. Additionally, the exponential distributions TED and CED are special cases of the GTED. We discuss the possible ways of estimating its parameters and introduce the normalized spacing for this purpose. Furthermore, we present methods for geographic aggregation and differentiation of the GTED and demonstrate the potential and universality of our simple approach by applying it to empirical data. The considerable improvement by the GTED in contrast to the TED is indicated by a large difference between the corresponding values of the Akaike information criterion.