A spatially varying model of the detection/location capabilities of the Southern Great Basin seismic network (SGBSN) has been derived that is based on simple empirical relations and statistics. This permits use of almost all the catalog data gathered; instead of ignoring data that are below the threshold of completeness, a spatially varying threshold model is developed so that subregions having lower completeness levels than the network as a whole can be outlined and the completeness level of each sub-region determined. Such a model is required to unambiguously identify regions that are aseismic due to natural processes rather than to limited detection and/or location capabilities. Accounting for spatial variations in detection/location threshold is also important for studies in which magnitude-frequency distributions are interpreted in terms of source scaling properties. The characteristics of the spatial distribution of earthquakes, where earthquake clusters and aseismic regions locate, appear to be stable at all magnitude levels so that inferences about where strain is being accommodated will be the same whether numbers of earthquakes or strain estimated from seismic moments are examined. For the southern Great Basin region these principal characteristics include clusters at the northern end of the Furnace Creek fault and in the Pahranagat Shear Zone, and a relatively large number of earthquakes in the northern and southeastern portions of the Nevada Test Site. These clusters cover regions much larger than the surface projections of any of the mapped faults. The extent to which seismicity is induced by nuclear testing is unclear. The predominantly aseismic regions include the area west of the Death Valley/Furnace Creek fault system and an almost complete absence of events at Yucca Mountain. Finally, a considerable number of isolated events in the SGBSN catalog cannot be correlated with mapped faults.

Seismicity and detection/location threshold in the Southern Great Basin Seismic Network

The incremental detection capability of a seismic station, i.e., the probability of detection as a function of event magnitude, is usually estimated either by using the exponential magnitude-frequency relationship of earthquakes (the recurrence method) or by comparing the station detections to an independent reference system (the direct method) . The recurrence method usually yields significantly lower 90 per cent thresholds than the direct method, while a hypothetical “true” threshold would lie between these values. This is explained theoretically as resulting from the global variance of observed signal amplitudes for a given seismic event. A maximum likelihood estimation procedure is developed for the direct method, and approximate confidence limits on the estimated detection thresholds are derived. The two estimation techniques are applied to determine the event detection capability of the NORSAR array for the Japan-Kuriles-Kamchatka region, and the results are related to previously published estimates.

On the estimation of seismic detection thresholds

: This report deals with the general problem of estimating the incremental detection capability of a seismic station, ie, the probability of detection as a function of event magnitude A maximum likelihood procedure is introduced to estimate the station detection thresholds by comparison with an independent reference station or network Approximate confidence limits for the estimated parameters are computed The results obtained by this (direct) method are compared to estimates derived from applying the exponential magnitude- frequency relationship of natural seismicity It is found that the latter (indirect) method in general gives significantly lower 90 percent detection thresholds It is concluded that, if a good reference system is available, the direct method will produce estimates that are more reliable than those provided by the indirect method

http://www.dtic.mil/cgi-bin/GetTRDoc?AD=ADA001100&Location=U2&doc=GetTRDoc.pdf

Estimation of Seismic Detection Thresholds

Based on 10 yr of observations reported in the International Seismological Centre bulletins (1971-1980), m b magnitudes for about 70,000 earthquakes have been recomputed using a maximum-likelihood estimation technique. Reportings from a network of 115 globally distributed stations were used in these calculations. Comparison to conventional m b estimates show that the network magnitude bias problem is quite significant at low and intermediate magnitudes. Recurrence statistics based on the revised m b estimates give b values consistent with those obtained from LASA and NORSAR data, and indicate that the number of earthquakes worldwide of m b ≧ 4.0 averages about 7,500 annually. This number is much lower than previous estimates. The teleseismic detection capability of the network (requiring at least four detecting stations) has been evaluated based on recurrence statistics. The estimated 90 per cent incremental m b threshold ranges from 3.9 to 4.5 in the northern hemisphere and from 4.2 to 4.8 in the southern hemisphere. This is consistent with results obtained by the “Networth” approach.

Study of magnitudes, seismicity, and earthquake detectability using a global network

This paper describes a methodology to assess the uncertainty in seismic hazard estimates at particular sites. A variant of the bootstrap statistical method is used to combine the uncertainty due to earthquake catalog incompleteness, earthquake magnitude, and recurrence and attenuation models used. The uncertainty measure is provided in the form of a confidence interval. Comparisons of this method applied to various sites in California with previous studies are used to confirm the validity of the method.

Application of the bootstrap method to quantify uncertainty in seismic hazard estimates

: The problem of estimating seismicity and the performance of a system which detects earthquakes is formulated in such a way that maximum likelihood estimation can be applied. The mean number of earthquakes which occur in a fixed time interval is assumed to be of the form exp (a-bm) where m is magnitude and a and b are constants. The probability of detection as a function of m is taken to be an error function. Procedures to obtain maximum likelihood estimates of a, b, and the error function mean and variance factors are derived, discussed, and applied to experimental data to check the relevance of the theoretical development.

Richard T. Lacoss

Papers

Estimation of seismicity and network detection capability