Improving the estimation of detection probability and magnitude of completeness in strongly heterogeneous media, an application to acoustic emission (AE)

Abstract

Reliable estimations of magnitude of completeness (M c ) are essential for a correct interpretation of seismic catalogues. The spatial distribution of M c may be strongly variable and difficult to assess in mining environments, owing to the presence of galleries, cavities, fractured regions, porous media and different mineralogical bodies, as well as in consequence of inhomogeneous spatial distribution of the seismicity. We apply a 3-D modification of the probabilistic magnitude of completeness (PMC) method, which relies on the analysis of network detection capabilities. In our approach, the probability to detect an event depends on its magnitude, source-receiver Euclidian distance and source-receiver direction. The suggested method is proposed for study of the spatial distribution of the magnitude of completeness in a mining environment and here is applied to a 2-months acoustic emission (AE) data set recorded at the Morsleben salt mine, Germany. The dense seismic network and the large data set, which includes more than one million events, enable a detailed testing of the method. This method is proposed specifically for strongly heterogeneous media. Besides, it can also be used for specific network installations, with sensors with a sensitivity, dependent on the direction of the incoming wave (e.g. some piezoelectric sensors). In absence of strong heterogeneities, the standards PMC approach should be used. We show that the PMC estimations in mines strongly depend on the source-receiver direction, and cannot be correctly accounted using a standard PMC approach. However, results can be improved, when adopting the proposed 3-D modification of the PMC method. Our analysis of one central horizontal and vertical section yields a magnitude of completeness of about M c ≈ 1 (AE magnitude) at the centre of the network, which increases up to M c ≈ 4 at further distances outside the network; the best detection performance is estimated for a NNE-SSE elongated region, which corresponds to the strike direction of the low-attenuating salt body. Our approach provides us with small-scale details about the capability of sensors to detect an earthquake, which can be linked to the presence of heterogeneities in specific directions. Reduced detection performance in presence of strong structural heterogeneities (cavities) is confirmed by synthetic waveform modelling in heterogeneous media.

Maximum Magnitude of Completeness in a Salt Mine

“…The catalog information was compiled from the data recorded by a microseismic network that was installed to monitor microcrack processes at the Morsleben salt mine, Allertal region, Germany. This site has been extensively monitored for AE studies (e.g., Spies and Eisenblätter, 2001;Becker et al, 2010;Maghsoudi et al, 2013Maghsoudi et al, , 2014.…”

“…M AE values of this catalog were arbitrarily scaled by a factor of 0.05 to range between 0 and 5, which could correspond to M w from −8 to −3, according to empirical energy relations (Eisenblätter and Spies, 2000;Kaiser, 2011). The M c using this M AE scale, derived from the directional probabilistic magnitude of completeness (PMC) approach, was found to be M c ≈ 1 at the center of the network and increases up to M c ≈ 4 at further distances outside the network (Maghsoudi et al, 2013(Maghsoudi et al, , 2014. Figure 1a illustrates the distribution of epicenters in the central part of the Morsleben mine.…”

“…M c is usually heterogeneous in time and space, as it depends on factors such as network geometry and instrumentation, noise level, and attenuation of the seismic signals with distance. Especially in mining settings, due to the effects of local heterogeneities on the seismic wavefield, the M c estimations may show strong spatial variations (e.g., Plenkers et al, 2011;Maghsoudi et al, 2013).…”

“…Schorlemmer and Woessner (2008) proposed an alternative method to infer M c , namely the probability-based magnitude of completeness, which does not require assumptions about the FMD. Extended approaches of this method have been applied to catalogs of acoustic emission (AE) events in a gold mine in South Africa (Plenkers et al, 2011) and in a salt mine in Germany (Maghsoudi et al, 2013).…”

“…In such conditions, if most (or all) sensors are saturated, and the remaining information is not sufficient (e.g., to locate or estimate the magnitude of the seismic event), this may not be included in the catalog, resulting in a catalog incompleteness for large magnitude events. In this work, we discuss this latter case by considering an AE monitoring network in a salt mine and the corresponding catalog, which we analyzed previously (Maghsoudi et al, 2013(Maghsoudi et al, , 2014. This data set is particularly well suited to test our hypothesis, because of the broad range of recordable magnitudes, the sensor characteristics, and the location procedure, which requires a minimum number of nonsaturated sensors.…”