Automatic inspection and analysis of digital waveform images by means of convolutional neural networks

Analyzing seismic data to get information about earthquakes has always been a major task for seismologists and, more in general, for geophysicists. Recently, thanks to the technological development of observation systems, more and more data are available to perform such tasks. However, this data “grow up” makes “human possibility” of data processing more complex in terms of required efforts and time demanding. That is why new technological approaches such as artificial intelligence are becoming very popular and more and more exploited. In this paper, we explore the possibility of interpreting seismic waveform segments by means of pre-trained deep learning. More specifically, we apply convolutional networks to seismological waveforms recorded at local or regional distances without any pre-elaboration or filtering. We show that such an approach can be very successful in determining if an earthquake is “included” in the seismic wave image and in estimating the distance between the earthquake epicenter and the recording station.

Early Result for Microtremor Characteristic Observation in Merapi and Merbabu

The arc magmatism and volcanic activity in Java are dominated by the subducting plate of Indo-Australian into the Eurasian plate. Merapi volcano is located in Central Java and known as one of the most active volcanoes in the world. Several studies have tried to estimate the magma reservoir zone in Mt. Merapi and suggested multiple layers of reservoirs with the shallow one at 1-2 km and a deeper at 6 -9 km or 15 km. The Low-Frequency Passive Seismic is one method to analyze the frequency spectrum below the recording station. Previous related studies show a promising a relation between hydrocarbon reservoir and higher amplitude at vertical component at a frequency between 0.1 – 6 Hz. An observation at the volcano sites have also been reported to display a different spectrum amplitude at the vertical component. This study exploited the same method in LFPS to analyze the frequency spectrum at Mt. Merapi and Mt. Merbabu. We use seismic data from the DOMERAPI temporary seismic network installed in the neighborhood of Merapi and Merbabu volcano. We analyze 53 broad-band seismometers data from October 2013 to mid-April 2015. We also add several stations from MERAMEX network instruments to compare spectrum analysis outside the Merapi and Merbabu volcano. We also removed some tele-seismic and regional events from the data to better analyze the LFPS signal. We have seen a higher amplitude in vertical component near Mt. Merapi and will proceed to analyze all stations.

Anthropogenic Noise and Its Footprint on ELF Schumann Resonance Recordings

A set of various short artificial disturbances from rifle firings, car engine operation, car radio, shakings of the apparatus, etc., were generated deliberately near our ELF recording stations in order to identify their footprint on the recordings of atmospheric electromagnetic radiation in the Schumann resonance (SR) band (from about 2–50 Hz). Such disturbances simulate anthropogenic noises from hunters, hikers, campers, etc., which may occur in a remote-isolated ELF recording station. We expect that our work will assist fellow scientists to differentiate between artificial signals created from anthropogenic activity and real signals attributable to geophysical phenomena.

A pragmatic approach to adjusting early instrumental local magnitudes for seismic hazard assessments in Australia

Prior to the development of Australian-specific magnitude formulae, the 1935 magnitude correction factors by Charles Richter—originally developed for southern California—were almost exclusively used to calculate earthquake magnitudes throughout Australia prior to the 1990s. Due to the difference in ground-motion attenuation between southern California and much of the Australian continent, many earthquake magnitudes from the early instrumental era are likely to have been overestimated in the Australian earthquake catalogue. A method is developed that adjusts local magnitudes (ML) using the difference between the original (inappropriate) magnitude formulae (or look-up tables) and the Australian-specific formulae at a distance determined by the nearest recording station likely to have recorded the earthquake. Nationally, these adjustments have reduced the number of earthquakes of ML ≥ 4.5 in the early instrumental catalogue by approximately 25% since 1900, while the number of ML ≥ 5.0 earthquakes has reduced by approximately 32% over the same time period. The reduction in the number of moderate-to-large-magnitude earthquakes over the instrumental period yields long-term earthquake rates that are more consistent with present-day rates, since the development of Australian-specific magnitude formulae (approximately 1990). The adjustment of early instrumental magnitudes to obtain consistently derived earthquake catalogue is important for seismic hazard assessments.

Challenges in the definition of input motions for forensic ground-response analysis in the near-source region

This article scrutinizes the determination of input motions for forensic ground-response analysis in the near-source region, based on recorded surface ground motions at strong-motion station sites, from the same event. The first part of the article draws upon observed ground motions from the 22 February 2011 6.2 Mw Christchurch earthquake to discuss key challenges of the problem associated with the strong spatial variation of ground motion in the near-source region. Effects from the complexity of the rupture, propagation of seismic waves through complex geological structures, and site characteristics are explored. It is argued that, because of the strongly varying source-path “signature” on near-source ground motions, “reference” input motions for ground-response analysis must be specific to, and have similar signature characteristics (be “compatible”) with, the target site which is subject to the analysis. The second part of the article presents a four-step procedure for the derivation of site-specific input motions involving (1) determination of the reference layer where the input motion is to be applied in the analysis, (2) record selection considering the appropriateness of the recording station site for deconvolution and its compatibility with the target site, (3) deconvolution of the selected record to remove local site effects from the recorded ground motion, and (4) scaling of the deconvolved motion to account for differences in the source-to-site distance between the recording station and the target site. As part of the proposed procedure, a novel (amplitude-duration) scaling method is presented. Results from one-dimensional (1D) effective-stress analysis of two target Christchurch sites using input motions from the proposed procedure are used to critically evaluate the procedure and discuss essential requirements for its successful application.

Determination of path attenuation and site characteristics of the North-west Himalayan region and adjoining regions within the Indian Territory using Generalized inversion method

North-west Himalayas and its adjoining regions have been experiencing deadly earthqaukes from time to time and are home for a large portion of population of Indian subcontinent. Knowledge of regional path attenuation and site parameters are prerequisite while attempting seismic hazard studies towards minimizing damages during future earthqaukes for a region. Present work focuses on the determination of path attenuation and site characteristics of earthqaukes recording stations, located in the north-west Himalayas and its adjoining regions, within India. It is done using two- step generalized inversion technique. In the first step of inversion, non-parametric attenuation curves are developed by constraining attenuation to be a smooth decaying function with hypocentral distance. Qs = (105 ± 11)f (0.94 ± 0.08) as S wave quality factor is obtained indicating that the region is seismically active having high degree of heterogeneities in the crustal medium. In the second step of generalized inversion, site amplification curve, at each recording station, is computed as the ratio of site spectral amplitude of horizontal and vertical components. In addition, based on Horizontal to vertical spectral ratio (HVSR) method, predominant frequency of each recording station is calculated. Values of predominant frequencies based on HVSR and generalized inversion are found matching for each of the recording station. Based on obtained predominant frequency, site class of 101 recording stations, which at present are absent, are determined in this work. Determined path attenuation as well as site parameters can be collectively used for developing regional ground motion models and subsequently for seismic hazard studies for the selected region.

Assimilation of atmospheric infrasound data to constrain tropospheric and stratospheric winds

<p>We use acoustical infrasound from explosions to probe an atmospheric wind component from the ground up to stratospheric altitudes. Planned explosions of old ammunition in Finland generate transient infrasound waves that travel through the atmosphere. These waves are partially reflected back towards the ground from stratospheric levels, and are detected at a receiver station located in northern Norway at 178 km almost due North from the explosion site. The difference between the true horizontal direction towards the source and the back-azimuth direction of the incoming infrasound wave-fronts, in combination with the pulse propagation time, are exploited to provide an estimate of the average cross-wind component in the penetrated atmosphere. <br>We perform offline assimilation experiments with an ensemble Kalman filter and these observations, using the ERA5 ensemble reanalysis atmospheric product as background (prior) for the wind at different vertical levels. Information from both sources is combined to obtain analysis (posterior) estimates of cross-winds at different vertical levels of the atmospheric slice between the explosion site and the recording station. The assimilation makes greatest impact at the 12-60 km levels, with some changes with respect to the prior of the order of 0.1-1.0 m/s, which is a magnitude larger than the typical standard deviation of the ERA5 background. The reduction of background variance in the higher levels often reached 2-5%. <br>This is the first study demonstrating  techniques to implement assimilation of infrasound data into atmospheric models. It paves the way for further exploration in the use of infrasound  observations (especially natural and continuous sources) to probe the middle atmospheric dynamics and to assimilate these data into atmospheric model products.  </p>

Optimal resolution seismic tomography: Lithospheric thinning beneath the British Tertiary Igneous Province and other new observations

<p>The maximum achievable resolution of a tomographic model varies spatially and depends on the data sampling and errors in the data. Adaptive parameterization schemes match the spatial variations in data sampling but do not address the effects of the errors. The propagation of systematic errors, however, is resistant to data redundancy and results in models dominated by noise if the target resolution is too high. This forces us to look for smoother models and thus limits the imaging resolution.<br><br>We develop a surface-wave tomography method that finds optimal lateral resolution at every point by means of error tracking. We first measure inter-station phase-velocities at simultaneously recording station pairs and compute phase-velocity maps at densely, logarithmically spaced periods. Unlike in the classical approach, multiple versions of the maps with varying smoothness constraints are computed, so that the maps range from very rough to very smooth. Phase-velocity curves extracted from the maps at every point can then be inverted for shear-velocity (V<sub>s</sub>) profiles. As we show, errors in these phase-velocity curves increase nearly monotonically with the map roughness. Very smooth V<sub>s</sub> models computed from very smooth phase-velocity maps will be the most robust, but at a cost of a loss of most structural information. At the other extreme, models that are too rough will be dominated by noise. We define the optimal resolution at a point such that the error of the local phase-velocity curve is below an empirical threshold. The error is estimated by isolating the roughness of the phase-velocity curve that cannot be explained by any Earth structure. A 3D V<sub>s</sub> model is then computed by the inversion of the phase-velocity maps with the optimal resolution at every point. The estimated optimal resolution shows smooth lateral variations, confirming the robustness of the procedure. Importantly, optimal resolution does not scale with the density of the data coverage: some of the best-sampled locations require relatively low lateral resolution, probably due to systematic data errors. We apply the method to image the Ireland’s and Britain’s upper mantle, using our large, new regional dataset. We report a pronounced thinning of the lithosphere beneath the British Tertiary Igneous Province, with important implications for the Paleogene uplift and volcanism in the region.</p>

Comparison of global lightning activity variations inferred from Q-bursts, Schumann resonances, and WWLLN-detected lightning strokes

<p>Measured time series of the extremely low frequency (ELF, 3 Hz-3 kHz) band electromagnetic field can be considered as a superposition of background and transient signals. Transient signals produced by exceptionally powerful lightning strokes far from the recording station are named Q-bursts. The direction of the source lightning stroke at the recording station can be calculated using the horizontal components of the Poynting vector. The source lightning stroke can be identified in the lightning database of the World Wide Lightning Location Network (WWLLN, wwlln.net) by the matching detection time and direction calculated from ELF measurements.</p><p>Schumann resonance (SR) peaks appear at ~8Hz, ~14Hz, ~20 Hz, etc., in the spectra computed from the background ELF timeseries. SRs are natural electromagnetic resonances with wavelengths comparable to the circumference of the Earth-ionosphere waveguide. Peak amplitudes and frequencies in the resonance spectrum detected in the ELF band at any given location on the Earth depend on the distribution and intensity of the global lightning activity which excites SR.</p><p>ELF measurements are routinely performed in the Széchenyi István Geophysical Observatory (NCK, 47°38' N, 16°43' E) near Nagycenk, Hungary. Vertical electric and the horizontal magnetic components of the atmospheric electromagnetic field are monitored by the Schumann resonance recording system. In this work, we study the variation of the number of lightning strokes with high charge moment change (CMC; indicated by the number of large amplitude Q-bursts recorded at NCK) and the variation of the number of lightning strokes with large peak current (indicated by the number of WWLLN-detected energetic lightning strokes). In addition to considering the total number of WWLLN-detected lightning strokes and Q-bursts, we analyze lightning strokes occurring  only in west, south, east, and north directions from NCK, corresponding predominantly to the three main lightning producing regions of the tropical lands in America, Africa, and Indonesia as well as to the Pacific Ocean. Time variations of the number of high CMC and large peak current lightning strokes during November, 2014 are compared with time variation of the cumulative SR intensity detected at NCK station in the vertical electric field component in the same month. Similarities and differences in the time variations of the considered quantities are discussed in order to show how these indicators mirror the changing distributions of the global lightning activity.</p>

A Subsurface Geologic Feature Inferred from Relocation of Local Earthquakes in Al-Refaei District, Southern Iraq

For the last two decades, the District of Al-Refaei, southern Iraq, experienced several small to moderate size earthquakes that seem to be clustered in a relatively small area. The initial locations of these earthquakes from the available local and international bulletins show scattered events without any clear structural patterns. A seven-elements seismic array was installed for the period 2014 to 2018 to monitor this activity. Each element of the array consists of three-component Geospace GS11d high-frequency (4.5 Hz) geophone, a solar system power source, and telemetry communications system to transmit data to a central recording station. During the period of monitoring, the array recorded more than 56 earthquakes having a range of magnitude between 1.5 ML to 4.7 mb. Large number of the low magnitude earthquakes were not detected by other monitoring agencies; however, they were recorded by the array. Fifty-six of the most clearly recorded earthquakes were screened for relocation and analysis. Out of the 56 screened earthquakes, 35 were detected by the array alone. The majority of the selected earthquakes having their initial locations within the array. This improves the relocation process and increase accuracy. The relocated earthquakes express a clear pattern of a linear feature which strikes in the northwest-southeast direction. The direction of the newly inferred pattern coincides with the general tectonic setting of Iraq as it is parallel to Zagros suture zone and the general stress regime of the Arabian Plate. The depth of the relocated earthquakes ranges from about 3.9km to 8.9km. This indicates that these earthquakes occur along a shallow subsurface fault that was not mapped before.