SUMMARY
We present results on radiated seismic energy during simulations of dynamic ruptures in a continuum damage-breakage rheological model incorporating evolution of damage within the seismic ...source region. The simulations vary in their initial damage zone width and rate of damage diffusion with parameter values constrained by observational data. The radiated energy recorded at various positions around the source is used to calculate seismic potency and moment. We also calculate the normalized radiated energy from the source, in a way that allows comparing between results of different simulations and highlighting aspects related to the dilatational motion during rupture. The results show that at high-frequencies, beyond the dominant frequency of the source ($( {f > 3{f}_d} )$, the damage process produces an additional burst of energy mainly in the Pwaves. This eccess of high-frequency energy is observed by comparing the radiated energy to a standard Brune's model with a decay slope of the radiated energy of n = 2. While the Swaves show good agreement with the n = 2 slope, the Pwaves have a milder slope of n = 1.75 or less depending on the damage evolution at the source. In the used damage-breakage rheology, the rate of damage diffusivity governs the damage evolution perpendicular to the rupture direction and dynamic changes of the damage zone width. For increasing values of damage diffusivity, dilatational energy becomes more prominent during rupture, producing a high-frequency dilatational signature within the radiation pattern. The high-frequency radiation pattern of the Pwaves includes two main lobes perpendicular to the rupture direction, reflecting high-rate local tensile cracking during the overall shear rupture process. Analysing the possible existence and properties of such high-frequency radiation pattern in observed Pwaves could provide important information on earthquake source processes.
Determining earthquake hypocenters and focal mechanisms requires precisely measured P wave arrival times and first‐motion polarities. Automated algorithms for estimating these quantities have been ...less accurate than estimates by human experts, which are problematic for processing large data volumes. Here we train convolutional neural networks to measure both quantities, which learn directly from seismograms without the need for feature extraction. The networks are trained on 18.2 million manually picked seismograms for the Southern California region. Through cross validation on 1.2 million independent seismograms, the differences between the automated and manual picks have a standard deviation of 0.023 s. The polarities determined by the classifier have a precision of 95% when compared with analyst‐determined polarities. We show that the classifier picks more polarities overall than the analysts, without sacrificing quality, resulting in almost double the number of focal mechanisms. The remarkable precision of the trained networks indicates that they can perform as well, or better, than expert seismologists.
Key Points
We train and validate convolutional neural networks to pick P wave arrival times and first‐motion polarities on 19.4 million seismograms
Arrival time picks are within 0.028 s of the analyst pick 75% of the time, and first‐motions are classified with 95% precision
The remarkable performance of the trained networks suggests that they can perform as well, or better, than human experts
The novel technique of distributed acoustic sensing (DAS) holds great potential for underwater seismology by transforming standard telecommunication cables, such as those currently traversing various ...regions of the world’s oceans, into dense arrays of seismo‐acoustic sensors. To harness these measurements for seismic monitoring, the ability to record transient ground deformations is investigated by analyzing ambient noise, earthquakes, and their associated phase velocities, on DAS records from three dark fibers in the Mediterranean Sea. Recording quality varies dramatically along the fibers and is strongly correlated with the bathymetry and the apparent phase velocities of recorded waves. Apparent velocities are determined for several well‐recorded earthquakes and used to convert DAS S‐wave strain spectra to ground motion spectra. Excellent agreement is found between the spectra of nearby underwater and on‐land seismometers and DAS converted spectra, when the latter are corrected for site effects. Apparent velocities greatly affect the ability to detect seismic deformations: for the same ground motions, slower waves induce higher strains and thus are more favorably detected than fast waves. The effect of apparent velocity on the ability to detect seismic phases, quantified by expected signal‐to‐noise ratios, is investigated by comparing signal amplitudes predicted by an earthquake model to recorded noise levels. DAS detection capabilities on underwater fibers are found to be similar to those of nearby broadband sensors, and superior to those of on‐land fiber segments, owing to lower velocities at the ocean‐bottom. The results demonstrate the great potential of underwater DAS for seismic monitoring and earthquake early warning.
Key Points
The noise content of underwater distributed acoustic sensing (DAS) along three different telecommunication cables is quantified and compared to adjacent broadband stations
Earthquake detection capabilities using DAS are similar to those of broadband instruments
Detection capabilities are mainly a function of the recorded noise, cable response, and apparent velocity
SUMMARY
The Earth’s ellipticity of figure has an effect on the traveltimes of seismic waves over teleseismic distances. Tables of ellipticity corrections and coefficients have been used by ...seismologists for several decades; however, due to the increasing variety and complexity of seismic phases in use, current tables of ellipticity coefficients are now outmoded and incomplete. We present a Python package, EllipticiPy, for the calculation of ellipticity corrections, which removes the dependence on pre-calculated coefficients at discrete source depths and epicentral distances. EllipticiPy also facilitates the calculation of ellipticity corrections on other planetary bodies. When applied to both Earth and Mars, the magnitudes of ellipticity corrections are of the order of single seconds and are significant for some seismic studies on Earth but remain negligible on Mars due to other greater sources of uncertainty.
Research on integrating statistical knowledge into deep learning models for earthquake forecasting has been limited. Traditional deep learning models require extensive parameter learning from ...scratch. This study proposes a Spatio-Temporal Convolutional (STC) model that incorporates spatio-temporal decay prior knowledge derived from the Epidemic-Type Aftershock Sequence (ETAS) into a convolutional kernel. This allows the STC model to have the prototype to learn the pattern of mainshocks to trigger aftershocks at the beginning of training, with only 4 neurons to fine-tune it. In California, the STC and the ETAS model are conducted for forecasting next-day earthquakes with magnitudes of M ≥3, M ≥4, and M ≥5. Both performances were assessed using the Receiver Operating Characteristic (ROC) curve, the Precision-Recall (PR) curve, and the Parimutuel Gambling Score (PGS). The evaluation results indicate that the STC model surpasses ETAS in forecasting next-day earthquakes not accidental. Furthermore, our analysis suggests that including earthquakes below the complete magnitudes can enhance the STC model's classification performance, as small earthquakes also contain information about future earthquakes.
SUMMARY
An unbiased estimation of the b-value and of its variability is essential to verify empirically its physical contribution to the earthquake generation process, and the capability to improve ...earthquake forecasting and seismic hazard. Notwithstanding the vast literature on the b-value estimation, we note that some potential sources of bias that may lead to non-physical b-value variations are too often ignored in seismological common practice. The aim of this paper is to discuss some of them in detail, when the b-value is estimated through the popular Aki’s formula. Specifically, we describe how a finite data set can lead to biased evaluations of the b-value and its uncertainty, which are caused by the correlation between the b-value and the maximum magnitude of the data set; we quantify analytically the bias on the b-value caused by the magnitude binning; we show how departures from the exponential distribution of the magnitude, caused by a truncated Gutenberg–Richter law and by catalogue incompleteness, can affect the b-value estimation and the search for statistically significant variations; we derive explicitly the statistical distribution of the magnitude affected by random symmetrical error, showing that the magnitude error does not induce any further significant bias, at least for reasonable amplitude of the measurement error. Finally, we provide some recipes to minimize the impact of these potential sources of bias.
SUMMARY
Many natural hazards exhibit inverse power-law scaling of frequency and event size, or an exponential scaling of event magnitude (m) on a logarithmic scale, for example the Gutenberg–Richter ...law for earthquakes, with probability density function p(m) ∼ 10−bm. We derive an analytic expression for the bias that arises in the maximum likelihood estimate of b as a function of the dynamic range r. The theory predicts the observed evolution of the modal value of mean magnitude in multiple random samples of synthetic catalogues at different r, including the bias to high b at low r and the observed trend to an asymptotic limit with no bias. The situation is more complicated for a single sample in real catalogues due to their heterogeneity, magnitude uncertainty and the true b-value being unknown. The results explain why the likelihood of large events and the associated hazard is often underestimated in small catalogues with low dynamic range, for example in some studies of volcanic and induced seismicity.