Auswahl der wissenschaftlichen Literatur zum Thema „Prompt elastogravity signals“

SUMMARY Since gravity changes propagate at the speed of light, gravity perturbations induced by earthquake deformation have the potential to enable faster alerts than the current earthquake early warning systems based on seismic waves. Additionally, for large earthquakes (Mw > 8), gravity signals may allow for a more reliable magnitude estimation than seismic-based methods. Prompt elastogravity signals induced by earthquakes of magnitude larger than 7.9 have been previously detected with seismic arrays and superconducting gravimeters. For smaller earthquakes, down to Mw ≃ 7, it has been proposed that detection should be based on measurements of the gradient of the gravitational field, in order to mitigate seismic vibration noise and to avoid the cancelling effect of the ground motions induced by gravity signals. Here we simulate the five independent components of the gravity gradient signals induced by earthquakes of different focal mechanisms. We study their spatial amplitude distribution to determine what kind of detectors is preferred (which components of the gravity gradient are more informative), how detectors should be arranged and how earthquake source parameters can be estimated. The results show that early earthquake detections, within 10 s of the rupture onset, using only the horizontal gravity strain components are achievable up to about 140 km distance from the epicentre. Depending on the earthquake focal mechanism and on the detector location, additional measurement of the vertical gravity strain components can enhance the detectable range by 10–20 km. These results are essential for the design of gravity-based earthquake early warning systems.

AbstractRapid and reliable estimation of large earthquake magnitude (above 8) is key to mitigating the risks associated with strong shaking and tsunamis1. Standard early warning systems based on seismic waves fail to rapidly estimate the size of such large earthquakes2–5. Geodesy-based approaches provide better estimations, but are also subject to large uncertainties and latency associated with the slowness of seismic waves. Recently discovered speed-of-light prompt elastogravity signals (PEGS) have raised hopes that these limitations may be overcome6,7, but have not been tested for operational early warning. Here we show that PEGS can be used in real time to track earthquake growth instantaneously after the event reaches a certain magnitude. We develop a deep learning model that leverages the information carried by PEGS recorded by regional broadband seismometers in Japan before the arrival of seismic waves. After training on a database of synthetic waveforms augmented with empirical noise, we show that the algorithm can instantaneously track an earthquake source time function on real data. Our model unlocks ‘true real-time’ access to the rupture evolution of large earthquakes using a portion of seismograms that is routinely treated as noise, and can be immediately transformative for tsunami early warning.

AbstractState‐of‐the‐art earthquake early warning systems use the early records of seismic waves to estimate the magnitude and location of the seismic source before the shaking and the tsunami strike. Because of the inherent properties of early seismic records, those systems systematically underestimate the magnitude of large events, which results in catastrophic underestimation of the subsequent tsunamis. Prompt elastogravity signals (PEGS) are low‐amplitude, light‐speed signals emitted by earthquakes, which are highly sensitive to both their magnitude and focal mechanism. Detected before traditional seismic waves, PEGS have the potential to produce unsaturated magnitude estimates faster than state‐of‐the‐art systems. Accurate instantaneous tracking of large earthquake magnitude using PEGS has been proven possible through the use of a Convolutional Neural Network (CNN). However, the CNN architecture is sub‐optimal as it does not allow to capture the geometry of the problem. To address this limitation, we design PEGSGraph, a novel deep learning model relying on a Graph Neural Network (GNN) architecture. PEGSGraph accurately estimates the magnitude of synthetic earthquakes down to 7.6–7.7 and determines their focal mechanisms (thrust, strike‐slip or normal faulting) within 70 s of the event's onset, offering crucial information for predicting potential tsunami wave amplitudes. Our comparative analysis on Alaska and Western Canada data shows that PEGSGraph outperforms PEGSNet, providing more reliable rapid magnitude estimates and enhancing tsunami warning reliability.

Abstract In the minutes following a large earthquake, robust characterization of the seismic rupture can be obtained from full wavefield records at local distances or from early signals recorded by regional broadband seismometers. We focus here on the latter configuration, and evaluate the individual and joint performances of the early low-frequency elastic phases (W phase) and the recently discovered prompt elastogravity signals (PEGS). The 2011 Mw 9.1 Tohoku–Oki earthquake is a natural target for this evaluation, because the high quality of global and regional networks enabled to gather the best PEGS data set so far. We first confirm that the well-established W-phase method, using records from global seismological networks, is able to provide a reliable centroid moment tensor solution 22 min after the earthquake origin time. Using regional stations, an accurate W-phase solution can be obtained more rapidly, down to 10 min after origin time. On the other hand, a PEGS-based source inversion can provide even earlier, starting 3 min after origin time, a lower bound of the seismic moment (Mw 8.6) and constraints on the focal mechanism type. However, relying solely on PEGS introduces uncertainties caused by the hindering seismic noise and trade-offs between source parameters that limit the accuracy of source determination. We show that incorporating even a few early W phase signals to the PEGS data set reduces these uncertainties. Using more complete W phase and PEGS data sets available 5 min after origin time enables to converge towards a result close to the Global Centroid Moment Tensor solution.

L'estimation précise et rapide de la magnitude des grands séismes est cruciale pour prévoir les tsunamis potentiels. Les systèmes traditionnels d'alerte sismique précoce s'appuyant sur les premières ondes sismiques (P) enregistrées, fournissent des estimations rapides de la magnitude (Mw) mais saturent généralement pour les événements de Mw ≥ 7,5, les rendant inadaptés pour l'alerte tsunami. Des systèmes alternatifs, s'appuyant sur la phase W ultérieure ou sur des signaux géodésiques, fournissent des estimations de magnitude non saturées plus précises, au prix d'une alerte beaucoup plus lente, et donc de temps d'avertissement beaucoup plus courts. Dans ce contexte, nous explorons le potentiel des signaux élasto-gravitationnels précoces (PEGS). Les PEGS se propagent à la vitesse de la lumière, sont sensibles à la magnitude et au mécanisme au foyer du séisme et ne saturent pas pour les très grands événements. Afin d'exploiter rapidement l'information contenue dans ces signaux de très faible amplitude, nous utilisons une approche d'apprentissage profond. Nous entraînons d'abord un Réseau de Neurones Convolutif (CNN) pour estimer la magnitude et la localisation d'un séisme à partir des PEGS synthétiques augmentés de bruit empirique (enregistré par de vrais sismomètres). Testé sur des données réelles le long de la zone de subduction chilienne, nous montrons qu'il aurait correctement estimé la magnitude du séisme de Maule de 2010 (Mw 8,8). Néanmoins, l'approche semble être limitée aux événements de Mw ≥ 8,7 dans ce contexte. Nous utilisons ensuite un Réseau de Neurones Graphiques (GNN) conçu pour améliorer les performances du CNN. Nous montrons que le GNN peut être utilisé pour estimer rapidement la magnitude des événements de Mw ≥ 8,3 au Pérou. Enfin, nous implémentons le modèle dans le système d'alerte précoce du Pérou (en complément du système actuel d'alerte précoce aux séismes) et testons son utilisation opérationnelle pour l'alerte tsunami en temps réel simulé
Accurate and timely estimation of large earthquake magnitudes is critical to forecast potential tsunamis. Traditional earthquake early warning systems, relying on the early recorded seismic (P) waves, provide fast magnitude (Mw) estimates but typically saturate for Mw ≥ 7.5 events, making them unfit for tsunami warning. Alternative systems, relying on the later W phase or on geodetic signals, provide more accurate unsaturated magnitude estimates, to the cost of much slower warning, and therefore much shorter warning times. In this context, we explore the potential of prompt elastogravity signals (PEGS). PEGS propagate at the speed of light, are sensitive to the magnitude and focal mechanism of the earthquake and do not saturate for very large events. In order to rapidly leverage the information contained in these very low-amplitude signals we use a deep learning approach. We first train a Convolutional Neural Network (CNN) to estimate the magnitude and location of an earthquake based on synthetic PEGS augmented with empirical noise (recorded by actual seismometers). Tested on real data along the chilean subduction zone, we show that it would have estimated correctly the magnitude of the 2010 Mw 8.8 Maule earthquake. Nevertheless, the approach appears to be limited to Mw ≥ 8.7 events in this context. We then use a Graph Neural Network (GNN) designed to improve the performance of the CNN. We show that the GNN can be used to rapidly estimate the magnitude of Mw ≥ 8.3 events in Peru. Finally, we implement the model in the early warning system of Peru (as a complement of the current earthquake early warning system) and test its operational use for tsunami warning in simulated real time