Preface to the special issue of Artificial Intelligence in Seismology

Seismology is a data-intensive and data-driven science.The rapid growth of seismometer density and data size calls for more efficient and effective processing tools.In recent years,artificial intelligence(AI)has been increasingly used in various areas of seismology.Among them,earthquake monitoring is likely the one most impacted(Kong QK et al.,2019;Mousavi and Beroza,2022).Popular seismic phase picking models and workflows like PhaseNet,EQTransformer,RISP,PALM,LOC-FLOW,QUAKE-FLOW(Zhu WQ and Beroza,2019;Mousavi et al.,2020;Liao SR et al.,2021;Zhou YJ et al.,2021;Zhang M et al.,2022;Zhu WQ et al.,2023)have been proposed and widely used.Also,AI algorithms for association(Ross et al.,2019;Yu ZY and Wang WT,2022),polarity determination and focal mechanism inversion(Ross et al.,2018;Zhang J et al.,2023;Li S et al.,2023),earthquake discrimination(Li ZF et al.,2018;Linville et al.,2019;Miao FJ et al.,2020)have emerged.

The China Seismological Reference Model project

The importance of developing high-resolution seismic models to improve understanding of tectonic processes and enhance seismic hazard mitigation programs,along with the rapid expansion of seismic coverage in China,called for a seismological reference model to be established in China.The China Seismological Reference Model(CSRM)project was initiated by the National Natural Science Foundation of China with two primary goals:(1)the CSRM would serve as a primary source for the current state of seismological research in China,and(2)the seismic data and constraints used to construct the CSRM would be used as a backbone open-access cyberinfrastructure for future research in seismology.The CSRM project was also intended to promote data exchange and scientific collaboration in seismology in China.Accordingly,two parallel efforts of the project are being pursued:(1)construction of the CSRM,and(2)development of a CSRM product center.The CSRM is jointly constrained by various types of seismic constraints extracted from the seismic data recorded at 4511 seismic stations in continental China following a top-down approach,with the seismic structures in the shallower part of the Earth constrained first.Construction of the CSRM involves three preparation steps:(1)building datasets of various seismic constraints from the seismic data,(2)developing a method to incorporate the constraints of surface wave observations from regional earthquakes into the inversion of the seismic structure,and(3)constructing high-resolution pre-CSRM seismic models of the velocity structure in the shallow crust and the Pn-velocity structure in the uppermost mantle.In the final process,the CSRM will be constructed by jointly inverting all the seismic constraints using the pre-CSRM models as starting models or a priori structures.The CSRM product center(http://chinageorefmodel.org)archives and distributes three types of products:CSRM models,the Level 1 original seismic data used to extract seismic constraints in the construction of the CSRM,and Level 2 data on the seismic constraints derived from the Level 1 data and the inferred earthquake parameters in the construction of the CSRM.The CSRM product center has archived 141 TB of Level 1 data from 1120 permanent broadband stations in the China Seismic Network Center and 3391 temporary stations from various institutions and data centers around the world,as well as 140 GB of Level 2 data on various seismic constraints and inferred event parameters from the construction of the CSRM.The CSRM is expected to provide significant insights into the composition and tectonic dynamics in continental China and to enhance the capability of various seismic hazard mitigation programs in China from near real-time rapid determination of earthquake parameters to an earthquake early warning system.The CSRM could also provide guidance for focuses in future seismological research and the design of future active and passive seismic experiments in China.Several focuses are suggested for future seismological research in China,along with the building of a national cyberinfrastructure to sustain and expand the operations of the CSRM project.

Implementation of a particle-in-cell method for the energy solver in 3D spherical geodynamic modeling

The thermal evolution of the Earth’s interior and its dynamic effects are the focus of Earth sciences.However,the commonly adopted grid-based temperature solver is usually prone to numerical oscillations,especially in the presence of sharp thermal gradients,such as when modeling subducting slabs and rising plumes.This phenomenon prohibits the correct representation of thermal evolution and may cause incorrect implications of geodynamic processes.After examining several approaches for removing these numerical oscillations,we show that the Lagrangian method provides an ideal way to solve this problem.In this study,we propose a particle-in-cell method as a strategy for improving the solution to the energy equation and demonstrate its effectiveness in both one-dimensional and three-dimensional thermal problems,as well as in a global spherical simulation with data assimilation.We have implemented this method in the open-source finite-element code CitcomS,which features a spherical coordinate system,distributed memory parallel computing,and data assimilation algorithms.

Recent advances in earthquake monitoring I:Ongoing revolution of seismic instrumentation

Seismic networks have significantly improved in the last decade in terms of coverage density,data quality,and instrumental diversity.Moreover,revolutionary advances in ultra-dense seismic instruments,such as nodes and fiber-optic sensing technologies,have recently provided unprecedented high-resolution data for regional and local earthquake monitoring.Nodal arrays have characteristics such as easy installation and flexible apertures,but are limited in power efficiency and data storage and thus most suitable as temporary networks.Fiber-optic sensing techniques,including distributed acoustic sensing,can be operated in real time with an in-house power supply and connected data storage,thereby exhibiting the potential of becoming next-generation permanent networks.Fiber-optic sensing techniques offer a powerful way of filling the observation gap particularly in submarine environments.Despite these technological advancements,various challenges remain.First,the data characteristics of fiber-optic sensing are still unclear.Second,it is challenging to construct software infrastructures to store,transfer,visualize,and process large amount of seismic data.Finally,innovative detection methods are required to exploit the potential of numerous channels.With improved knowledge about data characteristics,enhanced software infrastructures,and suitable data processing techniques,these innovations in seismic instrumentation could profoundly impact observational seismology.

Preliminary results of strong ground motion simulation for the Lushan earthquake of 20 April 2013,China

The earthquake occurred in Lushan County on 20 April, 2013 caused heavy casualty and economic loss. In order to understand how the seismic energy propagates during this earthquake and how it causes the seismic haz- ard, we simulated the strong ground motions from a rep- resentative kinematic source model by Zhang et al. （Chin J Geophys 56（4）：1408-1411, 2013） for this earthquake. To include the topographic effects, we used the curved grids finite difference method by Zhang and Chen （Geophys J Int 167（1）：337-353, 2006）, Zhang et al. （Geophys J Int 190（1）：358-378, 2012） to implement the simulations. Our results indicated that the majority of seismic energy con- centrated in the epicentral area and the vicinal Sichuan Basin, causing the XI and VII degree intensity. Due to the strong topographic effects of the mountain, the seismic intensity in the border area across the northeastern of Boxing County to the Lushan County also reached IX degree. Moreover, the strong influence of topography caused the amplifications of ground shaking at the moun- tain ridge, which is easy to cause landslides. These results are quite similar to those observed in the Wenchuan earthquake of 2008 occurred also in a strong topographic mountain area.

Recent advances in earthquake monitoring II: Emergence of next-generation intelligent systems

Seismic data processing techniques,together with seismic instrumentation,determine our earthquake monitoring capability and the quality of resulting earthquake catalogs.This paper is intended to review the improvement of earthquake monitoring capability from the perspective of data processing.Over the past two decades,seismologists have made considerable advancements in seismic data processing,partly thanks to the significant development of computational power,signal processing,and machine learning techniques.In particular,wide application of template matching and increasing use of deep learning significantly enhance our capability to extract signals of small earthquakes from noisy data.Relative location techniques provide a critical tool to elucidate fault geometries and seismicity migration patterns at unprecedented resolution.These techniques are becoming standard,leading to emerging intelligent software systems for next-generation earthquake monitoring.Prospective improvements in future research must consider the urgent needs in highly generalizable detection algorithms(for both permanent and temporary deployments)and in emergency real-time monitoring of ongoing sequences(e.g.,aftershock and induced seismicity sequences).We believe that the maturing of intelligent and high-resolution processing systems could transform traditional earthquake monitoring workflows and eventually liberate seismologists from laborious catalog construction tasks.

Morphology and possible origins of the Perm anomaly in the lowermost mantle of Earth

We have constrained a small-scale,dome-shaped low-velocity structure near the core-mantle boundary(CMB)of Earth beneath Perm(the Perm anomaly)using travel-time analysis and three-dimensional(3-D)forward waveform modeling of seismic data sampling of the mantle.The best-fitting dome-shaped model centers at 60.0°E,50.5°N,and has a height of 400 km and a radius that increases from 200 km at the top to 450 km at the CMB.Its velocity reduction varies from 0%at the top to–3.0%at 240km above the CMB to–3.5%at the CMB.A surrounding 240-km-thick high-velocity D''structure has also been detected.The Perm anomaly may represent a stable smallscale chemical pile in the lowermost mantle,although the hypothesis of a developing mantle plume cannot be ruled out.

USTC-Pickers:a Unified Set of seismic phase pickers Transfer learned for China

Current popular deep learning seismic phase pickers like PhaseNet and EQTransformer suffer from performance drop in China.To mitigate this problem,we build a unified set of customized seismic phase pickers for different levels of use in China.We first train a base picker with the recently released DiTing dataset using the same U-Net architecture as PhaseNet.This base picker significantly outperforms the original PhaseNet and is generally suitable for entire China.Then,using different subsets of the DiTing data,we fine-tune the base picker to better adapt to different regions.In total,we provide 5 pickers for major tectonic blocks in China,33 pickers for provincial-level administrative regions,and 2 special pickers for the Capital area and the China Seismic Experimental Site.These pickers show improved performance in respective regions which they are customized for.They can be either directly integrated into national or regional seismic network operation or used as base models for further refinement for specific datasets.We anticipate that this picker set will facilitate earthquake monitoring in China.

Preface to the special issue on earthquake sequences in western China in May 2021

In May 2021,multiple sequences of earthquakes struck western China.Among them,The M_(S) 6.4 Yangbi earthquake and the M_(W) 7.4 Madoi earthquake are the two most significant events.Especially,the Madoi earthquake is the largest earthquake in China after the 2008 Wenchuan earthquake.The two earthquakes have attracted broad concerns from both the public and the academic society(Liao et al.,2021,Zhang et al.,2022a).In this special issue,ERA compiled and published related research articles investigating timely advances of the two events.The special issue includes 7 papers,covering seismic observation,surface deformation investigation,seismic hazard survey,and geodynamical modeling.

Aseismic ridge subduction and flat subduction:Insights from three-dimensional numerical models

Flat subduction can significantly influence the distribution of volcanism,stress state,and surface topography of the overriding plate.However,the mechanisms for inducing flat subduction remain controversial.Previous two-dimensional(2-D)numerical models and laboratory analogue models suggested that a buoyant impactor(aseismic ridge,oceanic plateau,or the like)may induce flat subduction.However,three-dimensional(3-D)systematic studies on the relationship between flat subduction and buoyant blocks are still lacking.Here,we use a 3-D numerical model to investigate the influence of the aseismic ridge,especially its width(which is difficult to consider in 2-D numerical models),on the formation of flat subduction.Our model results suggest that the aseismic ridge needs to be wide and thick enough to induce flat subduction,a condition that is difficult to satisfy on the Earth.We also find that the subduction of an aseismic ridge parallel to the trench or a double aseismic ridge normal to the trench has a similar effect on super-wide aseismic ridge subduction in terms of causing flat subduction,which can explain the flat subduction observed beneath regions such as Chile and Peru.

Observation 20-s periodic signals on Mars from InSight,Sols 800-1,000

Seismometers of the InSight probe(Interior Exploration using Seismic Investigation,Geodesy and Heat Transport)currently operating on Mars have recorded not only seismic events but also high-frequency non-seismic periodic signals that appear to have been induced by variations in the Martian environment and the hardware.Here,we report an observation of a long-period signal with a dominant period of~20 s from Martian solar days(Sol)800 to Sol 1,000.This 20-s signal is detected mostly at quiet nighttime—from22:00 to 04:00 LMST(Local Mean Solar Time)—at the InSight landing site.The measurement of the particle motion suggests that this linearly polarized signal focuses on the horizontal plane with an angle of~30°from the north.By examining the temporal variation of the signal’s amplitude and polarization angle and its times of occurrence in relation to the planet’s atmospheric data,we suggest that this20-s signal may be relevant to wind and temperature variations on Mars.Furthermore,we study the possible influence of this 20-s signal on the noise autocorrelation and find that the stacked autocorrelograms can be quite different when the 20-s signal is present.

A method for inversion of layered shear wavespeed azimuthal anisotropy from Rayleigh wave dispersion using the Neighborhood Algorithm

Seismic anisotropy provides important constraints on deformation patterns of Earth＇s material. Rayleigh wave dispersion data with azimuthal anisotropy can be used to invert for depth-dependent shear wavespeed azimuthal anisotropy, therefore reflecting depth-varying deformation patterns in the crust and upper mantle. In this study, we propose a two-step method that uses the Neighborhood Algorithm （NA） for the point-wise inversion of depth-dependent shear wavespeeds and azimuthal anisotropy from Rayleigh wave azimuthally anisotropic dispersion data. The first step employs the NA to estimate depth- dependent Vsv （or the elastic parameter L） as well as their uncertainties from the isotropic part Rayleigh wave dispersion data. In the second step, we first adopt a difference scheme to compute approximate Rayleigh-wave phase velocity sensitivity kernels to azimuthally anisotropic parameters with respect to the velocity model obtained in the first step. Then we perform the NA to estimate the azimuthally anisotropic parameters Gc/L and Gs/L at depths separately from the corresponding cosine and sine terms of the azimuthally anisotropic dispersion data. Finally, we compute the depth-dependent magnitude and fast polariza- tion azimuth of shear wavespeed azimuthal anisotropy. The use of the global search NA and Bayesian analysis allows for more reliable estimates of depth-dependent shear wavespeeds and azimuthal anisotropy as well as their uncertainties.We illustrate the inversion method using the azimuthally anisotropic dispersion data in SE Tibet, where we find apparent changes of fast axes of shear wavespeed azimuthal anisotropy between the crust and uppermost mantle.

First-principles investigation of the concentration effect on equilibrium fractionation of Ca isotopes in forsterite

Previous theoretical studies have found that the concentration variations within a certain range have a prominent effect on inter-mineral equilibrium isotope fractionation(10^3 lna).Based on the density functional theory,we investigated how the average Ca–O bond length and the reduced partition function ratios(10^3 lnb)and103lna of 44 Ca/40 Ca in forsterite(Fo)are affected by its Ca concentration.Our results show that Ca–O bond length in forsterite ranges from 2.327 to 2.267 A with the Ca/(Ca+Mg)varying between a narrow range limited by an upper limit of 1/8 and a lower limit of 1/64.However,outside this narrow range,i.e.,Ca/(Ca+Mg)is lower than1/64 or higher than 1/8,Ca–O bond length becomes insensitive to Ca concentration and maintains to be a constant.Because the 10^3 lnb is negatively correlated with Ca–O bond length,the 10^3lnb significantly increases with decreasing Ca/(Ca+Mg)when 1/64<Ca/(Ca+Mg)<2/16.As a consequence,the 10^3lna between forsterite and other minerals also strongly depend on the Ca content in forsterite.Combining previous studies with our results,the heavier Ca isotopes enrichment sequence in minerals is:forsterite[orthopyroxene[clinopyroxene[calcite & diopside[dolomite[aragonite.Olivineand pyroxenes are enriched in heavier Ca isotope compared to carbonates.The 10^3lna between forsterite with a Ca/(Ca+Mg)of 1/64 and clinopyroxene(Ca/Mg=1/1,i.e.,diopside)is up to^0.64%at 1200 K.The large 103lnaFodiopsiderelative to the current analytical precision for Ca isotope measurements suggests that the dependence of10^3 lnaFo-diopsideon temperature can be used as a thermometer,similar to the one based on the 103lna of 44 Ca/40 Ca between orthopyroxene and diopside.These two Ca isotope thermometers both have a precision approximate to that of elemental thermometers and provide independent constraints on temperature.

Monitoring of velocity changes based on seismic ambient noise: A brief review and perspective

Over the past two decades,the development of the ambient noise cross-correlation technology has spawned the exploration of underground structures.In addition,ambient noise-based monitoring has emerged because of the feasibility of reconstructing the continuous Green’s functions.Investigating the physical properties of a subsurface medium by tracking changes in seismic wave velocity that do not depend on the occurrence of earthquakes or the continuity of artificial sources dramatically increases the possibility of researching the evolution of crustal deformation.In this article,we outline some state-of-the-art techniques for noise-based monitoring,including moving-window cross-spectral analysis,the stretching method,dynamic time wrapping,wavelet cross-spectrum analysis,and a combination of these measurement methods,with either a Bayesian least-squares inversion or the Bayesian Markov chain Monte Carlo method.We briefly state the principles underlying the different methods and their pros and cons.By elaborating on some typical noisebased monitoring applications,we show how this technique can be widely applied in different scenarios and adapted to multiples scales.We list classical applications,such as following earthquake-related co-and postseismic velocity changes,forecasting volcanic eruptions,and tracking external environmental forcing-generated transient changes.By monitoring cases having different targets at different scales,we point out the applicability of this technology for disaster prediction and early warning of small-scale reservoirs,landslides,and so forth.Finally,we conclude with some possible developments of noise-based monitoring at present and summarize some prospective research directions.To improve the temporal and spatial resolution of passive-source noise monitoring,we propose integrating different methods and seismic sources.Further interdisciplinary collaboration is indispensable for comprehensively interpreting the observed changes.

Shallow crustal velocity structures revealed by active source tomography and fault activities of the Mianning–Xichang segment of the Anninghe fault zone, Southwest China

The Anninghe fault is a large left-lateral strike-slip fault in southwestern China. It has controlled deposition and magmatic activities since the Proterozoic, and seismic activity occurs frequently. The Mianning-Xichang segment of the Anninghe fault is a seismic gap that has been locked by high stress. Many studies suggest that this segment has great potential for large earthquakes(magnitude >7). We obtained three vertical velocity profiles of the Anninghe fault(between Mianning and Xichang) based on the inversion of P-wave first arrival times. The travel time data were picked from seismograms generated by methane gaseous sources and recorded by three linearly distributed across-fault dense arrays. The inversion results show that the P-wave velocity structures at depths of 0-2 km corresponds well with the local lithology. The Quaternary sediments have low seismic velocities, whereas the igneous rocks,metamorphic rocks, and bedrock have high seismic velocities. We then further discuss the fault activities of the two fault branches of the Anninghe fault in the study region based on small earthquakes(magnitudes between ML 0.5 and ML 2.5) detected by the Xichang array.The eastern fault branch is more active than the western branch and that the fault activities in the eastern branch are different in the northern and southern segments at the border of 28°21′N. The high-resolution models obtained are essential for future earthquake rupture simulations and hazard assessments of the Anninghe fault zone. Future studies of velocity models at greater depths may further explain the complex fault activities in the study region.

Thermal structure of continental subduction zone: high temperature caused by the removal of the preceding oceanic slab

The thermal structure of the continental subduction zone can be deduced from high-pressure and ultra-high-pressure rock samples or numerical simulation.However,petrological data indicate that the temperature of subducted continental plates is generally higher than that derived from numerical simulation.In this paper,a two-dimensional kinematic model is used to study the thermal structure of continental subduction zones,with or without a preceding oceanic slab.The results show that the removal of the preceding oceanic slab can effectively increase the slab surface temperature of the continental subduction zone in the early stage of subduction.This can sufficiently explain the difference between the cold thermal structure obtained from previous modeling results and the hot thermal structure obtained from rock sample data.

Passive Seismic Structure Imaging of a Coal Mine by Ambient Noise Seismic Interferometry on a Dense Array

Active source seismic method is generally used to image subsurface structures for resource exploration,including oil,gas and coal.Although it can provide highresolution subsurface structures,due to some economic and environmental restrictions,it is not suitable in some cases.In recent 20 years,passive seismic survey based on ambient noise seismic interferometry(ANSI)has started to be widely used in imaging subsurface structures.In comparison,ANSI does not need active sources and can image subsurface structures at different depths as a lowcost alternative to active seismic exploration.

Thermodynamic properties of wadsleyite with anharmonic effect

The thermodynamic properties of crystals can be routinely calculated by density functional theory calculations combining with quasi-harmonic approximation. Based on the method developed recently by Wu and Wentzcovitch （Phys Rev B 79：104304, 2009） and Wu （Phys Rev B 81：172301, 2010）, we are able to further ab initio include anharmonic effect on thermodynamic properties of crystals by one additional canonical ensemble with numbers of particle, volume and temperature fixed （NVT） molecular dynamic simulations. Our study indicates that phonon-phonon interaction causes the renormalized phonon frequencies of wadsleyite decrease with temperature. This is consistent with the Raman experimental observation. The anharmonic free energy of wadsleyite is negative and its heat capacity at constant pressure can exceed the Dulong-Petit limit at high temperature. The anharmonicity still significantly affects thermodynamic properties of wadsleyite at pressure and temperature con- ditions correspond to the transition zone.

3D v_P and v_S models of southeastern margin of the Tibetan plateau from joint inversion of body-wave arrival times and surface-wave dispersion data

A new 3D velocity model of the crust and upper mantle in the southeastern （SE） margin of the Tibetan plateau was obtained by joint inversion of body- and sur- face-wave data. For the body-wave data, we used 7190 events recorded by 102 stations in the SE margin of the Tibetan plateau. The surface-wave data consist of Rayleigh wave phase velocity dispersion curves obtained from ambient noise cross-correlation analysis recorded by a dense array in the SE margin of the Tibetan plateau. The joint inversion clearly improves the Vs model because it is constrained by both data types. The results show that at around 10 km depth there are two low-velocity anomalies embedded within three high-velocity bodies along the Longmenshan fault system. These high-velocity bodies correspond well with the Precambrian massifs, and the two located to the northeast of 2013 Ms 7.0 Lushan earthquake are associated with high fault slip areas during the 2008 Wenchuan earthquake. The aftershock gap between 2013 Lushan earthquake and 2008 Wenchuan earthquake is associated with low-velocity anomalies, which also acts as a barrier zone for ruptures of two earthquakes. Generally large earthquakes （M 〉 5） in the region occurring from 2008 to 2015 are located around the high-velocity zones, indicating that they may act as asperities for these large earthquakes. Joint inversion results also clearly show that there exist low-velocity or weak zones in the mid-lower crust, which are not evenly distributed beneath the SE margin of Tibetan plateau.

Analytical and numerical simulations of uplift processes at the Tibet-Sichuan boundary

Previous studies have shown that the uplift of Tibetan plateau started in response to the collision of Indian plate and Eurasian plate. During this process, the crust of Tibetan plateau has been greatly thickened which leads to significant elevations. The elevation gradient is extremely large at the east boundary of Tibetan plateau where Longmenshan fault exists, dropping from 4500 to 500 m within a distance of 100 km, while it is more gentle at the south and north sides of Sichuan basin. Such a difference of elevation gradient has been explained with a crustal channel flow model. However, pre- vious crustal flow models consider the thickness of the lower crust as a constant which is highly simplified. Therefore, it is essential to build a more realistic crustal flow model, in which the thickness of the lower crust is variable and dependent on the inflow velocity of crustal materials. Here we build up both analytical and numerical models to study the mechanism and process of the uplift of Tibetan plateau at the eastern boundary. The results of the analytical model show that if the thickness of the lower crust can vary during the uplift process, the lower crustal viscosity of the Sichuan basin needs to be 1022 Pas to fit the observed elevation gradient. Such a viscosity is one-order magnitude larger than the previous results. Numerical model results further show that the state of stresses at the plateau boundary changes during uplift processes. Such a stress state change may cause the formation of different fault types in the Longmenshan fault area during its uplift history.