Conference Presentations
2023UGM Annual Meeting
- Boris Rösler: Insight into Earthquake Source Processes from Moment Tensor Catalogs (Talk)
Analysis of moment tensor catalogs gives insight into general properties of many earthquakes beyond what can be obtained by studies of individual earthquakes and allows estimating uncertainties in the determination of seismic source processes. Traditionally, uncertainties in moment tensors are derived from the misfit between observed and synthetic waveforms. However, the differences between moment tensors in the USGS and the Global CMT Project catalogs are typically an order of magnitude larger than the reported errors, suggesting that the reported errors substantially underestimate the uncertainty due to different inversion procedures. Differences between double-couple (DC) components decrease with magnitude, and correlation between non-double-couple (NDC) components increases, suggesting that seismic sources of large earthquakes are determined more reliably. A dataset compiled from three global and four regional catalogs shows that NDC components are essentially independent of magnitude for earthquakes over a large magnitude range, with a mean deviation from a DC source of about 20%. Additionally, there is essentially no difference in NDC components between earthquakes with different fault mechanisms and in different geologic environments. This consistency indicates that most NDC components, especially for smaller earthquakes, do not reflect real source processes and are likely to be artifacts of the inversion. Through comparison of moment tensors from three global catalogs, I find that NDC components of large earthquakes are more reliably determined, and that the largest NDC components are more likely to represent real source processes. The correlation of NDC components in different catalogs allows quantifying the noise contained in them and thus determination which catalog provides the most precise NDC components.
AGU Fall Meeting- Boris Rösler: Uncertainty Estimates for Moment Tensors and Quantities derived from them from Comparison of Global Catalogs (Poster)
Analysis of moment tensor catalogs gives insight into general properties of many earthquakes beyond what can be obtained by studies of individual earthquakes and allows estimating uncertainties in the determination of seismic source processes. Traditionally, uncertainties in moment tensors are derived from the misfit between observed and synthetic waveforms. However, the differences between moment tensors in the USGS and the Global CMT Project catalogs are typically an order of magnitude larger than the reported errors, suggesting that the reported errors substantially underestimate the uncertainty due to different inversion procedures. Differences between double-couple (DC) components decrease with magnitude, and correlation between non-double-couple (NDC) components increases, suggesting that seismic sources of large earthquakes are determined more reliably. A dataset compiled from three global and four regional catalogs shows that NDC components are essentially independent of magnitude for earthquakes over a large magnitude range, with a mean deviation from a DC source of about 20%. Additionally, there is essentially no difference in NDC components between earthquakes with different fault mechanisms and in different geologic environments. This consistency indicates that most NDC components, especially for smaller earthquakes, do not reflect real source processes and are likely to be artifacts of the inversion. Through comparison of moment tensors from three global catalogs, I find that NDC components of large earthquakes are more reliably determined, and that the largest NDC components are more likely to represent real source processes. The correlation of NDC components in different catalogs allows quantifying the noise contained in them and thus determination which catalog provides the most precise NDC components.
2022SSA Annual Meeting
- Boris Rösler & Seth Stein: Are most earthquakes' non-double-couple components artifacts? (Talk)
Earthquake moment tensors can be decomposed into double-couple components describing slip on planar faults and non-double-couple (NDC) components. NDC components can arise in three ways. Some appear to be intrinsic, indicating complex source processes differing from slip on a fault for earthquakes in specific geologic environments, notably volcanic areas. Others are additive, reflecting the combined effect of double-couple sources on multiple faults with different geometries. Alternatively, they may be artifactual results of the moment tensor inversion. Combining moment tensors from three global and four regional catalogs for 2016-2020 provides a dataset of NDC components of 12,856 earthquakes with 2.9 < Mw < 8.2 in various geologic environments. The NDC components vary only slightly with magnitude, with a mean deviation from a double-couple source of around 20%. The consistency suggests that most NDC components do not reflect rupture on multiple faults, which has been observed only for larger earthquakes. Similarly, there are only small differences in NDC components between earthquakes with different faulting mechanisms, or in different geologic environments. These consistencies suggest that most NDC components components are not intrinsic due to complex source processes that are often assumed to be most likely in volcanic and thus extensional areas. Hence it appears that for most earthquakes, especially smaller ones, the NDC components are artifacts, in accord with numerical experiments showing that they arise when the Earth model used for the inversion differs from that used to generate synthetic waveforms.
SAGE/GAGE Community Workshop- Boris Rösler & Seth Stein: Insight into Earthquake Source Processes from Large Moment Tensor Catalogs (Poster)
Analysis of large seismic moment tensor datasets gives insight into general properties of many earthquakes beyond what can be obtained by studies of individual earthquakes. Our studies to date yield three intriguing results. First, the differences between moment tensors in the USGS and Global CMT Project catalogs are typically an order of magnitude larger than the reported errors, suggesting that the reported errors substantially underestimate the uncertainty due to different inversion procedures. Second, a large dataset from three global and four regional catalogs shows that non-double-couple (NDC) components are essentially independent of magnitude for earthquakes with 2.9 < Mw < 8.2, with a mean deviation from a DC source of about 20%. There is essentially no difference in NDC components between earthquakes with different fault mechanisms, and in different geologic environments, and NDC components are only weakly correlated between catalogs. This consistency indicates that most NDC components, especially for earthquakes with Mw < 6.5, do not reflect actual source processes and are likely to be artifacts of the inversion. Third, numerical simulations of the effect of laterally varying Earth structure replicates general features of the pervasive NDC components in moment tensor catalogs, showing that these components are largely artifacts of the inversions not adequately accounting for the effects of laterally varying Earth structure.
AGU Fall Meeting- Boris Rösler & Seth Stein: Insight into Earthquake Source Processes from Large Moment Tensor Catalogs (Poster Talk)
Analysis of large seismic moment tensor datasets gives insight into general properties of many earthquakes beyond what can be obtained by studies of individual earthquakes. Our studies to date yield several intriguing results. The differences between moment tensors in the USGS and the Global CMT Project catalogs are typically an order of magnitude larger than the reported errors, suggesting that the reported errors substantially underestimate the uncertainty due to different inversion procedures. Additionally, the moment tensors of the GCMT Project have larger scalar moments on average. A large dataset from three global and four regional catalogs shows that non-double-couple (NDC) components are essentially independent of magnitude for earthquakes with 2.9 < Mw < 8.2, with a mean deviation from a DC source of about 20%. There is essentially no difference in NDC components between earthquakes with different fault mechanisms, and in different geologic environments, and NDC components are only weakly correlated between catalogs. This consistency indicates that most NDC components, especially for earthquakes with Mw < 6.5, do not reflect actual source processes and are likely to be artifacts of the inversion. Through the comparison of moment tensors from three global catalogs, we find that NDC components of large earthquakes are more reliably determined. Similarly, the largest NDC components are more likely to represent real source processes. Numerical simulations of the effect of laterally varying Earth structure replicates general features of the pervasive NDC components in moment tensor catalogs, showing that these components are largely artifacts of the inversions not adequately accounting for the effects of laterally varying Earth structure.
2021SSA Annual Meeting
- Boris Rösler & Seth Stein: Uncertainties in Seismic Moment Tensors Inferred from Differences Between Global Catalogs (Talk)
Catalogs of moment tensors form the foundation for a wide variety of studies in seismology. Despite their importance, assessing the uncertainties in the moment tensors and the quantities derived from them is difficult. To gain insight, we compare 5000 moment tensors in catalogs of the USGS and the Global CMT Project for the period from September 2015 to December 2020. The GCMT Project generally reports larger scalar moments than the USGS, with the difference between the reported moments decreasing with magnitude. The effect of the different definitions of the scalar moment between catalogs, reflecting treatment of the non-double-couple component, is consistent with that expected. However, this effect is small and has a sign opposite to the differences in reported scalar moment. Hence the differences are intrinsic to the moment tensors in the two catalogs. The differences in the deviation from a double-couple source and in source geometry derived from the moment tensors also decrease with magnitude. The deviations from a double-couple source inferred from the two catalogs are moderately correlated, with the correlation stronger for larger deviations. However, we do not observe the expected correlation between the deviation from a double-couple source and the resulting differences in scalar moment due to the different definitions. There is essentially no correlation between the differences in source geometry, scalar moment, or fraction of the non-double-couple component, suggesting that the differences reflect aspects of the inversion rather than the source process. Despite the differences in moment tensors, the reported location and depth of the centroids are consistent between catalogs.
EGU General Assembly- Boris Rösler & Seth Stein: Analysis of Differences in Seismic Moment Tensors between Global Catalogs (Talk)
Catalogs of moment tensors form the foundation for a wide variety of studies in seismology. Despite their importance, assessing the uncertainties in the moment tensors and the quantities derived from them is difficult. To gain insight, we compare 5000 moment tensors in catalogs of the USGS and the Global CMT Project for the period from September 2015 to December 2020. The GCMT Project generally reports larger scalar moments than the USGS, with the difference between the reported moments decreasing with magnitude. The effect of the different definitions of the scalar moment between catalogs, reflecting treatment of the non-double-couple component, is consistent with that expected. However, this effect is small and has a sign opposite to the differences in reported scalar moment. Hence the differences are intrinsic to the moment tensors in the two catalogs. The differences in the deviation from a double-couple source and in source geometry derived from the moment tensors also decrease with magnitude. The deviations from a double-couple source inferred from the two catalogs are moderately correlated, with the correlation stronger for larger deviations. However, we do not observe the expected correlation between the deviation from a double-couple source and the resulting differences in scalar moment due to the different definitions. There is essentially no correlation between the differences in source geometry, scalar moment, or fraction of the non-double-couple component, suggesting that the differences reflect aspects of the inversion rather than the source process. Despite the differences in moment tensors, the reported location and depth of the centroids are consistent between catalogs.
SAGE/GAGE Community Workshop- James Neely, Boris Rösler, Seth Stein & Bruce Spencer: Insight into Earthquake Stress Drops and Moment Tensors from Large Global Datasets (Poster)
Examining large datasets of source parameters can reveal how earthquakes differ by tectonic region and fault geometry. Although the specifics of the individual studies vary, they follow the same general steps: apply processing procedures and algorithms to seismograms, assume an earthquake source model, and then estimate the parameter of interest. However, these studies often fail to account for uncertainties introduced by processing procedures and model assumptions. Studies of large sets of earthquakes can overcome these limitations and provide insight into which observations are real and which are artifacts of the methodology. We compared multiple estimates of stress drop, scalar moment, and non-double-couple (NDC) components of moment tensors for a range of earthquakes. We found practically zero correlation between stress drops from two independent studies, indicating that reported stress drop differences between earthquakes may not reflect true differences. In a second study, we compared moment tensors in the USGS and the Global CMT Project catalogs and found that the differences are typically an order of magnitude larger than the reported errors, suggesting that the errors substantially underestimate the uncertainty. GCMT generally reports larger scalar moments than USGS, with the difference decreasing with magnitude. The differences appear to be intrinsic to the tensors, presumably in part due to different phases used in the inversions. A third study found that NDC components of moment tensors are essentially independent of magnitude, focal mechanism, and tectonic setting for earthquakes with 2.9 < Mw < 8.2. The consistency suggests that most NDC components do not reflect complex rupture processes. Although some earthquakes have real NDC components, it appears that for most earthquakes, especially smaller ones, NDC components are likely to be artifacts of the inversion.
AGU Fall Meeting- Boris Rösler & Seth Stein: Insight into Earthquake Source Processes from Large Global Datasets (Poster)
Most studies of earthquake source parameters give detailed information about individual earthquakes. A complementary approach is examining large datasets to gain insight into general properties of many earthquakes, rather than specifics for individual earthquakes. In the traditional formulation for inverse problems, such studies gain high stability - general properties - at the cost of low resolution - specifics for individual earthquakes. In one study we compared moment tensors in the USGS and the Global CMT Project catalogs. The differences are typically an order of magnitude larger than the reported errors, suggesting that the errors substantially underestimate the uncertainty. GCMT generally reports larger scalar moments than the USGS, with the difference decreasing with magnitude. This difference is larger and of opposite sign from that expected due to the different definitions of the scalar moment. Instead, the differences are intrinsic to the tensors, presumably in part due to different phases used in the inversions. A second study examines non-double-couple (NDC) components of moment tensors, which may reflect complex source processes for earthquakes in specific tectonic environments, the combined effect of double couple sources with different geometries, or artifacts of the inversion. A large dataset of moment tensors for earthquakes from three global and four regional catalogs shows that NDC components are essentially independent of magnitude for earthquakes with 2.9 < Mw < 8.2, with a mean deviation from a double-couple source of ~20%. The consistency suggests that most NDC components do not reflect complex rupture processes, which should be a greater effect for larger earthquakes because a significant NDC component requires substantially different geometry between portions of the rupture. Furthermore, there is essentially no difference in NDC components between earthquakes with different fault mechanisms, in different tectonic environments, or in different types of lithosphere. This consistency indicates that most NDC components do not reflect actual source processes, which would likely cause variability. Hence although some earthquakes have real NDC components, it appears that for most earthquakes, especially smaller ones, NDC components are likely to be artifacts of the inversion.
2020AGU Fall Meeting
- Vivian Tang, Boris Rösler & Suzan van der Lee: Dynamically Triggered Seismic Activity in Alaska (Talk)
Alaska is the most seismically active state of the USA. In the past 100 years, over nine thousand earthquakes occurred in Alaska with magnitudes greater than 4.5. The number of stations increased about fourfold after USArray (TA) deployed in Alaska. These new stations provide the opportunity to detect smaller earthquakes, including dynamically triggered earthquakes and tremor. Peterson and Christensen observed tectonic tremor in south-central Alaska between 1998 to 2001. Between 2006 and 2012, Gomberg and Prejean have found tremor spots in central mainland Alaska and in the islands of Unalaska and Akutan. Several studies also have reported dynamically triggered earthquakes in central Alaska. Because the current amount of seismic stations provides an overwhelming number of waveforms to analyze, we enlisted the help of over five thousand volunteer scientists through an online project, called "Earthquake Detective". In our project, we ask volunteer scientists to classify weak signals from small, potentially triggered seismic events by listening to relevant sections of seismograms that are converted to audible frequencies. By comparing the volunteer scientists' classifications with our own expert classifications for data from 5 of the 30 teleseismic earthquakes, we determined that volunteer scientists achieve over 90% accuracy in detecting earthquakes or tremor when at least 7 of 10 volunteer scientists agreed. Volunteer scientists detected earthquakes and tremor in central and southcentral Alaska within thousands of records of 30 teleseismic earthquakes with Mw > 7.5 from 2013 to 2018. These local earthquakes and tremor occurred during the first 30 minutes for surface wave propagation. However, some earthquakes were not associated with triggering. About 150 thousand earthquakes occurred in central and south-central Alaska from 2013 to 2018. Therefore, it is likely that some detections of earthquakes by volunteer scientists are part of this average rate of ~3 such earthquakes every hour. Nevertheless, the possibly triggered earthquake and tremor detections also emerged from volunteer scientists' classifications. These local earthquake and tremor signals were recorded by stations clustered in central and southcentral Alaska. The signals may originate about 100 km to 250 km away from NE Anchorage, which is close to the place where Peterson and Christensen reported their tremor observations.
2019SSA Annual Meeting
- Boris Rösler, Suzan van der Lee, Frank Elavsky & Manochehr Bahavar: New IRIS Data Product: Dynamic Surface-Wave Radiation Patterns (Talk)
We present the theory and implementation of a new data product at the IRIS DMC. The Surface Wave Radiation Pattern product shows the azimuthal variations of radiated seismic-wave spectral amplitudes for Rayleigh and Love waves computed for frequencies between 0.01 and 0.06 Hz. The amplitudes are obtained from a database of computed excitation coefficients in a layered, spherical Earth, using interpolated normal-mode frequencies to represent the fundamental-mode surface wave branch. The event-based product generates static plots for every known earthquake with magnitude Mw > 6.0 based on the source mechanisms reported by the Global CMT Project. For hypothetical earthquakes, dynamic radiation pattern plots are based on arbitrary source mechanisms chosen by the user. The database of the interactive product contains the spectral amplitudes of Green's functions computed for a particular hypocenter at virtual stations located around it. These spectral amplitudes are convolved with the components of a moment tensor to display the associated surface-wave radiation pattern. The dynamic plots are displayed in a web browser and updated without delay for every change in frequency of the seismic waves and in source mechanism except its depth. Since the spectral amplitudes of surface waves do not depend linearly on the source depth, a new data set is loaded from IRIS' servers for each change in source depth. The radiation pattern product supports researchers in estimating azimuthal variations in earthquake-generated ground motion amplitudes as well as in identifying nodes and anti-nodes of surface wave amplitudes to understand signal-to-noise ratios at seismic stations in the context of moment tensor inversions, wave propagation studies, and seismic tomography. The ground motion predicted by the surface-wave radiation pattern product is reflected in the ShakeMaps of shallow large earthquakes provided by the USGS. This is not surprising for the modeled areas of the ShakeMaps, but the semblance holds for the areas that are based on observations. The purpose of the surface-wave radiation pattern is to provide a reference for all researchers studying surface waves from a specific or hypothetical seismic event.
AGU Fall Meeting- Boris Rösler, Suzan van der Lee & Kevin Chao: Influences on Surface-Wave Induced Dynamic Stresses on Arbitrary Faults in a Layered Earth (Talk)
Propagating Rayleigh and Love waves from large earthquakes generate considerable dynamic stresses in the Earth's crust, even in the far field. These stresses occasionally trigger local seismicity on faults that are presumed to be critically stressed. Both volumetric and deviatoric stress variations in these dynamic stress tensors can affect the normal stress on a given fault, while shear stress variations on the fault are a result from deviatoric stresses from both types of surface waves, which often act on the fault at the same time. Reports of potentially dynamically triggered tremor or earthquakes on a given fault are frequently accompanied by estimates of dynamic stresses on this fault. However, these reports are often based on the peak stresses in a homogeneous Poisson solid. To complement these estimates, we calculate full dynamic stress tensors from 3-component surface-wave displacement fields for a realistic Earth model and project these stresses onto faults with known orientation. We quantify and visualize spatiotemporal variations in the shear and normal stresses calculated from simulated data for the Mw 8.2 earthquake near the coast of Oaxaca in Mexico in 2017 and the Denali earthquake in Alaska in 2002 and compare them to the locations of triggered seismicity recordings on the San Andreas fault and the Wasatch fault. We observe that 1) stresses generated at mid- and lower-crustal depths, where triggered events occur, are generally larger than those at the surface, 2) the orientation of the fault relative to the epicenter of the earthquake plays a major role in locally increasing Coulomb stresses, 3) triggering coincides with high Coulomb stresses over time which are dominated by the shear stresses and 4) combined stresses from simultaneous Rayleigh and Love waves create complex dynamic stress histories which follow the surface-wave radiation patterns.
- Vivian Tang, Boris Rösler, Jordan Nelson, JaCoya Thompson, Alice Lucas, Kevin Chao, Zhigang Peng, Michelle Paulsen, Laura Trouille & Suzan van der Lee: 5000 Ears and Eyes Detect and Classify Triggered Seismic Events in Alaska (Poster)
We are engaging citizen scientists in an experiment to test if many human ears can replace the process of a professional seismologist in identifying dynamically triggered seismic events. This is particularly important for data sets of seismograms that exploded in size during this century's big-data revolution, in particular through Earthscope. In this citizen seismology project we ask citizens to listen to relevant sections of seismograms that are accelerated to audible frequencies. This approach has three advantages: 1) The human ear naturally performs a time-frequency analysis and is capable of discerning a wide range of different signals, 2) Many human ears listening to the same data provides statistics that rank seismograms in order of their likelihood to contain a recording of a triggered event, which is helpful to researchers' analysis of this data, and 3) the volunteers' responses can be compared to the responses of a machine-learning algorithm to assess its performance. We offer two different interfaces: One audio only and one where the audio signal is accompanied by a visual graph of the time series. The events we are asking citizens to help identify are seismic events that are much smaller than those of which time series are typically analyzed, and include small earthquakes as well as tectonic tremor (series of overlapping deep low-frequency earthquakes). While much progress has been made in understanding how these events might be triggered by strong surface waves from large earthquakes far away, there is no consensus on its physical mechanism. The aim of our project is to receive the help of citizens to increase general knowledge of when seismic events occur and not occur. A better understanding of triggered seismic events is expected to provide important clues towards a fundamental understanding of how earthquakes might interact. Currently, over 2,000 volunteers have made about 58,000 classifications. We have analyzed the subjects that have been classified by the maximum number (10) of volunteers. Nearly 860 subjects are classified as earthquakes, 116 subjects are tremor, 1093 subjects are noise and 115 subjects are none of the above. By comparison to expert's classifications, we determined that volunteers achieve a higher reliability in detecting earthquakes and noise than tremor, as is the case for machines.
2018AGU Fall Meeting
- Boris Rösler, Suzan van der Lee & Kevin Chao: Surface-Wave Induced Dynamic Stresses on Arbitrary Faults in a Layered Earth (Talk)
Propagating surface waves from large earthquakes can generate considerable far-field dynamic stresses in the Earth's crust. These stresses can exceed triggering thresholds on critically stressed faults. Reports of observations of dynamically triggered seismic activity often point out a connection with high dynamic stresses from passing Rayleigh and/or Love waves. However, stresses related to radially propagating transversely polarized Love waves and elliptically polarized Rayleigh waves are not the sole contributors to dynamic stresses. Other stresses are related to the shape of the surface waves' fundamental-mode eigenfunctions as a function of depth and to changes in amplitude and sign of radiated surface waves with azimuth. From synthetic fundamental-mode surface wave displacement seismograms calculated for a one-dimensional Earth model, we obtain the strain tensor at virtual stations located at equal azimuth at varying radial distances from the epicenter. We compute the dynamic stress tensor, as a function of time, from the strain tensors using elastic constants from the same Earth model. We use the principal stress as a measure for the dynamic stress generated during passage of surface waves and compare the orientation of the principal stress axes with the orientation of far-field faults. We apply our approach to recent earthquakes such as the September 8, 2017 earthquake near the coast of Chiapas in Mexico with a moment magnitude of Mw = 8.2. Our results indicate that the stress generated by the passage of surface waves exceeds values for which earthquake can be triggered. The horizontal distribution of the principal stress follows the radiation pattern of Rayleigh and Love waves. Local earthquakes registered along the San Andreas Fault in California are assumed to have been triggered by Love waves due to the alignment of the fault along the direction of their maximum stress radiation. The stress generated by both surface waves has a maximum at depth reaching similar values as at the surface. This interesting result may favor the occurrence of triggered events in regions where the stress caused by the passage of surface waves derived from seismic stations located at the surface of the Earth is estimated as too low for triggering earthquakes.
- Vivian Tang, Boris Rösler, Jordan Nelson, JaCoya Thompson, Alice Lucas, Kevin Chao, Zhigang Peng, Michelle Paulsen, Laura Trouille & Suzan van der Lee: Earthquake Detective: Engaging Citizens in the Detection of Dynamically Triggered Seismic Events (Poster)
We are engaging citizen scientists in an experiment to test if many human ears can replace the process of a professional seismologist in identifying dynamically triggered seismic events. This is particularly important for data sets of seismograms that exploded in size during this century's big-data revolution, in particular through Earthscope. In this citizen seismology project we ask citizens to listen to relevant sections of seismograms that are accelerated to audible frequencies. This approach has three advantages: 1) The human ear naturally performs a time-frequency analysis and is capable of discerning a wide range of different signals, 2) Many human ears listening to the same data provides statistics that rank seismograms in order of their likelihood to contain a recording of a triggered event, which is helpful to researchers' analysis of this data, and 3) the volunteers' responses can be compared to the responses of a machine-learning algorithm to assess its performance. We offer two different interfaces: One audio only and one where the audio signal is accompanied by a visual graph of the time series. The events we are asking citizens to help identify are seismic events that are much smaller than those of which time series are typically analyzed, and include small earthquakes as well as tectonic tremor (series of overlapping deep low-frequency earthquakes). While much progress has been made in understanding how these events might be triggered by strong surface waves from large earthquakes far away, there is no consensus on its physical mechanism. The aim of our project is to receive the help of citizens to increase general knowledge of when seismic events occur and not occur. A better understanding of triggered seismic events is expected to provide important clues towards a fundamental understanding of how earthquakes might interact.