Periodic Reporting for period 2 - MILESTONE (Microseismicity Illuminates Subduction Zone Processes)
Reporting period: 2022-12-01 to 2024-05-31
The vast majority of catastrophic large earthquakes occur in subduction zones, and their timing and size is controlled by properties of the plate interface (coupling, fluid content). Processes in the downgoing slab of oceanic lithosphere (dehydration, stress distribution) are thought to also influence the plate interface. Both plate interface and slab are seismically active, and the constant background of 10s of thousands of microearthquakes every year contains abundant information on their state. Project MILESTONE is aimed at extracting this information by using state-of-the-art machine learning based tools to compile huge catalogs of microearthquakes from a set of subduction zones around the globe.
Microearthquakes on the plate interface appear to cluster around the edges of highly coupled regions, so-called asperities. This means that they can help us to map out the extents of possible future earthquakes, and different forms of seismicity (repeating earthquakes, swarms) can be used to infer the partitioning of energy between aseismic slip and stress buildup. We aim to move this direction of research forward by analyzing much larger set ofs events, collected at different plate interfaces, together with geodetic information. Microearthquakes inside the downgoing slab of oceanic lithosphere occur as a consequence of dehydration reactions along the path to higher pressure and temperature conditions. Locations and mechanisms of intraslab earthquakes can constrain slab geometries, temperature distribution, fluid budget and the evolution of the stress state in the oceanic lithosphere.
Results from MILESTONE have the potential to substantially improve earthquake hazard estimates along subduction zones. The obtained estimates of highly coupled regions can be used to generate shaking scenarios as well as tsunami simulations, which in turn can inform decision makers for concrete measures of disaster preparedness.
The overall objectives of the project are:
1) Understanding the relation between background microseismicity, interplate locking, megathrust earthquake ruptures and other plate interface processes. The use of "deep" seismicity catalogs of large subduction zone segments allows the delineation of stress accumulation at the downdip edges as well as along-strike separators of highly coupled regions on the megathrust (asperities).
2) Better defining what controls along-strike heterogeneity of interplate locking, and what physical processes are involved in its evolution. This can be achieved by looking for signatures of aseismic creep in the background seismicity (repeating earthquakes, swarms) and using these together with GNSS recordings to estimate the partitioning between seismic and aseismic motion along the megathrust as well as its temporal evolution.
3) Constraining the controls on style and abundance of intraslab seismicity and its variability along one subduction zone as well as between subduction zones.
4) Better understanding what mineral reactions may be involved in DSZ seismicity, and what the physical process of seismogenesis in the presence of fluids may be.
5) Better constraining the possible link between intraslab and megathrust seismic activity
Microearthquakes on the plate interface appear to cluster around the edges of highly coupled regions, so-called asperities. This means that they can help us to map out the extents of possible future earthquakes, and different forms of seismicity (repeating earthquakes, swarms) can be used to infer the partitioning of energy between aseismic slip and stress buildup. We aim to move this direction of research forward by analyzing much larger set ofs events, collected at different plate interfaces, together with geodetic information. Microearthquakes inside the downgoing slab of oceanic lithosphere occur as a consequence of dehydration reactions along the path to higher pressure and temperature conditions. Locations and mechanisms of intraslab earthquakes can constrain slab geometries, temperature distribution, fluid budget and the evolution of the stress state in the oceanic lithosphere.
Results from MILESTONE have the potential to substantially improve earthquake hazard estimates along subduction zones. The obtained estimates of highly coupled regions can be used to generate shaking scenarios as well as tsunami simulations, which in turn can inform decision makers for concrete measures of disaster preparedness.
The overall objectives of the project are:
1) Understanding the relation between background microseismicity, interplate locking, megathrust earthquake ruptures and other plate interface processes. The use of "deep" seismicity catalogs of large subduction zone segments allows the delineation of stress accumulation at the downdip edges as well as along-strike separators of highly coupled regions on the megathrust (asperities).
2) Better defining what controls along-strike heterogeneity of interplate locking, and what physical processes are involved in its evolution. This can be achieved by looking for signatures of aseismic creep in the background seismicity (repeating earthquakes, swarms) and using these together with GNSS recordings to estimate the partitioning between seismic and aseismic motion along the megathrust as well as its temporal evolution.
3) Constraining the controls on style and abundance of intraslab seismicity and its variability along one subduction zone as well as between subduction zones.
4) Better understanding what mineral reactions may be involved in DSZ seismicity, and what the physical process of seismogenesis in the presence of fluids may be.
5) Better constraining the possible link between intraslab and megathrust seismic activity
Progress for each of the five different work packages (WPs), as defined in the Description of the Action:
WP1 - Earthquake Catalog Creation
This WP was the main focus of project activities so far. We performed a large-scale testing and calibration exercise for different deep-learning based picking algorithms by handpicking a large benchmark dataset from Northern Chile and comparing the performance of the automatic picking algorithms against this benchmark. Moreover, we set up a testing environment for a suite of five different machine-learning based phase association algorithms, based largely on synthetic event data, and performed a benchmark study comparing their performance. As the latter is a new approach and of considerable interest to the community, we are currently preparing a publication on this benchmark study. Combining the best-performing picker with the best-performing associator with parameter settings that were found to be optimal, we obtained a functional workflow for assembling large high-quality earthquake catalogs from raw waveform data. We decided to first concentrate on data from the Chilean margin, and have compiled a preliminary long-term catalog for Northern Chile, and a shorter-term one for the entire margin. Although we still work on minor improvements, these catalogs can be published towards the end of 2024. Work on data from Costa Rica and Alaska has also started, which means that WP1 is somewhat behind schedule, but not dramatically so. Most of the six project publications so far resulted from previous or preliminary work related to this WP applied to the Chilean margin, often in combination with other observations obtained by collaborators.
WP2 - Tomographic Inversion
Work on WP2 started in late 2022; so far we have obtained a preliminary tomographic model for the Northern Chile region based on an early version of the regional seismicity catalog. We currently refine this model and optimize parameter choices, and also use the deep-learning based pickers from WP1 to obtain additional picks from available temporary networks. To help interpret our tomographic images, we already started to compile and modify petrophysical and thermal models (planned for WP5) together with our Chilean collaborators.
WP3 - Statistics and Imaging
The compilation of receiver function results has not started yet, but we undertook some methodological work on b-value analysis of a preliminary seismicity catalog of Central Chile. The developed technique, largely developed by collaborating scientists, provides a more rigorous approach for identifying areas or time intervals of robustly distinct b-value using a Bayesian approach.
Work on WPs 4 and 5 has not started yet
WP1 - Earthquake Catalog Creation
This WP was the main focus of project activities so far. We performed a large-scale testing and calibration exercise for different deep-learning based picking algorithms by handpicking a large benchmark dataset from Northern Chile and comparing the performance of the automatic picking algorithms against this benchmark. Moreover, we set up a testing environment for a suite of five different machine-learning based phase association algorithms, based largely on synthetic event data, and performed a benchmark study comparing their performance. As the latter is a new approach and of considerable interest to the community, we are currently preparing a publication on this benchmark study. Combining the best-performing picker with the best-performing associator with parameter settings that were found to be optimal, we obtained a functional workflow for assembling large high-quality earthquake catalogs from raw waveform data. We decided to first concentrate on data from the Chilean margin, and have compiled a preliminary long-term catalog for Northern Chile, and a shorter-term one for the entire margin. Although we still work on minor improvements, these catalogs can be published towards the end of 2024. Work on data from Costa Rica and Alaska has also started, which means that WP1 is somewhat behind schedule, but not dramatically so. Most of the six project publications so far resulted from previous or preliminary work related to this WP applied to the Chilean margin, often in combination with other observations obtained by collaborators.
WP2 - Tomographic Inversion
Work on WP2 started in late 2022; so far we have obtained a preliminary tomographic model for the Northern Chile region based on an early version of the regional seismicity catalog. We currently refine this model and optimize parameter choices, and also use the deep-learning based pickers from WP1 to obtain additional picks from available temporary networks. To help interpret our tomographic images, we already started to compile and modify petrophysical and thermal models (planned for WP5) together with our Chilean collaborators.
WP3 - Statistics and Imaging
The compilation of receiver function results has not started yet, but we undertook some methodological work on b-value analysis of a preliminary seismicity catalog of Central Chile. The developed technique, largely developed by collaborating scientists, provides a more rigorous approach for identifying areas or time intervals of robustly distinct b-value using a Bayesian approach.
Work on WPs 4 and 5 has not started yet
Expected results until the end of the project:
1) A collection of "deep" seismicity catalogs for the Chile, Alaska and Costa Rica subduction zones, with substantially lower completeness magnitudes and more accurate location estimates than previously existing ones. These catalogs form the base for all further results and enable the characterization of background seismicity in the different constituent parts of each subduction zone, changes along strike, dip or against time of single subduction zones as well as a comparison of activity between different subduction zones.
2) A framework of jointly analyzing or inverting GNSS data with microseismicity information to derive more highly resolved constraints on plate interface coupling, especially in the offshore region.
3) 3D models of seismic wavespeeds from selected settings in the three subduction zones, determined using local earthquake tomography based on subsets of the previously described seismicity catalogs. These velocity models will be used to constrain regional variations in (de)hydration of the downgoing plate, which in turn can control arc volcanism and the amount of aseismic deformation on the plate interface
4) A set of improved slab and plate interface geometries, as well as models of the thermal structure of the different subduction zones.
5) Novel models of plate interface coupling distribution based on the framework described in Point 2. These will serve as valuable inputs for hazard estimation and tsunami simulation efforts.
1) A collection of "deep" seismicity catalogs for the Chile, Alaska and Costa Rica subduction zones, with substantially lower completeness magnitudes and more accurate location estimates than previously existing ones. These catalogs form the base for all further results and enable the characterization of background seismicity in the different constituent parts of each subduction zone, changes along strike, dip or against time of single subduction zones as well as a comparison of activity between different subduction zones.
2) A framework of jointly analyzing or inverting GNSS data with microseismicity information to derive more highly resolved constraints on plate interface coupling, especially in the offshore region.
3) 3D models of seismic wavespeeds from selected settings in the three subduction zones, determined using local earthquake tomography based on subsets of the previously described seismicity catalogs. These velocity models will be used to constrain regional variations in (de)hydration of the downgoing plate, which in turn can control arc volcanism and the amount of aseismic deformation on the plate interface
4) A set of improved slab and plate interface geometries, as well as models of the thermal structure of the different subduction zones.
5) Novel models of plate interface coupling distribution based on the framework described in Point 2. These will serve as valuable inputs for hazard estimation and tsunami simulation efforts.