Periodic Reporting for period 4 - GEO-4D (Geodetic data assimilation: Forecasting Deformation with InSAR)
Berichtszeitraum: 2022-07-01 bis 2023-12-31
Earthquakes are one of the most destructive natural hazard on Earth and, despite tremendous progress on both fundamental aspects of the physics of earthquakes and on hazard mitigation, earthquakes remain unpredictable and pose a significant threat for a large part of the world population. One way to advance toward a more resilient society against earthquakes is to grow our understanding of tectonic faulting in order to better characterize the associated hazard.
In the past 20 years, tremendous efforts have been made in the development of ground deformation measurements in tectonically active regions, including ground- and space-based measurements such as GNSS or satellite imagery. Plate tectonics stress active faults by bending the crust until faults fail abruptly during an earthquake, radiating devastating seismic waves. However, recent observations actually revealed that faults can slip slowly, without radiating seismic waves, gently releasing stress. In parallel to these observations, our understanding of the mechanics of faulting made giant leaps toward more and more realistic models of loading and release of stress by slip on faults. We now know slow slip and earthquakes interact and both participate in the release of stress along faults, but the interplay between these and the underlying mechanisms controlling the mode of slip along faults are not understood.
In short, what are the physical mechanisms controlling whether a fault will generate slow harmless slip or a devastating earthquake? And more importantly, what are the most important data we should collect in order to grow our understanding and to improve the predictive ability of our models?
In Geo4D we have built tools to feed measurements of ground deformation into physics-based models of faulting, toward a data assimilation approach of ground deformation and earthquakes models. Similarly to what meteorologists do everyday, the question is whether we can build physics-based models that will be trained by incoming data. After an intense phase of data collection in the field in Turkiye and of processing of global data base of satellite imagery, we have produced a wealth of observations of slow slip on continental settings in various places, including in the Himalayas, Pakistan, Turkiye and the Philippines. Our data base has allowed us to describe the finest possible scales of fault slip along several faults in the world and to test data assimilation methods to see whether we can, or not, forecast slow slip. This task is a never ending one and while the objectives of the Geo4D project were met, we aim to pursue this endeavor.
In the past 20 years, tremendous efforts have been made in the development of ground deformation measurements in tectonically active regions, including ground- and space-based measurements such as GNSS or satellite imagery. Plate tectonics stress active faults by bending the crust until faults fail abruptly during an earthquake, radiating devastating seismic waves. However, recent observations actually revealed that faults can slip slowly, without radiating seismic waves, gently releasing stress. In parallel to these observations, our understanding of the mechanics of faulting made giant leaps toward more and more realistic models of loading and release of stress by slip on faults. We now know slow slip and earthquakes interact and both participate in the release of stress along faults, but the interplay between these and the underlying mechanisms controlling the mode of slip along faults are not understood.
In short, what are the physical mechanisms controlling whether a fault will generate slow harmless slip or a devastating earthquake? And more importantly, what are the most important data we should collect in order to grow our understanding and to improve the predictive ability of our models?
In Geo4D we have built tools to feed measurements of ground deformation into physics-based models of faulting, toward a data assimilation approach of ground deformation and earthquakes models. Similarly to what meteorologists do everyday, the question is whether we can build physics-based models that will be trained by incoming data. After an intense phase of data collection in the field in Turkiye and of processing of global data base of satellite imagery, we have produced a wealth of observations of slow slip on continental settings in various places, including in the Himalayas, Pakistan, Turkiye and the Philippines. Our data base has allowed us to describe the finest possible scales of fault slip along several faults in the world and to test data assimilation methods to see whether we can, or not, forecast slow slip. This task is a never ending one and while the objectives of the Geo4D project were met, we aim to pursue this endeavor.
We have initiated our efforts by first establishing a data processing strategy for InSAR images. Over the last decade, satellite constellations have been built and data now flow continuously. We have organized innovative strategies to process these data. In particular, our workflow now also includes AI tools we have developed to detect the finest scales of slow slip. Our processing strategy is operational and will be able to ingest data from upcoming missions.
In parallel, and in order to test our methodology, we have built a GNSS network to monitor ground deformation in Turkey, a region where we know earthquakes and slow slip events happen and can be measured with great fidelity. We are now operating a 19 sites network which allowed us to characterize with great details the spatio-temporal behavior of slow slip along the North Anatolian Fault. In addition, we have established ties with other groups working on this part of the world and we are jointly operating a network of 15 seismic stations and 2 creepmeters, making this fault section one of the densely instrumented in Europe. Our observations allow to dig deep in the mechanics of earthquakes and faulting.
We have explored the occurrences of slow slip events along multiple regions of the world, including Chile, Turkiye, Pakistan or California. Our findings suggests that these events occur globally and provide new hints on the mechanics of faulting. We have preliminary works showing great promises to include such behavior in data assimilation procedures to test whether such events can be included in forecast. We have also shown that slow slip events globally are related to metamorphic dehydration reactions in subduction zones.
In parallel, and in order to test our methodology, we have built a GNSS network to monitor ground deformation in Turkey, a region where we know earthquakes and slow slip events happen and can be measured with great fidelity. We are now operating a 19 sites network which allowed us to characterize with great details the spatio-temporal behavior of slow slip along the North Anatolian Fault. In addition, we have established ties with other groups working on this part of the world and we are jointly operating a network of 15 seismic stations and 2 creepmeters, making this fault section one of the densely instrumented in Europe. Our observations allow to dig deep in the mechanics of earthquakes and faulting.
We have explored the occurrences of slow slip events along multiple regions of the world, including Chile, Turkiye, Pakistan or California. Our findings suggests that these events occur globally and provide new hints on the mechanics of faulting. We have preliminary works showing great promises to include such behavior in data assimilation procedures to test whether such events can be included in forecast. We have also shown that slow slip events globally are related to metamorphic dehydration reactions in subduction zones.
We made significant progress in the following directions:
1. Geodetic data analysis: By exploring the continuous flow of data, we realized we needed additional tools in order to fully analyse these. We therefore have decided to move in the direction of machine learning and have made progress for the detection and quantification of ground deformation. Our methodology is ow able to detect millimeter slip events in noisy InSAR time series.
2. Systematic production of maps of surface displacements: We have mapped ground displacement over various actively deforming regions of the world and have significantly advanced our understanding of tectonic faulting in these regions. More exploration to come…
3. InSAR technical improvements: We have produced significant improvements in the processing of InSAR data.
4. Data Assimilation: After some preliminary tests conducted in the frame of the Geo4D project, we continue to test and explore the capabilities of data assimilation for fault slip forecasting.
1. Geodetic data analysis: By exploring the continuous flow of data, we realized we needed additional tools in order to fully analyse these. We therefore have decided to move in the direction of machine learning and have made progress for the detection and quantification of ground deformation. Our methodology is ow able to detect millimeter slip events in noisy InSAR time series.
2. Systematic production of maps of surface displacements: We have mapped ground displacement over various actively deforming regions of the world and have significantly advanced our understanding of tectonic faulting in these regions. More exploration to come…
3. InSAR technical improvements: We have produced significant improvements in the processing of InSAR data.
4. Data Assimilation: After some preliminary tests conducted in the frame of the Geo4D project, we continue to test and explore the capabilities of data assimilation for fault slip forecasting.