The seismic and petrological signature of magma migration beneath the Reykjanes Peninsula, Iceland
Lead Research Organisation:
University of Cambridge
Department Name: Earth Sciences
Abstract
The Reykjanes Peninsula is located in southwest Iceland, and if one includes the nearby capital city of Reykjavik, hosts more than 60% of the population of the entire country. From a geological perspective, the peninsula features a segment of the Mid-Atlantic Ridge system that forms the boundary between the North American and Eurasian tectonic plates. It is delineated by a series of segmented volcanic zones featuring crust that is heavily fractured and contains fissures associated with historic eruptive episodes. Intriguingly, evidence points to these eruptions occurring periodically with an interval of approximately 800 years. Dating of lava from the most recent eruption places it in the 13th century.
In late 2019, elevated earthquake activity along the Reykjanes Peninsula was detected, possibly associated with magma intrusion deep in the crust. Over the subsequent 12 months, this activity repeatedly waxed and waned, and included several sizable earthquakes, including a magnitude 5.7 that shook buildings in Reykjavik in February this year. In the last several month, this activity has gradually escalated, with thousands of earthquakes being detected every day, and was accompanied by the intrusion of an ~10 km long dike at a depth of 1.0-1.5 km. On 19th March 2021, this activity culminated in an eruption near Fagradalsfjall, located at the southern end of the dike, the first on the entire peninsula in ~800 years. This relatively small but spectacular eruption is ongoing (for instance, two new volcanic fissures opened up approximately 700m from the original eruption site on April 5th), and is accompanied by earthquake activity and volcanic tremor (continuous shaking associated with the passage of melt in the crust).
The goal of this project is to deploy an array of seismic stations in the region of elevated activity on the Reykjanes Peninsula. This will allow us to record ground motion associated with earthquakes that rupture on tectonic faults, which form part of the plate boundary, and magma migration in the crust. We will also use powerful microscopes to map out the compositional variation in samples of newly-erupted lava. The composition of the magma and the crystals tell us about the depths that the magma was stored before it moved towards the surface and erupted. Maps of individual crystals can be used as "geological stop-watches" to time how long magma took to rise before eruption and our new methods to do this will be applied to the ongoing eruption.
By combining evidence for magma movement from the earthquakes and lava, we can build models of how the magmatic system works and inform more effective future volcano monitoring strategies in Iceland and elsewhere. Such advances may help us to understand the periodicity of the volcanic activity, the potential for future eruptions and large earthquakes that may pose a hazard to surrounding population centres, and the nature of the interaction between the plate boundary system and the volcanic system, the latter fueled by elevated levels of melt (magma) in the lower crust and mantle beneath.
In late 2019, elevated earthquake activity along the Reykjanes Peninsula was detected, possibly associated with magma intrusion deep in the crust. Over the subsequent 12 months, this activity repeatedly waxed and waned, and included several sizable earthquakes, including a magnitude 5.7 that shook buildings in Reykjavik in February this year. In the last several month, this activity has gradually escalated, with thousands of earthquakes being detected every day, and was accompanied by the intrusion of an ~10 km long dike at a depth of 1.0-1.5 km. On 19th March 2021, this activity culminated in an eruption near Fagradalsfjall, located at the southern end of the dike, the first on the entire peninsula in ~800 years. This relatively small but spectacular eruption is ongoing (for instance, two new volcanic fissures opened up approximately 700m from the original eruption site on April 5th), and is accompanied by earthquake activity and volcanic tremor (continuous shaking associated with the passage of melt in the crust).
The goal of this project is to deploy an array of seismic stations in the region of elevated activity on the Reykjanes Peninsula. This will allow us to record ground motion associated with earthquakes that rupture on tectonic faults, which form part of the plate boundary, and magma migration in the crust. We will also use powerful microscopes to map out the compositional variation in samples of newly-erupted lava. The composition of the magma and the crystals tell us about the depths that the magma was stored before it moved towards the surface and erupted. Maps of individual crystals can be used as "geological stop-watches" to time how long magma took to rise before eruption and our new methods to do this will be applied to the ongoing eruption.
By combining evidence for magma movement from the earthquakes and lava, we can build models of how the magmatic system works and inform more effective future volcano monitoring strategies in Iceland and elsewhere. Such advances may help us to understand the periodicity of the volcanic activity, the potential for future eruptions and large earthquakes that may pose a hazard to surrounding population centres, and the nature of the interaction between the plate boundary system and the volcanic system, the latter fueled by elevated levels of melt (magma) in the lower crust and mantle beneath.
Publications
Greenfield T
(2022)
Deep long period seismicity preceding and during the 2021 Fagradalsfjall eruption, Iceland
in Bulletin of Volcanology
Kahl M
(2022)
Deep magma mobilization years before the 2021 CE Fagradalsfjall eruption, Iceland
in Geology
| Description | The main discoveries and achievements that have resulted from the work funded by the award are summarised as follows: (1) Using the seismic data recorded by our array, we have discovered several clusters of deep long period (DLP) earthquakes at 10-12 km depth that occurred during and after the March 2021 eruption of Fagradalsfjall on the Reykjanes Peninsula. Before the eruption, these events were located ~2 km northeast of the eruption site, whereas during the eruption, they occurred ~1 km to the southwest. This microseismicity (low magnitude earthquakes) is likely linked to magma migration associated with the deep volcanic plumbing system beneath Fagradalsfjall, but according to petrological modelling lies some 5 km shallower than the primary magma storage zone that feeds Fagradalsfjal. This suggests that the seismicity took place during the upward migration of melt, perhaps due to an encounter with a barrier to melt transportation, or exsolution of CO2 rich fluids. (2) Diffusion chronometry shows that there is a remarkable coincidence between the timing of shallow geophysical unrest (seismicity, geodetic) and overturn in the deep magmatic system - 85% of diffusion ages record near-Moho reorganisation of the magmatic system between the beginning of 2020 and the onset of eruption in March 2021. An acceleration in these processes also appears to coincide with the onset of deep seismicity in the lower crust. The remainder of the diffusion ages may record magmatic processes that are a key part of the build-up to eruption, but are not yet visible in the geophysical observations. (3) Microseismic detection in the upper 7 km of the crust (the brittle zone) has revealed the presence of over 125,000 earthquakes during the deployment period, with magnitudes ranging between -1 to 5.8. More than 70% of the earthquakes occur during four swarms where the earthquake rate exceeds 1000 earthquakes per day, and are likely related to the intrusion of a NE-SW oriented dyke. The large and prolonged swarm in February-March 2021 is due to the intrusion and propagation of a shallow dyke with associated triggering of seismicity in the surrounding regions due to the change in stress. The dyke first propagated north-east before heading south-west from its initiation point, eventually reaching 10 km in length. The eruption then occurred 2 km southwest of this initiation point. The observed seismicity dropped shortly before the eruption, presumably due to the release in stress as magma accumulated near the surface. (4) Shear wave splitting analysis using local earthquakes has been applied to analyse the influence of pre-existing structures and the contemporary stress field on seismic anisotropy in the crust. We find that anisotropy is confined to the top ~4 km of the crust with an average fast axis orientation of N024 E ± 33 degrees and average delay time of 0.10 ± 0.06s. The majority of measurements show good alignment between fast polarisation and the orientation of both mapped surface fractures and maximum horizontal stress. However, in regions with large N-S striking faults, the generation of anisotropy is controlled by damage zones around the faults. These results indicate a balance of both stress-induced and structural anisotropy mechanisms across the region. |
| Exploitation Route | The raw data collected during the course of the project, both seismic and petrological, will be made available to researchers after an initial embargo period. Both datasets capture a very unique train of events on the Reykjanes Peninsula that culminated in an eruptive sequence never before seen in SW Iceland. It is likely to be the focus of scientific study for the years and decades to come, thus making our datasets a valuable resource to those wishing to undertake further research in the area. Collaboration with a large proposal funded by Icelandic Council (RANNIS) to Eniko Bali and colleges at the University of Iceland to obtain further petrological insight to the precursory processes for the eruption. Our findings and our techniques are crucial for their future work. |
| Sectors | Education,Environment,Other |
| Description | Two of our stations in the Reykjanes array were set up to transmit data in real time to IMO (Iceland Meteorology Office), who are in charge or monitoring seismic and volcanic hazard in Iceland. This data contributed to their monitoring of the ongoing eruption and associated seismicity, which was only some 30 km away from Reykjavik and 10 km away from the international airport Keflavik. Such efforts are vital for forecasting how the hazard might develop, and the associated disaster management planning. Data from our stations have been used in student projects (e.g. final year/Masters project that are worth a substantial portion of their grade) that end up appearing in departmental blogs designed to reach the general public, and therefore have outreach/impact potential. To date, the data from our array has supported four of these projects, which is a substantial number. e.g. https://blog.esc.cam.ac.uk/masters-student-gets-detailed-snapshot-of-earthquake-tremors-in-reykjanes-peninsula-iceland/#more-2041 |
| First Year Of Impact | 2021 |
| Sector | Education,Environment,Other |
| Impact Types | Societal,Policy & public services |
| Title | Reykjanes Seismic Network |
| Description | This dataset consists of 3-component broadband seismic data recorded by an array of 19 instruments installed in the Reykjanes Peninsula by the University of Cambridge. |
| Type Of Material | Database/Collection of data |
| Year Produced | 2021 |
| Provided To Others? | Yes |
| Impact | This dataset has been used as the basis for a published paper, and has formed the basis for four P3 (masters) projects in the Department of Earth Sciences at the University of Cambridge. |
| URL | https://ds.iris.edu/mda/8F_2020/ |
| Description | Cambridge-CAS collaborative partnership |
| Organisation | Academy of Sciences of the Czech Republic |
| Country | Czech Republic |
| Sector | Academic/University |
| PI Contribution | We have a data sharing agreement with the Czech Academy of Sciences (CAS), since Cambridge and CAS installed complementary seismic arrays on the Reykjanes Peninsula to record seismicity associated with the eruption. We have shared our data with CAS for them to use in their analysis of earthquake source mechanisms and seismic tomography |
| Collaborator Contribution | CAS has supplied all of their data to us for use in our analysis of the microseismicity, including the deep events and earthquakes associated with the dike and eruption. We are also using their data for shear wave splitting analysis. The data from CAS approximately doubles the size of our dataset, and is extremely valuable for improving the project outcomes. |
| Impact | So far, we have a joint paper published in the Bulletin of Seismology based on the joint datasets, which looks at the deep crustal seismicity beneath the Reykjanes Peninsula. We also ran a joint workshop on the results of our collaboration in Cambridge in March 2023, in which 9 researchers from CAS attended |
| Start Year | 2021 |
| Description | Cambridge-Iceland collaboration |
| Organisation | University of Iceland |
| Country | Iceland |
| Sector | Academic/University |
| PI Contribution | We run multiple seismic networks in Iceland, including the one in the Reykjanes Peninsula that is the subject of this grant. We have developed a strong working relationship with the University of Iceland on this deployment, partly due to the Covid pandemic. We have supplied the equipment and a large portion of funds to enable the installation of this equipment, and played a major role in the fieldwork. We also take the lead in the data analysis and publication of results. |
| Collaborator Contribution | Our project partners undertook the initial core deployment due to COVID-19 restrictions. As a result of the emergency declared when the volcano erupted, they were able to liberate some government funding to pay for a portion of the installation and service costs. They also provide storage and technical assistance with the fieldwork, which is especially valuable. |
| Impact | Two papers from the collaborative research efforts funded by this project have been published, as indicated in the publications section of this submission. Additional papers are in preparation. The collaboration is multi-disciplinary, since it involves both seismology and petrology, which are highly complementary, since the eruptive products that are analysed are related to the seismicity, which tracks the ascent of melt through the crust. The collaboration also facilitated the streaming of data from two of our stations to IMO for the purposes of real time hazard analysis and forecasting. The two publications are listed below: M. Kahl, E.J.F. Mutch, J. Maclennan, D.J. Morgan, F. Couperthwaite, E. Bali, T. Thordarson, G.H. Guðfinnsson, R. Walshaw, I. Buisman, S. Buhre, Q.H.A. van der Meer, A. Caracciolo, E.W. Marshall, M.B. Rasmussen, C.R. Gallagher, W.M. Moreland, Á. Höskuldsson, R.A. Askew; Deep magma mobilization years before the 2021 CE Fagradalsfjall eruption, Iceland. Geology 2022;; 51 (2): 184-188. doi: https://doi.org/10.1130/G50340.1 Greenfield, T., Winder, T., Rawlinson, N. et al. Deep long period seismicity preceding and during the 2021 Fagradalsfjall eruption, Iceland. Bull Volcanol 84, 101 (2022). https://doi.org/10.1007/s00445-022-01603-2 |
| Start Year | 2021 |
| Title | QuakeMigrate v1.0.0 |
| Description | QuakeMigrate 1.0.0 Release Notes QuakeMigrate 1.0.0 is the culmination of almost 18 months of hard work, and is the first major, stable release of the package. The codebase has been largely re-written, delivering dramatic improvements in run-time and memory usage, many new features and far greater flexibility to the user. There have also been a large number of bug fixes relating to the stability and performance of the package. All users are encouraged to update to this release. We anticipate releasing v1.1.0 soon, with a significant expansion of tutorials and related documentation, and some further improvements to user-flexbility. (And maybe a couple more bug-fixes!) So please give us your feedback!! Thank you to everyone who has been part of the journey so far! With your help, we have tested the code thoroughly, but it is inevitable that we won't have covered every possible use-case and combination of parameters - so please continue to provide feedback. And please let us know when it works well, too! It is our favourite source of motivation. Highlights Cross-platform support: QuakeMigrate has now been successfully built and run on Linux, Mac and Windows! The core development of the code continues to be carried out largely on Linux, so additional feedback & testing from Windows and Mac users will continue to be appreciated. Documentation: We now have comprehensive documentation automatically built from the source code, plus installation instructions (for Linux, Mac and Windows) and a limited set of tutorials. More coming soon! Local magnitude calculation: support for built-in local magnitude calculation (and output of cut raw, real and Wood-Anderson simulated waveforms for each event). Single-phase support: QuakeMigrate now supports flexible multiple-phase stacking. For example, users processing Z-only nodal data may wish to only stack P-wave onset functions. The flexible implementation allows users to design onset functions for any phase of their choosing (with P and S the default for the time being). But if you want to design an onset function to identify and migrate/stack Vsh, Vsv etc. it is now easy to do! Publication-ready plots: significant improvements to the information content, clarity and aesthetics of automatic plots (including from the trigger, locate, local-magnitude calculation and auto-picking modules & sub-modules). Examples: Two new examples; another demonstration of using QuakeMigrate to detect and locate high-frequency basal icequakes (amongst crevassing noise!), plus detecting dike-induced earthquakes during the 2014 Bárðarbunga-Holuhraun dike intrusion - featuring an event rate of more than 150 earthquakes per hour (~ 4000 per day!). Tests: we have made our first steps on the journey to including a comprehensive suite of tests, to verify correct operation after installation. These are based on the included icequake_iceland and dike_intrusion example use-cases. More to come soon. Logging: flexible log output - currently split mainly between "info", "warning" and "debug" levels. We recommend sticking with the default - "debug" returns a lot of additional info!! Modularity: the picker and onset sub-modules are now built on an abstract base class, making it straightforward to build and share new alternatives that integrate with the core migration machinery of the package. License: QuakeMigrate 1.0.0 is released under the GPL v3 license. Users who have been using the development branch v1.0.0 serves as a citable release of the version of the code that has been available on the development branch since January 2020 - however note there have been a significant number of further enhancements and bug fixes over this period, particularly since May 2020. We intend to maintain a far more granular series of releases in future - thank you for your patience! Users upgrading from the 0.x.x series Users upgrading from the 0.x.x series will notice differences in the absolute value of the coalescence functions calculated, due to bug-fixes affecting the core migration and stacking functions. However, the relative values of the coalescence functions through time should remain largely unchanged, meaning trigger results and relative uncertainty estimates are still valid. Absolute uncertainty estimates, however, were previously being over-estimated, and trigger thresholds will need to be updated. To re-emphasise: there have been major bug-fixes and improvements reaching throughout the entire package. To benefit from these improvements we highly recommend re-processing the data through all stages (detect, trigger & locate), though we have tried to incorporate backwards-compatibility for all old output file formats wherever possible, for example to allow you to solely re-run locate(). Changelog This provides a limited summary of the major changes to the code since the previous release. For full details please see the linked pull requests, and commit messages therein. Change to top-level package name - QMigrate ? quakemigrate See #85 quakemigrate.core C libraries now build automatically when installing the package. C functions have been cleaned up and documented properly. See #21. quakemigrate.export - new! Add some export utilities to a new module (export). Currently includes functions to take the outputs of QuakeMigrate to: ObsPy catalogue; ObsPy catalogue to NLLoc OBS file; ObsPy catalogue to MFAST SAC input; QuakeMigrate to pyrocko Snuffler input. quakemigrate.io Refactor the I/O system entirely. This was in response to Issue #66. See 68c13f7 for full details. The re-write of the quakemigrate.io.data sub-module includes fixing bugs and making breaking changes related to the detrending and upsampling/decimation of waveform data. See c1ff447 and #103. Add support for reading response information (via a light wrapper of the obspy function) and functions for response removal & Wood-Anderson seismogram simulation, for use with the new quakemigrate.signal.local_mag module. Add a new, more comprehensive, transparent & flexible check_availability() function, to check what data is available given a set of provided quality criteria. Added functions for removing response and simulating WA response here, allowing for this functionality to be accessed from anywhere across the package. This includes the ability to output real and/or WA cut waveforms (as velocity or displacement seismograms), for example for spectral analysis. Response removal parameters are specified (along with the instrument response inventory) when creating the Archive object. The format of triggered event files, event files, pick files and station-availabiity files have been heavily overhauled. See #76. Fix to allow for numerical station names. quakemigate.lut - significant rewrite (see #54, #65) Significant changes to API - see examples. Users from the 0.x series will need to update their look-up table files (using an included utility function) or re-compute them. Fully-documented, including a tutorial in the documentation. Handling of a possible bug/ambiguity in the scipy RegularGridInterpolator API. The default for the fill_value parameter does not appear to be consistent with documentation. The traveltime lookup tables are now stored in a dictionary structure maps[station][phase] to enable migration of a flexible combination of seismic phases, and to make it possible to migrate onsets for a subset of the phases and stations contained in the LUT. See #75, #103. The 3-D grid on which the lookup tables are defined is now more intuitive to build. The user simply chooses the positions of a pair of opposite corners (the lower-left and upper-right) in geographic coordinates, the geographic and cartesian projections (using pyproj), and a node spacing along each axis. The number of grid nodes will be calculated automatically to span the volume of interest. User-specified units: the user must specify when making an LUT whether to use units of metres or kilometres; this will then be used consistently throughout the package. See #79. quakemigrate.plot - new! See #83. Extracted all base plotting methods to individual modules within the quakemigrate.plot module. No longer using a class to pass the information around. Revamp all of the figures produced by QuakeMigrate to include more useful information and to make better use of the available space. See 72b1c47 for details. quakemigrate.signal Refactor to be more flexible with the input data. QuakeMigrate now allows for single-component data to be used, or stacking to be performed on just one phase (e.g. just P or just S). The required changes reached quite deep into the package and have changed how Onset objects are created, but is ultimately very straightforward to use. See #103. Channel names can now be specified by the user, by default they are "Z" for vertical component, and "[N,1]" and ["E,2"] for horizontal components. Internally use obspy.Trace objects to store data up to the point of passing it to the C functions; this adds greater flexibility and more built-in methods for quality checking, filtering, re-sampling etc. than the previous framework using arrays. quakemigate.signal.onsets - significant re-write Extracted the embedded onset function generation from the core QuakeScan class to a new submodule, quakemigrate.signal.onsets. Various changes to parameter names - see examples. Created an Abstract Base Class - Onset. This class can be used as a base class for a class implementing an alternative onset function algorithm, while ensuring compliance with the embedded code in QuakeScan. Created a new class OnsetData to store the pre-processed waveforms used for onset calculation, the onset functions themselves, and all associated parameters and attributes. Significant expansion to the options available for data quality-checking; now exposed to the user, giving the flexibility to select which to use. The STA/LTA onset function remains the default. quakemigate.signal.pickers - significant re-write Extracted the embedded picking functions from the core QuakeScan class to a new module, quakemigrate.signal.pickers. General improvements to picking through a mix of fixes and features. Major improvement to clarity/style. Created an Abstract Base Class - PhasePicker. This class can be used as a base class for a class implementing an alternative phase picking algorithm, while ensuring compliance with the embedded code in QuakeMigrate. The user can provide a different onset function for phase picking using GaussianPicker than they used for migration in locate(). Bug fixes related to calculating the pick threshold; new option to use the median absolute deviation of the noise, rather than a percentile. See #116. Fitting a 1-D Gaussian to the phase onset function remains the default method of phase picking. quakemigate.signal.trigger Add dynamic trigger threshold method based on the median absolute deviation, a robust statistical estimator that is insensitive to extreme outliers. See #59. Added the ability to trigger events restricted to a specific region of the grid (specified as geographic coordinates, and illustrated on the trigger summary plot). Numerous bug-fixes related to the handling of overlapping triggers. Fix an indexing bug in trigger that caused the last event to be missed if it was a single sample in length. quakemigrate.signal.local_mag - new! See #71 A comprehensive suite of codes for local earthquake magnitude calculation by measuring displacement amplitudes from Wood-Anderson simulated seismograms. This can optionally be used as part of a locate run to automatically output ML estimates for each located event. General changes Update to various class attributes. Deprecation warnings + internal attribute re-mappings have been included to ease the transition. Add additional examples (Rutford icequakes + Iceland dike intrusion). This includes a script that performs data download from IRIS through an FDSN client. See #105. Option to locate events from a user-provided triggered event file, as well as the existing functionality to automatically read triggers between two timestamps from the default output directory structure. See #53. Re-write setup.py to automatically build the C extension library and link it to the rest of the package. Should be more robust on different platforms. Functions to read .dll or .so files (depending on operating system) have been added to automatically load the correct linked library. See #84. Add a system of module imports through __init__.py files to reduce the verbosity of input statements. New, intuitive directory structure for outputs, allowing straightforward batch processing of data archives. See #68. Option to run trigger() multiple times from the same detect() output by using different run sub-names. Also incorporated into the functions used by locate() to read triggered event files. Add readthedocs documentation with sphinx. Here we can host the documentation for the source code, instructions on installation, and tutorials on how to use the package. The internal reference frame has Z being positive down (the depth frame). Station elevations in the station input file are in the positive up (elevation/natural frame) and converted internally. More information on intermediate results is retained in the final output files. For example, the coalescence and normalised coalescence values for a triggered event, along with which value it was triggered against. Added tests based on the "Iceland Icequake" and "Volcanotectonic_Iceland" example use-cases to veryify correct operation ofter installation. Added Continuous Integration (Travis CI) allow us to catch breaking issues before new features/fixes are merged and, in particular, should help us keep on top of changes related to upstream changes to dependencies. License: QuakeMigrate v1.0.0 is released under the GPLv3 license. Optimisations Optimise compute by changing the flags passed to the the compiler. Bumped more of the compute() method to the C library to maximise efficiency. See a9ffdb6 Moved to logging with the native Python logging module, which allows for writing logs to file as well as stdout in a more concise manner. See #81 Thanks! Thank you to all those who have contributed to this release, whether by writing code, reporting bugs, or suggesting new features and improvements! This includes, but is not limited to (listed alphabetically): Pascal Audet Vivek Babu Conor Bacon Paul Derand GitHub user @fontiela Miriam Gauntlett Amy Gilligan Tim Greenfield Thomas Hudson Junior Kimata Ian Lee Andrew Owen Miriam Reiss Daniel Roberts Charlie Schoonman Nate Stevens Jonathan Smith GitHub user @bvicic GitHub user @SeisVincent Tom Winder |
| Type Of Technology | Software |
| Year Produced | 2021 |
| Open Source License? | Yes |
| Impact | This software is seeing rapid uptake by the seismology community around the world. |
| URL | https://zenodo.org/record/4442748 |