Using Full-Waveform Inversion to recover sub-surface physical properties at the Hikurangi margin
File(s)
Author(s)
Gray, Melissa
Type
Thesis
Abstract
The discovery of new slip modes at subduction zones since 2000 raises some interesting questions about how subduction megathrust faults behave. The Hikurangi margin offshore North Island New Zealand is the perfect natural laboratory to study different types of seismic behaviour, in- cluding slow-slip events (SSEs), as the margin experiences some of the shallowest (<2kmbsf) well documented SSEs worldwide. Theories to explain why some margins slip aseismically rather than in large stick-slip events often invoke the presence of high fluid pressure, conditionally stable fric- tional conditions (perhaps related to variations in lithology) and/or the subduction of ocean floor relief. As a plate boundary fault that undergoes slow slip has not yet been drilled it is currently difficult to directly test these hypotheses. Seismic reflection images allow us to investigate key areas at subduction margins, including the incoming section, trench and accretionary wedge, but these techniques often require extensive pre-processing and give limited information on lithology and fluid presence as they principally tell us about impedance contrasts rather than an individ- ual physical property. Full-Waveform Inversion (FWI) is a relatively new technique which uses the wave equation to generate synthetic seismic data, and iteratively updates a starting model to improve the match between the synthetic and observed data, with the aim of producing high- resolution models of subsurface physical properties including P-wave velocity. This technique has become standard practice in the oil industry for imaging shallow hydrocarbon reservoirs in mostly sedimentary settings, and is being increasingly applied to academic datasets acquired across various geological settings including subduction zones.
The aims of this research are twofold. Firstly, to test whether full-waveform inversion is a plausible technique at the Hikurangi margin for improving models of key features, and secondly to deter- mine what these recovered physical properties imply about the margin, and whether this affects seismicity. In the first science chapter, 2D FWI is applied to a 2D marine streamer dataset col- lected across the Hikurangi margin to recover a high-resolution model of P-wave velocity. While a previous, conventional velocity model created using pre-stack depth migration produced an ad- equate starting model, including variable velocities in the water column, precisely located seafloor
depths, and smoothing of abrupt boundaries needed to be carried out before FWI modelling. FWI modelling showed great improvement in model resolution compared to previous velocity modelling across the margin, resolving features to fault-block scale. Low-velocity-zones along faults in the model are identified as potential fluid conduits, which may be pathways for fluid release from the subduction interface at depth. Part way through this project a new 3D seismic experiment was funded by NERC (NZ3D-FWI). The second science chapter describes performing a suite of 3D synthetic FWI tests to optimise the location of the onshore seismometer array, to ensure that 3D FWI had the potential to recover high-resolution images of the principal targets of the survey: the subduction interface, a proposed subducting seamount, and the source region of shallow SSEs at Hikurangi. By investigating three potential land receiver layouts, the synthetic testing concluded with the selection of the optimal land-receiver layout which was used in the experiment. In the final year of the PhD, the margin was drilled as part of International Ocean Discovery Program (IODP) Expedition 372 and 375. Physical property data were acquired from logging-while drilling (LWD), wireline logging, and direct measurements made on the core. The third and final science chapter of this project uses physical property measurements from the IODP expeditions and de- rives relationships between P-wave velocity and other physical properties. These relationships are then used to transform the 2D FWI P-wave velocity model into models of porosity, S-wave velocity, Vp/Vs, bulk modulus and shear modulus. These models are used to examine whether sediments are over- or under- compacted in the incoming section and wedge, provide estimates of porosity within fault zones and elastic moduli within the accretionary wedge.
The model reveals the shallow structure of the overriding plate down to depths of ∼2km below the sea floor, including the fault plumbing system above the the origination zone of the SSEs. Hanging walls within the accretionary wedge appear more compacted than footwalls, and low velocity and corresponding low compaction and high porosity within the fault zones suggest they are acting as fluid pathways. Based on drilling data, the incoming section is composed of clastics, pelagics and volcaniclastics, and the FWI model is used to track these lithologies away from the drill site towards the trench. Sediment compaction estimates suggest that the different lithologies experience lateral compression that starts at ∼5.3-8km seaward of the trench. Pelagics which have been incorporated into the wedge appear to be under-compacted, likely due to their low permeability, while clastics appear over-compacted, suggesting that they are well drained. Porosity estimates indicate that ∼33% of fluid has been expelled from the clastic sediments and ∼20% from the pelagic layer across the P ̄apaku frontal thrust. In the derived porosity model, 45% of initial fluid is released in the 5km section seaward of the wedge, and this rises to ∼62% at 25km landward from the trench. Shear modulii estimates suggest a low shear modulus of (1-4GPa) within the wedge, which is an important input parameter for tsunami modelling in the region, which has previously been modelled with much higher (7GPa) rigidity. New data for 3D FWI modelling at Hikurangi has now been collected, which will allow the work in this thesis to be extended both deeper and in 3D, in order to further investigate the physical properties at Hikurangi and their role in seismicity at the margin. Overall, the research in this thesis highlights that FWI is a promising method for producing high-resolution models at subduction zones, and using newly acquired data specifically for FWI, these models have the potential to produce a high-resolution characterisation of the SSE source zone below Hikurangi.
The aims of this research are twofold. Firstly, to test whether full-waveform inversion is a plausible technique at the Hikurangi margin for improving models of key features, and secondly to deter- mine what these recovered physical properties imply about the margin, and whether this affects seismicity. In the first science chapter, 2D FWI is applied to a 2D marine streamer dataset col- lected across the Hikurangi margin to recover a high-resolution model of P-wave velocity. While a previous, conventional velocity model created using pre-stack depth migration produced an ad- equate starting model, including variable velocities in the water column, precisely located seafloor
depths, and smoothing of abrupt boundaries needed to be carried out before FWI modelling. FWI modelling showed great improvement in model resolution compared to previous velocity modelling across the margin, resolving features to fault-block scale. Low-velocity-zones along faults in the model are identified as potential fluid conduits, which may be pathways for fluid release from the subduction interface at depth. Part way through this project a new 3D seismic experiment was funded by NERC (NZ3D-FWI). The second science chapter describes performing a suite of 3D synthetic FWI tests to optimise the location of the onshore seismometer array, to ensure that 3D FWI had the potential to recover high-resolution images of the principal targets of the survey: the subduction interface, a proposed subducting seamount, and the source region of shallow SSEs at Hikurangi. By investigating three potential land receiver layouts, the synthetic testing concluded with the selection of the optimal land-receiver layout which was used in the experiment. In the final year of the PhD, the margin was drilled as part of International Ocean Discovery Program (IODP) Expedition 372 and 375. Physical property data were acquired from logging-while drilling (LWD), wireline logging, and direct measurements made on the core. The third and final science chapter of this project uses physical property measurements from the IODP expeditions and de- rives relationships between P-wave velocity and other physical properties. These relationships are then used to transform the 2D FWI P-wave velocity model into models of porosity, S-wave velocity, Vp/Vs, bulk modulus and shear modulus. These models are used to examine whether sediments are over- or under- compacted in the incoming section and wedge, provide estimates of porosity within fault zones and elastic moduli within the accretionary wedge.
The model reveals the shallow structure of the overriding plate down to depths of ∼2km below the sea floor, including the fault plumbing system above the the origination zone of the SSEs. Hanging walls within the accretionary wedge appear more compacted than footwalls, and low velocity and corresponding low compaction and high porosity within the fault zones suggest they are acting as fluid pathways. Based on drilling data, the incoming section is composed of clastics, pelagics and volcaniclastics, and the FWI model is used to track these lithologies away from the drill site towards the trench. Sediment compaction estimates suggest that the different lithologies experience lateral compression that starts at ∼5.3-8km seaward of the trench. Pelagics which have been incorporated into the wedge appear to be under-compacted, likely due to their low permeability, while clastics appear over-compacted, suggesting that they are well drained. Porosity estimates indicate that ∼33% of fluid has been expelled from the clastic sediments and ∼20% from the pelagic layer across the P ̄apaku frontal thrust. In the derived porosity model, 45% of initial fluid is released in the 5km section seaward of the wedge, and this rises to ∼62% at 25km landward from the trench. Shear modulii estimates suggest a low shear modulus of (1-4GPa) within the wedge, which is an important input parameter for tsunami modelling in the region, which has previously been modelled with much higher (7GPa) rigidity. New data for 3D FWI modelling at Hikurangi has now been collected, which will allow the work in this thesis to be extended both deeper and in 3D, in order to further investigate the physical properties at Hikurangi and their role in seismicity at the margin. Overall, the research in this thesis highlights that FWI is a promising method for producing high-resolution models at subduction zones, and using newly acquired data specifically for FWI, these models have the potential to produce a high-resolution characterisation of the SSE source zone below Hikurangi.
Version
Open Access
Date Issued
2019-10
Date Awarded
2020-03
Copyright Statement
Creative Commons Attribution NonCommercial NoDerivatives Licence
Advisor
Bell, Rebecca
Morgan, Joanna
Publisher Department
Earth Science & Engineering
Publisher Institution
Imperial College London
Qualification Level
Doctoral
Qualification Name
Doctor of Philosophy (PhD)