Search

Numerical models of subduction commonly use diffusion and dislocation creep laws from laboratory deformation experiments to determine the rheology of the lithosphere. The specific implementation of these laws varies from study to study, and the impacts of this variation on model behavior have not been thoroughly explored. We run simplified 2D numerical models of free subduction in SULEC, with viscoplastic slabs following (1) a diffusion creep law, (2) a dislocation creep law, and (3) both simultaneously, as well as several variations of model 3 with reduced resistance to bending. We compare the results of these models to a model with a constant-viscosity slab to determine the impact of the implementation of different lithospheric flow laws on subduction dynamics. In creep-governed models, higher subduction velocity causes a longer effective slab length, increasing slab pull and asthenospheric drag, which, in turn, affect subduction velocity. Numerical and analogue models implementing constant-viscosity slabs lack this feedback but still capture morphological patterns observed in more complex models. Dislocation creep is the primary deformation mechanism throughout the subducting lithosphere in our models. However, both diffusion creep and dislocation creep predict very high viscosities in the cold core of the slab. At the trench, the effective viscosity is lowered by plastic failure, rendering effective slab thickness the primary control on bending resistance and subduction velocity. However, at depth, plastic failure is not active, and the viscosity cap is reached in significant portions of the slab. The resulting high slab stiffness causes the subducting plate to curl under itself at the mantle transition zone, affecting patterns in subduction velocity, slab dip, and trench migration over time. Peierls creep and localized grain size reduction likely limit the stress and viscosity in the cores of real slabs. Numerical models implementing only power-law creep and neglecting Peierls creep are likely to overestimate the stiffness of subducting lithosphere, which may impact model results in a variety of respects. [ABSTRACT FROM AUTHOR]

An accurate assessment of the physical states of the Earth system is an essential component of many scientific, societal, and economical considerations. These assessments are becoming an increasingly challenging computational task since we aim to resolve models with high resolutions in space and time, to consider complex coupled partial differential equations, and to estimate uncertainties, which often requires many realizations. Machine learning methods are becoming a very popular method for the construction of surrogate models to address these computational issues. However, they also face major challenges in producing explainable, scalable, interpretable, and robust models. In this paper, we evaluate the perspectives of geoscience applications of physics-based machine learning, which combines physics-based and data-driven methods to overcome the limitations of each approach taken alone. Through three designated examples (from the fields of geothermal energy, geodynamics, and hydrology), we show that the non-intrusive reduced-basis method as a physics-based machine learning approach is able to produce highly precise surrogate models that are explainable, scalable, interpretable, and robust. [ABSTRACT FROM AUTHOR]

Numerical models of subduction commonly use diffusion and dislocation creep laws from laboratory deformation experiments to determine the rheology of the lithosphere. The specific implementation of these laws varies from study to study, and the impacts of this variation on model behavior have not been thoroughly explored. We run simple 2D numerical models of free subduction in SULEC, with viscoplastic slabs following: 1) a diffusion creep law, 2) a dislocation creep law, and 3) both in parallel. We compare the results of these models to a model with a constant-viscosity slab to determine the relative importance of each creep mechanism in subducting lithosphere and the impact of the implementation of different lithospheric flow laws on subduction dynamics. We find that dislocation creep dominates diffusion creep throughout subducting lithosphere with moderate (5 mm) grain size in the upper mantle. However, both diffusion and dislocation creep predict very high viscosities in the cold core of the slab. The resulting high slab stiffness causes the subducting plate to curl under itself at the mantle transition zone, affecting patterns in subduction velocity, slab dip, and trench migration over time. Peierls creep and localized grain size reduction likely limit the stress and viscosity in the cores of real slabs. Numerical models implementing only power-law creep and neglecting Peierls creep are likely to overestimate the stiffness of subducting lithosphere, which may impact model results in a variety of respects. Our models also demonstrate a feedback between effective slab length and subduction velocity. Analogue and numerical models with constant-viscosity slabs lack this feedback, but still capture the qualitative patterns observed in more complex models. [ABSTRACT FROM AUTHOR]

An accurate assessment of the physical states of the Earth system is an essential component of many scientific, societal and economical considerations. These assessments are becoming an increasingly challenging computational task since we aim to resolve models with high resolutions in space and time, to consider complex coupled partial differential equations, and to estimate uncertainties, which often requires many realizations. Machine learning methods are becoming a very popular method for the construction of surrogate models to address these computational issues. However, they also face major challenges in producing explainable, scalable, interpretable and robust models. In this manuscript, we evaluate the perspectives of geoscience applications of physics-based machine learning, which combines physics-based and data-driven methods to overcome the limitations of each approach taken alone. Through three designated examples (from the fields of geothermal energy, geodynamics, and hydrology), we show that the non-intrusive reduced basis method as a physics-based machine learning approach is able to produce highly precise surrogate models that are explainable, scalable, interpretable, and robust. [ABSTRACT FROM AUTHOR]

The presence of pre-existing rheological heterogeneities in the lithosphere plays a significant role during subsequent stages of deformation in essentially every geological process. Extensional basins located in foreland fold-and-thrust belts will alter the spatio-temporal evolution of its associated orogen. It remains unclear how far can horizontal stresses act and reactivate extensional structures, and it is difficult to interpret from the geological record how stresses were transferred across a heterogeneous crust. Here we examine the inversion of extensional basins in foreland fold-and-thrust belts by using three-dimensional analogue experiments that simulate first an extensional stage, followed by a shortening cycle. Our results show how extensional basins proximal to the orogenic front effectively localise deformation in the shape of thrusts and prevent stress transfer beyond their location. Basins that are located at large distances from the indenter also show evidence of mild inversion at early stages, but characterised only by basin infill contraction and uplift. When multiple extensional basins are present, although the degree and type of inversion will depend on their relative location and distance to the orogenic front, here were prove that the presence of additional extensional features in the vicinity of a basin can be a first order controlling factor in their reactivation history. We share additional insights of how a fold-and20 thrust belt evolves once the extensional basins have been incorporated by the advancing wedge and we hope that comparisons with natural examples will shed light on some still unanswered questions related to the process of basin inversion in orogenic belts. [ABSTRACT FROM AUTHOR]

[1] We investigate the deformation of orogenic wedges that form in the early stages of continent-continent collisions using a two-dimensional numerical model limited to the upper lithosphere. Our models show that deformation at the plate margins is influenced by rheology, surface processes, and the balance between inward mass flux and outward subduction flux, as controlled by the subduction load (which represents the effects of slab pull and resistive forces) and flexural downbending. We find three characteristic deformation modes: (1) near-pure subduction with little or no accretion; (2) frontal accretion with development of an accretionary wedge built up by offscraping of the sediment layer at shallow depth; and (3) independent frontal and basal accretion where a retrothrust allows stacking of basement nappes at crustal to mantle depths. Nearpure subduction is enabled for 'ordinary-rheology' materials, characterized by brittle and viscous material behavior (approximating a 'Christmas tree-type' depth profile), and almost zero friction along the subduction shear zone. Frontal accretion occurs when slightly increased friction along the subduction shear zone allows offscraping of the sediment layer from the subducting plate. Independent frontal and basal accretion develops in strong-rheology models with an almost fully brittle material behavior. Major surface erosion or a reduction of the subduction load promote the development of large basement nappes. Frontal accretion is favored by major sedimentation during convergence, a large backstop, and in the case of a lateral transition from a 'strong-rheology' to an 'ordinary-rheology' subducting plate. Our numerical models develop first-order characteristics as observed in natural orogenic wedges, for example upper crustal nappe stacks, frontal and basal accretion, or extension in the core of an orogen. Frontal and basal accretion are interdependent, and tend to stabilize the subduction system. Citation: Seizer, C., S. J. H. Buiter, and O. A. Pfiffner (2008), Numerical modeling of frontal and basal accretion at collisional margins, Tectonics, 27, TC3001, doi:10.1029/ 2007TC002169.

Continental rifts can form when and where continents are stretched. If the driving forces can overcome lithospheric strength, a rift valley forms. Rifts are characterised by faults, sedimentary basins, earthquakes and/or volcanism. With the right set of weakening feedbacks, a rift can evolve to break a continent into conjugate rifted margins such as those found along the Atlantic and Indian Oceans. When, however, strengthening processes overtake weakening, rifting can stall and leave a failed rift, such as the North Sea or the West African Rift. A clear definition of continental break-up is still lacking because the transition from continent to ocean can be complex, with tilted continental blocks and regions of exhumed lithospheric mantle. Rifts and rifted margins not only shape the face of our planet, they also have a clear societal impact, through hazards caused by earthquakes, volcanism, landslides and CO2 release, and through their resources, such as fertile land, hydrocarbons, minerals and geothermal potential. This societal relevance makes an understanding of the many unknown aspects of rift processes as critical as ever. [ABSTRACT FROM AUTHOR]