1:45 PM - 2:00 PM
[SCG52-13] Large subsidence in NW Colombia as a result of slow strain accumulation over an EQ cycle under a viscoelastic earth
Keywords:Viscoelastic earthquake cycle model, Subsidence, Slow and flat subduction, Caribbean NW Colombia, Seismic potential
Inversion of 3D GPS data in northwestern Colombia show a locked region that was interpreted as a locus for a future M8.0 earthquake/tsunami. The elastic coupling model reproduces horizontal SE velocities, but fail to explain the large subsidence rate presented by the GPS sites, both in their signs and magnitudes (Lizarazo et al., 2021).
To explain the vertical component, we applied a viscoelastic earthquake cycle model following Sagiya (2015) to evaluate the response of a lithosphere-asthenosphere system (2 layers) due to a cyclic rupture every 600 years in a very slow subduction setting (7 mm/yr). We used a grid search algorithm to optimize the dimensions and location of the source fault and we tested the dependency on the estimated rates to the assumptions on the viscosity value (η), recurrence interval (T) and elastic thickness (Te). Then, we perform the analysis with a multilayered model composed additionally by a thick sedimentary layer (12 km) and thick oceanic crust (16 km). The best source fault selection was based on a 3D RMSE criteria.
Results show that estimated rates for the last part of the earthquake cycle are independent of η and T, as long as it is much longer that the asthenosphere relaxation time. These two parameters only affect the coseismic offset size as well as the postseismic transient and its duration. Nevertheless, estimated velocities are sensitive to Te. A 200x150 km2 fault located at a depth of 14 km fully embedded in the thickest elastic lithosphere, reproduces nearly the same amount of horizontal deformation than the fault representative of the locked region (135x110 km2) under an elastic earth structure. At the same time, fitting of vertical motions is improved. By comparing 2-layered and multilayered viscoelastic earth structures, it can be observed that the sedimentary layer and thick oceanic crust contribution amounts only a 3.5% on the overall tridimensional deformation and therefore, they can be neglected.
Vertical motion during an earthquake cycle presents a coseismic uplift of ~0.7 m followed by a fast uplift of additional ~0.25 m during the first 50 years of the cycle for the coastal GPS sites closer to the locked region. This latter effect, contributes to increase interseismic subsidence rate that variates from 0.4 to 2 mm/yr depending on location. Since all the coseismic and postseismic deformation is fully recovered, there is no creation of marine terraces due to earthquake activity and it could be attributed to steady plate motion.
The present model independent of its simplicity, is able to explain that strain built over an earthquake cycle of 600 years can generate significant vertical deformation independent of the slow relative motion and geometry of slab in the dipping direction (Caribbean of Colombia: flat slab). This demonstrates that large subsidence observed in normal subduction zones (e.g. Northeast Japan) is not dependent on the subduction velocity but the recurrence interval to accumulate strain to release in future M>=8 earthquakes.
Further improvements can be introduced by considering the geometry of the oceanic plate in strike and dip directions that were fixed for the present analysis assuming average values.
To explain the vertical component, we applied a viscoelastic earthquake cycle model following Sagiya (2015) to evaluate the response of a lithosphere-asthenosphere system (2 layers) due to a cyclic rupture every 600 years in a very slow subduction setting (7 mm/yr). We used a grid search algorithm to optimize the dimensions and location of the source fault and we tested the dependency on the estimated rates to the assumptions on the viscosity value (η), recurrence interval (T) and elastic thickness (Te). Then, we perform the analysis with a multilayered model composed additionally by a thick sedimentary layer (12 km) and thick oceanic crust (16 km). The best source fault selection was based on a 3D RMSE criteria.
Results show that estimated rates for the last part of the earthquake cycle are independent of η and T, as long as it is much longer that the asthenosphere relaxation time. These two parameters only affect the coseismic offset size as well as the postseismic transient and its duration. Nevertheless, estimated velocities are sensitive to Te. A 200x150 km2 fault located at a depth of 14 km fully embedded in the thickest elastic lithosphere, reproduces nearly the same amount of horizontal deformation than the fault representative of the locked region (135x110 km2) under an elastic earth structure. At the same time, fitting of vertical motions is improved. By comparing 2-layered and multilayered viscoelastic earth structures, it can be observed that the sedimentary layer and thick oceanic crust contribution amounts only a 3.5% on the overall tridimensional deformation and therefore, they can be neglected.
Vertical motion during an earthquake cycle presents a coseismic uplift of ~0.7 m followed by a fast uplift of additional ~0.25 m during the first 50 years of the cycle for the coastal GPS sites closer to the locked region. This latter effect, contributes to increase interseismic subsidence rate that variates from 0.4 to 2 mm/yr depending on location. Since all the coseismic and postseismic deformation is fully recovered, there is no creation of marine terraces due to earthquake activity and it could be attributed to steady plate motion.
The present model independent of its simplicity, is able to explain that strain built over an earthquake cycle of 600 years can generate significant vertical deformation independent of the slow relative motion and geometry of slab in the dipping direction (Caribbean of Colombia: flat slab). This demonstrates that large subsidence observed in normal subduction zones (e.g. Northeast Japan) is not dependent on the subduction velocity but the recurrence interval to accumulate strain to release in future M>=8 earthquakes.
Further improvements can be introduced by considering the geometry of the oceanic plate in strike and dip directions that were fixed for the present analysis assuming average values.