Technical Guide¶
The USGS Slab Models project uses earthquake hypocenter data (among other constraining datasets) to model the geometries of active subducting slabs worldwide. By refining the Preliminary Determined Epicenters (PDE) catalog to a set of interface and intraslab events, a 3-dimensional slab model is generated via interpolation of those points onto a 3-dimensional grid. To further constrain the slab geometry, active-source seismic data interpretations, receiver functions, and seismic tomography models are incorporated where available.
Slab1.0, the predecessor to Slab2, constructed models for about 80% of subduction zones using a combination of global earthquake and moment tensor data and some active source data. Slab1.0 provided slab surfaces and contour files at a resolution of 20km. The 2-D to 3-D interpolation method used by Slab1.0 caused limitations on the accuracy of certain slabs and the ability to model others, hence Slab2 was created to overcome these hurdels. In addition to an increase in active-source data used, Slab2 also allows for input data in the form of seismic tomography, reciever functions, and many more to further constrain the slab surface. Since the publication of the 2018 Slab2 models (Hayes et al., 2018), the Slab2 database has been updated with new earthquake hypocenters and any new suite of Slab models genreated fall within the USGS Slab Models project.
Node Depth Constraint¶
The slab modeling code works by constructing a grid of nodes within a set search region, then searching for data points within 2.5 degrees around each node. If data exist in the search zone, a reference model is used to constrain the edges of the slab model. Next, a search ellipse is constructed around the node, with the long axis oriented along the strike of the reference model in that location, to more evenly focus the data selection. The ellipse is then extruded by a set thickness value to constrain data within a certain depth of the slab. Active-source (AS), Receiver Functions (RF), Control Points (CP), and Tomography (TO) data are all constrained by different individual search ellipsoids. For exact search ellipsoid dimensions, see Supplementary Materials for Slab2: A Comprehensive Subduction Zone Geometry Model. All data that fall within the search ellipse are then summed (weighted by uncertainty) and put into a probability density function to determine the best slab depth at that node. This depth value is representitive of the slab center location. Each node is assigned a depth to create the 3-D slab model.
Node Shifting¶
To account for the difference between interslab and intraslab events, a tapered shift function is applied, which shifts nodes from the slab center to the slab surface. The age of the oceanic lithosphere at the trench is used to calculate the slab thickness. A fraction of the slab thickness is used to determine the shift magnitude. This shift percentage is determined by the average PDF width vs slab thickness from the previous model iteration. The shift magnitude varies as a function of slab thickness below the seismogenic zone. The result yeilds a grid of nodes which have been shifted to represent the slab surface geometry.
Tomography Data¶
Below a certain threshold depth, teleseismic tomography models can be used to constrain the slab center location. The tomography section can convert tomographic velocity models to a grid of points representative of the slab center by identifying high velocity anomalies within a certain distance of a slab model guide. Due to the relatively large uncertainties inherent in tomography, this data has low influence on the resulting slab model geometry.
Surface Smoothing¶
Once slab depth values are assigned to all grid nodes, a continuous surface is interpolated between them using a least squares spline interpolation, specifically scipy.interpolate.LSQSphereBivariateSpline. The surface is smoothed by a smoothing parameter, determined by the apex of an L-Curve relating input node depth and smoothed surface depth. All surfaces are then reshaped by a clipping mask for the specific slab model (found in src/polygons/slab_polygons.txt).
Overturned Slabs¶
As overturned slabs cannot be defined by a 2-D array of nodes, nodes are instead defined vertically and search horizontally for overturned slab regions. The slab model is then defined in a vertical reference frame, which is then split by the shallowest vertical node. All points in the upper section are defined as a grid of depth values, just like all other slab models, while the lower grouping of nodes is defined by a point clound of coordinates (3-column ASCII file with lon, lat, depth).
Overturned slab models include the following:
izu ker man sol