Fault output

Fault output#

Introduction#

There are two primary methods for visualizing on-fault rupture dynamics:

  1. Evaluation of rupture characteristics at specific on-fault locations via ASCII receiver files

  2. Visualization across the entire fault surface with files that can be opened with ParaView

While threads or nodes can be allocated to write the ParaView output (see Asynchronous output), it is typically not required.

DynamicRupture Namelist OutputPointType Configuration#

The output type generated is determined by the OutputPointType variable in the DynamicRupture namelist. Here is an example of how to configure it:

&DynamicRupture
OutputPointType = 4
/

OutputPointType determines the type of output:

0 : No output
3 : ASCII fault receivers
4 : ParaView file
5 : Both ASCII fault receivers and ParaView file

Configuring ParaView output#

You can adjust the ParaView output using the Elementwise namelist. Here is an example of how to do this:

&Elementwise
printtimeinterval_sec = 1.0
OutputMask = 1 1 1 1 1 1 1 1 1 1 1 1 !described herafter
refinement_strategy = 1 ! or 2
refinement = 1
vtkorder = -1
timeseries = 'snapshot'
/

printTimeInterval_Sec#

  • Output is generated every printtimeinterval_sec.

refinement#

The refinement variable determines how the output mesh is created:

  • refinement = 0 outputs a single triangle for each mesh cell. The unknowns are calculated at the center of each cell.

  • refinement = 1 subdivides each triangle into 3 or 4 subtriangles, depending on the refinement_strategy. A higher refinement value will further subdivide each subtriangle.

  • refinement_strategy = 1 divides each triangle into 3 triangles, all sharing the triangle barycenter as a node.

  • refinement_strategy = 2 divides each triangle into 4 triangles.

OutputMask#

The OutputMask variable allows you to select specific unknowns for visualization. You can toggle the output writing of each unknown by changing its corresponding bit in the OutputMask array.

Here’s what each bit in the OutputMask array represents:

  1. SRs and SRd: Slip rates in along-strike and along-dip directions

  2. T_s, T_d: Shear stress in strike and dip directions, P_n: Normal stress

  3. u_n: Fault normal velocity (Note: SeisSol does not allow for fault opening (mode I))

  4. Mud: Current effective friction coefficient, StV: State variable for RS friction

  5. Ts0,Td0,Pn0: Total shear and normal stresses, including initial stresses

  6. Sls and Sld: Fault slip in along-strike and -dip directions

  7. Vr: Rupture velocity, computed from the spatial derivatives of the rupture time

  8. ASl: Accumulated slip

  9. PSR: Peak slip rate

  10. RT: Rupture time, the time at which the slip rate exceeds 1 mm/s, interpolated linearly between the time points of the friction solve. A point that slips from the start of the simulation on has a rupture time of 0.

  11. DS: Dynamic stress time. With LSW, the time at which ASl>D_c, interpolated linearly over the interval over which a sub-step integrates the slip. With RS, the time at which mu <= mu_w + 0.05 (f0 - mu_w), interpolated like the rupture time, but not before it. DS can be used to evaluate the process zone size.

  12. P_f and Tmp: Only with thermal pressurization, pore pressure and temperature

Initial fault tractions#

It is worth noticing that Ts0, Td0, and Pn0 outputted at t=0s are the initial tractions after a first step through the friction routines. In particular, if the shear traction read in the input files locally exceeds fault strength, the outputted traction at t=0s is reduced compared with the one read in the input files. To visualize the initial shear tractions (or any other parameters, e.g. d_c) given in the easi file, the script evaluate_ini_model_parameters_easi.py can be used. It depends on the easi python bindings.

./read_ini_fault_parameter.py output/data-fault.xdmf fault.yaml --ref_vector " -0.1,0,-1.0"

seissolxdmf python module#

You can read SeisSol ParaView files (XDMF/Hdf5 or XDMF/binary files, describing the fault outputs and the free-surface outputs and the volume wavefield outputs) using our Python module seissolxdmf. Find it on PyPi at: seissolxdmf.

Ascii fault receivers#

To generate ASCII receiver files, configure the Pickpoint namelist as in this example:

&Pickpoint
printtimeinterval = 1
OutputMask = 1 1 1 1 1 1 1 1 1 1 1 1 !described herafter
nOutpoints = 24
PPFileName = 'fault_receivers.dat'
format = 'csv'
/

format chooses between one text file per receiver ('csv', the default) and a single table holding all of them ('hdf5'), see HDF5 fault receivers below.

printtimeinterval determines how frequently the output is generated — every printtimeinterval (local) time step. Please note that using this output with local time-stepping may result in differently sampled receiver files.

The text files#

Every receiver has a file of its own, -faultreceiver-00001.dat and so on, numbered by its line in PPFileName. Its header names the columns and states where the receiver sits and the stress the fault starts out under there, rotated into the strike and dip directions:

TITLE = "Temporal Signal for fault receiver number 1"
VARIABLES = "Time" ,"SRs" ,"SRd" ,"T_s" ,"T_d" ,"P_n" ,...
# x1                    ...
# x2                    ...
# x3                    ...
# face-global-id        ...
# plus-cell-global-id   ...
# plus-face-side        ...
# minus-cell-global-id  ...
# minus-face-side       ...
# P_0                   ...
# T_s                   ...
# T_d                   ...

followed by a row per sample. T_s, T_d and P_n are the change of the tractions; adding T_s, T_d and P_0 of the header gives the total.

With fused simulations, a sample takes a row per simulation, and the second column, SimulationIndex, says which one, counted from zero. The lines P_0, T_s and T_d of the header then hold a value per simulation, in the same order. SeisSol used to write a column per quantity and simulation instead (SRs-1, SRs-2, …); postprocessing/science/widen_fused_receivers.py converts the files to that layout for tools that expect it.

OutputMask#

This is the same as for the ParaView output.

HDF5 fault receivers#

Setting format = 'hdf5' in the Pickpoint section writes all receivers into a single file, -faultreceivers.h5, instead of one text file each.

A row of that table is one receiver of one simulation, as a sample of a fused run is a row per simulation in the text files, and what their header states about a receiver becomes a column beside the table. The layout is the one of the off-fault receivers, described under Off fault receivers, with these columns:

/faultreceivers/Index            (receivers, 2)   group and row within that group
/faultreceivers/ReceiverId       (receivers,)     the receiver's number
/faultreceivers/SimulationIndex  (receivers,)     which fused simulation the row holds
/faultreceivers/FaceId           (receivers,)     the fault face, globally
/faultreceivers/PlusCellId       (receivers,)     the cell on the plus side
/faultreceivers/PlusFaceSide     (receivers,)     and its local face number
/faultreceivers/MinusCellId      (receivers,)     the cell on the minus side
/faultreceivers/MinusFaceSide    (receivers,)     and its local face number
/faultreceivers/Coordinates      (receivers, 3)   where the receiver sits
/faultreceivers/InitialStress    (receivers, 3)   P_0, T_s and T_d of the text header

The quantities of a sample are Time followed by the components the OutputMask turned on.

With local time stepping, the clusters a rank holds do not all cache the same number of samples between two writes, while the table has one sample axis for all of them. The count is therefore agreed across the ranks, and the receivers that cached fewer samples leave the rest of their column at a quiet NaN – which a reader can drop, unlike a zero it could not tell from a measurement.

samplechunk in the same section sets the storage chunking along the sample axis, as receiversamplechunk does for the off-fault receivers.

High-Order VTKHDF Output#

The high-order elementwise output can be enabled by setting vtkorder in the elementwise section to a positive value, corresponding to the order of the output polynomial per cell.

File groupings#

timeseries in the elementwise section overrides outputtimeseries for the fault output, so that it can be written as one file for the whole run while the other outputs stay one file per step, or the other way round. It takes effect only with a vtkorder set; see The IO Infrastructure for what the groupings are.

Cell fields#

Beside the quantities, every cell of the elementwise output carries:

fault-tag

the tag the mesh gave this fault face (see the fault tagging page). Several faces carry the same tag, and a mesh that tags nothing leaves it at its default.

global-id

an identifier of the face, unique across the mesh

partition

the rank that computed the cell

Contents