Difference between revisions of "HighF v14.12"

From SCECpedia
Jump to navigationJump to search
 
(72 intermediate revisions by 3 users not shown)
Line 1: Line 1:
== High-F 14.12 Verification and Validation Exercise ==
 
 
This page collects information about a proposed high frequency verification and validation exercise. Planning started in Dec 2014.
 
This page collects information about a proposed high frequency verification and validation exercise. Planning started in Dec 2014.
  
== Agreements Needed ==
+
== Goal ==
* Simulation Volumes
+
Most regional scale deterministic earthquake simulation verification and validation exercises have been done at frequencies 1Hz or lower [1] [2]. With the introduction of high performance CPU and CPU-GPU wave propagation codes, there has been significant recent progress running simulations up to 4Hz.  As we increase the capabilities of our deterministic earthquake simulation codes, we need to re-evaluate the codes performance at these higher frequencies that are now possible. The goal of this activity is to evaluate the performance of multiple wave propagation codes that are capable of simulation regional scale earthquakes at frequencies 4Hz+.
* Rotation Angle
+
 
* Projections used
+
== Approach ==
* Points per wavelength used
+
High frequency simulations introduce several code modifications intended to more accurately model the physics of high frequency earthquake wave propagation simulations. Modification to existing wave propagation codes to support 1Hz+ simulations include:
* Number and location of sites used for ground motion comparison
+
* Source descriptions with energy above 1Hz
* La Habra Point Source parameters
+
* Small scale heterogeneities in the 3D velocity models
* Currently available code capabilities in GPU codes
+
* Frequency dependent attenuation
** small scale heterogeneities
+
* Near fault plastic yielding
** FDQ
+
* Large displacement plastic yielding
** Plasticity
+
 
** dynamic rupture source
+
To help isolate the impact of each of these changes, we propose to start with simple (possibly unrealistic) high frequency (4Hz) ground motion simulations and compare the simulation results produced by different methods. Then we will introduce these modifications one at a time, as a way if isolating the impact of each change individually.
* V and V exercise stages - need to be reviiewed and approved
 
  
== Volume ==
+
== Problem Statement ==
 +
We have defined common input parameters for use by AWP-ODC, RWG, and Hercules. The problem statements includes the following parameters:
 +
#Simulation Region
 +
#Source Description
 +
#Station List
 +
#Velocity Model
 +
#Simulation Duration
 +
#Simulated Frequency Range
 +
#Sample per Wavelength
  
The four corners are:
+
== Simulation Region ==
*   -119.288842    34.120549
+
The simulation region four corners are given in the table below:
*  -118.354016    35.061096
+
*[[HighF v14.12#Simulation Box / Velocity Model]]
*  -116.846030    34.025873
 
*  -117.780976    33.096503
 
  
 
Volume size:
 
Volume size:
 
*180,000 m x 135,000 m x 61,875 m = 1.503e15 m3
 
*180,000 m x 135,000 m x 61,875 m = 1.503e15 m3
 +
 +
== Source Description ==
 +
We will use a point source, and a source time function for that point source produced by Zheqiang Shi (SDSU). This source time function has frequencies above 4Hz, scaled to Mw5.1 La Habra. The source is plotted and defined in the following files. These show the source trimmed, low-pass filtered (fc=5Hz, Butterworth, 2 forward passes, 4 poles), 4300 points, dt=0.001s (and plot of what is should look like).
 +
*[http://hypocenter.usc.edu/research/High-F/sv1_0291_5hz_trim.pdf Time domain Plot of Selected Source]
 +
*[http://hypocenter.usc.edu/research/High-F/spectra.pdf Source Frequency Domain Plot w/ and w/o Filtering]
 +
*[http://hypocenter.usc.edu/research/High-F/sv1_0291_5hz_trim.dat Data File For 5Hz Plot] (Source after filtering)
 +
 +
== 1D Velocity Model Definition ==
 +
The first velocity model will be modeled on a 1D model of the Los Angeles Basin used in on the [[Broadband Platform]] project. The model was smoothed to eliminate sharp contrasts between layers.
 +
*[http://hypocenter.usc.edu/research/High-F/bbp1d.pdf Diagram of smoothed 1D Model (pdf)]
 +
*[http://hypocenter.usc.edu/research/High-F/bbp1d_mid.pdf Diagram of smoothed 1D Model showing Layers (pdf)]
 +
*[http://hypocenter.usc.edu/research/High-F/bbp1d_int_500.rtf 1D model data file (20m resolution) (rtf)]
 +
 +
Definition of Los Angeles Basin 1D velocity models used in the Broadband Platform project:
 +
*[[Broadband_File_Format_Guide#Velocity_Model_File]]
 +
 +
== 3D Model Extraction Parameters ==
 +
 +
* 180km by 135km by 62km
 +
* 20m spacing
 +
* Points: 9000 by 6750 by 3100
 +
** Total: 188,325,000,000
 +
* Min Vs: 500m/s
 +
* Mesh file size: 2,259,900,000,000 => 2.1TB
 +
 +
== Station List ==
 +
We developed a list of ground motion observations for the La Habra earthquake. A description of the methods used to create this merged site list are provided at [[Selection of La Habra Ground Motion Observations]].
 +
 +
This site list is based on use of publicly posted data sets for the La Habra event. There should be on-scale accelerometer ground motion data at all of these sites for the La Habra mainshock. This list combine all channels available from SMO and SCECDC for the La Habra Event using only acceleration instruments within the simulation region:
 +
* Total SMO and SCECDC [HN] Stations in simulation region : 356
 +
**[http://hypocenter.usc.edu/research/High-F/merged_lahabra_sites.kml Merged (SMO + SCECDC) Site KML Map (KML File) ] 356 Accelerator Seismograms for La Habra Earthquake (in simulation region and less than 200km from epicenter))
 +
**[http://hypocenter.usc.edu/research/High-F/merged_lahabra_sites.txt  Merged (SMO + SCECDC) Site List (TXT File)] 356 Accelerator Seismograms for La Habra Earthquake (in simulation region and less than 200km from epicenter))
 +
 +
We expect all these recordings to be [ HN ] channel data, In SEED terminology,
 +
* [ H ] High Broad Band >= 80 to <= 250 Sample Rate (Hz) Corner Period: >= 10 sec
 +
* [ N ] Accelerometer (historically some channels from accelerometers have used instrumentation codes of L and G. The use of N is the FDSN convention as defined in August 2000)
 +
 +
 +
We have not yet plotted all data to visually confirm all records are above noise, and on-scale. We have not yet confirm all records are three component records.
 +
 +
[[Image:Merged smoscecdc labra sites.png|256px|thumb|right|Fig 1: 356 La Habra Earthquake ground motion accelerometer recordings for the La Habra Earthquake in Simulation Volume]]
  
 
== Simulation Box / Velocity Model ==
 
== Simulation Box / Velocity Model ==
Line 36: Line 81:
 
|-
 
|-
 
| Dimensions (km)
 
| Dimensions (km)
| 180 x 135 x 61.875
+
| 180 x 135 x 61.875 (31 km for 1d Model)
 
|  
 
|  
 
|-
 
|-
Line 42: Line 87:
 
|  -119.288842    34.120549, -118.354016    35.061096, -116.846030    34.025873, -117.780976    33.096503
 
|  -119.288842    34.120549, -118.354016    35.061096, -116.846030    34.025873, -117.780976    33.096503
 
|  
 
|  
 +
|-
 +
| Rotation Angle
 +
|  39.9
 
|-
 
|-
 
| UCVM Version
 
| UCVM Version
| 14.3.0
+
| Not used for first velocity model
 
| No heterogeneities
 
| No heterogeneities
 
|-
 
|-
 
| Velocity Model Versions
 
| Velocity Model Versions
| CVM-S4v26
+
| Simplified BBP 1D model defined above
 
|-  
 
|-  
 
|Miniumum Vs
 
|Miniumum Vs
Line 56: Line 104:
 
| Samples per wavelength
 
| Samples per wavelength
 
| 6 to 7 - AWP
 
| 6 to 7 - AWP
 +
| 10 to 12 - Hercules
 
|}
 
|}
  
Line 70: Line 119:
 
| La Habra
 
| La Habra
 
|  
 
|  
 +
|-
 +
| Mw
 +
| 5.1
 +
| Ref: http://www.scsn.org/2014lahabra.html
 +
|-
 +
| Moment
 +
| 5.764 E+23 Dyne-cm
 +
| Source: En-Jui
 
|-
 
|-
 
| Origin Time
 
| Origin Time
Line 79: Line 136:
 
| Source: En-Jui
 
| Source: En-Jui
 
|-
 
|-
| Strike/Dip/Rake 1
+
| Strike/Dip/Rake
| 134/55/155
 
| Source: En-Jui, initial
 
|-
 
| Strike/Dip/Rake 2
 
 
| 239/70/38
 
| 239/70/38
 
| Source: En-Jui, based on review of aftershock sequence
 
| Source: En-Jui, based on review of aftershock sequence
Line 91: Line 144:
 
|  
 
|  
 
|}
 
|}
 +
 +
USGS Origin Time: 2014-03-29 04:09:42.170 (UTC) [https://earthquake.usgs.gov/earthquakes/eventpage/ci15481673#executive USGS La Habra Page]
 +
 +
IRIS Origin Time: 2014-03-29 04:09:43.970 (UTC) [http://service.iris.edu/fdsnws/event/1/query?starttime=2014-03-28T00:00:00&endtime=2014-03-30T00:00:00&minmag=5&maxmag=5.2&orderby=time&format=xml&maxlat=34.366&minlon=-118.809&maxlon=-117.513&minlat=33.441&nodata=404 Iris La Habra Metadata]
 +
 +
=== Coordinate System ===
 +
 +
* X ~= 040 azimuth (positive from 0 to 135 km), units in m, m/s and m/s2
 +
* Y ~= 130 azimuth (positive from 0 to 180 km), units in m, m/s and m/s2
 +
* Z = UP (positive upward, toward the sky)
  
 
== Solver Parameters ==
 
== Solver Parameters ==
Line 104: Line 167:
 
|  
 
|  
 
|-
 
|-
| Simulation Length
+
| Simulation Length (Duration)
| 200 s
+
|100 s
 
|  
 
|  
 
|-
 
|-
 
| Delta T
 
| Delta T
| 0.005
+
| 0.001  - AWP
|  
+
| 0.005 - Hercules
 
|-
 
|-
 
| Plane Output Resolution
 
| Plane Output Resolution
Line 117: Line 180:
 
|-
 
|-
 
| I/O Print Rate
 
| I/O Print Rate
| every 10 steps
+
| every 25 steps - AWP
|  
+
| every 12 steps - Hercules
 
|-
 
|-
 
| Station List
 
| Station List
Line 136: Line 199:
 
* [http://hypocenter.usc.edu/research/High-F/High-F_Project_Plan_6Sept2012.pdf Original High-F Project Plan (pdf)]
 
* [http://hypocenter.usc.edu/research/High-F/High-F_Project_Plan_6Sept2012.pdf Original High-F Project Plan (pdf)]
 
* [http://hypocenter.usc.edu/research/High-F/High-F_Project_Plan_6Sept2012.docx Original High-F Project Plan (docx)]
 
* [http://hypocenter.usc.edu/research/High-F/High-F_Project_Plan_6Sept2012.docx Original High-F Project Plan (docx)]
 +
 +
== Agreements Needed ==
 +
* Simulation Volumes
 +
* Rotation Angle
 +
* Projections used
 +
* Points per wavelength used
 +
* Number and location of sites used for ground motion comparison
 +
* Proposed Station List to save results
 +
* La Habra Point Source parameters
 +
* Currently available code capabilities in GPU codes
 +
** small scale heterogeneities
 +
** FDQ
 +
** Plasticity
 +
** dynamic rupture source
 +
* V and V exercise stages - need to be reviiewed and approved
 +
 +
== Data Comparison ==
 +
 +
* [[HighF_v14.12_Data_Comparison]]
  
 
== Related Entries ==
 
== Related Entries ==
 
*[[High-F Project]]
 
*[[High-F Project]]
 
*[[CME Project]]
 
*[[CME Project]]
 +
*[[Seismtools]]
 
*[[AWP-ODC]]
 
*[[AWP-ODC]]
 
*[[Hercules]]
 
*[[Hercules]]
 +
 +
== References Used ==
 +
 +
* Withers, K.B., K.B. Olsen, and S.M. Day (2015). Memory efficient simulation of frequency dependent Q, Bull. Seis. Soc. Am., in press. (Vol 105).
 +
 +
* Lee, E.-J., Chen, P., & Jordan, T. H., 2014a. Testing waveform predictions of 3D velocity models against two recent los angeles earthquakes, Seismol. Res. Lett., 85(6), 1275–1284.
 +
 +
* Taborda, R. and J. Bielak (2013). Ground-motion simulation and validation of the 2008 Chino Hills, California, earthquake, Bull. Seismol. Soc. Am. 103, no. 1, 131–156, doi 10.1785/0120110325.
 +
 +
* Cui, Y., E. Poyraz, K.B. Olsen, J. Zhou, K. Withers, S. Callaghan, J. Larkin, C. Guest, D. Choi, A. Chourasia, Z. Shi, S.M. Day, J.P. Maechling, and T.H. Jordan (2013). Physics-based seismic hazard analysis on petascale heterogeneous supercomputers, Proc. Supercomputing Conference, 2013.
 +
 +
* Bielak, J., Karaoglu, H. and Taborda, R. (2011). Memory-Efficient Displacement-Based Internal Friction for Wave Propagation Simulation, Geophysics 76 (6): T131-T145.
 +
 +
* Taborda, R., J. López, H. Karaoglu, J. Urbanic, and J. Bielak (2010). Speeding up finite element wave propagation for large-scale earthquake simulations, Parallel Data Laboratory Tech. Rept. CMUPDL-10-109.
 +
 +
* Graves, R. W. (1996). Simulating seismic wave propagation in 3D elastic media using staggered-grid finite differences, Bull. Seismol. Soc. Am. 86, no. 4, 1091–1106.

Latest revision as of 17:30, 10 November 2017

This page collects information about a proposed high frequency verification and validation exercise. Planning started in Dec 2014.

Goal

Most regional scale deterministic earthquake simulation verification and validation exercises have been done at frequencies 1Hz or lower [1] [2]. With the introduction of high performance CPU and CPU-GPU wave propagation codes, there has been significant recent progress running simulations up to 4Hz. As we increase the capabilities of our deterministic earthquake simulation codes, we need to re-evaluate the codes performance at these higher frequencies that are now possible. The goal of this activity is to evaluate the performance of multiple wave propagation codes that are capable of simulation regional scale earthquakes at frequencies 4Hz+.

Approach

High frequency simulations introduce several code modifications intended to more accurately model the physics of high frequency earthquake wave propagation simulations. Modification to existing wave propagation codes to support 1Hz+ simulations include:

  • Source descriptions with energy above 1Hz
  • Small scale heterogeneities in the 3D velocity models
  • Frequency dependent attenuation
  • Near fault plastic yielding
  • Large displacement plastic yielding

To help isolate the impact of each of these changes, we propose to start with simple (possibly unrealistic) high frequency (4Hz) ground motion simulations and compare the simulation results produced by different methods. Then we will introduce these modifications one at a time, as a way if isolating the impact of each change individually.

Problem Statement

We have defined common input parameters for use by AWP-ODC, RWG, and Hercules. The problem statements includes the following parameters:

  1. Simulation Region
  2. Source Description
  3. Station List
  4. Velocity Model
  5. Simulation Duration
  6. Simulated Frequency Range
  7. Sample per Wavelength

Simulation Region

The simulation region four corners are given in the table below:

Volume size:

  • 180,000 m x 135,000 m x 61,875 m = 1.503e15 m3

Source Description

We will use a point source, and a source time function for that point source produced by Zheqiang Shi (SDSU). This source time function has frequencies above 4Hz, scaled to Mw5.1 La Habra. The source is plotted and defined in the following files. These show the source trimmed, low-pass filtered (fc=5Hz, Butterworth, 2 forward passes, 4 poles), 4300 points, dt=0.001s (and plot of what is should look like).

1D Velocity Model Definition

The first velocity model will be modeled on a 1D model of the Los Angeles Basin used in on the Broadband Platform project. The model was smoothed to eliminate sharp contrasts between layers.

Definition of Los Angeles Basin 1D velocity models used in the Broadband Platform project:

3D Model Extraction Parameters

  • 180km by 135km by 62km
  • 20m spacing
  • Points: 9000 by 6750 by 3100
    • Total: 188,325,000,000
  • Min Vs: 500m/s
  • Mesh file size: 2,259,900,000,000 => 2.1TB

Station List

We developed a list of ground motion observations for the La Habra earthquake. A description of the methods used to create this merged site list are provided at Selection of La Habra Ground Motion Observations.

This site list is based on use of publicly posted data sets for the La Habra event. There should be on-scale accelerometer ground motion data at all of these sites for the La Habra mainshock. This list combine all channels available from SMO and SCECDC for the La Habra Event using only acceleration instruments within the simulation region:

We expect all these recordings to be [ HN ] channel data, In SEED terminology,

  • [ H ] High Broad Band >= 80 to <= 250 Sample Rate (Hz) Corner Period: >= 10 sec
  • [ N ] Accelerometer (historically some channels from accelerometers have used instrumentation codes of L and G. The use of N is the FDSN convention as defined in August 2000)


We have not yet plotted all data to visually confirm all records are above noise, and on-scale. We have not yet confirm all records are three component records.

Fig 1: 356 La Habra Earthquake ground motion accelerometer recordings for the La Habra Earthquake in Simulation Volume

Simulation Box / Velocity Model

Parameter Value Notes
Dimensions (km) 180 x 135 x 61.875 (31 km for 1d Model)
Bounding Box (LL) -119.288842 34.120549, -118.354016 35.061096, -116.846030 34.025873, -117.780976 33.096503
Rotation Angle 39.9
UCVM Version Not used for first velocity model No heterogeneities
Velocity Model Versions Simplified BBP 1D model defined above
Miniumum Vs 500 m/s
Samples per wavelength 6 to 7 - AWP 10 to 12 - Hercules


Point Source Parameters

Parameter Value Notes
Event Name La Habra
Mw 5.1 Ref: http://www.scsn.org/2014lahabra.html
Moment 5.764 E+23 Dyne-cm Source: En-Jui
Origin Time 2014/03/29 04:09:42.97 Source: En-Jui
Origin Location -117.930; 33.922; 5.0km Source: En-Jui
Strike/Dip/Rake 239/70/38 Source: En-Jui, based on review of aftershock sequence
Rise Time 0.75 s

USGS Origin Time: 2014-03-29 04:09:42.170 (UTC) USGS La Habra Page

IRIS Origin Time: 2014-03-29 04:09:43.970 (UTC) Iris La Habra Metadata

Coordinate System

  • X ~= 040 azimuth (positive from 0 to 135 km), units in m, m/s and m/s2
  • Y ~= 130 azimuth (positive from 0 to 180 km), units in m, m/s and m/s2
  • Z = UP (positive upward, toward the sky)

Solver Parameters

Parameter Value Notes
Frequency 4.0 Hz
Simulation Length (Duration) 100 s
Delta T 0.001 - AWP 0.005 - Hercules
Plane Output Resolution 250m
I/O Print Rate every 25 steps - AWP every 12 steps - Hercules
Station List TBD
Software Version TBD


Fig 1:Location of La Habra Main Shock and aftershocks, used to select preferred fault plane solution. (Image Credit: En-Jui Lee)

High-F VandV Exercise Planning Documents (2015)

High-F Project Plan (2012)

Agreements Needed

  • Simulation Volumes
  • Rotation Angle
  • Projections used
  • Points per wavelength used
  • Number and location of sites used for ground motion comparison
  • Proposed Station List to save results
  • La Habra Point Source parameters
  • Currently available code capabilities in GPU codes
    • small scale heterogeneities
    • FDQ
    • Plasticity
    • dynamic rupture source
  • V and V exercise stages - need to be reviiewed and approved

Data Comparison

Related Entries

References Used

  • Withers, K.B., K.B. Olsen, and S.M. Day (2015). Memory efficient simulation of frequency dependent Q, Bull. Seis. Soc. Am., in press. (Vol 105).
  • Lee, E.-J., Chen, P., & Jordan, T. H., 2014a. Testing waveform predictions of 3D velocity models against two recent los angeles earthquakes, Seismol. Res. Lett., 85(6), 1275–1284.
  • Taborda, R. and J. Bielak (2013). Ground-motion simulation and validation of the 2008 Chino Hills, California, earthquake, Bull. Seismol. Soc. Am. 103, no. 1, 131–156, doi 10.1785/0120110325.
  • Cui, Y., E. Poyraz, K.B. Olsen, J. Zhou, K. Withers, S. Callaghan, J. Larkin, C. Guest, D. Choi, A. Chourasia, Z. Shi, S.M. Day, J.P. Maechling, and T.H. Jordan (2013). Physics-based seismic hazard analysis on petascale heterogeneous supercomputers, Proc. Supercomputing Conference, 2013.
  • Bielak, J., Karaoglu, H. and Taborda, R. (2011). Memory-Efficient Displacement-Based Internal Friction for Wave Propagation Simulation, Geophysics 76 (6): T131-T145.
  • Taborda, R., J. López, H. Karaoglu, J. Urbanic, and J. Bielak (2010). Speeding up finite element wave propagation for large-scale earthquake simulations, Parallel Data Laboratory Tech. Rept. CMUPDL-10-109.
  • Graves, R. W. (1996). Simulating seismic wave propagation in 3D elastic media using staggered-grid finite differences, Bull. Seismol. Soc. Am. 86, no. 4, 1091–1106.