Purpose: This page provides the details on the NZ 100m Simulation Validation project which will attempt to perform ground motion simulation validation for small and moderate Mw earthquakes, active shallow crustal and subduction.
This project will attempt to use the current workflow and science of the ground motion simulations although many aspects may not be optimal for areas which are currently exploratory (moderate Mw), or have not yet been explored (subduction). Table 1 below lists the issues related to each section which should be addressed and subsequently tested (potentially with lower resolution exercises) before the production run is commenced.
Table 1. List of issues to be addressed towards the validation exercise and suggested solutions.
| Item | Issue | Suggestion | Notes | Checkbox |
|---|---|---|---|---|
| Workflow | ||||
| Stability | Everyone generally works with something not on master, so there can be issues with compatibility between repos. | Can we pull everything relevant from master and check that everything actually works? | ||
| Newest IM calc changes | FAS, RotD50 and RotD100 additions had changes to several repos and changes many output files with with more components and potentially blank spaces. Would be good to know it all works on master. | Completing the above and doing proper verification of outputs should sort this. (BB: Can also consider not doing these to start with, then running them later) | Not done by default. Good to go without. | Done |
| Duplicate stations | Are we currently duplicating stations? | Install one that should have duplicates to check. | Done | |
| New vs old workflow | Results with new workflow may not be necessarily compatible with results from old workflow. Different fd statlist, different sim durations, VMs etc. | Rerun everything on new workflow. | Version 20.3.3.X: Simulation | Done |
| New stations | If we're running new simulations for all events, should we take the opportunity to add in any more stations? | (BB: Yes, good idea) | Decided to not add structural arrays. Added a few strong motion stations (based on scraping ftp and fdsn records). A lot were not new stations (installed ~2016?) but had no recordings previously. | Done |
| Sim binary versions | Should we use emod3d version 3.0.4 or 3.0.8? Should we use hb_high version 5.4.5 or 6.0.3? | Probably emod3d 3.0.8. Should probably do validation of old runs with hb_high 6.0.3 to see how much difference it makes for point sources and then decide. Could do both if we had to in the end. | hb_high 5.4.5.3 (velocity boundary fix) emod3d 3.0.8 | Done |
| HPC storage | We need to make sure we have enough space in terms of TB and inodes. | For 400m grid, we don't believe this will be an issue, can judge accordingly for 100/200m grid run after. Robin to clean up his run directory. | Done | |
| Auto submitter | Sometimes stops when there are still jobs to run. | Can someone check this? This one did EMOD3D, HF, and rrup but not BB, IM_calc and clean_up. Had to start the auto submitter again to do those. /nesi/nobackup/nesi00213/RunFolder/Validation/v20p3p11 | Only seemed to happen for 1 event runs. Run autosubmitter once more at end should mitigate this anyway. | Done |
| Corehour estimation | EMOD3D 3.0.8 has longer simulation times due larger time before rupture initiation. Causes time out sometimes. | Decide on how to revise CH est. | Claudio or other. | |
| Observed Ground Motions | ||||
| Events to simulate | Need to determine prospective events. | Determine range of Mw and depth, identify how many ground motions each have. (BB: Mw=3.5-7 (above 7 multi-fault is important, unless it would be possible to manually 'add' pre-existing faults into the workflow (e.g. Darfield and Kaik) want to avoid that for now); (BB: Depth - my gut feeling is shallower than 200km atm). |
| |
| Source | ||||
| Model choice for Mw range | Should we continue to use PS for Mw <= 5.0 and FF for Mw > 5.0? Do we want to go above Mw 7.0? | Probably same range for PS and FF. Don't know about going above Mw 7.0. (BB: I think as you have it at the moment is good. If it is possible we could do PS for Mw<5.5 and FF for Mw>F (i.e. have a 0.5unit overlap)) | I think go with PS: 3.5 - 5.0 FF: 5.1 - 7.0 Any "overlap" can be done ad hoc and swapped in post-processing. | Done |
| FF | What model of genslip do we use? What aspect ratio? | I think use latest version of genslip 5.4.2 and the non-square aspect ratio. Work to date hasn't been rigorous but it's probably the best option. | Robin - Awaiting email to Rob about subfaults above ground due to roughness. I did some testing on this. It seems to snap to something (probably top grid point) up to a certain amount above the ground surface (~500m above ground surface for 400m grid run). Therefore good to use genslip 5.4.2. | Done |
| Depth range | What range of depths do we want to do? | Plot CDF with depth + additional investigations. (BB: Depth - my gut feeling is shallower than 200km atm). | Done according to "Events to simulate" item. | Done |
| NZVM | ||||
| Max nz value | Maximum number of nodes on z direction currently set may be limiting for subduction earthquakes (I hit the limit when I did 50m grid runs). | Need to change in source code and recompile. Will depend on how deep we expect to go. | Robin and Jason - Changed and recompiled. | Done |
| NZVM version | We should use 2.03? The only difference between 2.02 and 2.03 is Waikato-Hauraki basin. | Should do a quick check to make sure everything is good. | Did not get round to it. Maybe I'll do it this weekend. | |
| Vs30 | ||||
Missing Kaiser Vs30 values | Kaiser et al. (2017) only has Vs30 values for 340 of ~380 stations. | Use our values for the others but trim them out in analyses in post-processing? (BB: Yes, I agree; that gives us the option post-analysis to play with both options). | Viktor Robin should be able to pull the real station values out of the nonuniform grid file Viktor produced. | Done |
Verification Subset
The following csv contains a set of 10 earthquakes to be used as the verification test subset of events, and a table and plot to summarize the events:
GeoNet_CMT_solutions_Verification_Subset.csv
Table 2. Summary of 10 verification subset events
| PublicID | Date | Latitude | Longitude | strike1 | dip1 | rake1 | Mw | CD | Notes |
|---|---|---|---|---|---|---|---|---|---|
| 2012p001403 | 20120101122700 | -43.4625 | 172.8325 | 222 | 53 | 101 | 4.8 | 4 | Standard SM in Cant, 4km centroid depth |
| 2014p211339 | 20140320044100 | -43.5949 | 172.1239 | 194 | 69 | 58 | 3.5 | 7 | Mw 3.5, smallest Mw considered |
| 3468575 | 20110221235100 | -43.566 | 172.6909 | 55 | 66 | 129 | 6.2 | 4 | Chch EQ, unmodified finite fault goes above ground surface |
| 2013p613797 | 20130816023100 | -41.734 | 174.1522 | 331 | 81 | 21 | 6.6 | 15 | Grassmere, largest ASC Mw considered |
| 2016p859628 | 20161114013000 | -42.367 | 173.7613 | 56 | 75 | 114 | 5.1 | 10 | Smaller Mw, dip-slip EQ with enforced square aspect ratio |
| 2017p636803 | 20170824144800 | -42.336 | 172.9411 | 217 | 81 | -137 | 4 | 42 | Mw 4.0 at ~40km |
| 2017p015402 | 20170106164600 | -37.7189 | 177.5335 | 353 | 70 | 71 | 4.9 | 60 | Mw 4.9 at ~60km |
| 2014p422047 | 20140606034600 | -39.46 | 175.2641 | 157 | 79 | 44 | 5 | 102 | Mw 5.0 at ~100km |
| 3063166 | 20090321202800 | -37.7863 | 176.7771 | 31 | 75 | 107 | 5.1 | 159 | Mw 5.1 at ~150km |
| 2012p923684 | 20121207181900 | -38.3501 | 176.0593 | 69 | 73 | 115 | 6.3 | 156 | Mw 6.3 at ~150km |
Figure 1. Plot of Mw-CD domain of 10 verification subset events
Run Instructions
Assuming the workflow is verified, the following is a quick list of the tasks required to run validation:
- Decide on range of Mw and centroid depth to consider.
- Obtain observed ground motions and screen for quality. (BB: Use standard NNet atm; we can re-pipe with a revised NNet later - as long as we keep sim results at all possible locations).
- Identify which events are worth simulating based on number of records (probably raw number of records here since CMG NN won't work well with extrapolated cases). (BB: Keep in mind improvements in NNet in near future may increase (or decrease) usable records, so be conservative and simulate more than may be necessary).
- Make sure compiled NZVM is right code version.
- Generate Srf and VMs using correct genslip version and NZVM model version.
- Move Srf and VMs to run directory and install using the correct gmsim version.
- Check root_params.yaml. Check task_config.yaml for emod3d, HF, BB, IM_calc, cleanup, rrup.
- Run auto submitter.
- Collect IM_calc files and perform analyses.