Validating OQ against Peer test case (Set1 Case 9a)

50 views
Skip to first unread message

Jeong Min Han

unread,
Aug 12, 2014, 7:39:35 AM8/12/14
to openqua...@googlegroups.com
Dear Dr Marco Pagani & Dr Damiano Monelli,
 
Thank you for your helpful comments on my issues with using OQ.
 
I have now a new set of problems encoutered while carrying out some PEER validation tests (Thomas et al., 2010; PEER report). It would be grateful if you could provide me any suggestion or correction.
 
The issue is regarding the PEER Set1 Test case 9a; this case considers a reserve fault dipping west wtih delta function at M6, and the Sadigh et al. (1997) GMPE with 3sigma.
 
I have attached both the plot (OQ hazard curves against PEER for all seven sites) and the input files (source model, gmpe model, configuration file).
 
My questions are:
1. When OQ hazard curves are compared to PEER results, I have marginal differences between OQ hazard curves and PEER. Can you please comment or provide recommendations on this discrepancy?
 
2. As you may have already noticed, I am getting mirrored hazard curves for site 2 and site 7 are mirrored (i.e. the hazard curve for site 2 should be for the hazard curve at site 7 and vice versa). This obviously explains that my source model is incorrect. More specifically, I have incorrect dipping direction wrong (my hazard curves are for a reverse fault dipping east). 
I used a simple fault source model which considers the coordinates for the top fault trace, seismogenic depth, dip and rake whereas there is no input for strike. Can you please comment on this matter?
 
Thank you for your extremely helpful comments in advance.
 
Kind regards,
 
Jeongmin Han
 
Hazard_Curve_PGA_Set1Case9a.pdf
set1case9_inputs.zip
Message has been deleted
Message has been deleted
Message has been deleted

Jeong Min Han

unread,
Aug 13, 2014, 6:03:17 AM8/13/14
to openqua...@googlegroups.com
I apology if I have not made myself clear.
Let me clarify my question particularly with the second question.
- How can I specify a strike in a simple fault model?

Marco Pagani

unread,
Aug 13, 2014, 6:16:55 AM8/13/14
to openqua...@googlegroups.com
Dear Jeong Min,

The strike direction is computed using the fault trace. If you want to
read more about this, the method we use is available here:
https://github.com/gem/oq-hazardlib/blob/master/openquake/hazardlib/geo/mesh.py#L519


In your case the strike direction should correspond to the direction you
take to move along the trace from the first point to the second one.

Please note that to the define the fault geometry we use the right hand
rule (see for example
http://books.google.it/books?id=TV-sAwAAQBAJ&pg=PA55&lpg=PA55&dq=geology+fault+strike++right+hand+rule&source=bl&ots=EWs5iiFOYO&sig=8lmF3jO3aAFeP2LIcTcFlgPVp9M&hl=en&sa=X&ei=gjrrU-PnKoWa1AWU9YCYDw&redir_esc=y#v=onepage&q=geology%20fault%20strike%20%20right%20hand%20rule&f=false)


Regards,

Marco



On 13/08/14 12:00, Jeong Min Han wrote:
> I apology if I have not make myself clear.
>
> Let me clarify my question particularly with the second question.
>
> - How can I specify the stike in simple fault model?
>
>

Jeong Min Han

unread,
Aug 14, 2014, 5:27:50 AM8/14/14
to openqua...@googlegroups.com
Dear Marco,
 
Thank you for your clarification.
 
Switching the coordinates of the fault trace solved this issue.
 
I am sorry for bothering you again but my problem still remains unsolved.
 
The case 9a examines the Sadigh et al. (1997) GMPE with epsilon = 3.
 
I have modified the default Sadigh et al. (1997) GMPE from gsim library to set the sigma equal to zero.
Just to make sure to check my modification on the Sadigh et al. (1997) GMPE, I have checked the modified Sadigh et al. (1997) GMPE against the bench mark table using check_gsim.py (checked by putting 0.001 for standard deviation instead of 0). The GMPE check results confirmed that my modification worked fine.
 
I corrected the source geometry (as attached figure) but hazard curves for most sites (i.e. site 1,2,4,6,7) show marginal differences with the PEER results except for the site 3 and 5 showing the identical match against the PEER results.
 
Can you please comment on this discrepancy (whether it is expected from the OQ) or do you have any other suggestion to correct this case?
 
Thank you very much for your time and helpful comments.
 
Kind regards,
 
Jeongmin Han
S1C9a_Rev1.pdf

Marco Pagani

unread,
Aug 15, 2014, 5:20:17 AM8/15/14
to openqua...@googlegroups.com
Dear Jeongmin,

As you correctly indicate the case 9a considers an epsilon value equal
to 3, that can be specified in the OpenQuake-engine .ini file (by
setting the 'truncation_level' parameter equal to 3).

Based on this, I don't understand why you had to modify the Sadigh et
al. (1997) in order to set the sigma value to 0. This is not in
agreement with the settings defined by Thomas et al. for Set 1 case 9a
nor with the information you provide in the initial part of your email.

Can you please provide further details so that I can understand better
and therefore provide more useful suggestions?

Best wishes,

Marco
--

*MARCO PAGANI* | Seismic hazard team coordinator | skype mm.pagani | g+
marco....@globalquakemodel.org | +39-0382-5169863

*GLOBAL EARTHQUAKE MODEL * working together to assess risk

*GEM -* globalquakemodel.org <http://www.globalquakemodel.org> | *T -*
@GEMwrld <http://twitter.com/GEMwrld> | *F -* GEMwrld
<http://www.facebook.com/GEMwrld>
Reply all
Reply to author
Forward
0 new messages