clipping of PGA

94 views
Skip to first unread message

athana...@gmail.com

unread,
Sep 9, 2020, 4:51:53 PM9/9/20
to OpenQuake Users
Hello to everyone,

I would like to ask if values of PGA are clipped at 1.0 g by default. 

I have realized some hazard calculations at a site (where PGA should be higher than 1.0 g) but the highest PGA I got was systematically equal to 1.0 g, indepedently of the element in the input I changed.

Meanwhile, PGV doesn't appear to have the same problem.

I have searched at the code, and the only relative thing I have found is in utils.py file in hazardlib, where it says that the mean value of PGA is clipped at 1.5 g:


def clip_mean(imt, mean):
    """
    Clip GMPE mean value at 1.5 g for PGA and 3 g for short periods
    (0.02 < T < 0.55)
    """
    if imt.period == 0:
        mean[mean > 0.405] = 0.405

    if 0.02 < imt.period < 0.55:
        mean[mean > 1.099] = 1.099

    return mean


I changed the value of 0.405, ran a calculation, but again the output value was 1.0 g.

Could you please tell me if there is another point in the code that refers to the same thing and maybe cancels the above-mentioned limit of 1.5 g? 

Many thanks in advance,
Athanasia

Michele Simionato

unread,
Sep 10, 2020, 1:04:04 AM9/10/20
to OpenQuake Users
On Wednesday, September 9, 2020 at 10:51:53 PM UTC+2 athana...@gmail.com wrote:
Hello to everyone,

I would like to ask if values of PGA are clipped at 1.0 g by default. 


I am not aware of any clipping and I have seen the engine produce values > 3g for PGA: it there must be something specific to your use case,
You should send a job.zip file containing your input files and then we can investigate.

           Michele 

Athanasia Houdoutoutoudou

unread,
Sep 10, 2020, 7:40:45 AM9/10/20
to OpenQuake Users
Thank you for the reply, I attach the link for the input.
Athanasia

Graeme Weatherill

unread,
Sep 10, 2020, 4:49:21 PM9/10/20
to openqua...@googlegroups.com
Dear Athanasia,

In the job.ini file the maximum of the PGA values you are setting in the "intensity_measure_types_and_levels" only reaches 1 g as a maximum. The range of values input here should be large enough to ensure that OQ can interpolate to the define the PGA with a given probability of exceedance in 50 years. If this is outside of the range specified in the job.ini file then OQ does not extrapolate and the results may be clipped. I would recommend increasing the range of PGA values set here. You can go up to 3 - 4 g if necessary. That should reduce the chances of clipping (unless your hazard is actually 3 - 4 g, in which case you may have other problems) and doesn't really add much additional computation cost.

Hope that helps.

Graeme
--
You received this message because you are subscribed to the Google Groups "OpenQuake Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to openquake-use...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/openquake-users/e9dd8a1a-070d-4d36-87d7-19dc8fa4321cn%40googlegroups.com.


Athanasia Houdoutoutoudou

unread,
Sep 11, 2020, 2:11:47 AM9/11/20
to OpenQuake Users

It worked perfectly after the addition of value 3! PGA was  increased to 1.1 g.
Thank you very much for the help, I would still be searching for the reason of clipping...!
Best regards,
Athanasia

Anirudh Rao

unread,
Sep 11, 2020, 12:45:40 PM9/11/20
to openqua...@googlegroups.com
Hi Athanasia,

The hazard probabilities of exceedance are computed only for the intensity levels you have defined in your job file using the intensity_measure_types_and_levels parameter. As Graeme mentioned previously, the highest value for PGA you had listed under intensity_measure_types_and_levels in your job file was 1.0 g, so the hazard curves were being computed only up to that level.

The clip_mean() function is not related to your issue; that is something used by the GMMs developed for the stable continental region of the Central and Eastern United States (CEUS) for the 2008 version of the US national seismic hazard models (NSHM). In the 1996, 2002, 2008, and 2014 versions of the US NSHM, ground motions at higher frequencies in the CEUS region were capped in order to avoid very large values (particularly near the New Madrid and the Charleston seismic zones) that would have been physically unrealistic. (This capping has been eliminated in the 2018 US NSHM with the development of new GMMs for the region). Within the OpenQuake engine, the "NSHMP2008" subclasses of the following GMMs thus include the clipping logic defined in the clip_mean() function you discovered:
  • Toro et. al. 1997
  • Frankel et. al. 1996
  • Campbell 2003
  • Atkinson and Boore 2006 (140 bar stress drop)
  • Atkinson and Boore 2006 (200 bar stress drop)
  • Tavakoli and Pezeshk 2005
  • Silva et. al. 2002
  • Somerville et. al. 2001
Best regards,
Anirudh

Best regards,
ANIRUDH RAO    seismic risk engineer    LinkedIn: Anirudh Rao    +39-0382-5169-936
GLOBAL EARTHQUAKE MODEL    working together to assess risk
GEM —  globalquakemodel.org     @GEMwrld      GEMwrld


Reply all
Reply to author
Forward
0 new messages