relating simulated ground motion to GMPE

Hyeuk Ryu

unread,

Jul 12, 2019, 9:20:44 AM7/12/19

to OpenQuake Users

Hi,

The OQ engine generates ground motion fields as many as the number of GMPEs when a logic tree of GMPEs is used.

When I checked csv and xml output from scenario hazard calculation, I noticed that they are not consistent or unclear how to relate gm value to each of GEMPs using the csv file.

"gmf-scenario_#.xml" has rupid value, which corresponds to ordinal column in the realizations_#.csv.

However gmf-data_#.csv contains eid, but converted rupid did not match to the rupid value in the gmf-scenario.xml. And the row is not in the same order of rupid or ordinal. So if the csv file is only used, then it's very likely to relate gm value to not the right gmpe. Let me know whether I misunderstood something.

Thanks.

Regards,

Hyeuk

gmf-data_17.csv

gmf-scenario_17.xml

realizations_17.csv

Michele Simionato

unread,

Jul 15, 2019, 4:19:40 AM7/15/19

to OpenQuake Users

You just discovered the reason why we completely removed the XML exporter in master. For legacy reasons (for people with an old XML parser) the name of the tag had to be `ruptureId` but it is actually not the `rupid` appearing in the CSV, rather it is the event ID appearing in the gmf_data dataset in the datastore. It is simply confusing. As I said, just upgrade to master which now is frozen and identical to what will be engine-3.6.

Michele

Hyeuk Ryu

unread,

Jul 16, 2019, 8:16:41 PM7/16/19

to OpenQuake Users

Hi Michele,

Thanks for your reply, but the point that I made was not on xml, but csv. Still I can not understand how to relate GMPE to gm value just by looking at both gmf-data#.csv and realizatios#.csv because eid was not referenced in the realizations#.csv. Also the ordinal should not be confused with the order of eid in the gmf-data#csv. Please see the following example where you can find the AtkinsonBoore2006 is at the 2nd in the realizatons file while the gmf value using the AtkinsonBoore2006 is at the 3rd in the gmf-data#.csv. FYI. I've used 3.7.0-git88e4182717. Please correct me if I misunderstood something.

# gmf-data#.csv

eid,sid,gmv_PGA

12884901888,0,3.671288E-02

12884901889,0,5.243856E-02

12884901890,0,3.433215E-02

12884901891,0,3.700328E-02

12884901892,0,4.202059E-02

12884901893,0,1.331436E-01

12884901894,0,3.095780E-02

# realizations#.csv

ordinal,branch_path,weight

0,[Allen2012],1.9200000e-01

1,[AtkinsonBoore2006],1.2500000e-01

2,[AtkinsonBoore2006Modified2011],1.1800000e-01

3,[BooreEtAl2014],1.0600000e-01

4,[ChiouYoungs2014],8.5000001e-02

5,[SomervilleEtAl2009YilgarnCraton],2.2800000e-01

6,[ZhaoEtAl2006AscSWISS05],1.4600000e-01

# correct matching

[Allen2012]->[[0.03671288]],0.192

[AtkinsonBoore2006]->[[0.03433215]],0.125

[AtkinsonBoore2006Modified2011]->[[0.05243856]],0.118

[BooreEtAl2014]->[[0.03700328]],0.106

[ChiouYoungs2014]->[[0.04202059]],0.085

[SomervilleEtAl2009YilgarnCraton]->[[0.13314363]],0.228

[ZhaoEtAl2006AscSWISS05]->[[0.0309578]],0.146

Michele Simionato

unread,

Jul 16, 2019, 11:56:45 PM7/16/19

to OpenQuake Users

You can see the relation between event ID and realization (i.e. the GSIM) with the command

$ oq show events

id,rlz

12884901888,0

...

For instance if you say that the value 0.03671288 should be associated to Allen2012, which is realization 0,

then the event 12884901888 should correspond to 0.

If you believe there is bug please send me the files you are using.

Michele

Hyeuk Ryu

unread,

Jul 17, 2019, 12:17:25 AM7/17/19

to OpenQuake Users

Hi Michele,

I confirm that it's not a bug. Thanks for the clarification. I did not know that there is such command showing rlz to event id.

Why is not it included in the list of outputs? Also can not find it in the manual.

Thanks.

Regards,

Hyeuk

Hyeuk Ryu

unread,

Jul 17, 2019, 12:23:28 AM7/17/19

to OpenQuake Users

Hi Michele,

One more thing. Why not use rlz instead of ordinal in the realizations.csv?

Thanks.

Regards,

Hyeuk

On Wednesday, 17 July 2019 13:56:45 UTC+10, Michele Simionato wrote:

Michele Simionato

unread,

Jul 17, 2019, 2:21:41 AM7/17/19

to OpenQuake Users

On Wednesday, July 17, 2019 at 6:23:28 AM UTC+2, Hyeuk Ryu wrote:

Hi Michele,

One more thing. Why not use rlz instead of ordinal in the realizations.csv?

Just legacy.

Michele Simionato

unread,

Jul 17, 2019, 2:33:07 AM7/17/19

to OpenQuake Users

On Wednesday, July 17, 2019 at 6:17:25 AM UTC+2, Hyeuk Ryu wrote:

Hi Michele,

I confirm that it's not a bug. Thanks for the clarification. I did not know that there is such command showing rlz to event id.
Why is not it included in the list of outputs? Also can not find it in the manual.

I am not sure if it should be in the list of outputs. After all, you are the only user that ever asked for it. I guess most people performs