Issues with the conditional spectrum calculator

[Sistla Saiteja]

Hi all,

I am using OpenQuake 3.20 and I tried using the conditional spectrum case_1 example (openquake\qa_tests_data\conditional_spectrum\case_1). I am facing two main issues, explained below:

1. I was able to replicate the results in the expected results folder. I performed the calculations by changing the conditioning period from 0.2s to 2s, with NoCrossCorrelation. I expected to see the same answer in both cases; however, this was not the case, and the difference in the results was about 100% at some periods. Is there something I am missing here?

2. The example used only area sources and active shallow crust TRT. My source model has point sources and characteristic fault sources, along with active shallow crust, volcanic, and subduction intraslab TRTs. I get an error (refer to the image below) when I run the conditional spectrum calculator. Nonetheless, when I just use the point sources and the active shallow crust TRT, the calculation completes, and the output shows: poe=0.1, imt=SA(0.2): 0.13024 vs 0.13818 (a difference of 5%, which I guess should be acceptable). Has anyone else faced issues when using characteristic fault sources, volcanic TRTs, and subduction intraslab TRTs? Any suggestions on how to get around this would be appreciated.

[Michele Simionato]

Looks like a bug. The only way we can help is if you send us the files you are using. Please run

$ oq zip job.ini

and send us the resulting job.zip

     Michele

[Sistla Saiteja]

Hi Michele,

The first issue I reported is not essentially a bug. My initial understanding was incorrect, and I stand corrected. When the conditioning period is changed from 0.2s to 1s, the M, D, and TRT contributions to the hazard change. Therefore, even though the cross-correlation is set to zero, the results obtained in both cases will not be the same. I apologize for the mix-up.

Regarding issue 2, I did some further investigation and narrowed it down. I modified the source model files in the case_1 example of the conditional mean spectrum example. The original example had two source models, source_model1.xml and source_model2.xml, both with area sources and active shallow crust TRT. I get an error (attached to the first email I sent in this thread) when I add an additional point source with volcanic TRT in source_model1.xml, leaving source_model2.xml unchanged. Please find a minimum working example of this case attached as job.zip.

PS: Leaving source_model1.xml unchanged, I also modified the TRT type from active shallow crust to volcanic in source_model2.xml. The conditional spectrum calculator works in this case with no issues.

I think the issue arises when multiple TRTs exist in a single source model file. Please let me know if there is an issue with the conditional spectrum calculator or if I got something wrong at my end, so that I can fix it.

Best regards,
Teja.

[Sistla Saiteja]

Although I did attach the job.zip file in my previous email, it does not appear here. Hence, I am sending it again.

[Michele Simionato]

Thanks for the detailed bug report. It is fixed in current master, see https://github.com/gem/oq-engine/pull/9862.

  Michele Simionato

[Sistla Saiteja]

Hi Michele,

Thanks for fixing this. I will try it out soon and let you know how it goes.

kind regards,

Teja

--
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/988a7c2e-e217-417a-8211-631cab46378bn%40googlegroups.com.

[Sistla Saiteja]

Hi Michele,

Thank you for resolving the issue with the conditional spectrum calculator handling multiple TRTs. It works well now. However, I am encountering a new problem when using the latest 2022 version of the New Zealand NSHM. The output CSV files from the Conditional Spectrum Calculator contain 'nan' values for both the mean and standard deviation.

Upon further investigation, I identified the likely cause. The 2022 NZ NSHM includes two main types of source models:

1. A distributed seismicity source model
2. A model where multiple rupture scenarios are considered within a TRT.

The analysis runs fine when using either source model individually. However, when I include both models together, the output CSV for the conditional mean spectrum (CMS) contains 'nan' values.  Could you please look into this? Unfortunately, I’m not certain if I can share the source files at the moment. I’m in the process of requesting permission and will try to provide you with a minimal working example if possible.