Hazard curves outputs and Postgresql database

[Paolo Zimmaro]

Hi all,

I'm using OPENQUAKE to perform few analyses with a logic tree similar to the one used for the SHARE model.

-I've noticed that if I use the computational settings of the SHARE model with my analysis, I obtain thousands of outputs, correspondent to all the possible combinations (realizations) of the logic tree leaves. I'm interested only in a final output that contains the final mean hazard curves for each site and for each IM with the assigned weights in the logic tree model. How I can extract only these outputs, avoiding the extraction of all the outputs?

In other words, how I can handle the outputs in order to extract only the mean hazard curves?

-The analyses took a long time to run (more than a week on my laptop), to speed up the computational time, I can run the analysis on a Cent OS Cluster based, but they cannot run analyses in which is needed a database (postgresq for the OQ engine), can you possibly suggest me alternatives (like text files or tables that can replace this database?).

Thank you in advance for your time.

Paolo Zimmaro

[Damiano Monelli]

Hi Paolo,

If you are interested in mean hazard curves, set 'mean_hazard_curves = true' in the configuration file. For each intensity measure type you will obtain

a 'hazard_curve-mean.xml' file.

Currently there is no way to avoid extracting all the results related to the different logic tree paths. OpenQuake generates them by default.

Regarding the computation time associated with large logic trees like the one of SHARE, I would suggest you to install the very latest OpenQuake

(https://github.com/gem/oq-engine/wiki/Installing-the-OpenQuake-Engine-Nightly). We just recently introduced a number of optimizations

that reduce computation time significantly (https://github.com/gem/oq-engine/pull/1423).

About replacing postgres with some other solution I cannot give any advice (I'm not a developer) but I fear is not a trivial task.

Best regards,

Damiano

--
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.
For more options, visit https://groups.google.com/d/optout.

[Michele Simionato]

Hi Paolo, the good news is that we are working on the issues that the engine has with SHARE-like with large number of realizations.

In particular the engine is generating more outputs than really needed. That issue will be fixed pretty soon. If you want to extract only the mean

curves we can easily add a command-line option for that.

The bad news is that currently the engine is tightly coupled to Postgres and that you cannot run it without it.

As Damiano said, you really want to upgrade to the nightly builds (see https://github.com/gem/oq-engine/wiki/Installing-the-OpenQuake-Engine-Nightly)

for the kind of computations you are doing.

[Paolo Zimmaro]

Thank you Damiano and Michele,

-Damiano, I've installed the OQ Engine Nightly, now the version that I'm using is: 1.0.0-2+dev1399981677-7539571 is this the correct one?

-Michele, can you let me know when you'll introduce the command line option to extract only the mean curves (or if the command is already implemented), this will help me a lot!
I'm performing new analyses, I'll let you know about the speed and if I'll have troubles with this release.

Furthermore, I have another question, can you implement, similarly with the Classical PSHA analysis a command to extract the mean results also from the disaggregation analysis? If it will be done, you can run only a disaggregation analysis and get both the results (PSHA and Disaggregation), instead of run for the same case two different analysies, PSHA and Disaggregation.

Thank you so much again.
Best regards

Paolo

[Michele Simionato]

Il giorno mercoledì 14 maggio 2014 01:15:02 UTC+2, Paolo Zimmaro ha scritto:

-Michele, can you let me know when you'll introduce the command line option to extract only the mean curves (or if the command is already implemented), this will help me a lot!
I'm performing new analyses, I'll let you know about the speed and if I'll have troubles with this release.

 

While it has taken a while, nowadays we have the option individual_curves=False. If you use a recent version of the engine and set this option in the job.ini file,  then the individual curves will not be saved in the database and they will not be exported. You will see only the mean curves and the quantile curves.