Specifying Scenarios for the Stochastic Assessment

[Godfred-oops]

Hello all,

I'm currently doing a test run for a stochastic assessment, simulating intensity at each site. Since it's a test run, I'm using two site locations, two IM realization per site, and two scenarios. For the two scenarios, I have a list of sources I'm simulating for, and I used the source index inside my configuration file (Please note, I'm using the backend code). The run is successful but the maps is not simulated for the two scenarios I specify.

Questions:
Am I simulating the maps/IM the right now? If not, what is the right way to do it.

Attached is the configuration file and the RupFile.json
RupFile.json
EQHazardConfiguration_stochastic.json

[jinyan1214]

Hello,

Thanks for your question and your interest in using this tool!

If I understand your question correctly, you have run scenario forecasting to get a number of earthquake rupture scenarios and get a RupFile.json. Then, you want to simulate IM maps for two scenarios in the RupFile.json.

To do this, please replace the "Generator" part in the your EQHazardConfiguration.json with below:
"Generator": { "RuptureFilter": "1-2,5", "method": "MonteCarlo" },

By doing this, the HazardSimulationEQ.py will simulate IM maps for the first two and the fifth rupture scenarios in your RupFile.json. You can modify the RuptureFilter to select the scenarios you want to simulate.

Note: It looks like the EQHazardConfiguration.json you attached comes from a very old version of the simulation code. Please use the latest simulation code at https://github.com/NHERI-SimCenter/SimCenterBackendApplications/tree/9126b14d2cdb23084b12fc43e7083ca9c9fd6f9e/modules/performRegionalEventSimulation/regionalGroundMotion

Hope this helps and please let us know if you have further questions.

[Godfred-oops]

In my case, someone did the scenario forecasting to get the rupture scenarios. Below is the attached RupFile.json. So if I get what you're saying. After forecasting the scenarios, then you run the HazardSimulationEQ.py file to get the IM maps.

Also can you please send a configuration file for running the latest simulation code.

RupSampled.json

[jinyan1214]

I am attaching the configuration file I use on my computer: EQHazardConfiguration.json

BTW, the function you desire is also available through the latest R2D user interface. So if you run a small example through the user interface, you will find one EQHazardConfiguration.json in your local work directory (by default, your Documents directory).

Hope this is useful.

[Godfred-oops]

Hello Jinyan,

I'm currently using the backend code to do the scenario downsampling for the scenario forecast I did, and I'm using Wang et., al, 2023 algorithm for the downsampling and specified I need 20 scenario but I'm getting 13.

What might be the issue?

Attached is the json file for the sampled scenario and Hazard Configuration
RupSampled.json
EQHazardConfiguration.json

[jinyan1214]

Hello,

Glad to know that you successfully ran the downsampling algorithm and obtained results.

The behavior you observed is normal in Wang et al. 2023's method. In this method, the number of selected scenarios depends on the regularization factor used for the LASSO regression. There isn't a way to know how many scenarios will be selected when a regularization factor is picked. So, in R2D, a series of regularization factors are used (by sklearn lasso_path), and each regularization factor leads to a number of selected scenarios. The regularization factor that results in a number that is closest to the predefined number (in your case 20) among all factors is selected as the final regularization factor. If you would like a larger number of scenarios, you can add "'LassoTuningParameter':[0.001, 0.01, 0.1]" next to "Model": "Wang et al. (2023)" in the hazard configuration file. This will allow you to use your own defined regularization parameter. You may need to go through a trail and error process to get a number that is closer to 20.

Hope this is useful.

Jinyan

[Godfred-oops]

Yeah

Thanks Jinyan!