Support for Spectral Amplitude at arbitrary damping ratios

Achieved via damping scaling factor models of Rezaeian et al. (2014, crustal) and Rezaeian et al. (2021, subduction), using the same GMM logic trees as currently implemented. For now we're assuming the crustal model will be applied to CEUS motions. Note that stable crust model is T dependent.

• Scenario GMM calculations are priority.
• Figure out how to apply conditional GMMs after or as part of Transforms.InputsToGroundMotions
• Coefficient interpolation support in UsgsDampingScaling.java
  • Verification that DSF results at intermediate damping ratios are reasonable (Andrew can do this)
• Spot-checking RezaeianDamping.java to see if there are any obvious issues with the functional form
• Implementation of Subduction DSF model (Rezaeian et al. 2021)
  • Needs special handling for T>5s and not sure what to do below T<0.1s
• Creation of a DampingScaling or UsgsDampingScaling or just Damping factory class
• Implementation of stable craton DSF model (for now, using the Active Crustal model of Rezaeian et al. 2014)
• Consideration of damping scaling factors of vertical ground motions? see Issue #121
• Resurrect damping sigma adjustment flag in config see #110 (comment 300240)
• Add main() method to DSF class
• Create unit tests against python results
• Config support DSF and sigma (boolean); link to DampingScaling
• Decide how to document implementation for end users once DSFs are exposed in web services and applications
• The DSF Implementation currently has a memory issue. Memory is not being deleted properly causing a swell of memory to be created. This significantly limits the amount of Imts you are able to run for your site.
• The DSF is not compatible with the web service. There needs to be a way to add a variable so the user can specify the Damping. this way the web service can be used for Damped Calculations.
• The nshmp-ws should allow the damping to be passed as a variable to allow for damped Gmm calculations. currently there is no way to specify a damping value.
• The public web service, local web service, and local command line 5% hazard curves are not equal. this is likely due to different nshms and default configs. While this isn't a huge issue more documentation on the default model and config used would be very helpful.
• Currently if a GMM supports any non-MPRS SA (IMT), it is filtered out of any GMM web service response. It would be preferable if UsgsDampingScaling supported interpolation between SAs. See GmmCalc.spectra().

added Enhancement GMM labels

mentioned in issue #138 (closed)

changed the description

marked the checklist item Figure out how to apply conditional GMMs after or as part of Transforms.InputsToGroundMotions as completed

marked the checklist item Figure out how to apply conditional GMMs after or as part of Transforms.InputsToGroundMotions as incomplete

changed the description

changed the description

marked the checklist item Resurrect damping sigma adjustment flag in config as completed

mentioned in merge request !253 (merged)

changed the description

@pmpowers @jaltekruse I've added the test inputs and results for the Rezaeian et al. (2014) DSF model below. The inputs are a modified version of what is used for the NGA-West2 GMMs (https://code.usgs.gov/ghsc/nshmp/nshmp-lib/-/blob/main/src/test/resources/gmm/nga-west-inputs.csv) that considers only M_w and R_rup sensitivity. The IMTs and damping ratios evaluated are:

imts = ['PGA', 'SA0P05','SA0P1','SA0P5','SA0P75','SA1P0','SA2P0','SA5P0','SA10P0']

damp_pct = [0.5,1.0,3.0,5.,10.,20.,30.]

Column 1 of the results csv contains the gmm input scenario number, REZ_14_DSF tag, and the imt and damping ratio evaluated. Column 2 is the median predicted DSF (not the log of the median), and column 3 is the log-standard deviation.

dsf-ngaw2-inputs.csv

dsf-ngaw2-results.csv

changed the description

Looking over the current RezaeianDamping_2014.java script, I'm not seeing any obvious red flags with the functional form, but I did see that there currently isn't any way to return the computed sigma of the DSF by itself (it's computed as part of updateSigma, which returns the standard deviation of the PSA after the DSF has been applied to it (and is eventually called by the apply method). For initial testing purposes, I think it'd be good to just return the median and sigma DSFs just to make sure that's doing what we expect it to do.

Simple method below for computing the sigma, although I'm not sure the best way to incorporate it into the apply method at the moment...

private static double computeSigmaDSF(
      Coefficients c,
      double dampingRatio) {
        
    double βTerm = log(dampingRatio / 5.0);
    double σlnDsf = abs(c.a0 * βTerm + c.a1 * βTerm * βTerm);
    return σlnDsf;
}

Also see ppt file from earlier meeting for implementation reference... nshmp-haz-DSF.pptx

changed the description

marked the checklist item Implementation of Subduction DSF model (Rezaeian et al. 2021) as completed

marked the checklist item Implementation of Subduction DSF model (Rezaeian et al. 2021) as incomplete

changed the description