Add support for PGV

• 2018 Conterminous U.S. GMMs (mark as beta)
• Generate CONUS dataset
• Draft ScienceBase page

assigned to @jaltekruse

changed title from Review GMM support for PGV to Add support for PGV

changed the description

Implement conditional PGV based on PSA(T_PGV(M)) [source: review preprint].

ln(PGV) = a_{1} + a_{2}(M-6) + a_{6}(8.5-M))^2 + f1(M)ln(PSA(T_{PGV})) + a_{4}ln(R_{RUP}+5exp(0.4(M-6)))

where

ln(T_{PGV}) = b_{1} + b_{2}M

ln(T_{PGV}) = -4.09 + 0.66M

and

f_{1}(M) = \begin{cases} a_{2} &\text{for } M < 5 \\ a_{2} + (a_{3} - a_{2})/2.5 &\text{for } 5 \le M \le 7.5 \\ a_{3} &\text{for } M > 7.5 \end{cases}

These formulas and the coefficients have been changed in the PEER report

Sigma:

\sigma^2_{lnPGV} = f^2_{1}(M) * \sigma^2_{lnPSA(T_{PGV})} + \sigma^2_{lnPGV|PSA(T_{PGV})}

@pmpowers

Proposed implementation notes... a GMM without native PGV support can implement the UsgsPgv calculator (set a private flag? e.g. pgvCalculatorIsSupported = true; or detect dummy PGV coefficients == 0?):

If imt is PGV and pgvCalculatorIsSupported {
  // initialize UsgsPgv calculator
  UsgsPgv usgsPgv = new UsgsPgv( Mw, rRup );
  
  /* package up the following in UsgsPgv instead of each individual Gmm? */
  // get Tpgv to calculate GM
  double pgvT = usgsPgv.TPgv;
  
  // get range of bounding Imts
  Range<Imt> imtBounds = Imt.fromPeriodtoRange(pgvT); // see next proposal
  //... detect if pgvT 'fuzzyEquals' an Imt... Imt.fromPeriod() for direct calc() rather than
  // interpolatedGmm...

  // as long as imtBounds has been set correctly (pgvT not equal any IMT), then:
  private final GroundMotionModel interpolatedGmm = new InterpolatedGmm(subtype, pgvT, imtBounds)
       // ??? Alternate form of InterpolatedGmm? (see next proposal)  --------> double

  // interpolate calcMean for bounding Ts
  //final GmmInput in;
  ScalarGroundMotion psa = interpolatedGmm.calc(in)

  // compute PGV
  ScalarGroundMotion pgv = usgsPgv.calcPGV(psa)
}

@pmpowers

Proposal for new Imt.java utility method:

• given target
  T_{PGV}
  , in seconds, return bounding Imts as Range<Imt> for use in INTERPOLATED_IMTS, which, instead of Map<Imt, Range<Imt>>, could be Map<Double<T>, Range<Imt>>, throwing an error if
  T_{PGV}
  is outside the range of SA Imts (run fromPeriod first to determine if target T has an 'exact' match Imt?):

public static Range<Imt> fromPeriodToRange(double targetPeriod) {
  boolean lowerFound = false;
  for (Imt imt : Imt.values()) {
    if (imt.name().startsWith("SA") {
      if (!lowerFound && saPeriod <= targetPeriod) {
        lowerImt = imt;
        continue;
      }
      if (lowerFound && saPeriod > targetPeriod) {
        upperImt = imt;
        return Range.closed(lowerImt, upperImt);
      }
    }
  }
  // error checking...
}

• this requires(???) alternate form for InterpolatedGmm():

  // current:
  InterpolatedGmm(Gmm gmm, Imt targetImt, Range<Imt> imtRange) {...}
    // may need to modify checkArgument(imtRange.contains(targetImt)) ???`
  // alternate:
  InterpolatedGmm(Gmm gmm, double targetPeriod, Range<Imt> imtRange) {...}

• method to obtain rangeImt bracketing TPgv should operate on subset of IMTs supported by gmm
• (future) if the number of available IMTs for the gmm is less than some threshold, alternatively use the coefficients for the PGA-based or PSA(T=1)-based models for PGV

mentioned in merge request !48 (merged)

mentioned in commit 63169af4

changed milestone to %nshmp-lib Review Ready - Sprint 2

added To Do label

removed milestone

removed To Do label

added To Do label

changed milestone to %nshmp-lib and -apps improvements