Inconsistent clamping of maximum ground motions.

[pmpowers-usgs]

clamp-notes.xlsx This is a recurring issue that was thought to be resolved, but apparently wasn't, as demonstrated during nshmp-haz verification. From the 2008 and 2014 OFR's:

The ground-motion distribution for peak ground acceleration is truncated at 3 g and for 0.2-s spectral acceleration at 6 g, when these values are less than the 3-sigma cutoff.

Although this guidance is only meant to cover the primary periods supported with each official map release (PGA, 5-Hz, and 1-Hz), per conversations with Steve Harmsen and code comments, OpenSHA dev implementations of the 2008 NSHMP, and the new nshmp-haz codes apply a model that clamps PGA at 3g and SAs<0.75s at 6g (see ExceedanceModel.NSHM_CEUS_MAX_INTENSITY; this link may break in the future if ExceedanceModel is refactored to ProbabilityModel).

In practice in the fortran codes, however, hazFXnga13l.f:

• applies a 3g clamp at 1-Hz (this gives rise to significant discrepancies in the immediate vicinity of New Madrid (see attached image).
• includes clamp arrays in Campbell and Toro GMMs that should be commented out in favor of the global array.

and hazgridXnga13l.f

• applies short period clamps inconsistent with other GMMs in the the three new (for 2014) models: AB06p, A08p, and Pez11.
• applies a clamp of 3g at 0.5s in Campbell

See attached spreadsheet for a summary.

clamp-notes.xlsx

[pmpowers-usgs]

Nothing anywhere indicates that the hazFXnga13l.f clamp at 1-Hz should be 3g (instead of 0 -- no clamp); changing to 0.0 and will revert if Harmsen or Frankel provide a compelling reason to revert back. Would then make the case to change all 1-Hz clamps in hazgridXnga13l.f

It remains to be seen if this has a significant effect on 2% to 10% in 50-year ground motions.

[pmpowers-usgs]

Clamping fixes were mistakenly made to parallel version of hazFX. 1.5s clamp also needs to be changed to 0.