[Shake-dev] developing new GMPE module for ShakeMap v3.5

[Bruce Worden]

Hi Georgia,

In general, the 3.2 and 3.5 results won't be the same because 1) the  
interpolation is now done differently, and 2) intensity is now handled  
by direct interpolation, rather than simply converting the PGA/PGV  
maps. We did not attempt to make the maps for the same events  
identical because the new approach yields better results.

In cases like scenarios, where there is no data, the two versions  
should be the most similar. To the extent that the GMPEs are the same  
(and one uses the Borcherdt-style site corrections and not the GMPE- 
native site corrections [which didn't exist in 3.2]), the PGM maps  
should be quite similar. The difference will be that in 3.2 the  
estimates are made on a coarse grid and interpolated by cubic spline  
to the output grid, and then site corrected. In 3.5, the estimates are  
made at every point in the output grid, and then site corrected. Thus,  
3.5 should better reflect the "shape" of the GMPE and of the source  
(if it is not a point source).

For MMI, if you are using the DefaultIPE (i.e., a GMPE combined with a  
GMICE rather than an explicit IPE), the difference will also  be in  
the way site corrections are done and the order they are applied. For  
3.2, site-corrected PGA and PGV maps are created (as described above),  
and then converted to intensity. In 3.5, PGA and PGV estimates are  
created at each point in the output grid, converted to intensity, and  
then site corrected. If you are using an explicit IPE, then the  
intensity map follows the 3.5 pattern for PGM maps: intensities are  
estimated at each point in the output grid, and then site corrected.

Bruce


On Dec 22, 2010, at 2:41 AM, Georgia Cua wrote:

> Hi Bruce,
>
> Here is the final version of the GMPE we are using in Switzerland,  
> using Borcherdt-style site corrections (SwissNGA_borcherdt.pm).
>
> I have another question..
>
> Do you expect that runs on the same event (say, a scenario, for the  
> simplest case) with versions 3.2 and 3.5 should be identical, or  
> just "close enough"?
>
> Thanks again,
> Georgia
>
> <SwissNGA_borcherdt.pm><ATT00001.html>