[CIG-SEISMO] PREM crust in SPECFEM3D GLOBE

[Jeroen Tromp]

Hi Arjun:

Indeed, the ocean is replaced by upper crust for PREM calculations. In 
SPECFEM3D_GLOBE, the effect of the ocean is approximated based on an 
ocean load approximation, as discussed in Komatitsch & Tromp (2002ab). 
In principle, one could use an acoustic solver in the ocean layer and 
couple it to an (an)elastic solver in the crust and mantle based on 
domain decomposition (this is how the outer core is coupled to the 
mantle and inner core), but at periods longer than ~17 s the ocean load 
approximation suffices. Of course you should be using a different code 
for spherically symmetric Earth models; for benchmarking purposes it is 
critical that you use exactly the same model with the same radii. 
Benchmarks with normal modes presented on the CIG web site and 
reproduced in the manual were performed in this fashion.

Best regards,

Jeroen Tromp

On 5/8/13 5:56 AM, A. Datta wrote:
> Hi there,
>
> I have a question about the implementation of the PREM crust in
> SPECFEM3D_GLOBE. As I understand, in PREM the top 3 km of the earth (and
> the crust) correspond to a water layer with rho = 1.02, Vp = 1.45 and Vs =
> 0
>
> But in SPECFEM, this is not the case. The PREM crust appears to uniform
> between 0 and 15 km. Here's an extract from the model_prem routine:
>
> else if(r > RMIDDLE_CRUST .and. r <= ROCEAN) then
>        rho=2.6d0
>        vpv=5.8d0
>        vsv=3.2d0
>        vph=vpv
>        vsh=vsv
>        Qmu=600.0d0
>        Qkappa=57827.0d0
> else if(r > ROCEAN) then
>        rho=2.6d0
>        vpv=5.8d0
>        vsv=3.2d0
>        vph=vpv
>        vsh=vsv
>        Qmu=600.0d0
>        Qkappa=57827.0d0
>
> RMIDDLE_CRUST corresponds to a depth of 15km and ROCEAN, to 3km. Even
> though depths above and below 3 km are considered in two separate if
> blocks, the model values assigned, as you can see, are the same. An unusual
> bit of code.
>
> I just wanted to confirm the crust in PREM as implemented in
> SPECFEM3D_GLOBE is indeed different from the actual PREM crust, and if so,
> it would be great to know why.
>
> Thank you very much,
>
> Arjun Datta
>