[MITgcm-devel] time interpolation in exf_set_gen

[Patrick Heimbach]

Hi Jean-Michel,

wow, thanks for getting to the bottom of this.
I agree, your change sounds good,
and hopefully makes the exf verifications more robust.

Cheers
-Patrick



On May 17, 2007, at 9:14 PM, Jean-Michel Campin wrote:

> Hi Patrick,
>
> I think I found the source of the differences between
> a) runoffperiod= 0 , no time interpolation
> b) runoffperiod= 1 month, do time interpolation , with 12
>  identicals reccords in runoffFile.
>
> It's a machine truncation problem, and I think it might be
> worth to change the current version:
>  genfld = fac * gen0 + (exf_one - fac)* gen1
> to:
>  genfld = gen1 + fac * ( gen0 - gen1 )
>
> reason is that fac + (exf_one - fac) is not exactly equal to 1,
> and for a lot of fields, gen0 & gen1 at 1 given point have the
> same magnitude (thinking of SWdown, LWdown, airTemp) so that
> the difference gen0-gen1 is often much smaller than gen0 or gen1,
> and the 2nd method will minimize the truncation error that
> we get on "fac".
>
> For this runoff problem (and I guess, it's not specific to the runoff
> but applies to any field when 12 identical months is compared
> to a single reccord case, period= 0), this change remove the  
> differences.
> Might also reduce the platform/compiler sensitivity of the output ?
>
> Cheers,
> Jean-Michel
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel

---
Dr Patrick Heimbach | heimbach at mit.edu | http://www.mit.edu/~heimbach
MIT | EAPS, 54-1518 | 77 Massachusetts Ave | Cambridge, MA 02139, USA
FON: +1-617-253-5259 | FAX: +1-617-253-4464 | SKYPE: patrick.heimbach