Non-integer timestep experience?

[Bjarne Büchmann]

Hi all,

 

I wonder if anybody has experience running with a noninteger (in seconds) timestep in GETM.

I am asking because we are setting up a hires model including deep fjords, expected resolution dx=dy=200m, depth down to about 900m.

I have a 600m “coarse” setup running as an intermediate step between outer basin-scale models and the inner hires model. And even for the intermediate model, I need to go down to dt=3s (CFL condition).

 

If you have experience with running with dt as a non-integer seconds value – or even fraction of a second, then I would very much like to hear your experience.

In principle, timestep (defined in time.F90) should be REALTYPE. I would still aim to get an ingeter number of time steps per single (or every few) minutes, but it would be nice to know that we can go lower than 1s if push comes to shove. (And that we don’t have to go all the way to 1s if CFL condition implies, say, dt<1.8s).

 

As always, any and all input is welcome. Thanks.

 

MED VENLIG HILSEN / BEST REGARDS
 
Bjarne Büchmann, PhD

Forsvarsministeriets Materiel- og Indkøbsstyrelse

Danish Ministry of Defence Acquisition and Logistics Organization
Joint GEOMETOC Support Centre

Lautrupbjerg 1-5 - DK-2750 Ballerup

TLF.   +45 72 81 56 24
MOBIL +45 40 78 73 20

 

---

This email was scanned by Bitdefender

[Ulf Graewe]

Hi Bjarne,
I run a Waddensea setup with a non-integer timestep:
&time
   timestep = 2.5,
   timefmt = 2,
   nlast = 10000,
   start = '2018-11-01 00:00:00',
   stop = '2018-12-01 00:00:00',
/
runs smoothly! (as smooth as a setup with extensive drying/flooding and high velocity gradients can run)

I once also run a setup with a fractional time step, such that dt*M was an integer fraction of the M2 tide. No issues with that!

However, I never checked in detail if there are odd things happening with the output, in case that mod(dt*M,3600) does not fit exactly.

Cheers,
ulf

This email was scanned by Bitdefender --

---
You received this message because you are subscribed to the Google Groups "GETM-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to getm-users+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/getm-users/e41fe8bffbd749bd8cab4a7bc4101508%40fcoo.dk.

[Knut]

Hi Bjarne,

yes it works, but needs some fixes waiting in the chain of branches to be merged. I will prepare a pull request and discuss with you on getm-devel.

Cheers, Knut

[Bjarne Büchmann]

Hi Ulf,

 

Knut wriote:

> yes it works, but needs some fixes waiting in the chain of branches to be merged.

 

Which code base did you run on (repo/branch info)?

 

MED VENLIG HILSEN / BEST REGARDS
 
Bjarne Büchmann

Joint GEOMETOC Support Centre

MOBIL +45 40 78 73 20

To view this discussion on the web visit https://groups.google.com/d/msgid/getm-users/88972f35-900d-c5d3-604d-da29e89d1c1e%40io-warnemuende.de.