Forum | increasing dflood causes model to crash at start

You are in the forum topic » increasing dflood causes model to crash at start

If this is your first visit, be sure to check out the forum rules. You need to register and be logged in before you can post. Go to the forum area 'Running models ' or to the 'top level forum'

Quickview

Author

Thread

tmashby

Member

Location:
Registered: 27/08/2008

increasing dflood causes model to crash at start
Recently, a colleague and I have found that sometimes when building and running an Isis model that increasing dflood can cause an otherwise ok model to crash right at the start of the run.

On the models where we have seen this, there are no sections where the water is even close to out of bank, let alone rising to a significant number of metres above the section.

In the most recent example I changed the value from default to 10 (chosen at random as ultimately the model will be linked to 2D) and the model crashed. Changing the value to 6 for example allows the model to run.

Any ideas why this should be?

02/03/2010 16:45

Link - Ip: Logged - Quote:

adamsk

Member

Location:
Registered: 27/08/2008 Warnings: 0

The only thing I can think of is that changing dflood will change the resolution of the section property calculation:
ISIS will calculate section properties (e.g. area, conveyance, width) at each elevation in the section data.
It will also calculate them at any intermediate point if the vertical distance exceeds 2% of the total section depth (including dflood).
Hence the above is dependent on dflood.
In your situation, your minimum resolution is increased with dflood=10m, and therefore you may have different conveyance, etc. values calculated at certain points which could be enough to tip the model over the edge.
Konrad

02/03/2010 17:39

Link - Ip: Logged - Quote:

adamsk

Member

Location:
Registered: 27/08/2008 Warnings: 0

3rd paragraph: That is to say "...vertical distance exceeds 2% between any two consecutive points (once they have been arranged in order of elevation)"

02/03/2010 17:41

Link - Ip: Logged - Quote:

PHA

Member

Location:
Registered: 27/08/2008 Warnings: 0

I would speculate that this is going to have the greatest impact at really low flows. As such, increasing base flows in the model could help avoid this problem (the failure to start with altered dflood that is).

Further though, it seems to me that the advice should be to have dflood as small as is reasonably possible in order to give best resolution to a model (just setting it to, say, 50 is a really bad idea (I've seen it done!)). Less of a concern for a model which is about to be linked to some floodplain, so long as you remember to turn dflood back down again once linked.

02/03/2010 17:56

Link - Ip: Logged - Quote:

adamsk

Member

Location:
Registered: 27/08/2008 Warnings: 0

I concur with PHA's opening sentence and maybe convert his speculation into hard evidence.

Indeed when we changed the default from 10% to 2% many years ago, we found that a few models would suddenly crash at low flows, hence the need for the pcmxvd parameter (sorry: I felt constrained to fit its essence - percentage of maximum vertical distance - into 6 letters).

The other thing I forgot to mention is that you could avoid this influence of dflood somewhat by adding extra cross section points in yourself, particularly if you have big vertical gaps at low depths (e.g. if you have a rectangular channel, add several extra points up the side of one of the walls).

02/03/2010 18:09

Link - Ip: Logged - Quote:

muhammad saddique

Member

Location:
Registered: 27/08/2008 Warnings: 0

03/03/2010 13:55

Link - Ip: Logged - Quote:

Robin Green

Member

Location:
Registered: 27/08/2008 Warnings: 0

On a similar topic, I have been given an ISIS-TUFLOW model whose dflood has been set at 75. This value seems to be well in excess of the standard range, but could anyone tell me what are the likely side effects of this - could the model be underestimating channel capacity for example? Or is the presence of the 2D domain sufficient to negate any significant impact?

Thanks in advance.

31/08/2010 15:31

Link - Ip: Logged - Quote:

adamsk

Member

Location:
Registered: 27/08/2008 Warnings: 0

Side effects would be that within the channel, it could overestimate conveyance locally due to coarser interpolation, but this would only be between specified cross-section z-values (or "y" as they are called in the table).
I shall reiterate: conveyance is calculated "exactly" at all user-input elevations within each cross-section.
It would be better IMO to use a higher pcmxvd to force coarse interpolation if needed, than a large dflood.
If the water did enter the "dflood zone", I'd check that it didn't go too far into it.
Alternatively, a high dflood may have been used to overcome a temporary instability.

31/08/2010 17:32

Link - Ip: Logged - Quote:

Quick reply

You need to login before you can post.
If you are not registered on the site, register here.

Forum topic area

You are in the forum topic » increasing dflood causes model to crash at start

Latest forum posts

Quick search

RSS Forum Feed