We use cookies to make your experience of our website better. You can delete or block cookies, but some parts of this website won't work without them. By using this website you accept our use of cookies.


25/02/2016 09:22:28

Dear All,

I have been asked to model a small field drain/ditch in order to assess the risk of flooding.  The ditch in question has a catchment of 0.65kmand therefore it is possible to calculate the peak flows.  However, the dimensions of the ditch are rather small- approx 0.60m with and 0.5m deep.  The model needs a 2d elemant so that we can define the flood extent.  My question is whether a linked ISIS-TUFLOW model would be appropriate in this case?  Obvioulsy the grid size will need to be appropraite for the scale of the assessment, but is there a 'limit' to how small a 1d2d model can be?  Or would it be preferable to model it purely in the 2d and use a z-line to define the ditch channel?

Any thoughts would be most appreciated.



25/02/2016 11:20:49

Hi Laura,

I am actually doing something very similar at the moment for a small watercourse at the top of a catchment. The US end is very much just a drainage ditch. 

The approach I am considering (as I am only interested in flood extents) is to use a 1D only model with extended flood plains (say 50m each side). I plan to run this with the inflow being provided by lateral inflow units from a ReFH boundary. This should produce a good representation of the in bank flows, which for my site I believe to be the majoirty due to the small nature of the problem. I will then drape the max stage results onto a TIN for the modelled watercourse, prior to intersecting with a LIDAR ground grid to produce a flood extents map.

I feel this method is justified as the catchment/ sourrounding are entirely rural, and the produced flood extents will just be used to inform where development may require extra mitigation/ may not be feasible.




25/02/2016 12:20:48

Hi Laura,

It is possible to model a ditch that is much smaller in width than the 2D cell size. For a 1D-2D model, it may necessitate representing part of the floodplain within your 1D cross-section to ensure that the model does not under-represent the amount of storage within the floodplain. Note that instead of digitising two HX lines (one each for the left and right banks) it is possible to digitise a single HX line to link to the 2D domain. This approach would be best suited where the left/right bank elevations are approximately equal.

For a 2D only model, along with a zln, you can make use of the Storage Reduction Factor and Cell Flow Width features within TUFLOW that allow for the storage of the cell and the flow width to be reduced. This would allow for a better approximation of the channel conveyance without having to reduce your cell size significantly. 

I do however, second Ben's comment that a 1D only approach shouldn't necessarily be disregarded. 2D and 1D/2D models are very powerful tools, however there are situations where a 1D model can more than meet the objectives of the study. Hopefully the above provides you with enough information to make the best decision for your project. 


Steph (TUFLOW UK Support)


25/02/2016 14:57:21

Thanks both- your comments are very much appreciated and definitely helpful.



Lei Yang

25/02/2016 15:32:54

Hi Laura, interesting question and interesting answers!

Generally speaking either a linked 1D-2D model or a 2D model can be used to conduct the small ditch and associated sub-catchment flood prediction. In some cases a 1D model could do as well.

However modelling such a small channel, the 1D-2D approach is likely to have instability issue which could easily induce huge amount of work on just maintaining a healthy run under each flood event.

If 1D isn’t the choice for your case, I suggest that you carry out a pure 2D modelling study while you will have much less instability issue and the inflow could be applied as a source term. The ditch dimension tells that the grid size of a 2D need to be 0.2m, in order to capture the channel’s geometry. The data availability becomes dominate rule to build such a fine grid model. The model is only reliable if the relevant resolution of data is available.

Hope this helps.


Sign in to create or reply to forum posts.

Privacy & Terms