View Full Version : Direct Rainfall Modelling in ICM

Duncan Kitts
July 16, 2013, 07:38 AM
Whilst talking about direct rainfall modelling using rainfall directly onto the 2D mesh in a training course, the discussion went to what approaches people take to avoid double-counting of rainfall? Within ICM there are a number of options. Most models have existing subcatchments which have a 1D representation of rainfall runoff, is the general approach when applying direct rainfall to delete these subcatchments (or remove the stormwater component?) or use the 'Outside Subcatchments' option in the 2D zone? If so what about the foul component, are these added to the model?

Also how do people represent buildings? Voids, Roughness Zones or porous polygons? If porous polygons are used, is direct rainfall applied to these or is the 'No Rainfall' option used? If the 'No Rainfall' option is used, or the buildings are represented as voids, is the rainfall that falls on the roofs of buildings represented in other seperate subcatchments represented the building footprints?

It seems, talking to people and looking at models that there is no common approach, would this be a fair assumption? ICM has a number of tools and a number of options to what approach can be taken, I was just wondering whether some are favoured over others?

July 18, 2013, 03:30 AM
You through a ball which is just what I was thinking during my two ICM modelling projects.
Direct Rainfall
From my view point, ICM model would be created based on existing Infoworks CS model normally. If we choose to delete the storm water of subcatchment (the subcatchments should be kept in a way to allow the wastewater discharge), we will need to deal with several issues carefully such as:

The foul system;
Roof runoff in combined and foul system (the concentration time of roof runoff is completely different to the time of runoff from 2D terrain surface).

My view on ICM in comparison to other well-known modelling tools is that the Area Take Off in subcatchment in ICM enables most appropriate and fast runoff calculations which form the value of choosing ICM. I prefer to use the ‘Outside Subcatchment’ option which simply do a good job and is least time consumed.
I reviewed an ICM model which covering a city in UK where voids were used to represent building. My inspection revealed that the voids had led to:

Generate a 2D mesh having tiny cells as small as 2cm;
Perform the model run for a 6 days on a high performance GPU pc for just 1440min simulation duration.

My conclusion is that voids should not be used to build city model that covers a large 2D area. During another project I built an ICM model in a city using Roughness Zone to represent the building and the model works perfectly.


Andrew Walker
July 24, 2013, 03:47 AM
Tiny elements need to be avoided at all costs. Even a few small elements in a large mesh have the potential to cripple your run-times (your example is a classic case).

There is a very easy way to detect all the small elements in a model. Validation will report on the smallest element you have, but a simple Theme will highlight them all in the GeoPlan for you. The theme shown below will paint all the small elements in different colours and place a large arrow against each. Even if the element is too small to see in the context of the GeoPlan Zoom Level, the arrow will still clearly show where the small element is.


The image below is from a real model where a small triangle has been trapped between the 2D boundary and the edge of mesh zones within the 2D area.


Note that this theme is based on a results item [2DTriangle.sim.area2d (Mesh element area)], but it only requries you to run the model for a single time step in order to provide the relevant results output to be themed (because 2DTriangle.sim.area2d is not a time varying output).

This Blog Article, written by Mike Reeves, is a very useful resource -> http://blog.innovyze.com/2013/07/23/2d-meshing-avoiding-small-elements-and-why-this-is-so-important-2/

July 24, 2013, 04:03 AM
Thanks Andrew. That was not our project and i was the model reviewer so i had no right to correct it. I knew that the 2D element in that model caused many problems because i am a professional modeller using various 2D modelling softwares with all kinds of expriences in terms of modelling.
I hope the modeller of that model could see the post and then learn the method that you just explained.

Kristian Ravnkilde
July 25, 2013, 03:12 AM
The other approach is to clean up the map layer/s you are using (preferably copies of them) to eliminate crossovers, nodes stupidly close together, etc before doing the meshing. MapInfo has simple tools (Snap/Thin), and I'm sure other GIS software has too.

Duncan Kitts
July 25, 2013, 07:36 AM
Thanks Andrew/Kristian, small triangles are a big issue in any 2D modelling but particularly in direct rainfall modelling. As I'm sure many have heard me rant about, it's best to proactive with dealing with small elements and think about the level of detail which is appropriate to your study area. The irregular meshing approach can represent a lot of detail (<0.01m) but it's just that to most applications this is too high (ie, at a city scale). Mike's blog post should be the first port of call for anyone doing 2D meshing with background layers.

Anyway, I don't want this thread to stray too far from the original topic as I think the approaches that are used to represent buildings and to avoid double-counting rainfall whilst conducting direct rainfall modelling is important and as far as I can tell there is little consistency.

Lei, whilst representing buildings as roughness zones what Mannings n value did you use and did you do anything to avoid rain falling directly onto these roughness zones representing buildings (ie, was each building represented as a subcatchment)?

August 2, 2013, 03:07 AM
The roughness manning n was 1.0 for building (it could be within a range of 0.8 to 2.0 from our exprience). All buildings were covered in subcatment.

March 20, 2017, 05:25 AM
Hi Duncan,
I realise this post is quite old, but I was searching for recommended values for the porosity of the polygons represented buildings? I am in favour of lower value such as 0.025, but it seems that in this case some of the water is trapped in the building and can't flow out. While bigger values like 0.4 the water flows almost free through the building.