Model size and complexity - Bigger is Better?

[Andrew Walker]

As computer manufacturers produce faster and ever more powerful PCs and Servers, so engineers and hydraulic modellers are able to produce larger and ever more complex models of their drainage networks. Whether or not you subscribe to the ‘Bigger is Better’ theory, there is always genuine interest in the size and complexity of models that have been produced and how they have grown as time goes by. With this in mind, it would be interesting to learn a little about the models you have created.

1. What type of study was it (i.e. Drainage Area Plan/Study, Surface Water Management Plan, Sewerage Master Plan, Integrated Catchment Study, etc, etc)?
2. How big was the model you created in terms of total catchment area and the number of pipes/manholes needed to meet the requirements of the study?
3. If the model contained 2D calculations, what area of the model was covered by the 2D zone and approximately how many elements were in the mesh?
4. Which modelling software did you use and when was the model created?

This is a topic that everyone can contribute to, so let’s hear about the models you remember well, whether that’s because of size, complexity or the shear difficulty of creating the model in the first place .

[Kristian Ravnkilde]

We have four models, 2 in IWCS and 2 in ICM (one fairly small). Key figures for the 3 largest are as follows:

	Upper Cole	Birmingham SWMP Northern		River Rea
	IWCS	IWCS		ICM
Nodes Total	5498	10798		Nodes Total	15214
Nodes Manholes	5491	10721		Nodes Manholes	14712
Pipes	5054	10792		Pipes	14556
Rivers	409			Channels	159
				Rivers	189
				Bridges	75
Pipe Length (m)	292328.53	643991.38		Pipe Length (m)	928246.19
Pipe Size (mm)	100 - 40000	100 - 7163		Pipe Size (mm)	100 - 28900
Subcatchments	4730	6176		Subcatchments	9752
Subcatchments Total Area (ha)	9487.65	5637.59		Subcatchments Total Area (ha)	18102.28
2D zone	1 no., 698ha.	1 no., 1354ha.		2D Zones	1no., 2956ha.
2D elements	163602	211027		2D Elements	510454

NB The largest "pipe sizes" are representations of open tanks and simplified overland flow channels.
Upgrading our PCs to decent Core-i7 with 16GB RAM has helped enormously with run times!

[Robert Dickinson]

Bigger is sometimes better! Modern models are also more complex internally (more processes) and often much longer in simulation time. It is not just the number of elements but what each element is simulating.

I have tried to develop a complexity index for models in the past to compare a 2010 model to a 1990 or earlier model. I tried to make it both additive (nodes + links + subcatchments) and multiplicative (duration + internal complexity such as pollutants and landuses). I am thinking in terms of SWMM 5 and InfoSWMM but generally it can be applied to ICM or IWCS models with some adjustment of the parameters.

The complexity index adds up the of raingages, subcatchments, junctions, outfalls, dividers, storages, conduits, pumps, orifices, weirs, outlets, control curves, diversion curves, pump curves, rating curves, shape curves, storage curves, tidal curves, time series, patterns, transects, hydrographs, aquifers, controls, climate objects and snowpacks objects. The complexity index is then multiplied by the number of pollutants for all subcatchments, junctions, outfalls, dividers, storages, conduits, pumps, orifices, weirs and outlets and it also includes the number of landuses times the number of subcatchment objects.

I would suggest the following changes to add in 2D elements + runoff surfaces

Index = (( 2D Objects + Nodes + Links +Link RTC Elements + Outfall Elements Subcatchments * Runoff Surfaces + Groundwater + RDII + DWF Elements + SnowMelt Elements)) * Pollutants * Duration

A 1000 day model would be 1000 times more complex than a 1 one day model, a model with groundwater and snowmelt would be three times as complex hydro-logically as a model without snowmelt and groundwater.

[Kristian Ravnkilde]

Maybe you and Andrew should agree what information to ask for, and then all respondents can (try to) provide it.

[Robert Dickinson]

Hi Kristian, I liked your informative list and I think your comment is more in line with what Andrew was looking for in this thread. I was just suggesting that model "bigness" or complexity is also a function of internal processes and duration as well the number of elements.

[David Garcia]

I am a firm believer in the "Bigger is Better" approach to building sanitary models. They are easier to build, calibrate and update. Our modeling methodology is based around building full pipe models with individual parcels as subcatchments. If you have decent GIS data available, then the build process is relatively simple, but can be time consuming. The two largest models I have built are for Little Rock Wastewater and the City of Hot Springs, both in Arkansas, United States. Here are their stats:

Little Rock:
Nodes: 27,350
Conduits: 37,348 (4"-72" pipe)
Total footage: 5,441,920 ft
Subcatchments: 1,655
Pump Stations: 5
Offline Storage: 78 MG

Hot Springs:
Nodes: 21,247
Conduits: 15,611 (1"-60" pipe)
Gravity sewer footage: 2,264,130 ft
Pressure sewer footage: 1,269,178 ft
Subcatchments: 24,504
Pump Stations: 270
Grinder pumps: 5,352