SWMM vs. MOUSE

A very interesting debate this has generated with some good comments on SWMM, XP- SWMM and MOUSE. To add my two cents worth, we have recently switched to the MOUSE model as the primary modeling tool in the City of Edmonton for sewerage and drainage work. Why? Because we wanted to be able to simulate real time control devices. We have a number of level activated structures in our system, and this is continually on the increase. Therefore we needed to have a tool to model this adequately.

That is the basic reason for the switch. Since then we are getting almost identical results from SWMM-EXTRAN simulation of our sewerage system as we achieve in MOUSE. We do not notice the major differences that have been commented on in earlier messages. However we were modeling with SWMM using a 1 second time step to achieve stable results. We have experimented with MOUSE and have been able to select a 10 second minimum time step and achieve stable results without compromising accuracy.

With SWMM-EXTRAN we were also getting the spikes which have been previously commented on. We do not have this occurence with MOUSE. However there is definitely not a smoothing function in MOUSE. We can achieve results which are anything but smoothed.

For those who are interested, CDM are working with DHI to produce an direct interface between SWMM RUNOFF Block results and the MOUSE model. We are currently testing this utility as we still produce hydrograph results using the SWMM binary interface file (although not from RUNOFF Block). Initial results appear satisfactory. This is also intended to be developed into a utility which can facilitate long term simulation within MOUSE, without actually simulating the normal daily base flow.

Those are my comments for what they are worth.

Chris Ward
City of Edmonton, Drainage Branch

I would like to know the following things:

1. Has anyone done a comparison between SWMM and MOUSE and if so, what are the key difference.

2. Can SWMM be used for the modeling of combined sewers containing storm overflow chambers?

Frank Fisl

I'm not very familiar with MOUSE however I can tell you that SWMM is certainly capable of simulating combined sewers and diversion structures. I have used it to simulate a combined sewer system that services and area of 25 sq. mi. and has over 100 diversion structures, 13 CSO pumping stations, and 5 collection pumping stations.

I have done a comparison of MOUSE to SWMM in the past year. My main concern was with the EXTRAN pipe model. Some of my complicated models have several hundred pipes and they give me a real problem with sudden spikes in the computed hydrographs and water levels. I know when and why the problem occurs - it happens as water levels step into surcharged flow - but I have been unable to satisfactorily solve it and this lead me to look at the potential of other models.

I tried XPSWMM and found their new surcharge algorithm to be significantly better than regular EXTRAN but the flows and water levels were much lower so I was worried that the problem may be merely masked - it is always better for lower flows in the regular SWMM, too. Also, I was worried why the answers were lower. Note - in some, simpler models the problem did not occur and SWMM matched XPSWMM.

The above is definitely not a numerical instability problem - it is a problem of EXTRAN not handling the suddenness of the step from open-channel pipe flow into surcharged pipe flow. I think that a real pipe might actually experience a sudden pressure spike as it stepped into surcharge and maybe EXTRAN is starting to simulate something that is actually there but it can't quite handle it - it computes such a severe spike that all the nodes and pipes have the Maximum value arrays filled with these spikes and thus the summary tables are rendered useless and I have to check every node and pipe's results and manually find the real peak values.

I could not do a direct transfer of a SWMM file to MOUSE; MOUSE has totally different input formats, so comparison of a large pipe model was not possible due to the workload required. Simple models were tried but the spike problem was not there and so all three SWMM, XPSWMM and MOUSE gave exactly the same answers.

I then cooked up a fake model that did have spikes similar to my real large model just with a manageable number of pipes. This model was very spikey in SWMM EXTRAN, less spikey in XPSWMM EXTRAN and perfectly smooth in MOUSE. MOUSE results are almost too good to be true. My problem now is that both XPSWMM and MOUSE matched each other well but were significantly lower than the regular SWMM EXTRAN (both flows and water levels). This means that either regular SWMM EXTRAN is over- estimating or the other two are under-estimating.

All of my trial work used the same runoff hydrographs for each model so only pipe routing was tested and there is no question of results being related to differences in inflow. It is possible that the inflows are being read or used differently.

I had a very good set of calibration data for my original (regular) SWMM model and this leads me to wonder if these results are the ones that are O.K. and XPSWMM and MOUSE are losing water in the process of getting smoother results. No proof just a suspicion and not a good one at that. This leads naturally to a point made well and often by Bill James and others that calibration is necessary and without it all models are wrong.

Seems to me that if you use the same inflow hydrographs that all pipe routing models should yield the same results but this is obviously not the case. Anyone know why not?

Since the spike problem is a function of flow and water level magnitude and since both XPSWMM and MOUSE computed lower values I can not conclude that they wouldn't have also computed spikes if they had gotten the same degree of surcharge. My theory is that XPSWMM might have been just as spikey because there was some spikeyness to its results at the lower values and MOUSE probably would remain smooth because the test run was perfectly smooth, that is it had no trend towards spikes.

This issue needs more work but I am out of time and money to continue. I would very much like to hear from anyone who has had the spike type problem and if they somehow solved it or even have any ideas I might try.

I did a complete review of the MOUSE manuals and it seems that the surface runoff computations are similar for one of the options - I was able to exactly duplicate a sample SWMM catchment calculations. My conclusion was that MOUSE can easily be made to duplicate SWMM, it has some features that SWMM doesn't and SWMM has some it doesn't with SWMM RUNOFF having the most. For surface runoff I think SWMM is technically superior especially considering the extensive North American experience with it, the ready access to the authors and other users plus the database of calibrated models that all cities have built up. There is no reason to even consider switching and much to be lost if you did.

As I rambled on about above, there may be a case to switching to the MOUSE pipe routing model to avoid problems with EXTRAN spikes but even this can not be clearly seen from my work. MOUSE does give infinitely better results as far as ease of use of the results goes (because there are no spikes) but if they do prove to be lower than SWMM because of an error you are stuck. At least with EXTRAN I have 22 years of experience with it and will eventually be able to examine the code and try to solve the one problem I have with it. Access to all of the SWMM users via this group plus Wayne Huber (a big plus)and the XPSWMM people will be a major help - maybe someone else will even solve it and share it with us all. Going offshore to solve even a serious problem with the spikes seems too risky for my taste, it was worth trying and may be a future consideration but certainly not something to rush into. My clients would likely kill me for even suggesting it.

I did find the MOUSE people to be very supportive and knowledgeable. They responded to every email immediately and completely. They are in Denmark so an email sent before 9:00 am would get a response before lunch. The feel of their manuals and general approaches was all very familiar, their English is better than mine and there were no difficulties in terminology at all. They are just like our group only different. They have even offered to work out a way to readily convert SWMM input files so I can compare my calibrated 700 pipe model but I have not had time to do my part as yet. If MOUSE pipe routing got stable hydrographs and tied into the SWMM RUNOFF block interface files and I could prove that the results were correct then I may have died and gone to modellers heaven. MOUSE costs are slightly higher than XPSWMM but not much and I bet they might dicker. They have an output processor, called MIKEVIEW, which is much like MTVE and XPSWMM's graphical routines which I think is the best of the bunch although some of MTVE's features are not there and I would ask the MOUSE people to add them.

I think that MOUSE can do your combined sewer problem but haven't tried any of their complicated stuff. This sort of modelling problem is always solveable with any model. If not exactly in the manual you just cook up an equivalent method of analysis or if worst comes to worst contact the authors and they build it in for you for a fee or for free and then they call it a new feature in the next release. MOUSE has several optional packages. All of the options would probaby more than double the cost.

I have always found that the cost of software was an insignificant part of the cost of doing a large study. Your labour charges will be at least 10 times the cost of software and usually much more. Software cost is a big problem however because you need to evaluate it before you can spec it in a proposal and most bosses will not authorise $5,000 to $20,000 gambles. I work for myself and am absolutely broke from software expenditures - could have bought a few cars. I note that the MOUSE people sent me two fully working evaluation copies for free but I must return them soon, so if you have any questions get back to me right away.

Reinhard Sprenger

I tend to agree with most of Richard's observations, particularly about the importance of calibration. However it is dangerous to perform simulations for model inflows well outside the range of those for which the model was calibrated. Of course, users generally want to be able to simualate large recurrence interval storms. From my experience, while EPA's EXTRAN may simulate less severe storms fairly well, it does tend to exaggerate surcharging for extremely intense events. This is frequently due to too long a time step. I have shortened my time step to as little as 1 second to maintain stability and a reasonable level of surcharge. But even then larger models may spike erratically. XP-SWMM's Version 4.0 (32-bit) solution techniques do seem to predict more realistic results. I have found the enhanced explicit solution technique to work particularly well for storms that caused EPA's EXTRAN to surcharge unrealistically.

Steve Sticklen

I have much experience of HydroWorks and SWMM, and some of MOUSE. HydroWorks and MOUSE are more or less similar. They are just two concurrent software of this sort. They are much user-friendly than SWMM. As to the numerical engine, I don't know which one is the best. However, I have never been disappointed with HydroWorks!

ir. Mingchuan Ruan
M.Ruan@ct.tudelft.nl

As the discussion indicates, both SWMM and MOUSE have dynamic hydraulic engines (in that they solve the full Saint Venant equations of gradually-varied, unsteady flow). For SWMM the engine is EXTRAN and for MOUSE the engine is MOUSE HD (hydrodynamic module). They differ in the numerical techniques used in the hydraulic engines: EXTRAN uses an explicit technique while MOUSE HD uses an implicit technique. Speed is the basic issue here with implicit being faster, generally. One reason MOUSE has a different solution technique (and other time- saving techniques) is to better accommodate continuous simulations which seem to be the norm in many countries.

Regarding the spiking discussion: MOUSE does not use any forced or artificial smoothing techniques. The bulk of the reason that EXTRAN and MOUSE HD seem to perform differently in the same situation is the difference in numerical solution techniques (explicit vs. implicit).

Both have the ability to model hydrology. SWMM does this with the RUNOFF block and MOUSE has an add-on module (NAM). RUNOFF and NAM share some approaches and differ in others.

In general terms (not knowing the specifics of your application), both will model combined sewers with storm overflow chambers.

MOUSE, developed by the Danish Hydraulic Institute in 1985 and subsequently improved, has several additional capabilities that have been developed largely in response to European efforts. These include the ability to model sediment transport and deposition in collection systems, ability to model water quality components in-system, ability to model Real Time Control techniques, and integration with receiving water quality and wastewater treatment works models. MOUSE also includes an add-on module to integrate with Geographic Information Systems (MOUSE GIS). MOUSE has a performance track record on par with SWMM having been applied on thousands of projects over the past 14 years (DHI has over 1000 licensed users of MOUSE worldwide).

DHI's user interface for MOUSE is called MIKE and includes extensive graphical pre- and post-processing capabilities to build model files and interpret model results. With the database and presentation capabilities available on the desktop today, the user interface for hydrologic/hydraulic models is increasingly important to the process of modeling. This is where proprietary models such as MOUSE have excelled over the last few years.

As Chris Ward alluded to, Camp Dresser & McKee is collaborating with DHI to develop a full version of their MIKE interface for SWMM which will include the GIS interface capabilities of MOUSE GIS. This product, MIKE SWMM, will be formerly introduced later this month (at WEFTEC) and will be available in December. MIKE SWMM will use the public domain version of SWMM and will be a full 32-bit interface (Windows 95&NT). We think this provides the best of both worlds for users who would like a state-of-the-art interface for the public domain version of SWMM as well as seamless linkage to MOUSE. CDM will continue to support and enhance the public domain version of SWMM as we have in the past.

Regarding the debate of proprietary vs. public domain engines: I propose that people on both sides of this issue are going to have to agree to disagree. European interests decided a long time ago that engines and user interfaces could be proprietary and that users would decide promote the best models with their purchase (survival of the fittest). North America has always viewed the engines as something to be advanced in the public domain, although user interfaces and model variants (XP) are often proprietary.

The applicability of a particular model depends on many things specific to the user and proposed application. A decision regarding public domain vs. proprietary must be made while weighing other factors and needs. The only universal truths are that the model must be proven (both MOUSE and SWMM are) and the model developed must be calibrated and verified.

Robert W. Brashear
brashearrw@cdm.com