simple RUNOFF question

Sipho,

You have asked some questions that probably deserve lengthy answers. However, given the nature of this form of communication, I will try to be brief.

The condition in the original post of only 1/4 of the impervious area being directly connected is generally much lower that what we normally see, particularly for an area that is 80 percent impervious. So, this component is not usually nearly so large. Dealing with NDCIA by reducing the amount of area over which infiltration can occur (i.e., reducing the infiltration parameters by the appropriate fraction) is normally a sound representation. Introducing another runoff plane to represent NDCIA is something that has been discussed in the past, and the conclusion has been that nothing much would be gained. It would be interesting to hear from the rest of the group if this conclusion has changed.

Prior to calibration, it is a good practice to take representative samples of, say, a few acres from the predominant land uses in your area and perform detailed calculations of DCIA and NDCIA. This process is usually performed with aerial photos. These detailed calculations serve as a guide for what reasonable ranges of parameters adjustments (DCIA and infiltration reduction factors) should be during the calibration. If the calibration data suggest adjustment outside of these reasonable ranges, then the next step is to more carefully examine the calibration data as opposed to changing parameters to values that are difficult to reconcile with what you can observe of your watershed. For example, you may estimate your DCIA to be 40 percent. However, the data from a single calibration event may suggest that your DCIA is 70 percent. In this case, it would almost certainly be the case that something is misleading with your calibration data, because your estimate (depending upon how carefully it was made) of DCIA is probably no worse than 10 percent (30 to 50 percent DCIA).

Dear Rob,

Such changes should be preferably implemented as a new user option, so that our existing models can reproduce their previous results and we do not have to recalibrate our existing models. A drawback of this approach is that the number of user defined options may grow enormously over time confusing the heck out of casual modelers. Every new capability that sacrifices model simplicity must be evaluated carefully before implementation. Recent popularity of the Linux operating system is a good example of what I am talking about. Simplicity my friend, simplicity!

Uzair M. Shamsi, Ph.D., P.E.
shamsiu@usfilter.com

It seems that the discussion is more about model calibration than simply RUNOFF.

Land use data is used to obtain initial estimates for hydrologic parameters, including %Imp, i.e. DCIA. If everything were perfect, we wouldn't have to modify any parameters. DCIA is one of the most influential calibration parameters for RUNOFF. As such, it will be modified to first calibrate to total runoff volume, and then perhaps for timing and/or magnitude of peak discharge.

Kaveh brings up a good point, that the "impervious" parameters (n, storage) are not used for NDCIA, and that there is no way to tell SWMM what NDCIA is. If the concern is obtaining the appropriate detention storage, as DCIA is decreased, impervious storage could be increased proportionally. (Detention storage is an estimate at best to begin with.)

It has been my experience that initial estimates for DCIA have been reduced by 50% or more during calibration. Most clients are satisfied with the modified parameters, as long as they can be justified (by the calibration) and explained .

And as we all know, an uncalibrated model isn't worth a hill of beans.

Regards,
Bill Walker
wkw@dhigroup.com

Point well taken. I think the implementation could be as simple as assigning a subcatchment ID to the NGTO parameter of the upstream subcatchment. In this case no extra parameters would be needed.

Sincerely,
Rob James
Computational Hydraulics Int.

Hello

Now that I am back from an extended trip I read through much of the above thread, which has certainly been central to the interests of my grad students. My humble and unimportant opinion is that the best source of info on how the overland flow kernel works in RUNOFF is the FORTRAN source code itself in the version you are using. But it takes weeks of study. Craig Kipkie did a careful analysis of v 4.4, Wayne's issue of Aug 8 1997. It probably has not changed much. Chris Kresin also looked at it. In Craig's dissertation, Chaps 3, 4, and 5 provide details of how the source code can be modified to include runoff from additional surfaces such as permeable pavers, which conceptually fit between DCIA and pervious areas. If Craig agrees I can post more of his dissertation on my website.

However there is a hugely detailed, illustrated explanation of these issues by Rob in the wizard-driven program known as "PCSWMM for Permeable Pavers", which has only recently been released. It is a rather big program and available free of charge (to nice folks like you). I have posted it at www.swapdrive.com. The username and password (both) are/is "comhydint" (for CHI, they need 6 characters). I should caution you that I have not attempted to download and install it from swapdrive, and there may be difficulties when several folks try it simultaneously (the total size of the zipped files is I think ca. 20 Mb). Perhaps some independent person who has run the program could comment on its value before you undertake the effort, especially if your link is slow. The program is also available on CDROM for nominal mailing and packaging costs, depending on need and numbers.

Finally at the risk of being self-promotional I hesitatingly remind interested folks who wish to investigate process performance without reading FORTRAN source code that at least one shell was specifically written to provide automatic ranking of the sensitivity of all parameters in all modules and all versions of SWMM. Maybe other decision support shells do this equally well. Finally please note that I have an interest in that shell (PCSWMM), so again be wary of postings like this, and seek informed outside opinions.

Thanks for an interesting and important thread.
Bill James

Hi,

I have been wondering about limitations of RUNOFF with respect to the design of small- scale runoff controls. For example, it seems that RUNOFF could have difficulties representing:

a) disconnected pervious areas (e.g. disconnected downspouts passed over grassed areas) (although I note Brett Cunningham's 'fix' of modifying the infiltration parameters to deal with this).
b) small infiltration devices (e.g. infiltration trenches)
c) small detention devices distributed around a catchment (e.g. small constructed depression areas). We might not want to model each device separately.

I think "we" need to either

1. Modify Runoff so that it can deal with such controls, and model each control separately and on a fine scale. Not very practical for a 20 square mile catchment, but potentially useful for small catchments

2. Figure out how to aggregate the effects of such controls.

3. 1 followed 2 based on simulation.

I am not happy with resorting to 'calibration' except perhaps insofar as it could be applied to working out aggregation parameters for application to new catchments. The whole point is that we want to use our fantastic modeling capabilities to DESIGN small-scale controls rather than evaluate them after the fact.

Has anyone done work in this area? Any more useful suggestions for 'fixes' for a, b, and c? Or do we just give up and say that Runoff was never intended to be applicable to small-scale controls, and just limit ourselves to designing conventional drainage systems (pipes and detention basins)?

Cheers,
Sandy Elliott
National Institute of Water and Atmospheric Research (NIWA)
Hamilton, New Zealand

Rob,

Surely as a first step the runoff from the NDCIA could be assigned to the pervious area within the same catchment. That would save setting up the connectivity between subcatchments. Admittedly this does not allow as much flexibility as the technique you suggested.

Cheers,
Sandy Elliott
National Institute of Water and Atmospheric Research (NIWA)
Hamilton, New Zealand

Sandy,

I believe that the issue of modeling controls for small areas within a model with coarser discretization is more than a RUNOFF limitation issue--it's a general modeling limitation issue. To model small features with a relatively high degree of confidence (and certainly to design them), the model needs to be at a scale that discretely represents the area upstream of the feature. Anything else is mostly likely too approximate for design. The theory behind RUNOFF holds up just fine for small areas.

There are several approximate methods aimed at addressing your concerns that are fine for watershed-level issues (see Don Waye's thread), but I do not see a substitute in the near future for design-level issues for RUNOFF or any other existing models.

Brett

Hello

For those who asked, as promised, and with thanks to Craig Kipkie for permission, I have rather hurriedly posted most of his dissertation on my web page:

www.eos.uoguelph.ca/webfiles/james

Click on "My postgrads" well down the 2nd column then scroll to T52, and click on the title.

Bill James

[hawstom]

Six years later, for subsequent readers, in version 5.0 of SWMM, it appears that options to improve the modelling of NDCIA have been added.

The current subcatchment element includes the following properties:

Subarea Routing: Choice of internal routing of runoff between pervious and impervious areas

Percent Routed: Percent of runoff routed between subareas (within the catchment).

Can anybody comment on how the Percent Routed property is used in the model?

Tom Haws
Gilbert, Arizona

[swmmguy]

swmm.