Over-Constrained Solution?

I am modeling a large, low-lying, flat watershed with 221junctions. This watershed has significant historical flooding problems, and these flooding problems are reflected in the SWMM model. Approximately 40 percent of the nodes are surcharged during the simulation.

I have experimented quite a bit with ISOL, SURTOL, ITMAX, KSUPER and several SWMM engines to obtain a reasonable continuity error. SURTOL is the most consistent and significant variable, but to "improve" continuity, I have to use a comparatively large SURTOL (when working with the 10-year SCS Type II storm).

I am observing the following relationship between SURTOL and continuity error:

SURTOL Cont. Error (%)
0.002 10.34
0.01 7.96
0.2 4.25
0.4 2.43
0.8 0.58

The flows and heads look reasonable (although I'm still trying to eliminate one very small surcharge flutter), so I'm wondering why the results appear to be improving as SURTOL is increased. Is this possibly an "over-constrained" solution scenario?

Thanks for any help.
John Paine
John_Paine@URSCORP.COM

John,

You have presented an interesting finding about large surcharged models in Extran. What you have done by increasing the SURTOL parameter is to minimize the usage of the Surcharge Algorithm in Extran. Extran for the ISOL=0 and ISOL=1 solution tries to balance the flows at the surcharged nodes by iterating until the sum of the flows at the surcharged nodes is within the allowed tolerance.

The allowed tolerance is the value of SURTOL times the average flow in the WHOLE network (surcharged and non-surcharged links). If the sum of the flows is not lower than the tolerance then Extran will continue iterating until the maximum number of iterations is reached. The equation that Extran uses for calculating the surcharged node depth is different than the node continuity equation for non-surcharged nodes.

It seems in your network the surcharge algorithm causes a significant continuity error if the tolerance is too small. As you increased the value of SURTOL the number of surcharge iterations decreased and the continuity error decreased because in this case the iterations either were not converging or were converging to a solution that caused the continuity error.

Usually in most networks you decrease the continuity error by lowering the value of SURTOL but it is interesting to know that sometimes an increase in the value of SURTOL improves the overall model. You could improve your table by adding the number of iterations alongside the percent continuity error.

Regards,
Bob Dickinson
DickinsonRE@CDM.COM

Bob:

Thanks for the input. I was able to eliminate the small flutter, and am pretty well convinced I have a 'good' model. After studying the results, everything seems to be working very nicely--no anomolies whatsoever. Sorry, but I didn't keep track of the number of iterations during the trials, but playing around with ITMAX seemed to make no difference. I tried dropping ITMAX as low as 10 for several runs, but ultimately it was SURTOL that allowed me to more or less dial in the continuity error.

I have seen in many large models that when lots of nodes go into surcharge, things can get numerically whacky, particulalry if the surcharged nodes are widely dispersed (separated by lots of non-surcharged nodes) throughout the model.

Thanks again for the help.
John Paine
John_Paine@URSCORP.COM

Bill,

We were faced this problem some time back with surcharged inlets, we found it is very hard to get good results with surcharged inlets, took us atleast one full month checking all the options and checking data of such a big project again and again. At last we found only way to solve the problem is add some storage points to delay the peak and see what's happening, then it worked, from that it seems like SWMM may not handle that many surcharged inlets. I suggest you too add storage and see what happens. Your problem reminding me of all my difficulties and research with different SWMM interfaces and version.

Don't know how much I helped you, Good Luck

Srinivas Ravulaparthi P.E.
sravu@TAPANAM.COM