Convergence of Horton's Equation in SWMM

I had previously been in touch with you regarding the calibration of SWMM. I have managed to successfully implement a local optimization technique. Now that i am looking a global optimizatin technique, i require a much larger number of iterations. Under some infiltration parameter values, the system will produce a warning message stating that the Horton's Infiltration Equation has not converged. Looking at the code of SWMM, this occurs whenever a particular parameter value is below 0.001*DELT. I have set up my program to neglect the output produced from such a set of data, but i was wondering what the implication of this non-convergence is? Does it reflect in the output in any way?

The default procedure in Subroutine HORTON when the Horton infiltration routine does not converge is to move out along the Horton infiltration curve by a time increment of 0.9*DT. This means that infiltration capacity is reduced by the appropriate amount along the exponential curve and that infiltration does occur. There are checks elsewhere to ensure that infiltration is not greater than the amount of water available (rainfall plus stored water).

To answer your question regarding the effect of the non-convergence messages -- it depends. If the simulation is already at the point where Horton infiltration is just equal to the minimum value, then there will be no effect at all. If there is no convergence when infiltration is just beginning and the simulation is "high" on the Horton curve (near fo = WLMAX), then infiltration might be overestimated (but again, it will not be more than what is available to be infiltrated). Or it might be slightly underestimated, if for some reason, we should really move down the time scale by a whole DT, not just 0.9*DT.

Non-convergence is usually caused by too big a time step. You can try reducing WET and WETDRY to numbers like 5 - 30 min and see what happens. A less likely option is that you have poor infiltration input parameters. For instance, be sure that the decay coefficient (DECAY) has units of 1/sec, not 1/hr. That is, DECAY should be a number on the order of 0.0005 per sec. Maybe your optimization routine is pushing DECAY out of its reasonable range?

The bottom line is that the final SWMM answers should be "reasonable" because of the built-in default procedure in Subroutine HORTON. If there are only a few non-convergence messages in a long simulation, the effect probably won't even be noticable. If there are many such occurences, then infiltration will likely be over-estimated somewhat, but you can't tell for sure without repeating the simulation with a shorter time step until the messages disappear.