I have a question(s) regarding calibration of SWMM models for use with
large, 5 and 10 year return, events. My context is basement flood relief
design work using 5 and 10 year return design events.

We always try to calibrate these models by instrumenting the sewer
systems for one or two summers. However, these sewers have very low
hydraulic capacities, usually one or two year return capacity. If we get
an intense rainstorm there is significant basement flooding and
frequently surface flooding because the inlets have approximately a 5
year capacity.

I feel that the large events are not useful for calibration because we
have no knowledge of the volume of water that went into flooding and the
duration of that flooding. Without knowing these two things there is no
basis to adjust any runoff parameters. I can approximate the observed
flooding with a wide range of parameter values with no basis to confirm
any specific magnitude for any input value. We have only water level
data with sometimes flow gages installed but so far the flow gages have
failed during every major storm for a wide variety of reasons.

We do get a large number of less intense storms that don't cause any
flooding and our models are well fit to observed data from these storms.

My question is "can I assume that the good fit to the lesser storms
implies a good degree of accuracy of runoff estimates to storms much
greater than the calibrated events?"

I think the issue can be split into two separate issues; runoff from the
impervious and pervious surfaces.

The lesser storms should be fairly accurate as to runoff from the
impervious surfaces because, at the lower rainfalls, the runoff is 100%
from the impervious areas and zero from the pervious surfaces. Can I
assume that the runoff from the impervious areas would still be accurate
for greater storms or would there be a factor not covered by the lesser
storms?

Since runoff from pervious surfaces is zero for lesser storms we have
essentially zero calibration data for the pervious surfaces.

Has anyone done calibration of SWMM models to severe storms where there
was no flooding and heavy enough rainfall that there was runoff from the
pervious surfaces? If so, could you share the resultant input parameters
with some description of the areas and models?