ponding over steel roof joists

huangyhg

ponding over steel roof joists
in the northern united states we have snow loads in the winter.  some argue that, while ponding would only occur during the summer months, the winter snow consideration equalizes the need to consider potential ponding.  (many variables to consider.)
consider a lower flat roof with drift loads caused by a higher sloped roof to the north (slope ends at lower flat roof)and a higher flat roof to the south.  the northern "valley" produces a 60 psf drift.  the southern wall produces a 90 psf drift.  low ends of both drifts have a common boundary.
sliding snow from the north is not reasonable but drainage is part of the design.  bottom of scuppers are 12" from top of drain.  roof insulation is 1/4:12 sloped from top of drain yielding an 8" difference between bottom of scupper and top of roof insulation at parapet.
some argue that we could have a winter thaw, an overnight freeze and a heavy snow to follow.  considering the drains are clogged, do you design for ponding in addition to the design drift loads?  where do we stop?  consider the collapse in russia.  cosider how you felt before that news.
what are your thoughts?
thank you!
check out our whitepaper library.
having lived in the northern us for many years, i would say that the thaw - freeze - snow scenario is not unusual at all, but quite common.  some of the worst blizzards i have experienced have been in the spring when it has been melting. however, flat roofs are less commmon for the reasons you list.  therefore, as a designer you need to be ultra conservative when designing a flat roof.
i ran across this same issue with some long span parallel chord wood roof trusses (45' span).  architect called out 1/4" per foot pitch to a mini-ridge (+6") at the truss midspan.  the trusses deflect 2" to 3" under the design snow load...  i will be marking up the truss shop drawings to prevent a ponding situation; ponding instability is scary.
the principle is the same for any flat roof system be it wood trusses or steel bar joists.  
i think in your situation, you would design for any code prescribed snow drifts and set the top of steel elevations at any interior girders/truss girders and exterior spandrel beams such that the final deflected shape of the roof structure yields at least 1/4" per foot slope.  of course, this implies that you know or at least estimate the stiffness of someone else's pre-engineered product (wood trusses or steel bar joists).  this is the intent of ibc 1611.2 as i read it.
apparently, the russians, poles, and germans either do not do this correctly or their design snow loads are too small.
make sure that you get reassurance from the plumbing designer that he is not specifing a "contolled flow" roof drain or the new "restricted flow" drains that are now being marketed.
ponding seems to be more critical on long spans in those areas where the roof dead and live load is minimal.