Flat Roof Snow Load (ASCE 7)

Snow load is where a weather number becomes a structural one, and it is the calculation on this site that most deserves the warning attached to it: this page is a teaching aid and the governing building code is the authority. Roofs get designed from this number and roofs collapse under snow every winter.

The page follows ASCE 7, the United States standard, and only ASCE 7. Its flat-roof relation is $p_f = 0.7\,C_e\,C_t\,I_s\,p_g$: take the ground snow load from the standard's map, cut it by 0.7, then adjust for wind exposure, for how warm the roof is from beneath, and for what the building is for. Canada's National Building Code uses a different equation with a different factor set and an added rain-on-snow term, $S = I_s[S_s(C_b C_w C_s C_a) + S_r]$, and its factors are not interchangeable with these. Naming the convention is half the honesty of the page.

What each factor is doing. The 0.7 is the basic ground-to-roof conversion, and it exists because a roof is windswept in a way that an open measurement field is not — some snow blows off. $C_e$, the exposure factor, refines that from 0.7 for a fully exposed roof in windswept terrain to 1.2 for a sheltered one, because a roof tucked among trees or taller buildings keeps everything that lands on it. $C_t$, the thermal factor, is 1.0 for a normally heated building, 1.1 where the space below is barely above freezing, 1.2 for an unheated or freezer structure, and 0.85 for a continuously heated greenhouse: heat escaping upward melts the underside of the pack. Note the uncomfortable corollary — insulating a roof properly increases the snow it will carry, and a great many older buildings have been quietly protected by their own heat loss. $I_s$ is not an engineering property at all; it is policy, scaling the design event's return period by risk category, from 0.8 for a structure of low hazard to life up to 1.2 for a hospital or emergency facility.

The ground snow load is the part nobody can substitute for. $p_g$ is a mapped, site-specific value derived from decades of measurements and fitted to a design return period, and the map has case-study regions — mountainous areas mostly — where it refuses to give a number and defers to local records. A value borrowed from the next state, or from an older edition, is not a conservative approximation; it is a different number. Maps change between editions, and not always downward.

Now the list of things this equation deliberately does not compute, every one of which is a separate calculation in the same chapter, and every one of which routinely governs the design. It does not do drift: the triangular surcharge that piles against a parapet, a roof step, a penthouse or a rooftop unit, which regularly exceeds this uniform load by a factor of two or more at the place it acts. It does not do sliding snow from an upper roof onto a lower one. It does not do the rain-on-snow surcharge that applies to low-slope roofs in certain ground-snow ranges. It does not do unbalanced loading on a gable or curved roof, nor the partial loading cases. It has no slope factor, because it is the flat-roof case — a sloped roof takes a further step, $p_s = C_s p_f$. And it does not check the code's minimum roof snow load, which for low-slope roofs sets a floor this equation can fall below.

Roof failures under snow are overwhelmingly drift, ponding or blocked-drainage failures rather than uniform-load failures. A number from this page is the first line of a snow load assessment. It is never the last one.