Roof snow load is not the snow that falls on it
The published figure for a location is a GROUND snow load: a statistical value, conventionally with a fifty-year return period, derived from long records at measuring stations. It is the starting point and not the answer.
Roof snow is that ground value modified for what the roof does to snow. An EXPOSED roof loses snow to wind and carries less; a sheltered one, surrounded by trees or taller buildings, keeps everything that lands and carries more. A warm roof melts its underside and sheds; an unheated one over an open structure does not. And a steep slippery roof sheds while a shallow or rough one holds.
Two roofs in the same town, at the same ground snow load, can therefore differ by a factor of two in what they must be designed for — and the sheltered, cold, shallow one carries more than the exposed, warm, steep one despite looking more protected. The exposure and thermal factors are the terms that do that work, and they are judgements about the site rather than measurements.
One consequence catches renovations. Converting a heated loft to an unheated one, or insulating at the rafter line so the roof runs cold, moves the thermal factor in the direction that INCREASES the design snow load on a structure nobody has strengthened.
- p_g
- ground snow load, a statistical value for the location
- C_e
- exposure factor — sheltered roofs collect more, not less
- C_t
- thermal factor — an unheated roof carries more than a warm one
- I
- importance factor, by consequence of failure rather than by size
Drift is local, large, and arrives without anyone touching the roof
Snow does not lie evenly. Wind moves it until it finds an obstruction, and it accumulates against that obstruction in a wedge whose peak intensity is several times the balanced load over the open roof.
Every step, parapet, rooftop unit, penthouse and adjoining taller wall creates one. The drift's size is governed by the length of roof the wind sweeps to supply it — the FETCH — and by the height of the obstruction, so a long low roof running up to a tall wall produces the worst case in the family.
The consequence that matters most is that a drift can appear on a structure nobody has altered. Build a taller extension beside an existing low roof and the low roof, untouched and unmodified, now has a drift load it was never designed for. The same happens when a plant screen, a lift overrun or a new parapet is added, and it is a recognised cause of collapse in the first heavy winter afterwards.
Sliding snow does the same thing by a different route: an upper roof discharging onto a lower one delivers both a load and an impact, concentrated in a strip along the line of discharge. Snow guards are a way of preventing that, and their spacing is a separate calculation about holding snow in place rather than about carrying it.
Uplift runs the load path backwards
Wind flowing over a roof separates at the edges and produces SUCTION over most of it — often exceeding the roof's own weight, which is why roofs leave buildings upwards rather than being pressed in.
The suction is not uniform. Corners see the highest values, edges next, and the field of the roof the least, and the ratio between corner and field is large. This is why roof coverings and their fixings are zoned, why a covering fails at a corner and then unzips, and why the tighter fixing schedule at a corner is a requirement rather than an abundance of caution.
Structurally the important consequence is that uplift reverses the direction of every force in the load path. Gravity loads travel down through members that simply BEAR on one another; uplift travels up through the same joints, and bearing transmits nothing in tension. Every connection from the roof covering to the sheathing, sheathing to rafter, rafter to wall plate, plate to stud, stud to sole plate, and sole plate to foundation has to be capable of carrying tension — and the path is only as strong as its weakest link, because it is a chain rather than a distribution.
That is why straps, hold-downs and hurricane ties exist and why they have to be CONTINUOUS. A well-strapped rafter over an unstrapped wall has moved the failure down one storey rather than prevented it, which is the commonest defect found after wind damage.
A collar tie and a rafter tie are different members
Two horizontal members in a pitched roof look similar and do opposite jobs, and interchanging them is one of the most persistent errors in light framing.
A RAFTER TIE sits at or near the bottom of the rafters, at ceiling level, and resists the outward THRUST a pitched roof generates. A pair of rafters leaning against each other pushes their feet apart; the tie closes the triangle and turns that push into tension in itself. Without it, the walls spread and the ridge drops.
A COLLAR TIE sits high, in the upper third, and does not resist thrust — the geometry gives it almost no leverage there. Its job is to stop the two rafter planes separating at the ridge under uplift or unbalanced load, which is a tension member under wind rather than under gravity.
So a roof with collar ties and no rafter ties is unrestrained against spread, and the symptom appears years later as bowed walls and a sagging ridge rather than as a sudden failure. Raising rafter ties towards mid-height to gain headroom increases the force in them sharply for the same reason the collar tie has little effect — the lever arm is shrinking — and codes limit how far they may be raised for exactly that reason.
A ridge beam and a ridge board are different structures
The member at the top of a pitched roof is either a structural BEAM that supports the rafters, or a BOARD that merely aligns them. They look alike on a section and behave completely differently.
With a structural ridge beam, each rafter is a simply supported member spanning from the wall to the beam. There is no thrust, because the beam holds the top up rather than the rafters holding each other. Ties are therefore unnecessary — and the beam itself carries half the roof's load along its whole length, plus everything that lands on it, down to posts and foundations at its ends.
With a ridge board, nothing supports the top. The rafters form a triangle with the ceiling joists or rafter ties, thrust exists, and the ties are structurally essential rather than optional.
The practical test is the ceiling. A cathedral or vaulted ceiling has removed the ties, so it needs a ridge beam and a load path for it — and the most common serious error in a loft conversion or a vaulted-ceiling alteration is removing ceiling joists that were acting as rafter ties without providing that beam. The roof does not fall in; the walls move out, slowly, and the ridge follows.
Raised heels and truss plates: two details that do more than they look
Where a truss or rafter meets the wall, the depth available for insulation pinches to almost nothing at exactly the point where the wall's thermal bridge already is. A RAISED HEEL lifts the top chord so full insulation depth continues out over the wall plate, and the difference it makes to a roof's real thermal performance is far larger than its cost.
It is also a ventilation detail, because it leaves room for the baffle and the intake path the ventilation paper describes. A standard heel with full-depth insulation blocks its own eaves vent; a raised heel does not. Two problems solved by a geometry decision made by whoever specifies the truss.
Metal connector plates are the other detail with more in them than appears. A plate's capacity comes from TOOTH withdrawal and from the steel's net section, both established by testing, and the required area is computed per joint by the truss designer's software against the forces in that specific truss.
Which is why a truss must never be cut, notched or drilled on site. The plate sizes and positions were computed for the geometry as designed, and an altered member has neither the section nor the connection the calculation assumed. It is also why plate embedment matters: a plate pressed into a wet, misaligned or gapped joint is not the connection that was tested, and it cannot be inspected afterwards without dismantling it. The figures on these pages are screening estimates, and the number that governs a real roof comes from the truss designer's own output for that truss.
Calculators that use this method
Basis
- ASCE 7, Chapter 7 — snow loads: ground snow maps, exposure and thermal factors, slope factor, drift geometry at steps and projections, and sliding snow.
- ASCE 7, Chapters 26 to 30 — wind loads: the zoning of a roof into field, edge and corner areas and the components-and-cladding pressure coefficients for each.
- International Residential Code R802 for rafter ties, collar ties and their permitted heights, and R802.3 for ridge board against structural ridge beam.
- ANSI/TPI 1, National Design Standard for Metal Plate Connected Wood Truss Construction, for plate capacity from tooth withdrawal and steel net section.
- SBCA/TPI BCSI, Building Component Safety Information, on truss handling, bracing and the prohibition on field modification.
- Building America and equivalent guidance on raised-heel truss geometry, insulation depth over the wall plate and eaves ventilation clearance.
- AWC Wood Frame Construction Manual for the continuous uplift load path from roof covering to foundation.