A uniform snow load is the easy part of roof snow design. The hard part is that wind does not leave snow where it fell. It scours exposed areas and deposits snow in the lee of anything that interrupts the flow: a higher roof next to a lower one, a parapet, a plant screen, a lift overrun. The result is a drift, a wedge of snow that can weigh several times the uniform load and sits on just a few metres of roof. StructLoads models that wedge as its own triangular surcharge, because spreading it across the whole roof understates the members along the step and overstates everything else.
Why drifts form where they do
Wind moving across a roof picks up loose snow from the windward areas, carries it, and drops it wherever the flow slows down. A sudden step down in roof level creates a sheltered zone just downwind of the step, and snow settles there until the drift is tall enough to stop accumulating. The longer the roof upwind, the more snow is available to be carried, so drift size grows with the upwind fetch.
The same process happens in reverse. Snow on a lower roof can be blown against the face of a taller wall, building a windward drift. Codes usually treat the leeward drift as the larger and the windward drift as a smaller check, but both exist.
The Wikipedia article on snowdrifts describes the same transport and deposition process in landscape terms. On a roof, the “landscape” is every change in height.
Where drifts appear on a building
The common locations are predictable and worth listing before any calculation:
- Roof steps, where a taller part of the building meets a lower roof: the classic drift case and usually the largest.
- Parapets, which trap snow along the roof edge on the windward and leeward sides.
- Rooftop plant and screens, which act as small obstructions that collect drifts around their base.
- Adjacent buildings, where a taller neighbour within a few metres can drift snow onto a lower roof that is not part of the same structure.
- Valleys and multi-span roofs, where snow slides and drifts into the valley between pitched spans.
The adjacent building case is the one most often missed, because the step is not on your drawings. A single storey extension built against a three storey existing building sits directly in the drift zone of the taller roof.
How Eurocode treats drift at a step
Eurocode 1 Part 1-3 on snow loads handles roofs abutting taller structures with a shape coefficient that combines two effects: snow sliding off the upper roof and snow drifted by wind. The drift coefficient grows with the width of the upper and lower roofs and with the height of the step, and is limited by the step height times the snow density over the characteristic ground snow load.
The coefficient is bounded at a minimum of 0.8 and a maximum of 4.0 in the recommended values, and the drift length is twice the step height, limited to between 5 and 15 m. National annexes can change these limits, and the UK annex has its own drift provisions, so the project annex should always be checked.
The practical result is that at a step of a few metres, the snow load right at the face of the step can be three to four times the uniform roof snow load, falling linearly to the uniform value over the drift length.
How ASCE 7 treats drift at a step
ASCE 7 takes a more physical route. It calculates a drift height from the length of the upwind roof and the ground snow load, then converts that height to a load using a snow density that also depends on ground snow load. The drift is triangular, with a width usually four times its height, and it sits on top of the balanced snow load on the lower roof.
Two limits apply. The drift height cannot exceed the clear height available above the balanced snow on the lower roof, so a short step fills up and the drift is truncated. And the drift width is capped at eight times the clear height in that truncated case. The snow load calculation under ASCE 7 explains the balanced snow load that the drift sits on.
Comparing the drift approaches
| Approach | Drift shape | What sets its size | Typical peak relative to balanced load | Verdict for the takedown |
|---|---|---|---|---|
| EN 1991-1-3 recommended values | Linear from peak at the step to uniform at drift length | Upper and lower roof widths, step height, ground snow | 2 to 4 times | Clear, bounded, check the national annex |
| ASCE 7 leeward drift | Triangle on top of balanced load | Upwind roof length and ground snow load | Often 2 to 5 times at the face | Larger fetch gives larger drift |
| ASCE 7 windward drift | Triangle against the wall | Lower roof length, three quarters of the leeward formula | Usually smaller than leeward | Check, rarely governs |
| Parapet drift (both codes) | Short triangle against the parapet | Roof length upwind of parapet | Moderate, short length | Governs edge beams and purlins |
| Uniform load only | Flat | Ground snow and exposure | 1 times | Unconservative at every step |
The table shows why uniform snow is never enough on its own. The drift does not change the total snow on the building by very much, but it concentrates it on a narrow strip, and the members in that strip are designed for a load several times the average.
What the drift does to members and columns
A drift lies along the step, so it loads the members that run parallel to the step much harder than those further away. Purlins or joists that span perpendicular to the step see a triangular load near one end, which shifts their reactions toward the step. Beams that run along the step carry the full drift on their tributary width. And the columns at the foot of the step collect the drift from both directions at the corners.
This is why a column load takedown that applies a uniform snow load over each column’s tributary area understates the step-side columns. The correct approach is to calculate the drift as a line of extra load along the step, assign it to the members by tributary width, and add it to the balanced snow load only where the drift falls.
Lightweight roofs feel this most. On a heavy concrete roof the drift is a minority of the total load. On a light steel deck where self-weight is 0.3 to 0.5 kPa, a drift of 2 kPa at the step face can more than double the load on the first few metres of roof.
Snow drift and roof collapses
Many historical roof failures under snow involved drifts, unbalanced loads or ponding rather than uniform snow alone. The 1978 Hartford Civic Center roof collapse is frequently cited in discussions of roofs under snow and ice, and its history illustrates how heavy winter loads on a long span roof expose weaknesses that ordinary conditions do not reveal.
The lesson for takedown work is narrower: the extra snow is real, it is predictable from the geometry, and it lands on specific members. A careful takedown puts it there.
Combining drift with other actions
The drift is part of the snow action, not a separate variable action, so it enters the load combinations as snow, with the snow partial factor and combination value. The governing combination at a step is often dead plus snow with drift, even in a building where wind or imposed load governs elsewhere.
The drift is not combined with full roof live load, because the two are mutually exclusive in practice: maintenance crews do not stand on a deep drift. It is combined with dead load, with any rain on snow allowance or ponding check where the code requires it, and with wind as an accompanying action where combinations call for it. The roof live load versus snow load comparison covers why the two are treated as alternatives.
Drift around rooftop plant and solar arrays
Rooftop plant creates small drifts on all sides, and plant screens create drifts against their face. For a single unit, the drift is often modest because the obstruction is small and the upwind fetch to it is short. For a long run of plant, an acoustic screen, or a large rooftop mechanical unit, the drift can approach the parapet case and should be checked.
Solar arrays are a specific case. Panels mounted close to a flat roof can trap snow beneath and between rows, and tilted panels shed snow onto the roof in front of them. Both effects can concentrate snow, and they are worth checking whenever an array is added to a roof that was designed before panels were considered.
Extensions next to existing buildings
The most common drift oversight in practice is a low extension built against a taller existing building. The new roof sits in the lee of the existing roof, and the drift depends on the length of that existing roof, which may be large. A single storey extension 4 m below a long existing roof can carry a drift at the junction that dominates its design.
The same applies in the other direction. A new vertical extension that raises part of an existing building creates a new step, and the remaining lower roof now has a drift case it was never designed for. Any takedown for a vertical extension should check the drift on the roofs it overshadows.
Checking drift on an existing roof
Existing roofs designed before modern drift rules often have no drift allowance at all. Older codes tended to specify a uniform snow load with, at best, a simple surcharge near steps. When an existing building is assessed, for a new use, a roof plant addition or a neighbouring extension, the drift case is often the first check that fails. The members to look at first are the purlins and beams within one drift length of each step, the edge members behind parapets, and the columns at the foot of each step. A survey that records the actual step heights, parapet heights and upwind roof lengths gives everything needed to apply either code’s method.
A worked example: a step between two roofs
Take a two storey office with a single storey wing alongside. The step height is 3.5 m, the upper roof is 24 m wide in the wind direction, and the lower roof is 12 m wide. Characteristic ground snow load is 0.6 kPa, with a uniform roof shape coefficient of 0.8, so the uniform roof snow load is 0.48 kPa.
Under the Eurocode recommended approach, the wind drift coefficient is the average of the upper and lower roof widths divided by the step height, (24 + 12) / 2 / 3.5, or about 5.1. That exceeds the upper limit, and the density limit (2 kN/m3 times 3.5 m over 0.6 kPa) is about 11.7, so the cap of 4.0 applies. The drift length is twice the step height, 7 m, within the 5 to 15 m bounds.
At the step face, the snow load is roughly 4.0 times 0.6, or 2.4 kPa, falling to 0.48 kPa at 7 m from the step. On a lower roof beam running along the step with a 6 m tributary width, the average snow over that width is about 1.6 kPa rather than 0.48 kPa. The step-side columns of the lower wing carry more than three times the snow load of the columns at the far edge.
Key takeaways: snow drift loads at roof steps and parapets
Wind moves snow from exposed roofs to sheltered zones beside steps, parapets and plant, and the drift can carry three to four times the uniform snow load over a strip a few metres wide. EN 1991-1-3 caps the step drift coefficient at 4.0 over a length of 5 to 15 m, while ASCE 7 sizes a triangle from the upwind roof length and ground snow load. Apply the drift as a surcharge only on the members it lands on, check extensions against neighbouring taller roofs, and remember that adding height to a building creates new drift cases on the roofs around it.
Quick answers
Snow does not stay where it falls. The members beside every step and parapet need to carry where it ends up.