Roof collapses under snow rarely involve a record snowfall. In the great majority of cases the ground snow load for the event was within, sometimes well within, what the structure was designed for — and the roof failed anyway because the snow did not distribute evenly across it. Wind moves snow from higher surfaces to lower ones and deposits it against obstructions, producing local accumulations several times the balanced depth. Design provisions address this explicitly, which means a drift-related collapse usually raises a question about the design or a later alteration rather than about the weather.
Balanced load is only the starting point
The design process begins with a ground snow load for the location, adjusted for exposure, thermal condition and importance to give a balanced roof load. That figure applies to a uniform accumulation across a simple roof. Load provisions then require several additional cases to be checked — drift at steps and projections, unbalanced load on sloped and gabled roofs, sliding snow from a higher roof onto a lower one, and rain on snow. A design that checked only the balanced case has skipped the conditions that most often govern.
Where drift forms and why it concentrates
Wind carries snow across a roof until an obstruction interrupts the flow, at which point it deposits. Steps between roof levels, parapets, mechanical units, penthouses, screen walls and adjacent taller buildings all create these conditions. The drift forms as a triangular accumulation against the obstruction, and its peak intensity can be several times the balanced load. The upwind fetch — the length of roof feeding snow to the obstruction — largely determines how much accumulates, which is why a long low roof next to a step is a characteristic problem.
Adjacent structures create drift on roofs that have no step
A building constructed later next to an existing one can impose drift loading on the existing roof that was never part of its design. So can a new rooftop unit, a screen wall or a solar array. The original structure may be entirely adequate and become inadequate through a change outside it. Establishing when neighbouring construction or rooftop additions occurred, relative to the design date, is therefore a standard line of inquiry in these matters.
Reconstructing the load that was actually present
Weather station records give depth and precipitation, but depth alone does not give load — snow density varies enormously between fresh, settled and rain-soaked snow. Water equivalent data, where available, is far more useful. Photographs taken before and during the collapse frequently show drift geometry directly, and post-collapse measurement of remaining snow depth and density on adjacent unfailed roofs gives a direct measurement of what was sitting there.
Rain on snow is the aggravating case
An existing snowpack absorbs rainfall and holds it, adding load without adding depth. A roof carrying a modest snow load can gain substantially in a few hours of rain, and drainage is generally blocked by the snow itself, so water does not run off. Where a collapse followed a rain event on an existing pack, the reconstruction has to account for the absorbed water, and depth-based estimates will understate the load considerably.
The collapse pattern usually points to the drift
Drift loading is spatially concentrated, so failures initiate where the drift was deepest — typically at the low side of a step or against a parapet — and propagate outward. A collapse that began adjacent to an obstruction, in a structure where the remainder of the roof stood, is consistent with drift rather than with general overload. Debris mapping and the extent of the standing portion together establish this.
Snow removal as a contributing factor
Clearing snow from a roof can create the very condition that causes failure if it is done unevenly, leaving a large accumulated load on one portion of a structure whose adjacent bays have been relieved. Removal equipment adds concentrated load, and workers cutting through to the membrane can damage it. Where removal was in progress or recently completed, the sequence and the resulting distribution become part of the analysis.
Separating the load question from the capacity question
Establishing that drift load exceeded the design balanced load does not conclude the analysis, because the structure's actual capacity may differ from its design capacity — through deterioration, corrosion, prior damage, modifications, or members that were never as built as specified. Both sides of the comparison need independent work, and a conclusion drawn from the load side alone is incomplete.
What to preserve and record
Snow depth and density measured on the collapsed roof and on adjacent standing roofs before it melts, which is a matter of hours to days. Photographs of the drift geometry from every available source, including security cameras and bystanders. The structural drawings and any records of rooftop additions, adjacent construction or alterations. And the framing members themselves, which establish whether capacity matched the design.
This article is general technical orientation, not a failure analysis, an engineering opinion, or advice on any specific matter. Determining the cause of a particular incident requires hands-on examination by a credentialed expert.