As a retired structural engineer, I’ve been asked about snow loads on roofs.
Being curious, I have conducted density tests. Based on these tests, using a digital fishing scale, supposedly accurate to a hundredth of a pound, and measuring the weight of snow in a large brass tube, it appears that the average density of the packed snow in the area I tested was 18.8 pounds per cubic foot (pcf). My rough coffee can measurement, described below, resulted a density of 19 pcf.
Thus, for snow piled six feet high, the snow pressure at the bottom would be six times 18.8 pcf or 112.8 pounds per square foot (psf). Because the local building code dictates that roofs shall be designed for a minimum snow load of 65 psf, the height at which the building code limit is exceeded is 65/18.8, or 3.5 feet.
The 18.8 pcf that I derived for our location could vary considerably, so a good rule of thumb on the conservative side for this measurement is to assume that the snow density is 20 pcf. This would result in a snow depth of 3.25 feet as the limit for the local design of roofs.
A simple means to measure snow density is to jam a coffee can (or any can with straight sides) into the snow, level off the top and put on a lid to prevent evaporation. Let the snow melt in the can and then measure the depth of the water.
Divide the depth of the water by the height of the can and multiply this figure by 62.4 — the density of water per cubic foot. As an example, if the water stands two inches high in a six-inch tall can, then two divided by six times 62.4 equals 20.8 pcf.
If a layer of snow on a roof is four feet deep then the snow load on the roof is four times 20.8 or 83.2 psf. If a flat roof is 20 feet square as an example, then the total weight of snow on the roof is 33,280 pounds.
In performing this test it is a good idea to take at least three samples with the first sample at the top of the snow layer, the second sample at the mid-height of the snow layer, and the last sample at the bottom of the snow layer. The average calculated density of the snow from the three samples should be fairly close to the density calculated at mid-height.
In 2007, the Structural Engineers Association of Colorado (SEAC) published a new study on “ground” snow loads for the state of Colorado. This study divided ground snow loads into three layers: 1) Fresh snow, 2) Settled snow and 3) Compacted snow (from a practical point-of-view, I would also consider that a fourth layer of ice is possible).
For simplicity, fresh snow is assumed to have a weight of 7.5 pounds pcf, settled snow is assumed to have a weight of 15.0 pcf, and compacted snow is assumed to have a weight of 22.5 pcf. Ice would have a weight of about 62.4 pounds pcf, or approximately 5.2 ppi. These figures translate into 0.625 pounds per inch per square foot of area (ppi/psf) for fresh snow, 1.25 ppi/psf, and 1.875 ppi/psf. Thus, the weight of three layers of snow consisting of four inches of fresh snow, 20 inches of settled snow and six inches of compacted snow would have a weight of 38.75 pounds over an area of one square foot.
In terms of an average weight per square foot, the snow load would be 38.75 pounds/2.5 feet, resulting in 15.5 pcf. This is the applied ground snow load.
Of course, the question arises, how does one distinguish between the different layers of snow? Determining the depth of fresh snow is rather easy. However, distinguishing the difference between settled snow and compacted snow is a little more difficult.
Quite obviously, settled snow will vary in density with depth; consequently, as the settled snow approaches the depth of the compacted snow, the settled snow tends to approach the density of the compacted snow. Generally speaking, as one digs downward it is possible that one will notice that, at a certain level, the snow is definitely more compact.
For all practical purposes, this is the point at which the settled snow layer ends and the compacted snow layer begins. I guess you could say that if it’s possible with a reasonable amount of effort to jam a shovel into the snow the layer is settled snow. If it takes considerable effort, the layer is probably compacted snow.
For a heated residence, the actual snow load on a roof is usually less than the ground snow load. This is because of some exposure to the sun and there is residual heat escaping from the roof that tends to melt some of the snow in contact with the roof, allowing this melted snow (water) to drain away (and quite often form icicles). The amount of residual heat, the pitch of the roof and the type of roof surface (metal, wood shingles, asphalt shingles, etc.) all have an affect on the actual snow load applied to the roof.
Recognizing this, the actual snow load applied to a roof is usually less than the ground snow calculated above. (This will not be true for unheated, flat roofs). Thus, a reduction factor of roughly 0.8 to 0.9 can be applied to a calculated ground snow load in order to arrive at a closer estimate for the actual snow load on a roof.
However, if the snow density measurements are taken on the roof instead of the ground, then these measurements more accurately represent the actual snow load on the roof. Thus, using a roof reduction factor of 0.9, a roof that is 20 feet square carrying the ground snow load of 38.75 psf, as calculated above, would have a total roof load of 13,950 pounds.
It must be noted, however, that snow densities can vary considerably depending on location, elevation, surrounding atmosphere, time of year, and how much it has been exposed to the sun. The time of year also dictates the density of snow, with March and April (and sometimes May) having heavier snow densities. Thus, a light, dry powdery snow may only weigh an ounce per cubic foot while a heavy wet spring snow may weigh as much as 35 pounds per cubic foot.
The saving grace with a heavy wet spring snow is that it may melt faster and not stay around for long; but, until it does melt it can cause a lot of damage.
With respect to designing structures to support snow loads, building codes dictate the minimum snow load to be applied to a roof. However, it must be emphasized that a building code only specifies the minimum snow based on limited research and data accumulated over the years for a general locality. Every design engineer and owner must recognize that there are exceptions to the loads indicated in the building codes and plan accordingly. Mother Nature does not recognize building codes, and just because a structure is located outside the jurisdiction for which a specific building code applies, doesn’t mean that snow loads greater than the minimum stipulated will not occur. Therefore, local conditions must be investigated, and if prudence dictates, a heavier snow load than specified in the building code should be applied, even if this increases the cost of the structure.
There are many things to consider in designing the roof framing for a structure besides the snow load. Obviously, the pitch (slope) and surface covering (metal, asphalt, wood shakes) have a lot to do with the buildup of snow on a roof. Also, does the configuration of the roof lend itself to drifting snow due to wind, or are there valleys or isolated pockets susceptible to filling up with snow and increasing the snow load above the design snow load? Are there higher roofs where snow can slide from and land on a lower roof or deck, thus, not only increasing the static load on the lower roof or deck, but also actually producing a dynamic impact load on the lower roof or deck?
Building codes address these problems by providing criteria and formulas for reducing the snow load on a steeply pitched roof, roofs with different surface coverings and for computing the increase in snow load due to drifting. In addition to the above considerations, a designer must take into account if the roof is a folded plate roof, a sawtooth roof, a barrel vault roof and whether or not it is a warm roof or a cold roof.
Regardless of the type of roof, the roof must be analyzed for partial loading, which consists of a full snow load on part of the roof and half a snow load on the remaining part of the roof. Also, the roof should be analyzed for unbalanced loading, which consists of a full snow load on half of the roof and no snow load on the remaining half of the roof. Suffice it to say, a steeply pitched metal roof that allows the snow to slide off unhindered works well in the Pagosa area.
Of course, the main question is how can one tell if a roof is about to fail? The simple answer is: You can’t. There are clues, however, that one can see and hear. The most obvious clue is that the snow is over three feet deep and sometimes one hears subtle snaps, crackles and pops within the structure. This doesn’t mean that the structure is going to immediately fail but only that some warnings are being provided.
In designing a structure, a “factor-of-safety” is employed in the design process. In general, the “factor-of-safety” for structures is about 1.5. This means that a structure designed to hold a 30 psf snow load would actually be capable of supporting a 45 psf snow load — but don’t count on it. There are many components that make up a structural framing system, all of them interconnected and any one of them could fail, from the members themselves to the connections that hold them together. Thus, a framing system could be slightly over-designed and/or overbuilt and capable of supporting 110 percent of the required load by code, or slightly under-designed and/or under-built and only capable of supporting 90 percent of the required snow load.
Because there are so many types of framing systems for roofs, it’s not possible in this article to describe all of them and what to look for as evidence of structural distress. However, there are some general clues indicating possible structural problems, such as difficulty in opening and closing doors and windows, cracks in exterior and interior walls (especially at the upper corners of window and door frames), sagging beams and bowing columns. Another structural component to watch is overhanging eaves where snow can build up and form ice dams. This is especially true if the eaves cantilever greater than about 18 to 24 inches from the exterior wall.
Timber (or wood) as a building material, has the advantage of being able to sustain an overload condition on a short-term basis and recover, provided the overload condition is not too excessive. Structural steel, on the other hand, will deform to a certain point and then fail by buckling, bending or twisting, but if the roof framing members and/or its supporting members can be observed, there should be ample time to address the problem before failure occurs. Metal buildings and sheds with semicircular roofs are subject to sudden buckling failure of the entire roof framing system due to a severe overload condition. Just because the structure has semicircular roof doesn’t mean that it can’t fail. And finally, concrete and masonry materials, being brittle, may or may not provide sufficient warning prior to failure depending on how they are reinforced.