University of Minnesota Extension publishes two answers to the same question. One page says a building designed for 20 pounds per square foot should handle four feet of dry snow or two feet of wet snow and ice. Another says the same 20 psf handles six feet of dry, fluffy snow or one foot of wet, heavy snow. Four feet against six, and two feet against one.
Run it against UMN’s own densities and the second page wins. The service puts light fluffy snow at 3 pounds per cubic foot and wet heavy snow at 21. Twenty divided by 3 is 6.7 feet; twenty divided by 21 is 0.95. The four-and-two version implies densities of 5 and 10, half what the same organization publishes for wet snow. The error sits on the more prominent page.
Why the depth rules cannot agree
UMN says worry past four to six feet. NDSU says worry past a couple of feet of compacted snow. SDSU and UW-Madison publish no depth trigger at all, which is the more defensible position of the two.
None of them is wrong. They are describing different snow. Six feet of Minnesota powder at 3 pcf is 18 psf. Two feet of compacted North Dakota snow at 20 pcf is 40 psf. Both land near an agricultural design load, from depths that differ by a factor of three.
Nobody defines light snow. UMN and FEMA call it 3 pcf. SDSU’s conversion works out to 6, NDSU says 7, and UW-Madison gives 5 to 20 before running its own example at 20. Three feet of light snow is 9 psf by one source and 60 by another, the difference between a normal winter and getting the animals out.
The conversion underneath all of it has been disowned by the agency that popularized it. The National Weather Service at La Crosse calls ten inches of snow to one inch of water an old rule of thumb, far from the norm, and offers 12 to 1 for the Upper Midwest. NWS technical literature puts real ratios between 3 to 1 and 100 to 1. Every inches-to-pounds table here is an average standing in for something that varies thirtyfold.
So weigh it
Minnesota’s 1983 Special Report 104 publishes the method, and nobody has improved on it since. Take a 3-pound coffee can, 6 inches across. Push it down through the full depth of the snow, empty it into a pail, repeat until you have the whole core. Do not compact the snow going in. Melt it. Weigh the water in pounds and multiply by 5.1, or measure its depth in inches inside that same can and multiply by 5.2. Either way the answer is pounds per square foot.
The report works it twice. Four pounds of melt is 20.4 psf. Four inches of liquid in the can is 20.8. For ice, measure the average thickness, multiply by 5.2, and add it. Ice is thickest at the low edges and in the valleys.
Where you sample decides whether the number means anything. Take it from the center third of a roof plane. Valleys and drift zones are not representative, and FEMA states flatly that measuring snow on the ground and assuming it matches the roof is incorrect.
SDSU publishes a variant needing no coffee can: weigh an empty container, clear a measured patch near the eave, fill it, divide weight by the area cleared. It comes with a warning worth repeating: leaning a ladder against an already overloaded building can be what finishes it.
The number you are comparing against is lower than you think
Agricultural buildings get a code exemption almost everywhere, which is why barns go down in storms that leave houses alone. Minnesota requires 42 psf for houses in the north and 35 in the south, while many ag buildings there went up at 20. North Dakota houses run 30 to 40 and its ag buildings 20 to 35. South Dakota used 20 psf for farm buildings in the past, newer ones 25 or 30. Wisconsin exempts ag structures from its Uniform Dwelling Code, and New York exempts non-residential farm buildings from the state code.
FEMA puts it plainly: many older agricultural structures were never required to be designed to a specific snow load. Drift loads did not enter BOCA until 1975. Unbalanced roof loads did not reach the UBC until 1988, so a pole barn put up forty years ago was never designed for snow the way snow is now understood.
Your building’s own number may exist on paper. SDSU notes that the specifications handed over at contract completion often state the design roof snow load, and UMN points at truss certificates. SDSU adds a unit warning: when a builder says a twenty-pound snow load, he means 20 psf.
The only published trigger
One source puts a number on the decision. SDSU says to consider clearing a post-frame roof when the estimated load roughly equals or exceeds the design load but is less than twice it, and when every question in its inspection walk comes back no. Past twice the design load, the implication is to stay off and get help.
The walk happens in three stations. From outside: is the roof lower in the middle than at the end walls, are the walls pushed out or buckling, is the building leaning, has the siding dented where a knee brace may have slipped. From the doorway: is the ceiling flat and level, or wavy and bulging. From the attic entrance, before going in: truss plates bulging or lifting, webs buckling, purlins bending or twisting. A yes at any station means do not enter and do not get on the roof.
Cornell adds what decides how much warning you get. Wood structures show stress before they fail; pre-engineered metal usually will not, so the warning-sign list does not apply to them.
Clearing it without finishing it
NDSU titled its piece “Roof Snow Removal Probably Not Necessary,” and FEMA notes that with minor accumulation the risk of being on a roof outweighs the benefit. When you do clear, the sources agree on method.
Work from the ground with a rake. Plastic or wood tools, never metal, because metal tears roofing. No snow blowers. Do not chip or scrape ice; cold shingles are brittle and the granules come off with it. Remove the same amount from both sides, in narrow strips rather than wide patches, and never pile what you remove onto a loaded part of the roof. Cornell presses that point: a roof carrying a load spread evenly can fail once the same load is concentrated.
Clear the snow away from the sidewalls too. Cornell notes that posts, siding and curtains were not designed to take a drift leaning against them.
Leave a layer. FEMA says at least 2 inches, written for membrane roofs; UW-Madison and Cornell give the principle without a number, that the bottom layer protects the roof surface from the tool.
Direction is the one place the sources genuinely conflict. FEMA says start at the ridge and work toward the eave; SDSU and Cornell both say start at the eave and work up. The conflict dissolves once you ask where the person is standing. FEMA is writing for a contractor on a low-slope commercial roof. SDSU and Cornell are writing for someone on the ground with a rake, who cannot reach the ridge anyway.
What makes a safe roof unsafe overnight
Rain on an existing snowpack is the fastest way to a collapse. FEMA notes that codes cover light rain on snow but not heavy rainfall, and that a short rain can saturate a pack and raise the load sharply. Price it with the same constant: every inch of water the snow absorbs adds 5.2 psf. On a 30 by 60 foot shed that is 9,360 pounds from one inch of rain.
Drift does the same thing without any new weather. A USDA NRCS design guide for a 40-foot post-frame building with a 24 psf balanced load puts the unbalanced leeward load at 41 psf for the first 8 feet off the ridge, rising to 45 on a 60-foot building. The windward side drops to 7. SDSU’s point follows directly: an unbalanced load may damage a building more than a larger balanced one, because the unloaded side of a truss can lift and compromise the connection at the post.
Sources
- University of Minnesota Extension, Roof collapses from snow and ice — https://extension.umn.edu/agriculture/farm-operations-and-systems/farm-safety/roof-collapses-from-snow-and-ice
- University of Minnesota Extension, Snow Loads on Buildings — https://blog-swine.extension.umn.edu/2019/03/snow-loads-on-buildings.html
- University of Minnesota Agricultural Extension Service, Special Report 104, Residential Roof Snow Load Concerns (1983) — https://conservancy.umn.edu/server/api/core/bitstreams/148af538-7fd2-47f9-bfb8-23bd055a3995/content
- SDSU Extension, Post-Snowstorm Building Inspection and Snow Removal — https://extension.sdstate.edu/post-snowstorm-building-inspection-and-snow-removal
- NDSU Extension, Roof Snow Removal Probably Not Necessary — https://www.ag.ndsu.edu/news/newsreleases/2020/jan-6-2020/roof-snow-removal-probably-not-necessary
- University of Wisconsin-Madison Division of Extension, Snow Loads on Farm Buildings — https://lincoln.extension.wisc.edu/2018/04/16/snow-loads-on-farm-buildings
- Cornell University, Heavy Snow Loads: Dos and Don’ts for Barn Snow Removal — https://eden.cce.cornell.edu/2025/12/04/heavy-snow-loads-dos-and-donts-for-barn-snow-removal
- Curt A. Gooch, Cornell University, Heavy Snow Loads — https://nydairyadmin.cce.cornell.edu/uploads/doc_512.pdf
- Gooch and Gebremedhin, Cornell University, Barn Roof Failures (1999) — https://eden.media.uconn.edu/wp-content/uploads/sites/1244/2015/12/Barn-Roof-Failures.pdf
- FEMA P-957, Snow Load Safety Guide, via the US Government Publishing Office — https://www.govinfo.gov/content/pkg/GOVPUB-HS5_100-PURL-gpo59955/pdf/GOVPUB-HS5_100-PURL-gpo59955.pdf
- USDA NRCS Pennsylvania, Design Guide 12, Prescriptive Design Specifications for Non-Diaphragm Post-Frame Buildings — https://www.nrcs.usda.gov/sites/default/files/2022-11/PA-12%20Design%20Guide%2012-%20Prescriptive%20Design%20Specifications%20for%20Non-Diaphragm%20Post-Frame%20Buildings.pdf
- NOAA National Weather Service, La Crosse WI, Why snow ratios vary — https://weather.gov/arx/why_snowratios
- NOAA National Weather Service, Western Region Technical Attachment TA1604 — https://www.weather.gov/media/wrh/online_publications/TAs/TA1604.pdf
- UConn Extension, Reducing Storm Damage to Your Barn — https://extension.uconn.edu/publication/reducing-storm-damage-to-your-barn/
Leave a Reply