U.S. Snowfall 1900-2019: A Decade-by-Decade Look

As anyone who has followed my blogs for WU over the past ten years has no doubt noticed, I am always interested in the actual data-derived records so far as weather events are concerned. In my previous post, I looked at record snowfalls for all the states and several cities for various periods of time (24 hours, monthly, etc.). In the conclusion to that blog, I wrote a bit on how climate change may be affecting snowfall in the United States. The basic conclusion was that no one really understands much about this.

So, as a follow-up to that, I decided to look at how much snow has actually been measured decade by decade at 40 different cities/sites across the contiguous U.S. since 1900. I only included places that see winter snowfall regularly (i.e., in most years), meaning along and north of a line from North Carolina through Tennessee, Oklahoma and the mountainous regions of the West.

A constraining factor in choosing the sites is that they must all have a continuous monthly snowfall record dating back to at least 1900, something that a surprisingly few do. There is no complete record for Nashville, Tennessee; Roanoke, Virginia; Sheridan, Wyoming; and Seattle, Washington, among other cities that would seem to be obvious choices. For instance, in the high mountain areas of the West there are virtually no sites with a continuous period of record (POR) back to 1900, aside from Flagstaff, Arizona; Donner Summit in the high Sierra of California; and Red Lodge, Montana (which I did not include because of its obscurity).

The data

Before discussing the trends and reliability of the data, below is the table, based on reports culled from the National Weather Service NOWData website.

Figure 2. Map of the 40 sites used in my data table.

Inherent problems with the data

As noted in my previous blog, the methods of snow measurement in the U.S. have changed over time. An article authored by Matt Kelsch of NCAR/UCAR and official COOP observer for Boulder, Colorado, explains how in the oldest of snow records (mostly predating 1950) a simple 10-to-1 snow-to-liquid ratio was often used to estimate the snowfall (i.e., 1” of melted precipitation = 10” of snow). As it turns out, the average ratio for the contiguous U.S. is roughly 12:1 or 13:1, and that ratio can vary greatly from place to place and storm to storm, even within a single storm.

At some point—and that point in time was different among the various weather observation sites—actual snowfall began to be measured using a stick-like ruler, with the snow measurements made either at the end of each snowfall or at one or more regular times each day (e.g., at 7 a.m. or 7 p.m.).

At some point (and this is the problem with my data: that this “point” in time varied from site to site between the 1950s and 1990s), snowboards came into use (see Mr. Kelsch’s description of these in his writeup).

Figure 3. An official snowboard, as pictured in the National Weather Service pamphlet Snow Measurement Guidelines for National Weather Service Snow Spotters. The poles on either side allow the snowboard to be found in case of very heavy snow between observations. Image credit: NWS.

The use of snowboards led to snowfall being more accurately measured, but it also increased the amount of snow attributed to any given storm. This is because snowfall measurements were now being made as often as every six hours (when the snow board would be cleared to make way for the next six-hour measurement) instead of just once or twice a day. Since deep snow settles as it falls, this method increases the amount of snow measured.

As an observer who has used both techniques during his now-29-year COOP tenure in Boulder, Mr. Kelsch estimates that for extreme snowfalls the use of six-hourly snowboard measurements can result in snow totals that are 15 to 20 percent greater than what is actually measured on the ground. The potential for confusion became evident after New York’s official Central Park site reported a 24-hour snowfall of 26.8” on January 22-23, 2016, a new all-time record for New York City. That total was adjusted upward even higher, to 27.5”, after an NWS review found and corrected an error in the transmitted snow report. However, local weather-minded residents living near the site in Central Park (and there are many of those!) measured only 18” to 22” on the ground at the end of the storm.

At Newark International Airport, observations from the same storm showed a preliminary record of 28.1”. That total was declared invalid by the NWS because the private contractor who measured the snowfall took snowboard measurements once per hour, as opposed to the standard six-hour interval. The revised total of 24.0” fell short of the record of 25.6” set on Dec. 26, 1947.

Another example: The great Blizzard of March 1888 brought Central Park 2.10” of melted precipitation, resulting in the official 21.0” snowfall reported. Since temperatures during the height of the blizzard were in the low teens, it is likely that the ratio was much greater than 10 to 1, and thus the actual snowfall considerably more than the 21.0” officially reported.

There is also the issue of observation sites moving from one location to another over time. This is one reason why Marquette, Michigan, is not in my list: their average annual snowfall almost doubled when the NWS office moved from the town to the hills several miles south.

The bottom line is that comparing old snowfall measurements with new ones is comparing apples to oranges and, unfortunately, makes looking for historical trends (especially when talking about climate change) a hapless enterprise.

Figure 4. Clearing the streets of snow in Manhattan following a blizzard in January 2016 when an all-time record of 27.5” of snow was measured in Central Park. Seven of New York City’s top 10 biggest snowstorms (and 12 of the 36 storms resulting in a foot or more of snow since 1870) have all occurred since 1994. Is this a result of climate change (warmer ocean temperatures increasing the available moisture in winter storms) or because of changes in the technique used to measure snow depth? Image credit: Michael Yamashita.

Some apparent trends in snowfall

Regardless of the caveats mentioned above it is still worth mentioning some general trends:

—The 1920s and 1930s seem to have been the least snowy decades since 1900.

—The 1960s and 1970s seem to have been the snowiest decades since 1900.

—Snowfall in the nine-year period of 2011-2019 means that the 2010s will be Boston’s snowiest decade, while the period was less snowy than any prior decade on record for Washington, D.C., and Raleigh, North Carolina. Snowfall averaged above normal from Philadelphia northward and below normal south of there. Snowfall was below average for the entire West, aside from the northern Rockies (the 2010s will be the snowiest decade on record in Great Falls, Montana).

Figure 5. Great Falls, Montana already was on track for its snowiest decade on record when a monstrous unseasonal storm buried portions of Montana in late September this past year. The city picked up 17.7” of snow in just 24 hours September 28-29, which was the greatest 24-hour snowfall for any month on record for the site. Image credit: NWS/Great Falls.

—The first half of the POR (1901-1960) saw only 11 sites with their snowiest decade and 27 sites with their least snowy decade. Conversely, the second half of the POR (1961-2019) saw 29 sites with their snowiest decade and 13 sites with their least snowy decade. This could be evidence that the change in the technique of measurement has led to an increase in reported snowfall amounts. Given that assumption, it is interesting that the most recent decade (actually just nine years: the seasons of 2010-2011 through 2018-2019) saw 7 of the 40 sites experience their least snowiest decade (tied with the 7 such during the 1920s but just short of the 8 such in the 1930s). Given the snowboard bias, this could indicate a significant decrease in snowfall amounts overall, especially in the West and Mid-Atlantic regions. Of course, some of the seven possible decadal records for least snow based on 2010-2019 may not be realized after snowfall from 2019-2010 is factored in.

—Only two of the 40 sites researched, Buffalo and Sault Ste. Marie, are lake effect-influenced stations. This, of course, is too small a sample to verify the often-cited claim that snowfall has increased for snow-belt regions due to the warming of the Great Lakes.

—It would appear that in the past decade (based on 2011-2019), colder places at northern latitudes or higher elevations are seeing an increase in average annual snowfall, whereas the places in more southern latitudes are seeing a decrease in such.

Conclusion

This is, I’ll admit, an unscientific survey, but no one (to my knowledge) has attempted to even research the subject in much detail. This is probably because, with the change in measurement techniques over time, it is not possible to conclusively say that any one part of the country has become snowier or less so over the past 120 years.

Christopher C. Burt

Weather Historian

KUDOS: Thanks to Matt Kelsch for his research into the history of snow measurement techniques and his assistance with this article.