Identifying Northern Great Plains Blizzards in the Past, Present, and Future

The Northern Great Plains (NGP) is the blizzard capital of the United States. Here, a relatively flat, treeless landscape combines with cold conditions and a wintertime storm track to promote the occurrence of blowing snow. The Grand Forks National Weather Service (NWS) County Warning Area (CWA) for example, experiences on average 2-3 blizzards per year. These events are impactful enough that a local newspaper names blizzards, and these events have been associated with spring flooding, economic losses, and even deaths.

How blizzards will change in a warming climate is uncertain. Prior literature suggests offsetting impacts such as varying changes in extratropical cyclone frequency/intensity, resultant snowfall, and temperatures. Diagnosing these events is also problematic from both a historical and future climate perspective. Like other extreme events, there is no guarantee that historical databases such as NOAA's Storm Data are complete. Further, the definition of these events is dependent on visibility. The process of blowing snow (and resultant impact on visibility) is not included in historical reanalyses or climate simulations. This makes identification of these events difficult.

In this study, a competitive neural network known as the Self Organizing Map (SOM) is used to objectively classify atmospheric patterns associated with known blizzard events over the NGP. This information is then used to retrospectively identify potential blizzard events in the North American Regional Reanalysis (NARR). After comparing historical blizzard and blizzard pattern climatologies, this classification scheme is then used to classify events within the Community Earth System Model (CESM) Mother of All Runs (MOAR) to understand how patterns may change in the future.

The objective classification scheme demonstrates that regional blizzards are forced by a number of meteorological patterns including Colorado Lows, Hybrid Lows, Alberta Clippers, and Arctic Fronts. Colorado and Hybrid Lows are most likely early/ late in the season (e.g. Oct.-Nov., and Mar.-Apr.) while Arctic Fronts and Alberta Clippers are more likely during the middle of winter (Dec.-Feb.). Objectively defined patterns in the SOM largely aligns with the subjective classes chosen by NWS forecasters.

The climate model grossly predicts the historical climatology if blizzard patterns with a rapid increase in potential blizzard events in Dec. before tailing off in Apr. Unlike the observations, the model fails to produce a lull during Feb., and it is demonstrated that this is largely due to the over prediction of all patterns. Moving to the future, various emission scenarios demonstrate that blizzard patterns will decrease in the future, with patterns becoming less likely at the beginning and end of the season due to warming temperatures. This in turn cases a shift in peak blizzard occurrence to two-four weeks later in the winter. In general, blizzard patterns become warmer and it is hypothesized that this will in turn lead to a further reduction in blizzards due to a warmer/wetter snowpack that decreases the likelihood of the blowing snow process.