In short, Cohen relates the amount of snow cover and its rate of advance in Siberia in October to temperatures during winter in North America. The more snow in Siberia and the faster it increases he suggests, the colder it’s likely to be in the eastern U.S. This October, snow came down fast and furious in Siberia and, thus, Cohen is bullish about a harsh winter for the eastern U.S.
Seeking more details about Cohen’s methods and forecast for this winter, we presented him with 7 questions which he graciously answered below.
Capital Weather Gang (CWG): Can you briefly summarize the relationship between Siberian snow cover in October and the winter outlook in eastern North America? Is it the total amount of Siberian snow cover that matters or the speed of the advance, or some combination?
Cohen: My colleagues, at AER and at selected universities, and I have found a robust relationship between two October Eurasian snow indices and the large-scale winter hemispheric circulation pattern known as the North Atlantic or Arctic Oscillation pattern (N/AO).
The N/AO is more highly correlated with or explains the highest variance of winter temperatures in eastern North America, Europe and East Asia than any other single or combination of atmospheric or coupled ocean-atmosphere patterns that we know of. Therefore, if we can predict the winter N/AO (whether it will be negative or positive) that provides the best chance for a successful winter temperature forecast in North America but certainly does not guarantee it.
[Of the two indices we’ve analyzed], the first and longer [more data points] index is simply the monthly mean snow cover extent (SCE) for the entire month [of October] as measured from satellites. This record dates back to at least 1972 and is available on the Rutger’s Global Snow Lab website.
The second index that we developed last year, with the support of NSF and NOAA grants, measures the daily rate of change of Eurasian snow cover extent also during the entire month of October, which we refer to as the Snow Advance Index or SAI.
One important difference between the two indices is that the original SCE index is not sensitive to the time of snowfall or snow cover advance during October in contrast to the SAI where snow cover at the end of the month contributes to higher values of the SAI than snow cover advance at the beginning of the month. This SAI does have the disadvantage that it is only available since 1997. However, it is much more highly correlated with the N/AO than the original snow extent index and therefore is more skillful or accurate in predicting winter temperatures in North America.
CWG: What did Siberian snow cover do this October? What recent years did it most and least closely resemble?
Cohen: Both snow indices where above normal this past October. However, the SAI was much higher relative to normal than the SCE. The SAI was the second highest value in the 16 years that we calculated the value, second only to 2009. And, therefore, this past October snow cover most closely resembles that of 2009.
But in my opinion the atmospheric evolution so far this cold season does not match that of 2009. Clearly, there are other important factors that influence the winter atmospheric circulation such as ENSO, even possibly melting sea ice and others mentioned in the CWG winter outlook that are not included in our conceptual model on how the winter N/AO is dynamically forced. Still, we anticipate that the N/AO will be negative as it was in 2009. In fact, the N/AO set a new record low in winter 2009/10 [but] that would be hard to match.
CWG: Based on this year’s Siberian snow cover in October, what are you expectations for temperatures in the Mid-Atlantic and Northeast, and specifically, the D.C. area? Are you able to break it down month by month or only the winter as a whole?
Cohen: Based on the SAI, but not exclusively, we expect that winter temperatures will average below normal across the Mid-Atlantic and the Northeast including the D.C. metro area.
We do provide monthly and three month average forecasts. The three month average and the monthly average temperature forecasts for Washington D.C. all predict below normal temperatures this winter. However, it is important to note that the model skill or accuracy is better for the seasonal average than any one month and I certainly would not expect consistently below average temperatures this winter but instead high intra-seasonal variability.
CWG: Does your method predict precipitation? Can you say anything about snow in D.C. this winter?
Cohen: Let me preface my answer by first stating that we do not forecast snowfall only precipitation. The skill or accuracy of the model or predictors that we look at is better for temperature than precipitation and better for all precipitation than just snowfall.
Keeping that in mind, when the SAI is high like this October, winter precipitation does tend to be higher or above normal in the Washington D.C. metro region. Also, when October Eurasian SCE is above normal, higher snowfall is favored in Washington D.C. but the relationship is not terribly strong or robust. But again all these relationships with precipitation and especially snowfall are much more tenuous than with temperatures.
CWG: Longer range forecasts seem to be suggesting the N/AO going negative to start December and, perhaps, a fast start to winter. Do your methods support that?
Cohen: I would say that we have certain expectations of how the atmosphere responds to the rapid buildup of Siberian snow cover in October that are consistent with a negative N/AO and cold temperatures in the Eastern U.S. [during] the latter half of November and the first half of December, but it is not a necessary condition. It is of my opinion, anyway, that this early in the cold season, the atmospheric conditions across Eurasia are more important than the atmospheric conditions in North America or the North Atlantic for what happens later in the winter even here in eastern North America.
But on a personal level, I would be lying if I did not admit to feeling comfort to seeing the winter start consistent with the winter forecast. [But] at least as of this writing, it seems to me that there is unusually high spread among the models so I take nothing for granted.
CWG: How do the borderline El Nino/neutral conditions affect the confidence in your forecast?
Cohen: As long as I have been a seasonal forecaster, I have always considered El Nino/Southern Oscillation (ENSO) as a better predictor of precipitation than temperature across the Eastern US. I think this is supported by the observational or statistical analysis as well as the skill or accuracy of the climate models. We use ENSO in our seasonal forecasts but it is only a secondary predictor for winter temperatures. I may be unique in my opinion as a seasonal forecaster, but my confidence in the temperature forecast this winter is little changed by the phase of ENSO. However, I do feel less confident in the precipitation forecast as ENSO forecasts waver between El Nino and neutral.
I will add one caveat, there have been recent modeling studies that demonstrate that El Nino modulates the strength and position of the Aleutian Low that then favors stratospheric warmings and subsequently a negative winter N/AO that are consistent with our own research on the relationship between snow cover and stratospheric warmings. So the influence of ENSO on winter temperatures in the Mid-Atlantic and the Northeast may be greater than I acknowledge or that is represented in our seasonal forecast model.
CWG: Consider our winter outlook. What aspects do you generally agree with and disagree with?
Cohen: I feel that the CWG looked at many of the important factors for the winter climate and therefore winter prediction. I couldn’t agree more with the CWG that this winter will be colder and snowier than last winter but the statistical odds strongly favor that outcome.
Based on my previous answer, I would not use ENSO [i.e. El Nino phase] as my dominant or primary winter temperature predictor for the Mid-Atlantic. This is a bit out of my expertise, but I would say something similar for the PDO. It is my impression that the PDO is a better predictor of climate in the Western U.S. than the Eastern U.S. Also I am a little hesitant to use an index that has “decadal” in the title to make a seasonal forecast. For me the key is the winter N/AO and in my opinion atmospheric dynamics will more strongly determine the phase of the winter N/AO more so than past recent history.
In conclusion, I feel that the predictors of this winter favor a more negative N/AO and colder temperatures and probably more snowfall in the Washington DC region than the CWG. We don’t approach the winter forecast as static and will monitor the progress of the forecast throughout the cold season and adjust accordingly.
And lest anyone think that I am overly confident in our winter forecast, trust me when I tell you that months of sleepless nights await me. We have been fortunate to have gotten the past three winters correct but my colleagues are fond to remind me that you are only as good as your last forecast.