A striking example of the accuracy of the new method happened in the winter of 2002-03. Based on extensive Siberian snow cover during the fall, Cohen correctly forecasted cold weather in the eastern U.S., while most other forecasters predicted warm weather for the northern half of the U.S., based mainly on El Niño conditions. Similarly, Cohen's forecast for the winter of 2003-04 indicated cold conditions in the northeastern U.S., while other forecasts anticipated an equal chance of warm, cold or normal weather. (see images 3, 4, forecasted temperature anomaly Jan-Feb-Mar 2004 and observed temperature anomaly Jan-Feb-Mar 2004, U.S. only)
Snow cover has a pronounced impact on large-scale waves in the atmosphere, so the advantage of factoring in snow cover to improve winter forecasts is not limited to the United States. Cohen issued a real-time winter forecast for Europe for 2003-04, which was significantly more accurate than those issued by the European forecast centers. The departure from normal temperature forecast for the entire Northern Hemisphere was compared to the observed departure from normal temperatures. The forecast showed an almost exact correspondence to observed temperatures across most of the land areas of the Northern Hemisphere with the exception of northwest Canada.
“Such an on-target forecast is highly improbable,” says Cohen, “With techniques that heavily depend on El Nino for information. Regions impacted by snow variability differ from those influenced by El Nino. The influence of snow cover extent has the potential to complement El Nino-derived forecasts, and to advance our understanding of climate variability and its application in prediction models.”
Observed Eurasian snow cover on October 1, 2003, is compared with observed Eurasian snow cover on November 1, 2003, as seen from NOAA satellites. Green represents land areas, dark blue is ocean, light blue is sea ice and Eurasian snow cover is colored in white.
October is the month where snow cover undergoes its greatest expansion in the Eurasian region. During October, Eurasian snow cover can increase by as much as 10-15 million square kilometers, which is greater than the total land area of the United States, including Alaska. NSF-funded research has shown that variability in the extent of Eurasian snow cover can be used to predict cold or warm winters across the entire mid-latitudes of the Northern Hemisphere.