Writing in July 2010 when the eastern U.S. has been having one heat wave after another, it is easy to forget the winter just past, but winter 2009/10 was one of record snow in much of the mid-Atlantic region and very cold conditions across eastern North America. There was also widespread cold and snow in northwestern Europe.

During one storm, Senator Inhofe's family built an igloo that they named 'Al Gores' new home'.
At the time, as record snowfall hit Washington D.C., this was a big deal much exploited by climate deniers and others to ridicule global warming and climate science. Of course that confused climate and weather and took a rather regional perspective - failing, for example, to note the lack of snow in the Pacific Northwest or the warm temperatures in the subpolar and polar regions. Nonetheless it is always a good idea to try to explain anomalous seasonal weather both to search for any predictability and also to educate the wider public about causes and to offset attempts to exploit weather events to derail efforts to tackle climate change.

In a just published paper in Geophysical Research Letters (PDF) we attempt to offer an explanation of the northern hemisphere snow anomalies of winter 2009/10. The simple answer is a combination of an El Niño event and a strongly and persistently negative North Atlantic Oscillation. El Niños are warmings of the tropical Pacific Ocean that occur every few years as a result of natural atmosphere-ocean interaction in the tropical Pacific region and are predictable up to a few seasons in advance. The North Atlantic Oscillation (NAO) is a seesaw of atmospheric pressure above the North Atlantic Ocean between the Icelandic Low and the Azores high and arises from purely atmospheric dynamics and is not predictable beyond the week or two timescale of weather forecasting.


Observed relation of U.S. and northern hemisphere snow anomalies to ENSO and the NAO

It turns out that the snow data was a bit of a mess - there was no long and continuous data sets of snowfall updated to the current month. Instead we used two data sets: one was for snowfall at weather stations in the US and went from 1950 to 1999 and the other was hemispheric, satellite-derived snow water equivalent (SWE) from 1979 to 2007. We took the December through March values of these and correlated and regressed them with the Nino3 index of tropical Pacific sea surface temperatures (SSTs, the Nino3 index is SST anomaly averaged 5S-5N, 150W-90W) and the NAO and with a new NINO-NAO index (the normalized NINO3 SST index minus the normalized NAO index). Figure 1 (link) shows the relations for U.S. snowfall and Figure 2 for hemispheric SWE.

For U.S. snowfall an El Niño is associated with generally high snowfall in the southern Rockies, in the southeast and along the east coast and less snowfall to the north of these regions. A negative NAO has stronger correlations with snowfall and brings increased snowfall to the mid-Atlantic states. The correlation with an El Niño and a negative NAO (i.e. with the NINO-NAO index) shows a north-south dipole with snowier conditions to the south and less snowy to the north. The relations with satellite-derived SWE (see Figure 2) are consistent with snowfall over North America and also show that a negative NAO brings increased snow to northwest Europe.


Causes of the relation between snow anomalies and ENSO and the NAO

We then examined the causes of the snow anomalies. They could arise from change in total precipitation or from a change in the proportion falling as snow instead of rain. We know from prior work that El Niño brings more precipitation to the southern U.S. (coast to coast) and less to the Pacific Northwest so the increased snow in the southwest and less in the Pacific Northwest is partly explained by this. A negative NAO has little correlation to total precipitation amount over the U.S. and actually causes reduced precipitation over northwest Europe. However a negative NAO also brings cold to the North Atlantic sector from eastern North America into Europe. Hence, in these regions, we expect more of the total precipitation to fall as snow. Figure 3 (link) shows the relation of lower atmospheric temperature to the NINO3, NAO and NINO-NAO indices and the cold over the southern U.S. and Europe are very clear (with warm in the subolar regions). Figure 3 also shows a measure of the location of storm tracks (the upper tropospheric variance of meridional velocity) and demonstrates that both El Niño and a negative NAO cause the northern hemisphere winter storm track to shift south which enables snowstorms to strike across the southern and central U.S.


Explaining the winter snow anomalies of the winter 2009/10

In winter 2009/10 the Nino3 SST anomaly averaged about 1°C, a decent sized El Niño. The NAO index for December 2009 through February 2010 was a whopping -2.38 standard deviations (making this a highly unusual event). Using these values we can calculate the NINO-NAO index value for 2009/10 and then use the relations in Figures 1 and 2 to compute expected snowfall and SWE for the winter. These are shown in Figure 4 (link). As stated above, we do not have continuous snow data sets that both cover past decades that are also updated to current. Instead, to verify our expected snow anomalies against, we show the snowfall anomaly from a recent data set (the 2009/10 values minus the average of the three previous winters) and satellite-derived snow cover anomalies (which, note, are zero in regions that are always snow covered in winter). The expected values match the observed snow anomalies quite well confirming that the snow anomalies of the past winter were caused by the combination of an El Niño event with the negative NAO.

Of the two climate phenomena, the NAO is the dominating influence in the eastern U.S. and northwest Europe and, since it cannot be predicted beyond the timescale of weather forecasting, this means that snowy winters like this past one will remain a seasonal surprise. In summary, what happened this past winter is just another example of the kind of seasonal climate anomaly that can arise from purely natural variability of the atmosphere-ocean system. There seems no evidence that it had anything to do with climate change and should not be exploited to make arguments, one way or another, about the reality of that (which we do not doubt) or how to tackle it.

The work was carried out by Richard Seager, Yochanan Kushnir, Jennifer Nakamura, Mingfang Ting and Naomi Naik, all of the Lamont Doherty Earth Observatory, Earth Institute, Columbia University.

A research team at NOAA have come to similar conclusions using different analyses. See http://www.climatewatch.noaa.gov/2010/articles/forensic-meteorology-solves-the-mystery-of-record-snows.




Figure 1. The correlation of snowfall with the NINO3 index (top left), the NAO index (bottom left) and the standardized NINO3 minus standardized NAO (NINO-NAO) index and the regression of snowfall on the NINO-NAO index (bottom right). All indices and the snowfall are for the winter (December to March) mean. Units for the regression are inches.

Figure 1


Figure 2. Same as Figure 1 but for snow water equivalent. Units for the regression are mm.

Figure 2


Figure 3. The regression of the 850mb temperature on the NINO3, NAO and NINO-NAO indices (color) with the mean 850mb temperatare contoured (0°C isotherm bold, negative contours dashed, left column), and the regression of 300mb submonthly transient eddy meridional velocity variance, v'2, (right column) all for the December through March seasonal mean. Patterns are plotted only where significant at the 5% level. Units are deg C for temperature and m2/s2 for velocity variance.

Figure 3


Figure 4. Attribution of snowfall (top left, inches) and snow water equivalent (top right, mm) for December 2009 through February 2010 based on the regression patterns shown in Figures 1 and 2 and the seasonal mean observed NINO-NAO index. The bottom left panel shows the difference in observed station snowfall between the December 2009 to February 2010 season and the three previous December through February seasons (from NOHRSC data) and the bottom right panel shows the observed, satellite-derived, snow cover anomaly in percent for February 2010. (Snow cover anomalies are zero in subpolar and polar regions that are always snow covered).

Figure 4

REFERENCES