Presentation Narrative for AAG National Conference, March 14-19, 2004
Session: US Regional Climatology
Speaker: Melissa Malin

A Synoptic Climatological Approach to the Identification of January Temperature
 Anomalies in the United States

Slide 1: Today I am presenting preliminary results of research that was conducted at the University of Delaware by myself and four colleagues, all of whom work in climatology though only some of us in synoptics. Despite our more variable backgrounds, we became interested in this topic after reading a Bulletin of the American Meteorological Society article in 2002 that questioned the January thaw as a statistical phenomenon.

Slide 2: To give you a brief background on the January thaw, it appears to have originated out of New England weather folklore over a century ago and since gained scientific recognition, even outside of New England.  It is considered a singularity that the American Meteorological Society defines as a “…characteristic meteorological condition that tends to occur on or near a specific calendar date.” It is a particularly popular singularity in that it is associated with anomalous warming that comes around the coldest time of year. It is also a particularly controversial singularity as to date a causal mechanism for the thaw has not been identified, or at least agreed upon. Historically, it has been linked to atmospheric, oceanic, and even extra-terrestrial forcings.

Slide 3: With that in mind, our goals were to see if first, we could identify a thaw signal across the country and if so, assess the inter- and intra- regional variability of the thaw. The key here is that we’re going to use a synoptic climatological approach to explain the thaw through an assessment of changes in air mass frequency. So it might be that when the thaw occurs very cold air mass types are less frequent or very warm air mass types are more frequent.

Slide 4: We selected five regions across the country, each made up of five spatially variable stations with meteorological records containing less than 3% missing data.

Slide 5: Our data were for a period of 52 winters that we called December 1st – February 28th, as we were mainly interested in January temperature anomalies. Our weather data came from the National Climatic Data Center and our air mass data from the Spatial Synoptic Classification. It identifies days as one of the seven air mass types you see here on the left. The three air masses on the right are subcategories. You can consider these the extremes of the air mass types of their namesake. For example, the Dry Polar- represents the coldest and driest quadrant of the Dry Polar air mass type. We subcategorized air masses if more than 20% of the days were classified as that particular type. This was beneficial to us to see if it was the purest days of the most frequent air mass types that were associated with anomalous temperatures.

Slide 6: To start, we plotted average daily temperatures at each station. Here, you see the data for Philadelphia, Pennsylvania. These data were standardized using a five-day moving window. Window 1 represents the average temperature of December 1st – December 5th. Each window was identified by the date of its third day, so window 1 was identified as December 3rd.

Slide 7: We then fit a second-order polynomial curve to the data to act as a winter trendline. If at any point in the season a window temperature was greater than two standard deviations from this trendline, we considered it a singularity.  Here in Philadelphia, you see a singularity at January 24th and 25th.

Slide 8: To show you an example of singularities detected in the western US, here you see the data for Cheyenne, Wyoming where singularities occurred at January 2nd – 4th and January 16th – 18th.

Slide 9: After doing this for each station, we detected cohesive January thaw signals in every region of the country. By cohesive I mean that the date of the thaw was similar at every station in a region (with an exception for the Great Plains where we found a north-south dichotomy with when the thaw occurred). Perhaps more interesting, the thaw appeared to be moving systematically across the country, again, with a bit of an exception in the Great Plains. Now, I can show you when the thaw occurred across the country.  I should point out that though a majority of thaw dates in the Great Plains were between December 26th and 28th, at more southern stations several thaw dates were between the 17th and 22nd of January.

Slide 10: We also found cohesive January freeze signals in every region but the East. A January freeze is the term we would use for a singularity that fell below our lower bounds, or a negative temperature anomaly. The freeze appeared to move across parts of the country, though unlike the thaw not across the entire US.

Slide 11: For our synoptic analysis, we plotted individual air mass frequencies at each station and fit a linear trendline to the data. This was under the assumption that air masses display a general frequency trend across a season. So, it might be that warm air mass types are generally less frequent across the winter season or cold air mass types are more frequent. Here in Philadelphia, the Moist Polar + (MP+) air mass type is generally less frequent, likely a result of the source regions in the North Pacific and North Atlantic. The circulation patterns required to advance these air masses into the Mid Atlantic are rare and this air mass type often modifies before reaching Philadelphia. What we were looking for here were differences in air mass frequency at the date of the singularity to the trendline.  After finding these differences for each air mass, we did the exact same thing for temperature (taking the difference of the temperature at the date of the singularity to the temperature trendline).

Slide 12: We compared these differences by performing correlation analyses for each region and then for the entire US. Here you see the differentials acquired for all of the eastern region January thaws. Highlighted in orange are temperature differences, measured in degrees. (These are all positive values because they represent positive temperature anomalies.) To the right are air mass frequency differences in percentages. These are positive or negative depending on whether or not the frequency was greater than or less than the trendline value. To give you a quick example,  in row 1 Providence, Rhode Island has a January thaw at window 52. At this time the temperature was 1.16º above average. At the same time, the warm Dry Moderate (DM) air mass type was approximately 4% more frequent.  Two columns over, you see that the cold Dry Polar (DP) air mass type was approximately 17/ 18% less frequent at this time.  Comparing these values statistically produces correlation coefficients, r values, that range between –1 and 1, where ±  0.8 is generally considered a strong correlation. That isn’t what we found in the East, as you can see here. In fact, the January thaw was weakly correlated to changes in air mass frequency in this region.

Slide 13: Overall, our synoptic analysis indicated that the January thaw could not be correlated to changes in air mass frequency in the East and West regions. However, in the Midwest, the January thaw could be correlated to a decrease in frequency of the very cold Dry Polar- (DP-) air mass type. Here, an r value of –0.87 is pretty good (negative because of a positive temperature anomaly and a decrease in frequency). The thaw was also correlated to an increase in frequency of the warm Moist Moderate (MM) air mass type. These results intuitively make sense. In the Great Plains we found something similar, where the thaw was correlated to a decrease in frequency of all cold air mass types though less correlated to warm air mass types when they were more frequent. In the Mountain region, the January thaw is moderately correlated to the very warm, wet Moist Tropical (MT) air mass. However, I have the r value starred because it is in the wrong direction. So, where you would expect the thaw to be correlated to this air mass when it is more frequent, we actually found the opposite. For the entire US, the January thaw was only correlated to the DP- when it was less frequent. As no one condition really stands out here I pose the question, “Is it perhaps that when the thaw is occurring when cold air mass types are less frequent?” It sounds reasonable and certainly is the condition that we found the thaw most correlated to.

Slide 14: Our results were better for the January freeze. Overall, our synoptic analysis indicated that the freeze could not be correlated to changes in air mass frequency in the Midwest, supporting the freeze as a more western phenomenon. In the Great Plains the freeze was correlated to the cold DP- air mass when it was more frequent and the warm MM when it was less frequent. (Here, there is a negative r value because of a negative temperature anomaly and an increase in air mass frequency.) R values of approximately ± 0.7 are quite good.  In the Mountain region, we found the freeze was correlated to an increase in frequency of the same cold air mass type, the DP-. In the west, though the air mass types are different the same idea applies. The freeze is correlated to a more frequent cold air mass type and a less frequent warm air mass type. In this case, r values of approximately ± 0.9 indicate very strong correlations. So, for the January freeze I pose the question, “Is it perhaps an increase in frequency of cold air mass types in combination with a decrease in frequency of warm air mass types that cause the singularity?”

Slide 15: To conclude, these preliminary research results seem to be strong support for the existence of thaw and freeze events across the country. Our analysis of the spatial variability of these events indicate both singularities have cohesive intra-regional signals and somewhat consecutive east to west inter-regional signals. The thaw signal breaks up in the middle of the country and the freeze signal is western-biased. However, this trend was generally detected. To explain this trend is beyond the scope of this research. Perhaps these events are linked to circulation. Regardless, it is something well-worth looking into in the future. Our assessment of the thaw and freeze, synoptically, uncovered that very cold air mass types had the best correlation to the thaw, in particular the coldest days in a season, represented by the DP-. I highlight that these results were better for the January freeze and when in combination with warm air mass types responding in an opposite fashion. At the moment this is an ongoing investigation, where we have acquired from a website Spatial Synoptic Classification-identified monthly averages of air mass frequency. We can use these averages in our synoptic analysis of differentials and find the difference between air mass frequency at the date of the singularity to these values. By then performing our correlation analysis we can compare the two methodologies and hopefully find the conditions necessary to produce the January thaw and freeze events that, in the US, are occurring year after year.