Recent January DCA Temperatures Appear Inconsistent with 1981-2010 Normals

[RodneyS]

If the 1981-2010 temperature normals for Reagan National Airport are taken at face value, DCA temperatures on average bottom out during January 12th-17th (at a mean of 35.7 degrees), then rise gradually (to 35.9 degrees on January 21st, 37.2 degrees on February 5th, and peak between July 12th-21st at 80.0 degrees). However, January daily temperatures during this period -- and even more so during the most recent 30 years, 1984-2013 -- appear inconsistent with those normals. During the past 30 years, the first 14 days of January at DCA have each averaged above the 1981-2010 normals, whereas January 19th-23rd have each averaged below those normals. This may be illustrated by comparing January 13th to January 21st average temperatures during the past 30 years at DCA, as well as looking at the historical DC temperature record for those two days and the coldest average day of the year.

For the first 112 years of official DC daily temperature history -- 1872-1983 -- January 13th was the coldest January day, at 33.3 degrees on average; and January 21st was the second warmest January day, at 36.8 degrees on average. The January temperature peaked on the 22nd at 37.2 degrees on average during 1872-1983, with the annual mean temperature bottoming out on February 5th at 32.6 degrees on average.

However, during 1981-2010, the mean temperature at DCA on January 13th was 38.7 degree on average, decreased sharply to an annual minimum of 32.2 degrees on average on the 21st, and rose about halfway back to 35.3 degrees on average on February 5th. Further, the disparity with 1872-1983 has widened during 1984-2013, with January 13th the warmest January day of this most recent 30-year period at 40.3 degrees on average, January 21st the coldest day of the year at 32.1 degrees on average, and February 5th the coldest February day (29 years through 2012) at 35.9 degrees on average.

To recap, January 13th has gone from being the coldest January day at 33.3 degrees on average during 1872-1983 to the warmest January day at 40.3 degrees on average during 1984-2013 -- an increase of 7.0 degrees. January 21st has been the mirror image of January 13th -- going from the second warmest January day at 36.8 degrees on average during 1872-1983 to the coldest day of the year at 32.1 degrees on average during 1984-2013 -- a decrease of 4.7 degrees. Finally, February 5th has gone from being the coldest day of the year at 32.6 degrees on average during 1872-1983 to merely the coldest February day at 35.9 degrees on average during 1984-2012 -- an increase of 3.3 degrees.

The increase in temperature on February 5th between these two periods has been pretty clearly within the range of what would be expected in an era of warming temperatures, as February temperatures 1984-2012 have been 2.9 degrees warmer on average than during 1872-1983. But what about the January temperature record, and in particular, the contrast between January 13th and 21st during these two periods? Is that merely an aberration, or is a more fundamental change taking place that is not reflected in the current temperature normals?

[RodneyS]

If the 1981-2010 temperature normals for Reagan National Airport are taken at face value, DCA temperatures on average bottom out during January 12th-17th (at a mean of 35.7 degrees), then rise gradually (to 35.9 degrees on January 21st, 37.2 degrees on February 5th, and peak between July 12th-21st at 80.0 degrees). However, January daily temperatures during this period -- and even more so during the most recent 30 years, 1984-2013 -- appear inconsistent with those normals. During the past 30 years, the first 14 days of January at DCA have each averaged above the 1981-2010 normals, whereas January 19th-23rd have each averaged below those normals. This may be illustrated by comparing January 13th to January 21st average temperatures during the past 30 years at DCA, as well as looking at the historical DC temperature record for those two days and the coldest average day of the year.

For the first 112 years of official DC daily temperature history -- 1872-1983 -- January 13th was the coldest January day, at 33.3 degrees on average; and January 21st was the second warmest January day, at 36.8 degrees on average. The January temperature peaked on the 22nd at 37.2 degrees on average during 1872-1983, with the annual mean temperature bottoming out on February 5th at 32.6 degrees on average.

However, during 1981-2010, the mean temperature at DCA on January 13th was 38.7 degree on average, decreased sharply to an annual minimum of 32.2 degrees on average on the 21st, and rose about halfway back to 35.3 degrees on average on February 5th. Further, the disparity with 1872-1983 has widened during 1984-2013, with January 13th the warmest January day of this most recent 30-year period at 40.3 degrees on average, January 21st the coldest day of the year at 32.1 degrees on average, and February 5th the coldest February day (29 years through 2012) at 35.9 degrees on average.

To recap, January 13th has gone from being the coldest January day at 33.3 degrees on average during 1872-1983 to the warmest January day at 40.3 degrees on average during 1984-2013 -- an increase of 7.0 degrees. January 21st has been the mirror image of January 13th -- going from the second warmest January day at 36.8 degrees on average during 1872-1983 to the coldest day of the year at 32.1 degrees on average during 1984-2013 -- a decrease of 4.7 degrees. Finally, February 5th has gone from being the coldest day of the year at 32.6 degrees on average during 1872-1983 to merely the coldest February day at 35.9 degrees on average during 1984-2012 -- an increase of 3.3 degrees.

The increase in temperature on February 5th between these two periods has ben pretty clearly within the range of what would be expected in an era of warming temperatures, as February temperatures 1984-2012 have been 2.9 degrees warmer on average than during 1872-1983. But what about the January temperature record, and in particular, the contrast between January 13th and 21st during these two periods? Is that merely an aberration, or is a more fundamental change taking place that is not reflected in the current temperature normals?

The beat goes on.  January 13th reached 61 degrees this year, and averaged 46.5, vs 29.0 on January 21st.   January 13th -- once the coldest January day -- has now averaged 40.5 since 1984, vs 31.8 on January 21st.  Why so warm in early January in recent years, and why do the January cold waves now occur around the 21st?

[jnis]

Interesting analysis...that is a tough question? Thanks for sharing.

[LongRanger]

what you see is due to smoothing of the data so the normal means plot along a sine-wave-like curve

the same smoothing had been done on the 1971-2000 data as well, if it had not a January thaw would have appeared in a plot of the POR means

[RodneyS]

what you see is due to smoothing of the data so the normal means plot along a sine-wave-like curve

the same smoothing had been done on the 1971-2000 data as well, if it had not a January thaw would have appeared in a plot of the POR means.

Thanks for your comment, but I think the question remains whether something has fundamentally changed in recent years. Consider that, during the first 112 years of DC temperature records, January 13th had a lower average daily temperature than January 21st 70 times (62.5%), with three ties (2.7%).  However, during the most recent 31 years, January 13th had a lower average temperature than January 21st only five times (16.1%), with no ties.  Think of this situation in a sports context:  Suppose a baseball or hockey team had a winning percentage of .638 over its first 112 games (counting each of the three ties as half a win), and then plummeted to a winning percentage of only .161 over its most recent 31 games.  Would you say that it's more likely that those sharply varying winning percentages can be explained by: (1) random fluctuations, or (2) a fundamental change in the team's performance?

[MN Transplant]

Thanks for your comment, but I think the question remains whether something has fundamentally changed in recent years. Consider that, during the first 112 years of DC temperature records, January 13th had a lower average daily temperature than January 21st 70 times (62.5%), with three ties (2.7%).  However, during the most recent 31 years, January 13th had a lower average temperature than January 21st only five times (16.1%), with no ties.  Think of this situation in a sports context:  Suppose a baseball or hockey team had a winning percentage of .638 over its first 112 games (counting each of the three ties as half a win), and then plummeted to a winning percentage of only .161 over its most recent 31 games.  Would you say that it's more likely that those sharply varying winning percentages can be explained by: (1) random fluctuations, or (2) a fundamental change in the team's performance?

 

Rodney, have you split it out before and after the move to the river location?

[LongRanger]

random fluctuations since averages of smaller sample sizes always have more outliers than averages of larger samples

[RodneyS]

 Rodney, have you split it out before and after the move to the river location?

I have now done that.  First, a slight error crept into my previous calculations, as during the most recent 31 years, January 13th had a lower average temperature than January 21st six -- rather than five -- times.  Thus, the percentage changes slightly, from 16.1% to 19.4%.  Sorry for the error, but 19.4% remains a long ways from 63.8%.  Second, regarding your question, DCA became the official DC weather location on July 1, 1945, and so the first 74 Januaries reflect the previous 24th and M Street, NW location.  During those 74 years, January 13th had a lower average temperature than January 21st 48 times (64.9%), with one tie (1.4%).  In the next 38 years -- prior to the most recent 31 years -- January 13th had a lower average temperature than January 21st 22 times (57.9%), with two ties (5.3%).  So, the percentage did lessen somewhat immediately following the location change, but not all that much.   Overall, since the location change, January 13th has had a lower average temperature than January 21st 28 times (40.6%), with two ties (2.9%).  

[RodneyS]

random fluctuations since averages of smaller sample sizes always have more outliers than averages of larger samples

 

In that case, you are definitely in the minority among sports fans.    But let me ask you this:  What would it take to convince you that something more than random fluctuations were taking place?

[JBG]

The increase in temperature on February 5th between these two periods has been pretty clearly within the range of what would be expected in an era of warming temperatures, as February temperatures 1984-2012 have been 2.9 degrees warmer on average than during 1872-1983. But what about the January temperature record, and in particular, the contrast between January 13th and 21st during these two periods? Is that merely an aberration, or is a more fundamental change taking place that is not reflected in the current temperature normals?

I am quite sure, if we had good records going back to the beginning of the "Common Era" (or for Christians the time of Christ) there'd be thousands of fluctuations during moving 30-year periods. That is why I have long suggested using 50 year base periods, now that we have good records at most stations. I think there is a lot of statistically insignificant "noise." Add to that the effect of "smoothing."

One note in the other direction. The 30 year averages, for the longest time, gave KNYC a mean of 78 for a week or so in mid-to-last July. That has now been "smoothed" to make the 77 figure hold as the mean from about July 10 to August 10. And the mean high temperature has been "reduced" from 85 to 84. Do these changes mean anything?

[RodneyS]

I am quite sure, if we had good records going back to the beginning of the "Common Era" (or for Christians the time of Christ) there'd be thousands of fluctuations during moving 30-year periods. That is why I have long suggested using 50 year base periods, now that we have good records at most stations. I think there is a lot of statistically insignificant "noise." Add to that the effect of "smoothing."

One note in the other direction. The 30 year averages, for the longest time, gave KNYC a mean of 78 for a week or so in mid-to-last July. That has now been "smoothed" to make the 77 figure hold as the mean from about July 10 to August 10. And the mean high temperature has been "reduced" from 85 to 84. Do these changes mean anything?

January 21st did manage to nose out January 13th in DC for a warmer average temperature this year, 36.5 to 35.5.  So, the updated DC numbers for the two days are as follows: During the first 112 years of official DC temperature records, January 13th had a lower average daily temperature than January 21st 70 times (62.5%), with three ties (2.7%).  During the most recent 32 years, January 13th had a lower average temperature than January 21st seven times (21.9%), with no ties.  January 22nd has also grown much colder in DC since 1984, although this year it was a balmy 42.5.  So, perhaps that's some support for your position.  Still, it's striking how few cold waves have occurred in DC in mid-January for more than 30 years, while they have been common in the latter third of the month during that same period. 

 

Regarding your point about smoothing of temperature data, that's actually my point as well.  In other words, the "normal" temperatures reflect underlying assumptions about temperature trends, which may not be valid in all instances.   

[WEATHER53]

DCA has a lot of problems including being statistically non functional.

[Roger Smith]

Is a sports analogy valid here? The weather data in question are sets that are very close together. Large variations in frequency could be anticipated when looking at higher and lower data points among such sets. The only real test would be random variability as in coin toss frequencies and while the change looks rather large, it probably falls within about a 2-s.d. range of random expectation. So I don't know if we can really deduce much of a "reason" for these results, it certainly isn't as indicative of a background explanation as that much of a change in a sports team's performance. I am going to look into it further since I am always intrigued by singularities that move gradually in a time series from my research perspective.

 

One example would be the January thaw. If there were such an animal and it was expected around Jan 21, then if it had an external forcing cause, you could expect that to occur 1.5 days earlier every century, on average, in a constant Gregorian time frame. That is due to precession of the earth's axis in space. Let's say there are weak but significant resonances between the Sun and the mass of the galactic equator or various other fixed sources. Then these would occur 1.5 days later every century, or .015 days earlier every year. Nothing there that we could apply to changes in a 30-year normal set, but it's the kind of shift that I look for in weather data sets when assessing signal strength. Some other background factors move a lot faster, so I'll have a look at what actual differences you're noting and see if I can see anything.

 

The other factor to consider is that micro-climates have quite large signals in data, if you re-site an airport weather station it can lead to "normal" shifts of up to 3 F deg (!) and not every type of weather makes the same shift when you re-site. Just the very local radiation budget is quite significant. I did an experiment as part of my own education on boundaries of urban heat islands and found that the "normal" monthly temperature at a normally situated weather station would vary by that 3 F deg amount if you just moved the instruments as little as 30 feet and changed the surrounding environment by as little as having a couple of small trees nearer to the thermometer, or an unused unheated building closer. It was the equivalent of driving thirty miles north or south to a similar site at similar elevation to your control site.

[Roger Smith]

Okay, had a look at my Toronto data (1841 to 2013) which should run fairly similar in trends. The same things can be seen in this data as what you've reported. But as I looked at 1984 to 2013, I have the advantage of having columns of daily data and could easily see that four years (1995, 2005, 2006, 2013) had massive spikes on the 12th and 13th. Take those years out of the data and the other 26 years show almost no anomalous warmth on 12th-13th. The two curves (30-year and 26-selected-years) are almost identical except that there is a 3.5 F drop for the 12th and 13th (and 1.5 on the 30th). So the large spike in the data for 1984-2013 are almost entirely due to the anomalous warmth of four of the thirty years.

 

As to any other explanations, nothing comes to mind. Just seems that the recent 30 years were dominated by these four very warm episodes as well as the super cold episode of Jan 1994 which leaves an imprint a few days later (15th in my Toronto data is a full 9 deg colder than 13th in all years 1984-2013, and 5 deg colder in the 26 years not counting the four warm ones). 1984 also had a very cold interval around the mid-month dates.

 

Longer term, what I noticed by dividing the data into large time segments and fiddling with boundaries was that from about 1841 to 1872 (before your data set starts) the data are much colder in the first half of January than since 1872, whereas the data in second half have not warmed that much if any. Also, the so-called January thaw can be seen drifting later through the data sets from 17th-19th in the middle of the 19th century, to 21st-22nd in the 20th century, but more like 23rd-24th in more recent times, also that signal has weakened somewhat relative to all data and shows signs of a split, as though one component of it wants to stay fixed on 21st but other components are drifting forward in time.

 

Does this help in any way?

[RodneyS]

Okay, had a look at my Toronto data (1841 to 2013) which should run fairly similar in trends. The same things can be seen in this data as what you've reported. But as I looked at 1984 to 2013, I have the advantage of having columns of daily data and could easily see that four years (1995, 2005, 2006, 2013) had massive spikes on the 12th and 13th. Take those years out of the data and the other 26 years show almost no anomalous warmth on 12th-13th. The two curves (30-year and 26-selected-years) are almost identical except that there is a 3.5 F drop for the 12th and 13th (and 1.5 on the 30th). So the large spike in the data for 1984-2013 are almost entirely due to the anomalous warmth of four of the thirty years.

Thanks for the response, Roger.  However, DC temperature data show much more than a handful of anomalous warm days recently on January 13th.  I've already compared that day to January 21st for the periods 1872-1983 and 1984-2015.   Here now is the same comparison for January 13th and January 22nd:  During the period 1872-1983, January 13th averaged 33.3 degrees (coldest January day, and third coldest day of the year), compared with 37.2 degrees on January 22nd (warmest January day).  January 13th was the colder day 69 of those 112 years, or 61.6% of the time.   During the period 1984-2015, January 13th averaged 40.3 degrees (warmest January day), compared with 32.1 degrees on January 22nd (coldest day of the year).  January 13th was the colder day just 7 of those 32 years, or 21.9% of the time.   

[RodneyS]

Thanks for the response, Roger.  However, DC temperature data show much more than a handful of anomalous warm days recently on January 13th.  I've already compared that day to January 21st for the periods 1872-1983 and 1984-2015.   Here now is the same comparison for January 13th and January 22nd:  During the period 1872-1983, January 13th averaged 33.3 degrees (coldest January day, and third coldest day of the year), compared with 37.2 degrees on January 22nd (warmest January day).  January 13th was the colder day 69 of those 112 years, or 61.6% of the time.   During the period 1984-2015, January 13th averaged 40.3 degrees (warmest January day), compared with 32.1 degrees on January 22nd (coldest day of the year).  January 13th was the colder day just 7 of those 32 years, or 21.9% of the time.   

I just ran a two-tailed T-Test for the periods 1872-1983 and 1984-2015 to determine the probability that the temperature differences between January 13th and January 22nd in DC for each period are attributable to random factors, or statistical noise.  For the first period, the probability is 0.375%, or about one chance in 267.  For the second period, the probability is 0.088%, or about one chance in 1132.  The results for the second period are particularly interesting, as the 1981-2010 "normals" for DC indicate that temperatures bottom out in mid-January.  Thus, if those normals are taken at face value, January 22nd should be slightly warmer than January 13th, not eight degrees colder. 

[RodneyS]

On 1/22/2013 at 8:31 AM, RodneyS said:

If the 1981-2010 temperature normals for Reagan National Airport are taken at face value, DCA temperatures on average bottom out during January 12th-17th (at a mean of 35.7 degrees), then rise gradually (to 35.9 degrees on January 21st, 37.2 degrees on February 5th, and peak between July 12th-21st at 80.0 degrees). However, January daily temperatures during this period -- and even more so during the most recent 30 years, 1984-2013 -- appear inconsistent with those normals. During the past 30 years, the first 14 days of January at DCA have each averaged above the 1981-2010 normals, whereas January 19th-23rd have each averaged below those normals. This may be illustrated by comparing January 13th to January 21st average temperatures during the past 30 years at DCA, as well as looking at the historical DC temperature record for those two days and the coldest average day of the year.

For the first 112 years of official DC daily temperature history -- 1872-1983 -- January 13th was the coldest January day, at 33.3 degrees on average; and January 21st was the second warmest January day, at 36.8 degrees on average. The January temperature peaked on the 22nd at 37.2 degrees on average during 1872-1983, with the annual mean temperature bottoming out on February 5th at 32.6 degrees on average.

However, during 1981-2010, the mean temperature at DCA on January 13th was 38.7 degree on average, decreased sharply to an annual minimum of 32.2 degrees on average on the 21st, and rose about halfway back to 35.3 degrees on average on February 5th. Further, the disparity with 1872-1983 has widened during 1984-2013, with January 13th the warmest January day of this most recent 30-year period at 40.3 degrees on average, January 21st the coldest day of the year at 32.1 degrees on average, and February 5th the coldest February day (29 years through 2012) at 35.9 degrees on average.

To recap, January 13th has gone from being the coldest January day at 33.3 degrees on average during 1872-1983 to the warmest January day at 40.3 degrees on average during 1984-2013 -- an increase of 7.0 degrees. January 21st has been the mirror image of January 13th -- going from the second warmest January day at 36.8 degrees on average during 1872-1983 to the coldest day of the year at 32.1 degrees on average during 1984-2013 -- a decrease of 4.7 degrees. Finally, February 5th has gone from being the coldest day of the year at 32.6 degrees on average during 1872-1983 to merely the coldest February day at 35.9 degrees on average during 1984-2012 -- an increase of 3.3 degrees.

The increase in temperature on February 5th between these two periods has been pretty clearly within the range of what would be expected in an era of warming temperatures, as February temperatures 1984-2012 have been 2.9 degrees warmer on average than during 1872-1983. But what about the January temperature record, and in particular, the contrast between January 13th and 21st during these two periods? Is that merely an aberration, or is a more fundamental change taking place that is not reflected in the current temperature normals?

Time for an update.  The 1984-2013 trend that I observed six years ago regarding January DCA temperatures peaking on or about the 13^th and bottoming on or about the 21st has continued.  If that trend during the period 1984-2013 were only a temporary aberration, since that time January 13^th would be as likely to be colder than January 21^st as it would be to be warmer.  However, during 2014-19, January 13^th DCA temperatures averaged 39.1 degrees, whereas January 21^st DCA temperatures averaged only 34.4 degrees.  That brings the average DCA temperature during 1984-2019 to 40.1 degrees on January 13^th vs. only 32.6 degrees on January 21^st.   A statistical test (two-tailed T-test) indicates that it is highly improbable that this 7.5-degree difference for that 36-year period is a random variation, with the odds of that being only about 1 in 2200.

Moreover, in the not-too-distant future, NOAA's National Climatic Data Center (NCDC, now part of the National Centers for Environmental Information in Asheville, NC) will begin to calculate its new daily 30-year temperature normals for the period 1991-2020. That period is now only one year shy of completion for January 13^th and January 21^st, and the average DCA temperature during 1991-2019 is 41.3 degrees on the former day and 32.6 degrees on the latter day.   A two-tailed T-test of this 8.7-degree difference for that 29-year period shows that difference is even more unlikely to be a random variation, with the odds of that being only about 1 in 8500.

So, how will the NCDC handle the fact that January temperatures at DCA now seem to peak around January 13^th and bottom around January 21^st? Does anyone here know someone at NCDC that works on temperature normals or know someone else (including yourself) that is knowledgeable in this area?

[RodneyS]

On 1/22/2019 at 9:41 AM, RodneyS said:

Time for an update.  The 1984-2013 trend that I observed six years ago regarding January DCA temperatures peaking on or about the 13^th and bottoming on or about the 21st has continued.  If that trend during the period 1984-2013 were only a temporary aberration, since that time January 13^th would be as likely to be colder than January 21^st as it would be to be warmer.  However, during 2014-19, January 13^th DCA temperatures averaged 39.1 degrees, whereas January 21^st DCA temperatures averaged only 34.4 degrees.  That brings the average DCA temperature during 1984-2019 to 40.1 degrees on January 13^th vs. only 32.6 degrees on January 21^st.   A statistical test (two-tailed T-test) indicates that it is highly improbable that this 7.5-degree difference for that 36-year period is a random variation, with the odds of that being only about 1 in 2200.

Moreover, in the not-too-distant future, NOAA's National Climatic Data Center (NCDC, now part of the National Centers for Environmental Information in Asheville, NC) will begin to calculate its new daily 30-year temperature normals for the period 1991-2020. That period is now only one year shy of completion for January 13^th and January 21^st, and the average DCA temperature during 1991-2019 is 41.3 degrees on the former day and 32.6 degrees on the latter day.   A two-tailed T-test of this 8.7-degree difference for that 29-year period shows that difference is even more unlikely to be a random variation, with the odds of that being only about 1 in 8500.

So, how will the NCDC handle the fact that January temperatures at DCA now seem to peak around January 13^th and bottom around January 21^st? Does anyone here know someone at NCDC that works on temperature normals or know someone else (including yourself) that is knowledgeable in this area?

Time for another update.  Once again this year at DCA, January 13th was much warmer than January 21st (48.5 degrees vs 31.5).  So, for the most recent 30 years (1991-2020) at DCA, January 13th has averaged 41.6 degrees vs. 32.6 degrees on January 21st.  A two-tailed T-Test shows that the odds of this being a random variation are less than 1 in 18,500.  So, how will the NCDC handle this inconvenient fact when it calculates its new daily 30-year January temperature normals at DCA?  The 1981-2010 January temperature normals there show temperatures bottoming out in mid-January, not January 21st.  Further, even if the new DCA normals were to reflect a bottoming out on January 21st, the January 13th normal would likely be calculated using standard methodology as only 0.2-0.3 degrees higher than January 21st, and not the actual 9 degrees higher.  Could someone at NCDC think way outside of the box and develop a radically new January temperature model, showing an annual thaw at DCA occurring around January 13th? And if s/he did, would s/he be given an award or fired?  A thorny problem for NCDC, it seems to me.

[RodneyS]

On 1/22/2020 at 12:48 PM, RodneyS said:

Time for another update.  Once again this year at DCA, January 13th was much warmer than January 21st (48.5 degrees vs 31.5).  So, for the most recent 30 years (1991-2020) at DCA, January 13th has averaged 41.6 degrees vs. 32.6 degrees on January 21st.  A two-tailed T-Test shows that the odds of this being a random variation are less than 1 in 18,500.  So, how will the NCDC handle this inconvenient fact when it calculates its new daily 30-year January temperature normals at DCA?  The 1981-2010 January temperature normals there show temperatures bottoming out in mid-January, not January 21st.  Further, even if the new DCA normals were to reflect a bottoming out on January 21st, the January 13th normal would likely be calculated using standard methodology as only 0.2-0.3 degrees higher than January 21st, and not the actual 9 degrees higher.  Could someone at NCDC think way outside of the box and develop a radically new January temperature model, showing an annual thaw at DCA occurring around January 13th? And if s/he did, would s/he be given an award or fired?  A thorny problem for NCDC, it seems to me.

I have now done some additional analysis that shows that in several other eastern cities January temperatures during the most recent 30 years have peaked on or about January 13th and bottomed on or about January 21st:

Raleigh-Durham: 46.9 vs 37.7 degrees

Richmond: 43.6 vs 34.0 degrees

Philadelphia: 38.4 vs 29.3 degrees

New York:  37.3 vs 28.9 degrees

Boston: 34.1 vs 26.2 degrees

Portland: 26.4 vs 19.5 degrees

 

[RodneyS]

On 1/22/2020 at 12:48 PM, RodneyS said:

Time for another update.  Once again this year at DCA, January 13th was much warmer than January 21st (48.5 degrees vs 31.5).  So, for the most recent 30 years (1991-2020) at DCA, January 13th has averaged 41.6 degrees vs. 32.6 degrees on January 21st.  A two-tailed T-Test shows that the odds of this being a random variation are less than 1 in 18,500.  So, how will the NCDC handle this inconvenient fact when it calculates its new daily 30-year January temperature normals at DCA?  The 1981-2010 January temperature normals there show temperatures bottoming out in mid-January, not January 21st.  Further, even if the new DCA normals were to reflect a bottoming out on January 21st, the January 13th normal would likely be calculated using standard methodology as only 0.2-0.3 degrees higher than January 21st, and not the actual 9 degrees higher.  Could someone at NCDC think way outside of the box and develop a radically new January temperature model, showing an annual thaw at DCA occurring around January 13th? And if s/he did, would s/he be given an award or fired?  A thorny problem for NCDC, it seems to me.

Recent DCA temperatures offer more support for the "January 13th thaw-January 21st freeze" hypothesis, with the former day in 2022 averaging 41.5 degrees and the latter day averaging 23.0 degrees.  The new NCDC 1991-2020 DCA normals now reflect a bottoming out of yearly temperatures in the January 18th-22nd time frame (at 37.0 degrees), with January 13th 0.2 degrees higher.  The previous 1981-2020 normals showed a bottoming out right around January 13th (at 35.7 degrees), with January 21st 0.2 degrees higher.  While this update is a step in the right direction by NCDC, it does not come close to reflecting the reality that during 1991-2020 at DCA January 13th temperatures averaged 9.0 degrees higher than January 21st (41.6 vs. 32.6).  

 

What convinces me that this differential is not merely a weird aberration, but rather an unexplained phenomenon, is that I first noticed it nine years ago, and it has continued.  Specifically, during the last nine years (2014-2022), January 13th temperatures at DCA have averaged 6.7 degrees higher than January 21st (40.4 vs. 33.7).   While this differential is not quite as high as during 1991-2020, my first calculation of the differential was for the period 1984-2013, when it averaged 8.2 degrees (40.3 vs. 32.1). So, the differential actually increased when that 30 year period was moved forward by seven years to 1991-2020, and while it has been somewhat lower during the most recent nine years, it is still for those nine years way beyond what would be expected if 1984-2013 had been an aberration.

 

What is particularly interesting about all of this is that for the first 112 years of official daily temperature recordkeeping in DC (1872-1983), there tended to be a January 13th freeze and a January 21st thaw.  Specifically, during that time frame, January 13th averaged 3.5 degrees cooler than January 21st.  So, while for the most recent 39 years in DC, January 13th has averaged 7.0 degrees warmer than for the first 112 years (40.3 vs 33.3), January 21st has averaged 4.3 degrees cooler (32.5 vs 36.8).  Why the big shift? Inquiring minds want to know.

[MN Transplant]

Does this hold for longer periods, like the 12th-14th and 20th-22nd?

[wasnow215]

Also it seems to me true winter went from Dec 10 to March 10. Now it seems like Jan 10 to April 1. Or so. For temperatures. 

[RodneyS]

2 hours ago, MN Transplant said:

Does this hold for longer periods, like the 12th-14th and 20th-22nd?

Yes, although the differences are not quite as extreme.  See below table, showing the average daily temperatures for January 12-14 combined and January 20-22 combined for the periods 1872-1983, 1984-2022, 1984-2013, 1991-2020, and 2014-2022:

Period          Jan 12-14    Jan 20-22   Difference

1872-1983     33.8             36.4            +2.6

1984-2022    39.3             33.6             -5.7

1984-2013     38.6             33.0             -5.6

1991-2020     40.5             33.7             -6.8

2014-2022     41.4             35.6             -5.8      

[Roger Smith]

I will look at some of my data sets to see what happens at other locations in this time frame. Did you spot any other similar anomalies during the rest of the year? I will post my findings in a few minutes, I already have the averages available for various periods, just need to check them. 

[Roger Smith]

In the Central England Temperature data, the period 12th to 14th Jan has become milder in recent decades while the 19th to 21st have become colder. For 1991-2020 averages, the 12th-14th were 5.2 C, 5.4 C and 5.2 C then the coldest day of the month is the 22nd (3.9 C). This is similar although less amplified than the signal you found. However for 1981-2010 the warmest interval was 19th-21st. This has disappeared from the data, and since the two intervals have twenty shared years (1991-2010), the differential represents a shift from the 1980s to the 2010s. This also seems in line with your findings if not directly overlapping them. For Toronto, which has data from 1840 to present, I found that averages in January followed a similar trend to what you've described. Will post this and edit in a graph as it's on a different device. 

This is a graph of mean daily temps (in C) for Toronto for both 1981-2021 and 1991-2021. It can be seen that there is a spike in mean temperature around the 12th, and considerably lower values around the 21st. So this phenomenon is not confined to DCA. I would expect that I will find it also in the NYC data that I have (looking at that next). Will speculate on a cause after thinking about it for a while. This 12th to 21st drop in mean temperature does not show up very prominently before 1980 and there was a "January thaw" signal in earlier decades that has a remnant in that smaller spike around the 23rd. 

[RodneyS]

1 minute ago, Roger Smith said:

In the Central England Temperature data, the period 12th to 14th Jan has become milder in recent decades while the 19th to 21st have become colder. For 1991-2020 averages, the 12th-14th were 5.2 C, 5.4 C and 5.2 C then the coldest day of the month is the 22nd (3.9 C). This is similar although less amplified than the signal you found. However for 1981-2010 the warmest interval was 19th-21st. This has disappeared from the data, and since the two intervals have twenty shared years (1991-2010), the differential represents a shift from the 1980s to the 2010s. This also seems in line with your findings if not directly overlapping them. For Toronto, which has data from 1840 to present, I found that averages in January followed a similar trend to what you've described. Will post this and edit in a graph as it's on a different device. 

Thanks, Roger.  You will note above that in my post of February 20, 2020 I found a similar differential for six other cities in the Mid-Atlantic and Northeast US.  There are other temperature differentials in that region during the course of the year, but the January 12-14 to January 20-22 differentials seem to be the most extreme by far.

[Roger Smith]

This is the same graph as above, with the 1840-1980 averages added. Of course that interval was generally a bit colder too due to climate change and urban heat island growth. However, the trends are much different with respect to the 12th spike in temperature: 

 

MEAN DAILY TEMPS (TORONTO CITY) red _ 1991-2021 ... blue _ 1981-2021 ... green _ 1840-1980

... note that the 1840-1980 temperatures show basically a gradual decline interrupted by the "January thaw" signal. 

[Roger Smith]

Since you have already compared the DCA data to other U.S. locations I won't bother to crunch the numbers for NYC then. Just thinking about any possible cause for this, especially as it clearly was not a component of the climate before 1980. (it may have been for some interval in the past, will check on that). Also it seems to transfer across the Atlantic or occurs almost simultaneously on both sides of the Atlantic. That would point to an external source, if it were a teleconnection, you would expect a lag of several days. The slightly earlier pulse at Toronto is probably due to that city being further west and the effect likely has to do with waves coming in from the Pacific, so the direction of travel of these signals may be WNW-ESE. Perhaps there is a sympathetic response wave in the eastern Atlantic. I have noted in some of my research that a spike in temperatures in eastern North America often appears either simultaneously or within 3-5 days in the British weather data. There are numerous (non-random in frequency) examples of similar dates of warm and cold records. But it is not a guaranteed 1:1 relationship. For example, the 1936 heat wave in North America never showed any sign of transferring across the Atlantic at all. 

If it is an external signal, the only plausible place to look would be in the Sun-galaxy interactions. A solar-planetary tidal phenomenon, while plausible for rotating sectors of about this magnitude, would not stay centered on any given part of the calendar year. A Sun-galaxy interaction would do so. January 12-13 is roughly two weeks after the Sun has crossed the galactic equator and there does seem to be a bit of a temperature spike in late December, but what would cause a two week lag for this? Possibly it has some effect on the Pacific circulation which takes some time to organize a response to send downstream. It seems like a process like that should be more spread out at random over 2-3 weeks however. Anyway, continuing to think about this. Something happens in this recent era that did not happen routinely in the past. As you say, it is getting beyond the bounds of being "one of those things" that statistical variability can produce, and more towards a semi-permanent feature. 

[RodneyS]

Interesting analysis, Roger.  I think this would be a great topic for a research-oriented meteorologist or PhD candidate to investigate.  

[Rhino16]

42 minutes ago, RodneyS said:

Interesting analysis, Roger.  I think this would be a great topic for a research-oriented meteorologist or PhD candidate to investigate.  

Still quite a few years away, but… I was thinking the same thing earlier.