
I’ll have a lot more on this study later, but for now just a short rebuttal.
I believe this study is hopelessly flawed due to the fact that the authors take the data from the weather stations at face value without considering bias due to measurement error or siting error, both of which are rampant in the US surface station network.
Read my report at left.
While not all situations with poorly sited weather stations affect trends, a weather station like this one at the University of Arizona’s parking lot in front of the atmospheric science department is represenative of the kinds of problems that would lead to an increased number of new high temperature records set.

Plus then there’s the error problem. For example we saw this summer that Honolulu set new record highs, but they turned out to be in error. The kicker is that NOAA let the records stand anyway! The problem is that a number of climate stations are at airports. Watch this NWS employee say on record that these airport weather stations are “placed for aviation purposes…not necessarily for climate purposes.”
So take this NCAR study with a grain of salt, since the authors did not address any of these issues.
From NCAR: Record High Temperatures Far Outpace Record Lows Across U.S.
BOULDER—Spurred by a warming climate, daily record high temperatures occurred twice as often as record lows over the last decade across the continental United States, new research shows. The ratio of record highs to lows is likely to increase dramatically in coming decades if emissions of greenhouse gases continue to climb.
“Climate change is making itself felt in terms of day-to-day weather in the United States,” says Gerald Meehl, the lead author and a senior scientist at the National Center for Atmospheric Research (NCAR). “The ways these records are being broken show how our climate is already shifting.”
This graphic shows the ratio of record daily highs to record daily lows observed at about 1,800 weather stations in the 48 contiguous United States from January 1950 through September 2009.
Each bar shows the proportion of record highs (red) to record lows (blue) for each decade. The 1960s and 1970s saw slightly more record daily lows than highs, but in the last 30 years record highs have increasingly predominated, with the ratio now about two-to-one for the 48 states as a whole. [ENLARGE] (©UCAR, graphic by Mike Shibao.) News media terms of use*
The study, by authors at NCAR, Climate Central, The Weather Channel, and the National Oceanic and Atmospheric Administration (NOAA), has been accepted for publication in Geophysical Research Letters. It was funded by the National Science Foundation, NCAR’s sponsor, the Department of Energy, and Climate Central.
If temperatures were not warming, the number of record daily highs and lows being set each year would be approximately even. Instead, for the period from January 1, 2000, to September 30, 2009, the continental United States set 291,237 record highs and 142,420 record lows, as the country experienced unusually mild winter weather and intense summer heat waves.
A record daily high means that temperatures were warmer on a given day than on that same date throughout a weather station’s history. The authors used a quality control process to ensure the reliability of data from thousands of weather stations across the country, while looking at data over the past six decades to capture longer-term trends.
This decade’s warming was more pronounced in the western United States, where the ratio was more than two to one, than in the eastern United States, where the ratio was about one-and-a-half to one.
The study also found that the two-to-one ratio across the country as a whole could be attributed more to a comparatively small number of record lows than to a large number of record highs. This indicates that much of the nation’s warming is occurring at night, when temperatures are dipping less often to record lows. This finding is consistent with years of climate model research showing that higher overnight lows should be expected with climate change.
In addition to surveying actual temperatures in recent decades, Meehl and his co-authors turned to a sophisticated computer model of global climate to determine how record high and low temperatures are likely to change during the course of this century.
The modeling results indicate that if nations continue to increase their emissions of greenhouse gases in a “business as usual” scenario, the U.S. ratio of daily record high to record low temperatures would increase to about 20-to-1 by mid-century and 50-to-1 by 2100. The mid-century ratio could be much higher if emissions rose at an even greater pace, or it could be about 8-to-1 if emissions were reduced significantly, the model showed.
The authors caution that such predictions are, by their nature, inexact. Climate models are not designed to capture record daily highs and lows with precision, and it remains impossible to know future human actions that will determine the level of future greenhouse gas emissions. The model used for the study, the NCAR-based Community Climate System Model, correctly captured the trend toward warmer average temperatures and the greater warming in the West, but overstated the ratio of record highs to record lows in recent years.
However, the model results are important because they show that, in all likely scenarios of future greenhouse gas emissions, record daily highs should increasingly outpace record lows over time.
“If the climate weren’t changing, you would expect the number of temperature records to diminish significantly over time,” says Claudia Tebaldi, a statistician with Climate Central who is one of the paper’s co-authors. “As you measure the high and low daily temperatures each year, it normally becomes more difficult to break a record after a number of years. But as the average temperatures continue to rise this century, we will keep setting more record highs.”
An expanding ratio
The study team focused on weather stations that have been operating since 1950. They found that the ratio of record daily high to record daily low temperatures slightly exceeded one to one in the 1950s, dipped below that level in the 1960s and 1970s, and has risen since the 1980s. The results reflect changes in U.S. average temperatures, which rose in the 1950s, stabilized in the 1960s, and then began a warming trend in the late 1970s.
Even in the first nine months of this year, when the United States cooled somewhat after a string of unusually warm years, the ratio of record daily high to record daily low temperatures was more than three to two.
Despite the increasing number of record highs, there will still be occasional periods of record cold, Meehl notes.
“One of the messages of this study is that you still get cold days,” Meehl says. “Winter still comes. Even in a much warmer climate, we’re setting record low minimum temperatures on a few days each year. But the odds are shifting so there’s a much better chance of daily record highs instead of lows.”
Millions of readings from weather stations across the country
The study team analyzed several million daily high and low temperature readings taken over the span of six decades at about 1,800 weather stations across the country, thereby ensuring ample data for statistically significant results. The readings, collected at the National Oceanic and Atmospheric Administration’s National Climatic Data Center, undergo a quality control process at the data center that looks for such potential problems as missing data as well as inconsistent readings caused by changes in thermometers, station locations, or other factors.
Meehl and his colleagues then used temperature simulations from the Community Climate System Model to compute daily record highs and lows under current and future atmospheric concentrations of greenhouse gases.
About the article
Title: “The relative increase of record high maximum temperatures compared to record low minimum temperatures in the U.S.”
Authors: Gerald A. Meehl, Claudia Tebaldi, Guy Walton, David Easterling, and Larry McDaniel
Publication: Geophysical Research Letters (in press)

Hrmph. Cryptic but easier than I thought. I had made this rather crude file where I just glued all the “station information” onto each data record for each reported temperature. Horridly wasteful of disk space (then again, if it’s using a dimes worth of space I’ll be amazed… disk has gotten dirt cheap…).
Anyway, I did this for a quick “one off” report or three… and now it’s become a central feature of several investigative / reporting approaches. I’d be embarrassed about the crude approach if it were not just so darned useful for things like this. A “one off” how many airports? Question.
So I can find all records in a year with the UNIX “grep” command. I look for any leading 12 characters (country code, station ID, mod flag) then the four digits of the year I care about. I chose 2008 as a recent year. I then send those records into a ‘cut’ command that only passes one character, the “AirStation Flag”. While some airports are not flagged with A for Airstation, many, perhaps most, are. We will have a (probably small) undercount of airports. We then pass those into “wc -l” that counts the number of lines (and so, the number of airports that reported. Each site reports one time per year with 12 monthly average values on the line). A very similar thing is done, but we give ‘grep’ the “-v” flag for “everything OTHER THAN A” to get the non-airports.
The results are:
[chiefio@tubularbells tmp]$ grep …………2008 v2.step2.inv.id.withlat | cut -b 148 | grep A | wc -l
593
[chiefio@tubularbells tmp]$ grep …………2008 v2.step2.inv.id.withlat | cut -b 148| grep -v A | wc -l
612
[chiefio@tubularbells tmp]$
So we have 1205 total stations that make it all the way through the “STEP – UHI correction and pitch out short records” and the “STEP1 splice and dice”.
Of those, about 1/2 are Airports.
Back in 1989, before The Great Dying of Thermometers, we have:
[chiefio@tubularbells tmp]$ grep …………1989 v2.step2.inv.id.withlat | cut -b 148 | grep A | wc -l
1430
[chiefio@tubularbells tmp]$ grep …………1989 v2.step2.inv.id.withlat | cut -b 148 | grep -v A | wc -l
2943
[chiefio@tubularbells tmp]$
so: 4373 total, of which 1430 are AirStations. About 1/3.
I would consider a rise in the percentage of thermometers at airports from 33% to 50% a significant change…
“Many such journeys are possible. Let me be your gateway.” TOS “The City on the Edge of Forever”…
We’ve been told by the trolls we can’t go just by what’s happening in the USA. So, according to them, this work is meaningless.
That ought to have said”
“STEP 2 – UHI correction and pitch out short records”…
To make up for the “correction note” here is 1959. now you can plot airport growth vs ‘records set” 😉
[chiefio@tubularbells tmp]$ grep …………1959 v2.step2.inv.id.withlat | cut -b 148 | grep A | wc -l
1215
[chiefio@tubularbells tmp]$ grep …………1959 v2.step2.inv.id.withlat | cut -b 148 | grep -v A | wc -l
2997
[chiefio@tubularbells tmp]$
So 4212 and about 29%. Looks to me like a consistent pattern. Also you get to add in the conversion from prop jobs to jet exhaust and the rapid growth in volume of traffic and airport size over time too…
But you can take solace in the fact that GIStemp uses airports as rural reference stations for the UHI calculation, so we’re really just adding a higher percentage of “rural” to the record /sarcoff>
http://chiefio.wordpress.com/2009/08/23/gistemp-fixes-uhi-using-airports-as-rural/
Just for giggles,I went to the Weatherundergound .com site for LaGrande, Oregon.
I live one block from Central school which has a MesoWest remote (on top of the
school roof no less,) and the local Airport, which is the location of the ASOS.
Ok,the Airport sits at the very southern end of the Grande Ronde Valley.At the mouth of Ladd Cayon some oh,three miles away.Currently the wind is calm at the Airport.(rare enough) and the Temp is 28F. Here,on the roof of Central School , it is 35F.a difference of 7F. This is not unusual.My wife and I love this neighborhood because it is somewhat protected and warmer,the Grade School,High School,and Middle School, all are within a four block radius of my home all with physical plants, acres of pavment, along with- less than 1/4 mile the Grande Ronde Hospital-also with Physical plant.
While this is a rough study of UHi, I think it does ,modestly illustrate why we need to
sort out UHi from the chaff.Unlike the NCAR report.
E.M.Smith (20:58:30) :
Interesting & revealing chief 🙂
I wouldn’t be embarrassed about quick & dirty though.
Back in the day, I’d do a Q&D proof of concept in QB4 & later refine in assembler & C. Often I’d leave a QB front end for ease of modification.
DaveE.
Bruce (19:51:52) :
The study says record highs are not increasing. Just record lows are descreaing because it isn’t cooling at night.
Isn’t that the classic UHI signature as the heat radiates from ashpalt, brick and buildings?
While this may be the case, that will get lost in an explanation to the average person.
What will not be lost with the average person is the clear fact that winters are becoming longer and harsher, and summers are mild with shorter growing seasons.
Anyone hearing this news about heat records will intuitively know there’s something not quite right with it because it doesn’t add up with what’s happening in the real world they live in every day.
(but I think trolls will be dancing under their bridges with glee over the study)
David Alan (16:56:20) :
I am not surprised at the timing of this report about extreme temperature change…. Guess he couldn’t wait 2 more months.
Copenhagen?
Well now I’ve done it. I’ve made them into standard commands now…
$ airport 1949
907
$ nairp 1949
2368
total of 3275 and percent of 27.7 %
$ airport 1969
1379
$ nairp 1969
3450
total of 4829 and percent of 28.6%
$ airport 1979
1175
$ nairp 1979
3188
total of 4363 and percent of 26.9%
(Hmmm Arab Oil Embargo and major recession… did some airports go out of business? )
$ airport 1999
885
$ nairp 1999
1539
total of 2424 and percent of 36.5%
So that completes the environment for the “records set” graph up top. A correlation graph would be very interesting… and the trend continues even from 2008 to 2009. As of now, more than half of all the thermometer records that make it to the “anomaly map” step are at / from Airports.
Got to make it hard to find a really “rural” location for the GIStemp UHI method (that adjusts stations based on ‘nearby rural’ stations up to 1000 km away and often airports… that now will be lots of airports…)
$ airport 2009
575
$ nairp 2009
540
total of 1115 and percent of 51.6%
Ya think that might be an issue? … Ya think?
As a couple of people have already said–they did leave out the 30’s, the Dust Bowl years, of the United States.
That is a serious vulnerability of the study. It could make it difficult to convince people it is warmer now than it has ever been before in the United States since everyone knows about the dust storms of the 1930’s.
It is a curious red flag that that time period is not included in this work.
“Pay no attention to those black blizzards behind the curtain…”
Dammit!!
I think I’m the only one to comment on this & I’ve tried to find out!
Did the stations they focused on even HAVE records prior to 1950
DaveE.
Let me clarify.
They focused on weather stations that have been operating since 1950.!!
No mention of whether they had been operating before 1950!
That would explain roughly 1:1 high & low records in the ’50s.
More cooling in ’60s & ’70s, obviously!
After that, equally obvious.
DaveE.
“yonason (17:04:17) :
No, I don’t believe them
http://wattsupwiththat.com/2009/10/18/a-cold-start-to-fall-over-4500-new-snowfall-low-temp-and-lowest-max-temp-records-set-in-the-usa-this-last-week/#more-11803
It’s very disturbing how many of them there are, and in what positions of influence. Are the malicious, or just delusional? And why?”
Maybe they want to ensure future funding…
http://www.ucar.edu/governance/meetings/apr09/presentations/ncar_budget.pdf
I’ve made some Mathematica calculations of the “record high vs record low” frequencies – showing that especially the frequency of the cold records is expected to brutally decrease if the underlying warming trend (regardless of the cause) is being accumulated over an increasing period of time. See my comments at
http://motls.blogspot.com/2009/11/warm-records-vs-cold-records.html
These results are no “warning signs”. They’re just straightforward implications of mathematics that can translate one way of parameterize the overall warming into another way. The map is almost always nonlinear and it tells us nothing about the causes of the “non-random” changes. (They can still arise as noise but not white noise.)
Aırport monıtorıng statıons.
Yes, when we take off, the performance calculatıons requıre the temperature above a vast expanse of concrete – because that ıs what we are usıng to take off on. Probably a completely dıfferent temperature to normal ambıent.
Plus, there ıs an awful lot of heat generated around an aırport. You can feel great gusts of manmade warm aır bıllowıng across the aırcraft stands.
.
yonason (19:24:46) :
Layne Blanchard (15:55:00) :
I posted the data from Britain going back to 1800, above. Here it is again.
http://carbon-sense.com/2009/10/01/british-record/
There is no difference between 19th and 20th century over the whole 200 years.
.
.
But ıf you look at the Wıkı versıon of the CET, there suddenly ıs a great rıse ın temperature. How come?
http://en.wikipedia.org/wiki/File:CET_Full_Temperature_Yearly.png
.
The Hadley CET graph also shows a sudden rıse.
http://hadobs.metoffice.com/hadcet/
Here ıs some of the raw data for the CET. Now the scales may be a bıt smaller, but can you see a great rıse ın temperature ın the 1980s??
http://www.cru.uea.ac.uk/cru/climon/data/cet/
So how dıd the Hadley offıcıal graph suddenly develop that great rıse ın temperature??
.
Same issue in Holland, but vastly greater: http://benlanka.tweakdsl.nl/climate/datumrecords.png shows the distribution of high and low day records (min, max, average daily temperature).
To attribute this phenomenon to measurements gone awry is obvious nonsense. That would, e.g., leave only one explanation for the melting of glaciers worldwide: that the gnomes have reduced the freezing point of water.
Also worldwide the temperaturemeasurements would have gone awry only in the last two decades or so, how strange.
If one doubts the temperature measurements one can look at proxies, no? Proxies like changing ecosystems or melting ice might do well. Even in the USA, where climate change seems to lag the rest of the world (in Holland on the contrary it’s going faster).
This is pure propaganda and someone should sew them for it.
3 out of 4 data sets show NO PROOF of any warming during the past 15 years:
http://noconsensus.wordpress.com/2009/11/12/no-warming-for-fifteen-years/
Ralph (02:06:54) :
Aırport monıtorıng statıons…Probably a completely dıfferent temperature to normal ambıent.
Absolutely correct, and that’s common knowledge in the aviation community.
For the anecdotal file, typical daily temperatures off-airfield here (Kirkuk, Iraq) are 5-to-8 degrees C *cooler* than temperatures on the ramp and on the runway.
Anthony: I have the record hi and lo temperatures for each day of the year, it covers years 1891 to 2009 for Norfolk, Nebraska on a xl spreadsheet. Quite interesting.
If you could send me your E-mail I could forward it.
This morning the local television news reported the current temperature at the station was 31 degrees, but they also noted that the airport–perhaps 15 miles away–was reporting 42 degrees. All surrounding counties were reporting temperatures in the low 30s. Either the airport was experiencing a bubble of warmer air or the integrity of their thermometer is suspect.
It is not clear to me how the 1800 sites were selected for the survey. I see from AWs project page that ther are around 9000 US sites in total. Is this yet another case of cherry picking compounding naive acceptance of records at face value.
On a completely separate point, but raised by the computer modelling used above, a question to all the climate professionals and statisticians out there: Has a linear trend ever been observed historically over any extended period of time, based on true records?
I suspect the answer is no, and if this is the case, what justification is there for using them predictively?
‘Sew’, Ron? Nice one 🙂
Eindhoven just broke the november second decade record, +17.3° up from +17.1° C in 1995; more stations will follow tomorrow.