A New Analysis of U.S. Temperature Trends Since 1943

August 6th, 2012 by Roy W. Spencer, Ph. D.

With all of the hoopla over recent temperatures, I decided to see how far back in time I could extend my U.S. surface temperature analysis based upon the NOAA archive of Integrated Surface Hourly (ISH) data.

The main difference between this dataset and the others you hear about is that trends are usually based upon daily maximum and minimum temperatures (Tmax and Tmin), which have the longest record of observation. Unfortunately, one major issue with those datasets is that the time of day at which the maximum or minimum temperature is recorded makes a difference, due to a double-counting effect. Since the time of observation of Tmax and Tmin has varied over the years, this potentially large effect must be adjusted for, however imperfectly.

Here I will show U.S. temperature trends since 1943 based upon 4x per day observations, always made at the same synoptic times 00, 06, 12, and 18 UTC. This ends up including only about 50 stations, roughly evenly distributed throughout the U.S., but I thought it would be a worthwhile exercise nonetheless. Years before 1943 simply did not have enough stations reporting, and it wasn’t until World War II when routine weather observations started being made on a more regular and widespread basis.

The following plot shows monthly temperature departures from the 70-year (1943-2012) average, along with a 4th order polynomial fit to the data, and it supports the view that the 1960s and 1970s were unusually cool, with warmer conditions existing in the 1940s and 1950s (click for large version):

It’s too bad that only a handful of the stations extend back into the 1930’s, which nearly everyone agrees were warmer in the U.S. than the 40’s and 50’s.

What About Urban Heat Island Effects?

Now, the above results have no adjustments made for possible Urban Heat Island (UHI) effects, something Anthony Watts has been spearheading a re-investigation of. But what we can do is plot the individual station temperature trends for these ~50 stations against the population density at the station location as of the year 2000, along with a simple linear regression line fit to the data:

It is fairly obvious that there is an Urban Heat Island effect in the data which went into the first plot above, with the most populous stations generally showing the most warming, and the lowest population locations showing the least warming (or even cooling) since 1943. For those statisticians out there, the standard error of the calculated regression slope is 29% of the slope value.

So, returning to the first plot above, it is entirely possible that the early part of the record was just warm as recent years, if UHI adjustments were made.

Unfortunately, it is not obvious how to make such adjustments accurately. It must be remembered that the 2nd plot above only shows the relative UHI warming of higher population stations compared to the lower population stations, and previous studies have suggested that even the lower population stations experience warming as well. In fact, published studies have shown that most of the spurious UHI warming is observed early in population growth, with less warming as population grows even larger.

Again, what is different about the above dataset is it is based upon temperature observations made 4x/day, always at the same time, so there is no issue with changing time-of-observation, as there is with the use of Tmax and Tmin data.

Of course, all of this is preliminary, and not ready for peer review. But it is interesting.

19 Responses to “A New Analysis of U.S. Temperature Trends Since 1943”

Toggle Trackbacks

  1. Kasuha says:

    Very interesting analysis, thank you.

  2. Dave says:

    Have you corrected for possible data inhomogeneities? A lot of these airport sites were originally placed on rooftops, before moving to the ground. I would presume that could impart a non-climatic bias? Anthony Watts (and NOAA) has demonstrated that rooftop sites have a significant warm bias compared to surrounding ground-based sites.

  3. Stephen Wilde says:

    Good to see this:

    “previous studies have suggested that even the lower population stations experience warming as well. In fact, published studies have shown that most of the spurious UHI warming is observed early in population growth, with less warming as population grows even larger.”

    which is a point I’ve made several times in the blogs without realising there were studies in support.

    I’d be interested to see links to such reports.

    In fact I would go so far as to suggest that most of the rural sites suffer from UHI effects too since they often offer a focal point for further development nearby.

  4. Eric Barnes says:

    Thanks Dr. Roy. 🙂
    I’m guessing that the most warming occurs at UTC 0600 and the least at UTC 1800?

  5. P. Solar says:

    Dr. Spencer, could you link the data that went into this? I have a few ideas that I’d like to check out in relation to other datasets.

    This looks interesting.

  6. Terry says:

    I remember Gavin Schmidt once saying that the temperature trends for the globe could be adequately quantified using a little as 100 ideal high quality stations. It seems to me that using these stations might be a good start. Unfortunately we are stuck with having to use other methods for older records. I think this is certainly worth pursuing a lot further.

  7. Tim Thomson says:

    I just heard you on C to C and you were asked about the consistency of solar output. NASA along with monitoring surface temperatures, monitors the total irradiance of the sun impinging on the earth (TSA). If you plot TSA with the 5-year twa temperature you see a strong correlation and you see that TSA is ahead of surface temperature.
    I am an independent scientist of some repute. If you are interested in the data, please contact me.


  8. Monty says:

    Nice data mining, though I do see an opportunity here for a little more given the recent claims of increased variability.

    Could you plot a couple of those times separately? I’d really like to see, say 3AM vs 3PM, but midnight vs. noon would be fine. It would also be interesting to see the yearly standard deviation of each time of day data set (calculated about each years average, of course).

  9. Dr. Spencer, did a great job on Coast to Coast last night. It was just great!

  10. Fernando C Vidaurri PhD PE says:

    Recent article by James Hansen et al Perception of climate change. http://www.pnas.org/cgi/doi//10.1073/pnas.1205276109

    They chose a base perod of 1951 to 1980 and found that recent temperatures in 2011 and 2012 are above the base, hence, with a lot of graphs and statistical hand waving,global warming.

    In your studies the 1960’s and 1970’s were unusually cold. 20 out of 30 years in the Hansen baseline were colder, no wonder everything else looks statically warmer.

    Also, never mind the warmer earlier non-recent periods pointed out in Esper et al, Orbital forcing of tree-ring data Nature, Climate Change, Letters, 8 July 2012.

  11. Jay Kay says:

    Dr. Spencer, you had a blog entry 2 or 3 years ago that looked at UHI as a function of population density increase rather than simply population density.

    Since then, I’ve looked for that blog but haven’t been able to find it. It would be interesting to dig that study up.

    It seems to me that for a relatively static population density, one would not see any UHI effect on temperature change but would see an effect if the population density increased during that period.

  12. The temperature trend for that entire period is strongly linked to the PDO ,especially when the PDO switched from a cold phase prior to 1978, to a warm phase post 1978. Look again when the PDO , reverted back to a cold phase,in year 2009. The temperature trend versus the PDO phase, is very well correlated.

    Remember when the PDO is in a warm phase the number of El Nino ,events versus La Nina events increases, and vice versa, for when the PDO is in a cold phase.

    The volcanic eruptions also had some impact in the early 1980’s and early 1990’s. Throw in the Urban Heat Island effect , and you have the entire cause for the temperature trend during this period of time. I should say up to year 2005.

    The solar activity not being a factor due to the steady rhythmic 11 year cycle, against a background of above average solar activity in general, especially since the Dalton Minimum; during the time period 1943-2005,and even priotr to 1943.

    CO2 , obviously not a factor ,since it shows a continuous rise throughout that time period ,while the temperatures vary both up and down.

    Volcanic activity although having some influence was generally very low during that time period, probably having a net warming effect.

    UV light was strong, the solar wind was strong ,along with solar irradiance during that time period ,in contrast to when the Dalton Minimum ended around 1830. Also volcanic activity was much higher prior to 1830, only to decrease since then.
    Those factors probably brought the temperatures into the range Dr. Spncer shows from 1943- 2005, up from the range the temperatures were in during the Dalton Minimum. So the evidence is solar/volcanic action did have an effect, but the effects were felt prior to 1943,from the period 1830- 1943. The same for volcanic activity. Effects from these two forces on the climate played out by 1943 ,due to the fact they both reached a level of more or less consistency ,a few decades after the Dalton Minimum ended. Solar activity increasing and then leveling off at a high level, while volcanic actiivty was decreasing and leveling off at a low level,some decades after the Dalton Minimum ended.

    So all we have left post 1943 as possible potential forcing mechanisms on the climate, is essentially changes in the PDO, with the associated ENSO changes,the UHI effects, and maybe some very temporary effects from the isolated volcanic eruptions in the early 1980’s and early 1990’s. Hence the strong correlation between the PDO, versus the temperature trend ,with the UHI effects and isolated vlocanic activity thrown in for good measure.

    Note, solar actiivty went from an active phase to an inactive phase during the month of Oct.2005. I think lag times are involved here due to the oceanic effects, nevertheless the temperature trend since 2005,or even a few years prior stops going up, and then since 2009, the trend is down somewhat.

    Now I do realize this is for the United States,but the same trends in general can be seen for the N.H.,and the globe to a lesser extent(s.h. being mostly water not responding as much). These trends in the N.h. and the globe to a lesser extnet, evident in the 1943-present period, and the period from the end of the Dalton Minimum to 1943.

    Now that solar activity has entered a prolong solar minimum period(VERY LIKELY) ,it wiLl once again become a much greater forcing mechanism on the climatic system of the earth going forward. Especially once this weak solar maximum ,of solar cycle 24 ,passes on by.

    Therefore with a cold PDO in place ,add to this the lower solar activity ,with a possible increase in volcanic activity ,(which is associated with prolong solar minimum periods, or at least the data indicates this is so) along with a decrease in UV light(less ozone,different distribution ,a more -AO circulation), a decrease in the solar wind,(more cosmic rays more clouds,that is what the data indicates)the ingredients in my opinion are there for a cool down. How much will depend upon the degree of magnitude change ,along with the duration of time that goes by, for all these forces involved in the changing of the climate.

    It is going to be a most interesting decade in my opinion.

  13. AlaskaHound says:

    This preliminary study is quite interesting Roy.
    Thanks for the good work.
    On an unrelate4d topic, what data do we have for the vast Indian Ocean and its coupling effects with the troposphere?
    Is there an IO?



  14. Eric Barnes says:

    Being curious I downloaded ghcn daily and put it into an sql database.

    The below url has average max, min and mean temps in tenths of a degree C along with station counts. I only included long running stations that reported at least 344 daily values for 85 of the years from 1930 to 2011.

    After looking at it I might try to get the stations to be the same throughout the reporting period. They are close but there’s an annoying drop around 1950.


  15. I must say Dr. Spencer’s approach to the climate situation is a very sensible approach ,in that he rules nothing in or out.

    That is the correct way to approach this, given the extreme amounts of uncertainty.

    My take, probably viewed as unlikely by many, is probably due to the lack of conviction that the sun has enough variation to begin with, along with a major lack of confidence, in the associated secondary effects that may result.

    This is why ,with a high probability for the first prolong solar minimum since 1790-1830, this decade and period ,I think will shed much more light on this and just how correct or not correct,this school of thinking is.

    I want to know ,one way or the other.

    I think it is easier to change the amount of energy coming into earth’s climatic system, then it is to change the amount of energy leaving earth’s climatic system. With that said what is the source of energy for the climate system of the earth? The sun.

    Again, this is a great opportunity because we will have the contrast in place; weak solar activity ,colder conditions, versus an increase in CO2, warmer conditions.

  16. Harvey says:

    There could be a significant UHI effect, even for places that have had little growth in population.

    Many temperature gages are at airfields/airports, and those have changed a lot in the last 50 years.

    Typical aircraft being flown 50 years ago had large wheels able to handle rough grass, and were difficult to land in a cross wind. So many airfields were large square grass areas. Think WW2 big open grass airfields.

    Today’s aircraft have smaller wheels suitable and really need a clean and smooth hard surface runway, hence concrete or bitumen.

    So a typical airfield, even if it has remained a small operation serving the same population, has a lot more concrete and bitumen than it would have had 50 years ago.

    Of course if it has grown into a regional airport with 20 jets a day pumping out heat, rather than one DC3 a day during the 1950s, it will be a lot worse.

  17. John says:

    This is nice!

Leave a Reply