Guest Post by Willis Eschenbach
A while back in the US there was an ad for a hamburger chain. It featured an old lady who bought a competitor’s hamburger with a great big hamburger bun. But when she opened it up she asked …
I got to thinking about this in the context of whether there is any real danger in a degree or two of average temperature rise, or whether it’s a big bun with no beef. In my previous post, “Lies, Damned Lies, Statistics … and Graphs”, I closed by saying:
My conclusion? Move along, folks, nothing to see here …
A commenter took exception to this, saying
When talking about global average temperatures, tenths of a degree really do matter.
Now, if tenths of a degree changes over a century “matter” for the globe, they certainly must matter for parts of the globe.
So here’s your pop quiz for the day: Which US State warmed the most, which cooled the most, and by how much?
To answer this, I used the USHCN State Temperature Database. Here are my findings:
Figure 1. Temperature trends by state, USHCN data. Seven states cooled, and forty-one warmed.
The state that warmed the most was North Dakota (top center), which warmed 1.4°C per century. The state that cooled the most was Alabama (middle of three dark blue states, lower right). It cooled by 0.3°C/century.
To compare with my previous post, here’s a similar graph, of the decadal changes in North Dakota by month.
Figure 2. North Dakota decadal average temperatures by month, 1900-2009. Red line is the average for the decade 2000-2009. Photo is an old North Dakota farmhouse.
As with the US, for much of the year there is little change, and the warming is in November to February. Note that unlike the US, during that four months, the temperature of North Dakota is below freezing (32°F) …
Now, if tenths of a degree “matter”, if they are as important as the commenter claimed, we should have seen some problems in North Dakota. After all, it has warmed by 1.6°C since 1895. That’s almost three times the global average warming.
But somehow, I must have missed all of the headlines about the temperature calamities that have befallen the poor residents of the benighted state of North Dakota. I haven’t seen stories about them being “climate refugees”. I didn’t catch the newspaper articles about how it has been so hard on the farmers and the frogs. I am unaware of folks moving in droves to Alabama, which has cooled by -0.4° since 1895, and thus should be the natural refuge of those fleeing the thermal holocaust striking North Dakota.
In fact, I don’t remember seeing anything that would support the commenter’s claims that tenths of a degree are so important. North Dakota has warmed near the low end of the range forecast by the IPCC for the coming century, and there have been no problems at all that I can find. So I have to say, as I said before,
My conclusion? Move along, folks, nothing to see here … where’s the beef?
APPENDIX: R Code for the US Map
(I think this is turnkey. Sometimes WordPress puts in extra line breaks. If so, it is also available as a Word document here.)
The code requires that you download the USHCN Temperature Data cited above and save it as a “Comma Separated Values” (CSV) file. I downloaded it, opened it in Excel. I split it using “Text to Columns …” into the following columns, as detailed in the USHCN ReadMe file:
FILE FORMAT:
STATE-CODE 1-3 STATE-CODE as indicated in State Code Table above. Range of values is 001-110.
DIVISION-NUMBER 4 DIVISION NUMBER. Value is 0 which indicates an area-averaged element.
ELEMENT-CODE 5-6
02 = Temperature (adjusted for time of observation bias)
YEAR 7-10 This is the year of record. Range is 1895 to current
year processed.
JAN-VALUE 11-17 Monthly Temperature format: Range of values -50.00 to 140.00 degrees Fahrenheit. Decimals retain a position in the 7-character field. Missing values in the latest year are indicated by -99.90.
FEB-VALUE 18-24
MAR-VALUE 25-31
APR-VALUE 32-38
MAY-VALUE 39-45
JUNE-VALUE 46-52
JULY-VALUE 53-59
AUG-VALUE 60-66
SEPT-VALUE 67-73
OCT-VALUE 74-80
NOV-VALUE 81-87
DEC-VALUE 88-94
If that is too complex, the CSV file is here.
Here’s the R code:
# The code requires that you download
# the USHCN Temperature Data
# and save it as a "Comma Separated Values" (CSV) file.
# I downloaded it, opened it in Excel, and used
# "Save As ..." to save
# it as "USHCN temp.csv"
#Libraries needed
library("mapdata")
library("mapproj")
library("maps")
# Functions
regm =function(x) {lm(x~c(1:length(x)))[[1]][[2]]}
#Read in data
tempmat=read.csv('USHCN temp.csv')
# Replace no data code -99.9 with NA
tempmat[tempmat==-99.9]=NA
# split off actual temps
temps=tempmat[,5:16]
# calculate row averages
tempavg=apply(temps,1,FUN=mean)
# calculate trends in °C by state
temptrends=round(tapply(tempavg,as.factor(tempmat[,1]),regm)*100*5/9,2)
# split off states from regional and national
statetrends=temptrends[1:48]
#calculate ranges for colors
statemax=max(statetrends)
statemin=min(statetrends)
statefract=(statetrends-statemin)/staterange
#set color ramp
myramp=colorRamp(c("blue","white","yellow","orange","darkorange","red"))
# assign state colors
mycol=myramp(statefract)
# names of the states (north michigan is missing for ease of programming)
myregions=c("alabama", "arizona", "arkansas", "california", "colorado", "connecticut", "delaware",
"florida", "georgia", "idaho", "illinois", "indiana", "iowa", "kansas", "kentucky", "louisiana", "maine",
"maryland", "massachusetts:main", "michigan:south", "minnesota", "mississippi", "missouri", "montana", "nebraska",
"nevada", "new hampshire", "new jersey", "new mexico", "new york:main", "north carolina:main", "north dakota",
"ohio", "oklahoma", "oregon", "pennsylvania", "rhode island", "south carolina", "south dakota", "tennessee", "texas",
"utah", "vermont", "virginia:main", "washington:main", "west virginia", "wisconsin", "wyoming")
# draw map
par(mar=c(6.01,2.01,4.01,2.01))
return=map('state',regions=myregions, exact=T,projection='mercator',fill=T,
mar=c(5.01,8.01,4.01,2.01),col=rgb(mycol,maxColorValue=255),ylim=c(10,60))
# set up legend boxes
xlref=-.48
yb=.37
ht=.05
wd=.08
textoff=.025
# assign legend labels
mylabels=round(seq(from=statemin,by=staterange/12,length.out=13),2)
#draw legend
myindex=0
for (i in seq(from=xlref,by=wd,length.out=12)){
xl=i
xr=xl+wd
yt=yb+ht
rectcolor=myramp(myindex/11)
rect(xl,yb,xr,yt,col=rgb(rectcolor,maxColorValue=255))
text(xl,yb-textoff,mylabels[myindex+1],cex=.65)
myindex=myindex+1
}
text(xl+wd,yb-textoff,mylabels[myindex+1],cex=.65)
# add annotations
text(0,1.08,"US Temperature Trends (°C/century)")
text(0,1.03,"USHCN Dataset, 1895-2009",cex=.8)
Discover more from Watts Up With That?
Subscribe to get the latest posts sent to your email.
C. Shannon (02:53:56) : And I would be willing to bet that North Dakota isn’t the most extreme case globally.
I’ve just started a series of “Monthly Cumulative Anomaly” investigations by country. It’s interesting… As Willis found here, the results vary by month and by location rather a lot. In some cases you have 2 months “warming” by 3 or even 5 C, but the month between them cooling…
It is “young code” and needs a better QA series done on it prior to touting it as ultimate truth, but the results are still interesting. IMHO, doing differential anomalies by month pretty much shows that “Climate Change” is an instrument and processing issue. The way different months change is very “un-physical”…
For example in Australia:
http://chiefio.files.wordpress.com/2010/04/australia_dmt.png
we have, since the 1950s, June cooling while November warms dramatically. Yet December doesn’t change much.
This Russian Asian sector graph:
http://chiefio.files.wordpress.com/2010/04/russiaasian_dmt_full.png
Has a (volatile) basically flat Jan, Feb falling like a rock, Nov rocketing up, but Sept basically flat.
I’m still trying to ‘work out the kinks’ in how best to approach this process and how best to display the results. (It’s in the ‘early investigative stage’). But it is an intriguing “Dig Here!” as the results are very unlike what CO2 would be expected to cause…)
Willis Eschenbach (10:40:00) :
“Steve Hempell (20:32:14) Like to see this for the European countries. Any hints on how to do this if I want to tackle it?”
I don’t know if there is anything comparable to the USHCN that gives the European climate records country by country … let me know if you find something.
The GHCN uses the “Country Code” as the first three digits. You can break out any country via those first three digits. USA is 425. Australia is 501, New Zealand 507. Also, the first digit is “region” so “3” gets South America, “4” gets North America. “5” is the Pacific Basin. “6” is Europe.
There are some “odd bits” like Russia gets two “country codes” (635 and 222) as half is in Europe (Region “6”) and half is in Asia (Region “2”).
You can see the world broken out this way, by continent, then by country, with an anomaly graph for each here:
http://chiefio.wordpress.com/2010/04/11/the-world-in-dtdt-graphs-of-temperature-anomalies/
I’ve started “playing” with the cumulative anomalies by month, but I’m not happy with the visualization yet. You can see some of it here:
http://chiefio.wordpress.com/2010/04/15/dmtdt-climate-change-by-the-monthly-anomaly/
and here:
http://chiefio.wordpress.com/2010/04/18/australian-anomaly-walkabout/
I have to think that using something like R and a better idea about graphical presentation would work better (he hinted broadly at Willis 😉
Willis,
Thanks for a very interesting post but I failed your pop quiz. In fact, I wasn’t even close!! Out of curiousity, I decided to examine the Al data. (Source: ftp://ftp.ncdc.noaa.gov/pub/data/cirs/drd964x.tmpst.txt.) I calculated and normalized the annual temperature. I was quite surprised by the step-wise change in Det T that became apparent in the graphical display. Although the 1895 -2009 trendline has a cooling slope of -0.0066 s.d.’s/yr (-0.00747⁰F/yr), the step-wise change makes it rather meaningless. One might draw an eye-ball trendline from 1895 to 1957 and it would have a warming slope; then draw an eye-ball trendlne from 1958 to 2009 and it would also have a warming slope.
I have no idea why the AL state-wide data would have a step-wise change in 1958. There are currently many surface temperature stations in AL that are incuded in NCDC’s data set. I would have to do more research to see that number has changed and in particular, if the number changed in 1958.
Oh well …
So, sphaerica:
“But to repeat myself, the problem with urederra’s original statement, and Merrick’s interpretation, is that CO2 does not get it’s “heat” from the sun. CO2 is primarily transparent to the wavelengths of electromagnetic radiation received by the earth from the sun. This energy heats the surface.
The surface then emits infrared radiation, which is absorbed by greenhouse gases like CO2 and H2O. It is quickly re-emitted in all directions. What goes down warms the ground. What goes sideways warms the air. What goes up warms the air above, or eventually makes it back into space.”
Please explain to me how that differs from what I explained?
And Vincent:
“Well then, perhaps you can explain why the models all predict the main effect of GHG’s to be less cold winters, not hotter summers?”
Perhaps you misunderstand the difference between absolute heating and differential heating. I can explain that, if you like, but please explain to me why you would take a statement you found somewhere that suggests warming is more pronounced in winter and use that as evidence to contradict a physical law that was originally deduced by Newton and can be (and has been) experimentally verified in labs (as opposed to AGW)?
And at the same time, perhpas you can explain why, if as you say all the models predict less cold winters and not hotter summers, that all I hear all summer long on hot days “global warming, global warming, global warming!” and when it’s bitter cold, like last winter, all I hear is “that’s weather, not cimate”?
Re: My 04/18 15:57:04 comment
I mentioned in my prior comment that after failing Willis’ pop quiz; I decided to examine the AL data and was quite surprised to see that a step-wise reduction in temperature occurred in 1958.
When considering the entire 1895-2009 interval:
• TAvg = 63.32⁰F
• s.d. = 1.13⁰F
• BFSL Equation: T = -0.0074*yr+77.79
When considering only the 1895-1957 interval:
• TAvg = 63.81⁰F
• s.d. = 1.06⁰F
• BFSL Equation: T = 0.0237*yr+18.24
When considering only the 1958-2009 interval:
• TAvg = 62.73⁰F
• s.d. = 0.92⁰F
• BFSL Equation: T = 0.0278*yr+7.513
Using the 1895-1957 and 1958-2009 BSFL equations, the calculated 1957 and 1958 endpoints are 64.92⁰F and 61.95⁰F, respectively. In other words, the step-wise discontinuity at the 1957 and 1958 endpoints is -2.97⁰F. Furthermore, a warming trend exists before and after the discontinuity with slopes of +0.0237/yr and +0.0278/yr, respectively.
I’m not suggesting that considering the trends before and after the discontinuity is better way to interpret he change in temperature between 1895 and 2010, I’m suggesting that when discontinuities as large as 3⁰F are observed, there is something strange about the data. I suppose that something strange may have happened in AL in 1957-58; I’m more inclined to think that there is something strange in NCDC’s data set for AL. (I’m certain that the strange discontinuity has nothing to do with the fact that my wife and I were married in ME on 06/29/1957!!)
Can anyone tell me if there is a NCDC subdirectory that contains the specific surface station temperature data sets that NCDC uses to calculate the state-wide monthly temperature data set? If such a subdirectory doesn’t exist, is there a listing of the specific surface stations that NCDC uses?
Willis,
Re: My 04/18 15:57:04 and 04/19 12:40:51 comments
Having examined the AL temperature data, I decided to examine the ND data. The graphical display of the data is not unusual. However, although less apparent because the data is ‘noisier’, I can discern two step-wise changes in the ND data; one in 1917-18 and the other in 1980-81.
When considering the entire 1895-2009 interval:
• TAvg = 39.57⁰F
• s.d. = 1.94⁰F
• BFSL Equation: T = 0.0135*yr+26.323
When considering only the 1895-1917 interval:
• TAvg = 37.99⁰F
• s.d. = 1.47⁰F
• BFSL Equation: T = 0.0043*yr+29.732
When considering only the 1918-1980 interval:
• TAvg = 39.51⁰F
• s.d. = 1.51⁰F
• BFSL Equation: T = -0.0038*yr+46.89
When considering only the 1981-2009 interval:
• TAvg = 40.94⁰F
• s.d. = 1.98⁰F
• BFSL Equation: T = -0.0482*yr+137.43
Using the 1895-1917, 1918-1980 and 1981-2009 BSFL equations, the calculated 1917 and 1918 endpoints are 37.98⁰F and 39.60⁰F, respectively; the 1980 and 1981 endpoints are 39.37⁰F and 41.95⁰F, respectively. In other words, the step-wise discontinuities at the 1917 and 1918 endpoints and 1980 and 1981 are +1.62⁰F and +2.5⁰F, respectively . Furthermore, a slight cooling trend exists after both discontinuities with slopes of -0.0038/yr and +0.0482/yr, respectively.
Again, I’m not suggesting that considering the trends before and after the discontinuities is better way to interpret the change in temperature between 1895 and 2010, I’m suggesting that when discontinuities as large as 1.5⁰F to 2.5⁰F are observed, there may be something strange about the data.
hmccard (15:56:03) : edit
Yeah, I know, but you gotta live with the data you’re offered. I didn’t look at Alabama, but I did see the big step change in North Dakota. Kinda odd, because it is the average of a whole raft of stations …
I haven’t looked at the homogeneity algorithm they use yet, but I’ve suspected that it might accidentally amplify a problem in a few stations into a larger problem … or not.
Willis,
I have calculated the difference between monthly raw and TOBS GHCN datasets (http://www.ncdc.noaa.gov/oa/climate/ghcn-monthly/index.php) for a number of surface stations. The graphical results displayed step-wise changes typically in the range from 0.5⁰F to 1.5⁰F. My analysis shows that the adjustments are made on a seasonal basis, i.e., DJF, MAM, JJA and SON.
IMO, the TOBS adjustments are step-wise aperiodic for a given surface station, two or three occuring during the last century, and somewhat random for different stations, Randomizing the TOBS adjustments allows NCDC to claim that the bias introduced is small. That doen’t mean that the bias is small for any given station. My analysis also showed that the bias can be quite large for a set of closely-spaced stations. I did’t examine the bias related to a set of stations separated by 500km or 1200km.
I assume that NCDC’s state-wide temperature dataset is based on their adjusted datasets for whatever surface stations they selected. Perhaps the TOBS adjustments for AL were not that random after all. If I knew which surface stations NCDC used in their state-wide dataset, I could probably check some of the TOBS adjustments.
Willis,
Fyi – I used the USHCN v2 1895-2009 monthly temperature datasets for the 15 Alabama surface stations ( http://cdiac.orl.gov/ftp/ushcn_v2_monthly/) and calculated the state-wide averages for the raw and TOBS-adjusted temperature data, Traw and TTOBS . I then compared these annual datasets to the annual dataset that I calculated using the USHCN State Temperature Database (USHCN STB) (ftp://ftp.ncdc.noaa.gov/pub/data/cirs/drd964x.tmpst.txt) that you referenced in your post. A graphical comparison of the annual temperatures here shows they are temporally similar but there are differences.
For brevity’s sake and specificity, I refer to the difference between annual temperature from the two USHCN datasets as difTraw = USHCN v2 Traw – USHCN STB Tavg and difTTOBS = USHCN v2 TTOBS – USHCN STB Tavg; also ΔTOBS = USHCN v2 Traw – USHCN v2 TTOBS . A graphical comparison of the temperature differences may be found here.
It seems quite strange to me that:
1. For the 1895-2009 interval:
a. difTraw > 0 most years (104 out 114 years); average = +0.38⁰F; slope = +0.0041⁰F/yr
b. difTTOBS > 0 most years (84 out 114 years); average =+ 0.27⁰F; slope = +0.0051⁰F/yr
c. ΔTOBS > 0 most years (87 out 114 years) but average = +0.11 ⁰F; slope =- 0.0001⁰F/yr
2. A step-wise change in temperature occurred in 1957-58
3. For the 1900-1957 interval:
a. difTraw > 0 most years (105 out 114 years); average = +0.29⁰F; slope = +0.0048⁰F/yr
b. difTTOBS > 0 most years (84 out 114 years); average = +0.02⁰F; slope = – 0.0012⁰F/yr
c. ΔTOBS > 0 most years (90 out 114 years) but average = +0.27⁰F; slope = +0.0060⁰F/yr
d. TTOBS tracked Tavg quite closely during this interval as a result of the TOBS-adjustment
4. For the 1958-2010 interval:
a. difTraw > 0 most years (114 out 114 years); average = +0.49⁰F; slope = +0.0108⁰F/yr
b. difTTOBS > 0 most years (109 out 114 years); average =+ 0.47⁰F; slope = +0.0177⁰F/yr
c. ΔTOBS > 0 most years (90 out 114 years) but average = +0.02⁰F; slope = -0.0069F/yr
It is clear to me that the USHCN v2 and USHCN STB databases for Alabama are different. Apparently, NCDC made some additional adjustments to the STB dataset for Alabama.
Oh well…
Willis,
Apparently the hyperlinks to my graphs were dropped in my last message. I’ll try again’
AL annual temperatures: http://www.imagenerd.com/show.php?_img=al_temps-SFLIO.png
AL temperature differences: http://www.imagenerd.com/show.php?_img=dif_in_al_temps-vgMTl.png
Sorry about that ….