From environmentalist Jennifer Marohasy’s blog in Australia, please pay her a visit here – Anthony
There has been criticism of the potential for official weather stations in the USA to record artificially high temperatures because of the changing environments in which they exist, for example, new asphalt, new building or new air conditioning outlets. Meteorologist, Anthony Watts, has documented evidence of the problem and Canadian academic, Ross McKitrick, has attempted to calculate just how artificially elevated temperatures might be as a consequence.
A reader of this blog, Michael Hammer, recently studied the official data from the US official weather stations and in particular how it is adjusted after it has been collected. Mr Hammer concludes that the temperature rise profile claimed by the US government is largely if not entirely an artefact of the adjustments applied after the raw data is collected from the weather stations.
Does the US Temperature Record Support Global Warming?
By Michael Hammer
IN the US, the National Oceanic and Atmospheric Administration (NOAA) collects, analyses and publishes temperature data for the United States. As part of the analysis process, NOAA applies several adjustments to the raw data.
If we consider, the above graph, which shows, their plot of the raw data (dark pink) and the adjusted data (pale pink), it is obvious that the adjustments have little impact on data from early in the 20th century but adjust later temperature readings upwards by an increasing amount. This means that the adjustments will create an apparent warming trend over the 20th century. [Click on the above chart for a better larger view, this chart can also be viewed at http://cdiac.ornl.gov/epubs/ndp/ushcn/ndp019.html .]
NOAA state that they adjust the raw data for five factors. The magnitude of the adjustments are shown in Figure 2.
Figure 2. Form of individual corrections applied by NOAA. The black line is the adjustment for time of observation. The red line is for a change in maximum/minimum thermometers used. The yellow line is for changes in station siting. The pale blue line is for filling in missing data from individual station records. The purple line is for UHI effects (this correction is now removed). [Click on the chart for a better larger view or visit the same website as for Figure 1.]
It is obvious that the only adjustment which reduces the reported warming is UHI which is a linear correction of 0.1F or about 0.06C per century, Figure 2. Note also that the latest indications are that even this minimal UHI adjustment has now been removed in the latest round of revisions to the historical record. To put this in perspective, in my previous article on this site I presented bureau of meteorology data which shows that the UHI impact for Melbourne Australia was 1.5C over the last 40 years equivalent to 3.75C per century and highly non linear.
Compare the treatment of UHI with the adjustments made for measuring stations that have moved out of the city centre, typically to the airport. These show lower temperatures at their new location and the later readings have been adjusted upwards so as to match the earlier readings. The airport readings are lower because the station has moved away from the city UHI. Raising the airport readings, while not adding downwards compensation for UHI, results in an overstatement of the amount of warming. This would seem to be clear evidence of bias. It would be more accurate to lower the earlier city readings to match the airport readings rather than vice versa.
Note also the similarity between the shape of the time of observation adjustment and the claimed global warming record over the 20th century especially the steep rise since 1970. This is even more pronounced if one looks at the total adjustment shown in Figure 3 (again from the same site as Figure 1). As a comparison, a recent version of the claimed 20th century global temperature record downloaded from www.giss.nasa.gov is shown in Figure 4.
Figure 3. Magnitude of the total correction applied by NOAA
[Click on the charts for a larger/better view.]
Figure 4. Temperature anomaly profile from NASA GISS
Since the total corrections for the US look so similar to the claimed temperature anomaly, it begs the questions as to what the raw data looks like without any corrections. Does it show the claimed rapidly accelerating warming trend claimed by the AGW advocates? To determine this I took the raw data from the USHCN graph shown in Figure 1 and plotted this using a 5 year mean (blue trace), matching the smoothing in the NASA GISS profile shown in Figure 4. The result is shown in Figure 5. Please note that while the plot is one that I generated, the data comes directly from the raw data from Figure 1 published by NOAA.
Figure 5 Plot of raw temperature data versus time (from fig 1) 5 point smoothing. Vertical axis degrees Fahrenheit. Red line is a linear trend line. Green line is a 2nd order (parabolic) trend line.
Clearly the shape of this graph bears no similarity at all to the graph shown in Figure 4. The graph does not even remotely correlate to the shape of the CO2 versus time graph. The warming was greatest in the 1930’s before CO2 started to rise rapidly. The rate of rise in 1920, the early 1930’s and the early 1950’s is significantly greater than anything in the last 30 years. Despite the rapid rise in CO2 since 1960, the 1970’s to early 1980’s was the time of the global cooling scare and looking at the graph in Figure 5 one can see why (almost 2F cooling over 50 years).
A linear least squares trend line, created using the Excel trend line function (Red trace) shows a small temperature rise of 0.09C per century which is far less than the rise claimed by AGW supporters and clearly of no concern. However, the data shown in figure 5 bears little if any resemblance to a linear function. One can always fit a linear trend line to any data but that does not mean the fitted line has any significance. For example, if instead I fit a second order trend line (a parabolic) the result is extremely different. That suggests a temperature peak around 1950 with an underlying cooling trend since. Which trend line is the more significant one? If there was really a strong underlying linear rise over the time period it should have shown up in the 2nd order trend line as well. This suggests that it is questionable whether any relevant underlying trend can be determined from the data.
It would appear that the temperature rise profile claimed by the adjusted data is largely if not entirely an artefact arising from the adjustments applied (as shown in Figure 3), not from the experimental data record. In fact, the raw data does not in any way support the AGW theory.
Based on this data, the US temperature data does not correlate with carbon dioxide levels. The warming over the last 3 decades is completely unremarkable and if present at all is significantly less than occurred in the 1930’s. It is questionable whether any long term temperature rise over the 20th century can be inferred from the data but if there is any it is far less than claimed by the AGW proponents.
The corrected data from NOAA has been used as evidence of anthropogenic global warming yet it would appear that the rising trend over the 20th century is largely if not entirely an artefact arising from the “corrections” applied to the experimental data, at least in the US, and is not visible in the uncorrected experimental data record.
This is an extremely serious issue. It is completely unacceptable, and scientifically meaningless, to claim experimental confirmation of a theory when the confirmation arises from the “corrections” to the raw data rather than from the raw data itself. This is even more the case if the organisation carrying out the corrections has published material indicating that it supports the theory under discussion. In any other branch of science that would be treated with profound scepticism if not indeed rejected outright. I believe the same standards should be applied in this case.
*********************
Notes and Links
Interestingly, there was an earlier version of the NASA GISS data shown in Figure 4 which was originally published at http://www.giss.nasa.gov/data/update/gistemp/graphs/FigD.txt While this site has now been taken down the data was apparently archived by John Daly and available at his website http://www.john-daly.com/usatemps.006. The data is presented in tabular form rather than graphical form but appears to be either identical or extremely similar to that shown in my Figure 5.
Other contributions from Michael Hammer can be read here: http://jennifermarohasy.com/blog/author/michael-hammer/
[scroll down, click on the title for the full article]
Ross McKitrick, Ph.D. – Quantifying the Influence of Anthropogenic Surface Processes on Gridded Global Climate Data
http://www.heartland.org/events/NewYork08/newyork2008-video.html
Anthony Watts – http://wattsupwiththat.com/
Discover more from Watts Up With That?
Subscribe to get the latest posts sent to your email.
OK, a few technical issues on the TOBS adjustment. The usually cited authority for this is the paper of Karl et al 1986, and some other papers at about that time, referred to in the CDIAC and NOAA docs I cited above. Again I note that these protocols have been current for over twenty years. They are based on much publicly available information, and haven’t been discredited in that time. If you want to start discrediting them now, you’ll have to study the basis for them rather carefully.
So Chris Wright, here’s your link.
And Konrad, why TOBS adjustments after 1985? The reason is that they adjust to be correct for a specific reading time, for consistency over the record. I think it is midnight. There’s no “correct” time; the best you can get is consistency. If the current reading time is not the long term standard, then there will be ongoing adjustments. This illustrates the illusoriness of the fuss about the magnitude of the adjustments. If they had chosen something close to current practice as the standard, then the current adjustment would be much less. But it wouldn’t make any real difference.
And Tim Clark, how does TOBS cause multiple counting? Evan has been explaining it well. Let me try a specific example. Suppose the instrument is read and reset at 9am each day. Suppose Monday is a very cold morning, bottoming at 6am. Then that bottom is the minimum recorded for Sunday. And the 9am temp (also very cold) is the minimum recorded for Monday, replacing whatever low was reached on Tuesday morning. That cold Monday morning is recorded twice. Maximum’s won’t be split in the same way, so it is a cold bias. If instead the reading time had been 5pm, the opposite process would lead to a warm bias. If the reading time changes over the years, there is a trend. And there’s no point in speculating about whether the reading times have changed. It’s recorded fact; Karl and others have looked at the data, done their calcs, and these have stood for over twenty years.
So Robert Cook, the things you want shown are in Karl’s papers and elsewhere. There’s no claim that EVERYONE changed their times etc – just enough to make the difference. It isn’t true that UHI is now not corrected. It’s just that the correction is made within the general framework of consistency adjustment. This is just nomenclature; the methodology was always similar. UHI correction was always negative and decreasing.
And Michael Hammer – ditto about UHI “deletion”. It isn’t true that the corrections are “all one way”. There are two, TOBS and SHAP, that have lead to significant increases. On the CDIAC graph that you cited, one, UHI, has lead to decreases. FILNET gave a rise from about 1900 to 1910, and has been fairly neutral since. And MMTS was a small one-off recalibration for the introduction of thermistors in the early 90’s. As with all these changes, what do you suggest? That the data should seriously be used uncalibrated?
Robert A Cook PE (10:49:02) :
Also, I’d expect there to be some regional variations. E.g. Mostly sunny vs. mostly cloudy. Frequent/Occasional frontal passages. Sea breeze/Inland. Big shifts in hours of daylight through the seasons (Key West and Hawaii vs northern tier and Alaska).
I’d expect the biggest impact to be on days with inconveniently timed frontal passages. Every year we have a few days where the high temperature for the day is at midnight and a cold front comes through between then and noon with snow and firece Canadien winds and what limited sun we have, even if the clouds break, has no chance to warm us up to better the temperature at 0000.
ohioholic (23:26:01) :
I’m not confident in the following, but it might be right.
The 30 year window has been used in the past going back at least to days when computers were programed by cards, and probably back when computers were often college students on summer break. As such, it made sense to compile typeset by Liotype machine a 30 year reference.
Since then, people learned about the PDO and AMO, the satellite record is just about 30 years, and even a PC ready for dumping in China can produce “last 30 years” data with publication-ready output, either on paper or .pdf file.
One problem with the “last 30 years” or “last 60 years” to try to get a whole PDO cycle” is that research papers will all be referring to different “last 30 years” and it will be as hard to compare current data with the average, just as we need to know the monthly vintage when referring to GISSTEMP with all those backfilled data.
No wonder the models had to make up something that isn’t happening, water vapor feedback; they were hindcast to events that didn’t happen. I’m sorry I’m laughing myself sick. Sick at my uncapped and rising energy bills. I’m going to have to trade my soul to pay the bill. Funny, real funny.
=======================================
Nick Stokes (14:03:22)
Thanks for your link to the Karl et al paper. It will take me some time to get through all of it, but I have already found some areas I want to look into further. I hate to say this but some of the work is model based and the paper’s conclusions mention climate change. Red flag to a bull etc.
With regard to the other USHCN adjustments shown in figure 2, I still see concerns that have not been answered.
A. While the red plot showing adjustments for instrumentation changes seems correct in showing a positive adjustment for MMTS introduction, it should also show a negative adjustment of similar magnitude representing the introduction of latex paint for Stevenson screens around 1979.
B. The yellow plot for siting issues should be inverted, and include a sudden negative adjustment for the short cable issue at the introduction of MMTS.
C. The light blue plot for infilling correction seems to be indicating that data infilling (creation) has a constant negative bias that needs to be corrected for. If this plot is not random in sign and centered around the x axis, questions need to be asked about infilling methods.
D. The purple UHI adjustment should show a significant negative adjustment at the time of rural station drop out in the 1990s.
I have to agree with those such as Evan who say that without reference to time of observation data for individual stations, a generalized TOB algorithm has little efficacy. If TO is recorded as well as Tmax and Tmin, why is this data not being utilized? I am beginning to suspect that physical experimentation for TOB such as Anthony did with shelter coatings may be needed.
I do not see that Michael Hammer is in any way out of line in questioning these adjustments. Much of the history known about USHCN stations is not evidenced in the adjustment plots.
In Australia, many if not most weather stations changed from daily reading/resetting to half hourly. This was in the period approx 1989-1996. The half hour recording apparatus was largely replaced by one-minute recording for the high quality network about year 2000.
An instrument that gives a more-or-less continuous trace over 24 hours is not affected by TOBS. Simple daily inspection will show Tmax and Tmin, whenever they occur. Therefore, the last 20 years of data sent from Australia to GISS has no reason to have TOBS applied. I am currently studying how often it is.
Also, given the small number of days each year that a TOBS effect operated on mercury thermometers, coupled with the fact that most obs were then averged to get Tav from Tmax+Tmin/2, the error carried forward was likely to be halved. Putting these together and running scenarios, the likelihood of the TOBS error being of the size shown for the USA is probably incorrect for Tav. We shall never know until a count of the days when TOBS effect operated eavh year is revealed. If it happened on 5% of the days, the annual error is then 1/20 of the measurement error, roughly. There appears to be no justification whatsoever for its continuation once automatic, multiple daily recording was introduced.
I know that this repeats the sentiments of others who have posted above and my own old posts on CA. I have yet to see an explanation for the shape or magnitude of the USA corrction curve shown above.
The final point is that there would be more than one station observer in the USA who knew of the TOBS effect when recording daily readings and who had the nous to read the device again to get a correct reading for each day, or at least note the second reading in the metadata. Until the metadata examination results refute this possibility, one has to accept it as plausible. In those days people seemed to be more concerned with doing their reporting accurately and noting exceptions, because it was “their” data and their reputations relied on it.
OT perhaps, but still Australia centric. This is what the New South Wales govn’t is doing in terms of “informing the public”…
http://savepower.nsw.gov.au/
Little black balloons of CO2 floating up into the sky….
Although I am all in favour of saving power, consuming less (Originally from the UK which has always had high energy costs and I have lived in an area of New Zealand that has/had the highest unit price for electricity – it also happens to be one of the poorest regions there too) I learnt not to waste stuff.
fwiw, the land grant universities in your state may very well have a parallel non adjusted dataset that could be compared to the GISS datasets. When I’ve done this with the MO sites for the University of MO, I can’t find any trend at all.
“Little black balloons of CO2 floating up into the sky….”
Let’s see, the atomic mass of N2 is what? And O2 is what? So balloons filled with CO2, with it’s greater mass, is “floating up into the sky”. They must be using one of the IPCC’s models to get that result.
Barring rare events, such as a frontal passage that may invert the expected time of high (mid-afternoon) and low (sunrise) temperatures, there’s no way that reading at nearly the same time of day a Tmax/Tmin instrument is going to lead to great confusion. At, say, 9am one gets yesterday’s high and today’s low as a rule, rather than the 9am teperature per se. The whole idea that switching, say, to 10:30am would lead to substantial bias in monthly averages–one which evolves over time in the aggregate, to boot–needs to be critically re-examined.
Bill D (22:43:09) : large lakes
I appreciate the additional information and clarification.
I have been following the warming debate with interest especially Anthony’s survey of weather stations. Prof. Vincent Courtillot thinks that we have very different weather patterns in different parts of the world, however even though I am a skeptic on AGW, in my short life time of 62 years I am pretty sure we have seen either warming or significant changes in weather patterns here in the UK. 40 years ago wide spread snow in winter was common here even in the south of England and Wales but we have hardly seen snow for the last 20 years. So has there been a distinct warming that perhaps is now entering a cooling phase or is it just there are different patterns in different parts of the world, even in different parts of the USA?
Hi, wattsupwiththat.com – da best. Keep it going!
Have a nice day
@ur momisugly Nick Stokes (14:03:22)
Thanks for the link to the Karl 1986 paper. But this paper describes a method for estimating TOBS and does not contain a study of actual TOBS trends. However, it does contain a table that shows a continuous trend from AM to PM measurements from 1931 to 1984 (percentage of AM readings goes from 14 to 42). Unfortunately they do not give the source of this data, which is surprising in a scientific paper. Do you have a link to actual scientific studies of the TOBS trends? Many thanks.
Chris
It is interesting the the derived “raw” data in figure 5 show a peak in 1930 which corresponds to the “dust bowl”. The dust bowl was caused by drought and bad farming practices. Maybe it was quite hot too.
Chris Wright,
The most comprehensive review of world adjustment practices is this 1998 review paper. Unfortunately it is behind a paywall, and also does not have the detailed information on actual TOBS trends that you are seeking. It does describe the GHCN adjustments, which are limited to homogeneity, and do not include TOBS. For TOBS specifically, the 1986 paper does seem to be the most quoted reference.
Here is a paper by Karl et al on the MMTS adjustment.
“jtom (21:47:51) :
“Little black balloons of CO2 floating up into the sky….”
Let’s see, the atomic mass of N2 is what? And O2 is what? So balloons filled with CO2, with it’s greater mass, is “floating up into the sky”. They must be using one of the IPCC’s models to get that result.”
Well, as N2 makes up 78% of the atmosphere, I used that in an e-mail to the person at the “contact us” link at that site. So I e-mail the contact stating that N2 has an atomic weight of about 15, whereas CO2 is 44. Pointed out the fact that CO2 cannot “float up into the sky as if it is lighter than air and in a black balloon”, political spin and misinformation. Pointed out the fact that mercury (In the ecobulbs) is infinitely more poisonous than CO2 (Reference to Japan and people eating fish contaminated with mercury).
No response.
Can anybody get me a copy of :
Mitchell, J.M., Effect of changing observation time on mean temperature, Bull. Am. Meteorol. Soc., 39, 83-89, 1958.
I can’t retrieve it from BAMS. The search function will not bring it up. It appears to be the first cited paper.
For Chris Wright (again)
I think this 2003 review of TOBS is what you are looking for. Unfortunately, it is also behind a (fairly small) paywall.
Has anyone attempted to extrapolate the adjustments into the future? Just by eyeballing the charts it looks like the adjustments could becomes several degrees by 2100. Makes one wonder about the methods being used.
Nick Stokes (05:50:59) :
For Chris Wright (again)
I think this 2003 review of TOBS is what you are looking for. Unfortunately, it is also behind a (fairly small) paywall.
For your money, this might be a better investment.
Pielke, Roger A., Christopher A. Davey, Dev Niyogi, Souleymane Fall, Jesse Steinweg-Woods, Ken Hubbard, Xiaomao Lin, Ming Cai, Young-Kwon Lim, and Hong Li (2007), Unresolved issues with the assessment of multidecadal global land surface temperature trends, J Geophys Res, 112, D24S08.[CrossRef]
Richard M (07:27:20) :
Has anyone attempted to extrapolate the adjustments into the future? Just by eyeballing the charts it looks like the adjustments could becomes several degrees by 2100.
—
Richard, Richard…. Haven’t been keeping up with the news?
MIT did just that study a few weeks ago, and came up with a whirling prediction circular pieplate (er, template) that ALL corrected UHI thermometers will read +9 degrees by the year 2010.
Come on. Ya gotta keep up better!
@ur momisugly Nick Stokes (05:50:59)
Nick, thanks for your efforts. If there are any papers available that are free that would be great but, failing that, I wonder if you could briefly describe the basic explanation for this trend. You would certainly expect random changes, depending on who happens to be running the station, but there would have to be some systematic change that led to a consistent trend over fifty years. I would be intrigued to know what it was.
Many thanks,
Chris
Chris W,
Here is another recent paper with lots more detail (free). And here’s what they say about the history of TOB:
I would also love to know the justification for TOB trends. My current CAD modeling can find no justification for TOB adjustment outside of half day step changes once off for change in TO. I get the same Tav Monthly for TO at 0700, 1700 and 2400 reading and reset for MMTS using typical summer diurnal profiles and winter profiles. WUWT? If I introduce extreme (Tmax, Tmin randomly overlaping TO) red noise to my pseudo diurnal profiles, TOB is random in sign. Even though I have now read the Karl et al 1985 paper, I am not convinced of the need for TOB adjustments. If TOB adjustments are preformed without reference to the TO recorded for individual stations, I suspect these adjustments belong in the wicker filling cabnet. If these adjustment are preformed over the top of data infilling add a can of lighter fluid and a struck match…