In my post on the Mohonk Weather Station, the question came up about “raw” temperature data. Tom in Texas complained that he’d looked at data from the observer B91 forms and that it didn’t match what was posted in published data sets.
Neither NOAA nor NASA serve weather station data “raw”.
We’ve all seen examples posted here of how GISS adjusts data. But, it is not only NASA GISS that does this practice, in fact, NOAA adjusts temperature data also, and it is by their own admission. For example here is a NOAA provided graphs showing the trend over time of all the adjustments they apply to the entire USHCN dataset.
As illustrated in the graphs above, in simplest terms NOAA adds a positive bias to the raw data reported by weather station observers with their own “adjustment” methodology.
It is important to note that the graph on the bottom shows a positive adjustment of 0.5°F spanning from 1940 to 1999. The agreed upon “global warming signal” is said to be 1.3°F (.74C) over the last century.
The NOAA source for these graphs is: http://cdiac.ornl.gov/epubs/ndp/ushcn/ndp019.html
Discover more from Watts Up With That?
Subscribe to get the latest posts sent to your email.
“Can anyone point me to a reference for long term temperature records based on rural areas only where there is no UHI effect and presumably no need for for adjustment to the data?”
One possibility to answer your question is to refer to David Archibald’s rural data set that has over 100 years of history. Although he maintains his data set is “representative of the US temperature profile away from the urban heat island effect,” he has only four stations in his data set. In his data set, the U.S. has not returned to the high temperatures of the1930s. If you add Montevedio, MN (another rural station out of UHI range) to the data set, you continue to get the same answer. Of course, this data set is only the United States. If you add Antarctica, you still do not change the data set trend. If you want to add Arctic locations, you must be careful of micro siting issues, but you do not change the conclusion by adding Greenland sites. (Are we cherry picking?) Now if you want to add other continents, you have quality issues to address first. Longevity, free from micrositing issues, consistently rural all can be challenges.
John Philip
Imagine I take the time once a day by looking at 100 clocks and taking the average. Further imagine that I know one clock stopped for an hour then restarted and is now reading one hour slow. I must correct for this in my averaging, and apply the same adjustment every day unless the clock is put right. Now imagine more clocks go wrong, most in the same direction – the size of the adjustment will increase over time.
Your analogy is thus is that the entire AGW signal for the US derives from a correction for a linearly increasing number of clocks going wrong, with more and more running slow each year over a 40 year period.
If we alter your analogy to allow the clocks to keep correct time, replacing the bias with a change in the time of reading, we still have to show why, each year over a 40 year period, an increasing number of temperature readers decided to change their reading times.
Neither scenario seems remotely probable in a professional observation environment.
Legitimate or not you will find all the surface data trends higher than even RSS.
Yes, and the satellite reading should indicate a c. 30% greater trend than the surface data. That may not look like much on a graph where the “bumps” are in the same place, but it is highly significant.
I don’t think anyone disputes a rise from 1979 – 1998. It is the degree of change that is at issue. I don’t doubt the shape of the graph or the spikes and troughs. I question the degree of “tilt”.
For a heat sink to exaggerate a warming trend, there has to be a real warming trend in the first place for it to exaggerate. (It would also exaggerate a cooling trend, as the warm trend [sic] bias “unwinds” on the way down.) Note that the measured drop in temperatures from 2007-2008 was actually greatest for GISS.
Over the last decade of (relatively) flat temperatures, I would expect the heat sink effect — assuming the siting situation has been stable over the last decade — to have been mostly a push. Of course, one might not assume the siting situation (or the “encroaching concrete” situation) to have been stable.
Would somebody please explain “Time of Observation Bias” TOB.
Now to me a “bias” is a definite lopsided offset. If it were a random discrepancy of either sign, then it would not be a bias.
So if they are able to evaluate this TOB to where they could correct for it; why not just fix the measurement system or schedule to simply eliminate the TOB.
Stories of how climatology “science” is conducted sound to me to be somewhat like Miro$oft Windows; often described as the world’s largest computer virus.
With so many layer upon layers of band-aid fixes du jour, is seems thoroughly deserving of the “Format C:\” update.
So it seems that climate data gathering consists of multiple layers of proxies,
and fixes, and adjustments, and bias corrections. Does anybody even remember what the original AlGorythm was; or does it not matter after you add on sufficient layers of corrections; and TOB factors ?
Tom in Texas (11:44:02) :
Anyone know if satellite data is available for localized areas, or just NH, SH, and global?
Try UAH MSU global anomaly maps, available for the 2000-2008 period.
George E. Smith (13:02:26) :
Would somebody please explain “Time of Observation Bias” TOB.
At any weather station using a min/max thermometer set, the readings would occur at one particular time each day. Depending on what that time of day is, that station’s readings would usually be affected more by either previous day lows, or previous day highs (exactly by the lows and highs of the last 24 hours before the reading). This will often cause a Time of Observation Bias, or TOB.
If readings are taken near the times of daily highs, or daily lows (at 7 am, for example), those highs, and lows, often affect the readings of two days. In any case when the daily mean temperature are calculated from the min/max readings, annual averages of the effects of TOB on recorded temperatures can be more than 0.5-0.6°C (about 1°F) at many locations, and sometimes can even reach a magnitude of 1.2°C (2°F).
An another type of TOB is the effect of any change in the daily mean temperature calculation method. For example, the Hungarian Weather Service defines the daily mean as the average of 4 readings, at 01, 07, 13 and 19UTC. However, this method was only introduced in 1965. Since the beginning of Hungarian meteorological observations (which are started in Budapest in 1780) the daily mean temperature had been calculated from only 3 readings, as a weighted average:
Tmean = ( T_07h + T_14h + 2*T_21h )/4If you try to calculate average temperature for any day, you will get different values with each method. The long-term result is about 0.2-0,3 deg. C positive bias, which can create an artifical/enhanced warming trend.
You can read a very informative summary of TOB here. I would say it is a really good summary, maybe beacuse it is ‘only’ a Summary, not a Summary for Policymakers…
To John Philip:
Regarding “the origin of the idea that the lower troposphere should be warming at 30% above the surface rate,” I have heard Christy explain it, but my link to his presentation on the subject is no longer working. Although I am not confident of the 30% figure, the concept is quite straightforward: CO2 traps radiated energy in the troposphere which heats up the troposphere; in turn, the troposphere heats up the surface, so a signature of CO2 induced global warming is the troposphere heats faster than the surface while the stratosphere cools. I have not seen this concept rejected by Global Warming Pessimists.
Regarding heating at “a higher rate . . . over the tropics,” there is quite a bit of backpedaling on this concept in the GWP community. Although the IPCC graphs certainly do seem to confirm this concept, many GWP speakers are claiming that these graphs are being misinterpreted and that lack of tropical troposphere does not undermine the models. I am intrigued by your reference to Santer. I read his paper and was not convinced. Perhaps my suspicions are on full alert when I again hear the phrase “we have adjusted the observed data and now there is no conflict with model outputs.” But more important, others have tried to replicate Santer’s methodology for more recent data, and the results do not hold.
Regarding the GISS adjustments to deliver a non-urban trend, I find it remarkable that you can read the GISS procedure and can come the conclusion that it is reliably doing what it says it is doing. However, Hansen apparently comes to that conclusion so maybe you can too. However, I assure you that I do not share that confidence. The problems with the Night Light methodology and quality control are so numerous that the length of such a post would be unwieldy.
Regarding Hadley sea surface temperature data and GISS satellite derived data, it is not surprising that from 1980 to 2006 these two measures for the oceans showed increases over the positive phases of the PDO and AMO. What is more interesting to me is that for much of 2008, the temperatures were as low as (or even lower than) temperatures in 1980 (according to satellite data). It is hard to me to believe that CO2-induced AGW is such a pressing issue when “variability” can wipe out almost thirty years of temperature increases.
John Philip (01:25:12) :
But wait a minute, doesn’t UAH rely on GISS data to “fill in” for cloud covered areas, and doesn’t GISS rely on RSS data to fill in for oceans, polar regions and holes in Russia and Africa?
It seems to this casual observer that interlocking data sets would logically walk somewhat in lockstep.
@ur momisugly Jeremy Thomas (04:34:29) and Philip_B (17:32:29)
Actually, the full text is available at that site. (thanks to Philip_B for the link). The main thrust of the paper is the development of a computer subroutine which calculates the TOB for a set of input parameters. There isn’t any direct information on how TOB could have a positive trend over decades. But there were several interesting comments. They stated that there had been a general change in the rules regarding times of observations some time around the 1930’s. It seems that many stations are operated by volunteers and it is conceivable that they switched over to the new regime over quite a few decades, giving rise to a long term positive trend. But it does seem a bit unlikely, nevertheless.
The paper also stated that at many stations the observation times changed several times in a decade, maybe to fit in with the lives of the current observers. This effect would be random and probably would not contribute to any long term trend.
Overall, I’m not convinced by the TOB argument. Have there been any good studies that were able to establish a long term trend due to TOB? That I would like to see.
Chris
Weather Station Data, Raw or Adjusted?
Wrong question!
Let’s start with correct measurements before we discuss the “raw or adjusted” question.
I am still thinking of David Archibald’s temperature forecast for May 2009 he made some time ago! Will he get it right?
The Canadian Prairie Winter Temps anomalies
have dropped a whopping 6.6 Degree in just three years time according to this publication:
http://globalfreeze.wordpress.com/2009/03/25/canadian-prairie-winter-temperature-anomalies-drop-by-66-71-degrees-c-in-just-three-years/
It is unbelievable that the AGW scare still has such a grip on our politicians.
If any legislation is accepted, we should send them out of office for reasons of fraud and incompetence.
At Gizmac, a site that informs their readers about technical gadgeds and developments, promotion of AGW religion is a red threat.
Today they are very worried about the number of people believing AGW has natural causes.
Let’s make them clear thos people are right:
http://www.gizmag.com/climate-change-belief-research-great-cause-for-concern/11322/
IIRC, the raw satellite data has to be calibrated against something. I remember (prb’ly David Smith at Climateaudit) that radiosonde measurements were used for that since they get temps at mid-tropospheric heights. So in some respects even sat temps aren’t independent.
It goes back to Pielke Sr’s recommendation — use ocean heat content for the “real” temps. Of course, measuring this isn’t easy.
“”” Adam Soereg (14:54:08) :
George E. Smith (13:02:26) :
Would somebody please explain “Time of Observation Bias” TOB.
At any weather station using a min/max thermometer set, the readings would occur at one particular time each day. Depending on what that time of day is, that station’s readings would usually be affected more by either previous day lows, or previous day highs (exactly by the lows and highs of the last 24 hours before the reading). This will often cause a Time of Observation Bias, or TOB.
If readings are taken near the times of daily highs, or daily lows (at 7 am, for example), those highs, and lows, often affect the readings of two days. In any case when the daily mean temperature are calculated from the min/max readings, annual averages of the effects of TOB on recorded temperatures can be more than 0.5-0.6°C (about 1°F) at many locations, and sometimes can even reach a magnitude of 1.2°C (2°F).
An another type of TOB is the effect of any change in the daily mean temperature calculation method. For example, the Hungarian Weather Service defines the daily mean as the average of 4 readings, at 01, 07, 13 and 19UTC. However, this method was only introduced in 1965. Since the beginning of Hungarian meteorological observations (which are started in Budapest in 1780) the daily mean temperature had been calculated from only 3 readings, as a weighted average:
Tmean = ( T_07h + T_14h + 2*T_21h )/4
If you try to calculate average temperature for any day, you will get different values with each method. The long-term result is about 0.2-0,3 deg. C positive bias, which can create an artifical/enhanced warming trend.
You can read a very informative summary of TOB here. I would say it is a really good summary, maybe beacuse it is ‘only’ a Summary, not a Summary for Policymakers… “””
Adam, thanks for that explanation; I’m not surprised. Any daily reporting based on a min max thermometer is going to be wrong as far as representing the average temperature for that day at that site; and making three or even four measurments doesn’t really solve the problem.
Also it is not correct to describe the effect as a bias; which implies that it always errs on one side or other of being correct.
The problem is not with the time of reading; it is simply that such data sampling protocols violate the Nyquist sampling theorem; and they do so by at least a factor of two or more; and in that case, it is not possible to recover the true average temperature; because the aliassing noise due to gross undersamplng corrupts the zero frequency signal which is the average being sought.
But TOB and temporal aliassing noise are the least of our worries, because the spatial violation of the Nyquist sampling theorem is by orders of magnitude; not factors of 2-4, so the set of reporting stations, in no way reflects a correct sampling of the entire data field spatially.
Which is why GISStemp and its lookalikes in ground observations, simply report GISStemp anomalies; and in no way refelct the real average earth surface or lower troposphere temperature.
Climatologists need to burn their statistical mathematics text books, and buy a good book on sampled data sytem theory.
I use “Digital and Sampled-Data Control Systems.” by Julius T. Tou of Purdue University. It’s mcGraw-Hill publication but maybe a bit dated (1959) there are more modern texts my colleagues use.
Until the climate data sampling problem is corrected; these groups like GISS and Hadley, will continue to report absolute rubbish disguised as science.
But Adam I do appreciate you taking the time to explain this TOB concept to us.
George
TO: beng (09:18:35) : “. . . the raw satellite data has to be calibrated against something.”
It may be intuitive to think so; however, the explanation is that they compare atmospheric measurement against measurements from outer space, and that is how they determine atmospheric temperatures. This issue has been discussed a couple of times on this blog, and the scientific (albeit confusing in my view) explanation can be found on the UAH site.
TOBs bias explained On One Side of a Postcard:
If my observation time is, say, 2 AM (i.e., near Tmin), and it is 0°C at 2AM on Day 1 and 10°C at 2AM on Day 2 (24 hours later), my Tmin for BOTH days is going to be 0C:
TMin will be at 2:00 AM for Day 1 (O°C) — and 2:01 (one minute later) for Day 2 (O°C)!
Reverse effect if TOBS is near Tmax.
Juraj V. (00:58:02) :
—
Here is comparison between the Bratislava airport measurement on the city outskirts (550,000 inhabitants) vs Lomnicky Peak in the High Tatras (star observatory on the peak, definitely rural):
—-
Thank you! The city went up from 10.50 C in 1951 to near 11.6 at the end of 2006, but the mountain observatory stayed the same.
—
TOBS bias? So, why is the BIAS continuing to change? Once “adjusted” – always and “conveniently” upwards by the way – why is increasingly upwards? Philips “answer” makes no sense mathematically.
Doesn’t the raw data hold time of the reading? A time bias may occur – but once it occurs at a station, then the bias does not increase linearly for the rest of the century. If 1200 stations are changing their reading time over fifty years, then (by about 15 years) half are increasing, half decreasing and the “bias” should stop increasing.