The graph below organizes US HCN stations into two groups. Those that took morning readings (i.e. reset the thermometer in the morning or night) on July 15, 1936 and those which took afternoon readings (reset in the afternoon) on that date.
Both groups show identical downward trends in US maximum summer temperatures since 1910. If TOBS were real, the 1936 morning reset temperatures would be warming much faster than the 1936 afternoon reset temperatures. In actuality, the morning temperatures are cooling slightly faster than the afternoon temperatures.
This shows that TOBS has little or no effect on the temperature, and is simply another excuse to cool the past and turn a long term cooling trend into a warming trend.
How does it make sense that afternoon temps are cooler than morning? Is something flipped?
It is not morning vs. afternoon temperatures. It is the time when the thermometer was supposedly reset.
The delta is due to the fact that it is a different set of stations.
Now THAT’S a cool graph! Checkmate for Karl, et al 1986.
Having a TOBS adjustment doesn’t seem to make sense if the times when the temperatures are recorded are all the same for a given station. For example, if TOBS is set for 2pm, then the temperature will continue to rise after the initial observation is taken. Likewise, if the TOBS is 8pm, then the temp reading at this time will be lower than the actual ‘peak’ temperature for the day. However, this only matters for the very first measurement. After that you are actually recording the true high temperature reading, but just for the previous day. That said, you should have a very nice dataset if you just drop the first day. It may not be the EXACT measurement for that day, but there should be nothing wrong with the relative trend of the data for the daily collections taken as a whole.
If I remember correctly what I learned in my Intro to Met undergrad class, the thermometers record the Max temp for the previous period–usually 24 hours. Some station managers reset the thermometer (i.e. manually shook the mercury back down towards the bulb) in the morning, some did it in the afternoon. Tony, please correct me if I am wrong. So, the only issue is whether the max temp recorded for the recording period included the current or previous day’s max temp.
IF you reset the thermometer at the max of day 1, day 2 is cooler, then day 1 would be double counted. TOBS assumes all the station operators in the past were irresponsible morons.
Only if you changed when the thermometer was reset from one day to the next. if you always reset it at the same time, the maximums would just be a day off.
No, resetting near the max or min will cause major problems in the temperature record. It causes either the high or low to be sometimes double counted.
Hey philjourdan, Steve is right on this — and I know because I initially thought the same as you about this. Here’s a quick explanation. Suppose that today’s high was 100 degrees at 15:00 and you check the readings right at the same time, 15:00. You write down the readings and reset the thermometer. The temperature at that moment is still 100 degrees, so as soon as you finish resetting the thermometer, it goes back up to 100 degrees. Even if tomorrow’s high temperature is only 85 degrees, when you go to read the thermometer (tomorrow at 15:00) you will read a high temperature of 100 degrees. Why? Because 100 degrees is the highest temperature reached since you last reset the thermometer at 15:00 yesterday. You have double counted the high temperature. Now, imagine the same scenario, but with the temperatures reversed, i.e., with today’s high temp at 85 degrees, and with tomorrow’s high temp at 100 degrees. You read 85 degrees and reset the thermometer, which immediately goes back to 85 degrees. Tomorrow is a scorcher and the temperature gets to 100 degrees, that is what will be your high reading for the day. In other words, you will NOT double count the high temp.
If you reset your thermometers at the time of the high point for the day, you will double count that high unless tomorrow is even hotter. In the same way, if you reset your thermometer at the time of the low point for the day, you will double count that low, unless tomorrow is even lower.
Yea, Steve’s answer made me think about it. I was caught up in the “am/pm” issue. In other words, I was thinking that a person would reset it at 8am or 8pm (probably far away from the min and max). But I then saw if you reset it around the maximum (or minimum) you double count.
I am just slow at times. 😉
If it were a problem, then temperature records should be full of double counted values. Since every time a cooler day followed a warmer day, there would be a duplicate max recorded (for PM measurement).
Well, is the record full of duplicated values or not? There should be a means of testing it statistically.
For example, PM measurement gets the daily LOW correct. For every pair of days where day 2 has a lower low, one might expect day 2’s high to be lower, and hence a duplicated high should be recorded. What fraction of the time does a station record a duplicated high in these cases?
It won’t necessarily be an exact duplicate.
TOBS assumes all the station operators in the past
I mostly agree.
My father was an aeronautical-engineer. He maintained his own min-max thermometer records on K&E orange 1×1 millimeter 11×14 graph paper from 1960 to the late 1990’s. His time of observation was nominally when he came home at 6pm. But when the weather turned ugly, he might reset the thermometer before going to bed to preserve the next day’s lower high.
People don’t do this sort of thing, year in and year out to record crap. They take an interest in it. It is a matter of pride — or they wouldn’t do it at all.
Jason, that makes my head spin. Yikes.
Are these hourly temp observations or daily maximum temperatures?
Prior to the advent of Electronic Temperature Stations all the old readings were Daily Max/Min temps.
The whole point about the madness of TOBs is that they have no idea if the “correct conditions” for a value to be carried over to the next day existed or not. So they adjust them all anyway.
Afternoon readings would favour carrying over the high during the change from summer highs to winter lows, ie descending temperatures. Morning readings would favour carrying over colder readings during winter to summer period, ie ascending temperatures.
I’ve become even more skeptical of this “adjustment” after looking through several hundred of the original scanned records. In all of the records I’ve looked at I’ve seen 3 instances with days having duplicate temps. And 2 of those were the low. Many of the record keepers were very diligent about documenting when they changed the expected time of measurement, you can see handwritten notes to that effect. Are there made up temperatures? I have no doubt, but no adjust will ever fix that. The whole idea that we can know the “average temperature” to within a tenth of a degree is completely unbelievable.
TOBS won’t necessarily create an exact duplicate. Only if the reset was right at the max temperature.
And not even really, then, IF you are counting actual reporting periods vs days.
And to throw a little more fuel on the fire…are the times local or GMT (as that is what it was known back then…), standard or daylight?
Until modern electronic thermometers were used no station, anywhere was recorded at an ‘optimal’ time.
Basically, the current temperature record is only correct for the station it was made at, as long as the time the observations were recorded was consistent.
No amount of fudging the data is going to make a better fit for another purpose.
No. The only concern is that the thermometer is reset as far away from the max and min temperature as possible. Consistently wrong would make the problem worse.
Something else that doesn’t quite fit for me is trying to look at these as ‘daily’ min/max temps. I think that is why some suggest using TOBS. If Monday had a recorded high of 100, and this temp is recorded at the actual daily max, and let’s say Tuesdays has a real max of 95, then the record would say that Tuesday’s temp is recorded way too high (100 from Monday). Is the problem then that we are associating ‘days’ with these temperature periods instead of perhaps 24-hour periods? Instead of looking at a Monday vs Tuesday scenario, what if this was just looked at in terms of: for this 24-hour period the max/min temps were x/y, and for the next consecutive 24-hour period the max/min temps were … In this scenario it shouldn’t matter what time of day the temperature was recorded, as long as the record has the temperature recording at the same time.
The only way to get an accurate reading is to reset the thermometer as far away from both the max and min as possible,.
That is correct if you reset both thermometers at the same time. But if you have a weather station that is occupied during the whole day, that would be silly. You could reset the Max-thermometer in the morning and the Min-Thermometer in the evening.
Anerican climatologists assume, that American weathermen were (and still are) stupid. So they do the TOPS adjustment. I assume it is the other way around.
Ok, damn, keep forgetting to think about it with the reset on the thermometer. So, really we are looking at double counting only on those days with a cooler max temp than the previous day, correct? Does the TOBS fudge factor also take this rising/falling temperature trend into consideration?
That is difficult during seasons or even days of transition – guessing when that will be.
If you adjust a station’s temperatures for a change in TOBS, you create a discontinuity.
If you extrapolate a trend across a discontinuity, you create bollocks.
How many of the Tmax temps are exactly duplicated over two consecutive days? This condition would have to exist for a max Temp to be double counted.
So the most difficult was reading the min temperature.
Nick Stokes has an interesting post…
http://moyhu.blogspot.com.au/2014/07/tobs-pictured.html
He sees what he wants to see, but it’s obvious (to me at least) that using MinMax thermometers is a bogus way to measure temperatures from jump streen.
Ha Ha …. went over to Stoke’s site and the words horses arse come to mind.
Yeah. He’s quite the character, almost a parody of a climate scientist. I think he missed his calling as a lawyer. As a scientist, he’s so far from objective it’s not even funny. His comments at Climate Audit are a marvel.
I have been pondering TOBS and got to thinking about the other weather data recorded at the same time. For example, where I live, particularly in winter, the actual time of day that the maximum or minimum can vary well away from the ‘typical’ 2/3 pm for the max and 4/5 am for the minimum.
I’ll give an illustration of this effect. Day one is after a strong mid-latitude storm. The barometric pressure is rising, the wind is strong from the NW and the dew point is falling. The overnight low will be near the dew point temperature but will come it a bit above that dew point temperature from the wind (you’ll not have any dew/frost on the ground either). If the ultimate peak of the barometric pressure comes in during this 24 hour period, the high for the day will be around 2 pm, depending on zenith angle and cloud cover. As we go into the next day, it might be cooler at MN than it was at 5 am. This would be the proper overnight low for the 24 hour period, but that may not be the recorded low for that day. As time passes, the peak of the barometric pressure passes and the winds die down and the dew point has reached its minimum. The station’s hourly curve will fit the ‘norm’.
As more time passes, the pressure falls and the wind begins to shift to a southeasterly flow and picks up. This will bring in moisture and the dew point rises. Timed right and with a strong enough fetch, the temperatures rise steadily over 48 hours. The recorded ‘overnight’ low now happened 30 hours ago and your ‘afternoon’ high happened at midnight. Over time, these excursions will ‘average’ out if the station is well maintained and not moved. In any other situation, all bets are off.
Personally I think any TOBS adjustment needs to be studied in the context of the actual realized weather, including barometric pressure changes, dew point temperature changes, wind vector, cloudiness and the changes in cloudiness, and sunshine hours.
Actually, Steven is well aware that the traditional times of the minimum and maximum are often irrelevant as he lived in Colorado where the weather changes in a heart beat. It actually happens that way in most places. not all the time, but frequently enough that even a regimen of when to reset the instruments will leave false readings.
I can see that, phil :). I do wonder whether the desk jockeys at NCAR do, inside those nice offices in Boulder and their colleagues elsewhere.
As you point out, even a midnight observation time could lead to double-counting minimum temps if there is an unusual weather event. There is no single observation time that would provide perfect Min and Max data every day over decades.
The standard is for recordings taken at midnight, which doesn’t make sense to me when a huge majority of the readings are being done in the mid- to late- morning. Why not make define the standard for USHCN at 7AM, at least for this sort of climate-centric trend analysis?
By keeping the standard at midnight, the are making POSITIVE adjustments (to correct those double-counted morning lows) to data being collected in 2014, which is just ludicrous. The more opportunities for ‘adjustments’ in the procedure, the less reliable the data is, at least as far as I’m concerned.
IMO the problem is easier to imagine with a picture of the Min/Max thermometer.
http://upload.wikimedia.org/wikipedia/commons/4/46/Max_Min_Thermometer.JPG
If TOBS were real, the 1936 morning reset temperatures would be warming much faster than the 1936 afternoon reset temperatures.
Why?!?!
Here is why as I understand it.
You have created two classes of thermometers:
A) Those with morning, (or evening)? (see note 1) recordings as of July 15, 1936
B) Those with afternoon recordings (closer to the max of the day) , as of July 15, 1936.
Over time, those Class B stations will change policy to morning readings.
Over time, the Class A stations will stay as morning readings.
The TOBS adjustment theory is that the afternoon readings have a slight bias toward double counting hot days. Therefore, afternoon readings should have a warm bias. Over time, the Class B stations become morning observations and according to the theory, should cool by about 0.3 deg C (+/- an unspecified uncertainty). The Class A stations will not experience that change in TOBS. Therefore, the Class B stations should see a lower temperature gradient than the Class A stations.
But, Class A and Class B are not statistically different according to the chart at top. Furthermore, you are only looking at the recorded maximums, which should highlight the bias, if any, in the afternoon maximum readings.
This is a persuasive demonstration that TOBS adjustments are unwarranted on a large scale.
Would you confirm that the maximums that you use are unadjusted Raw maximum temperatures?
What are the number of stations in each of the classes?
We are saying the same thing. If the US was warming, the 1936 morning thermometers would show more warming than the 1936 afternoon thermometers, for the reasons you described.
All temperatures I used were raw maximum daily temperatures.
There are about 300 morning stations and 900 afternoon stations.
Where’s the metadata about station reset times? I can’t see it in the ghcnd reame.txt ftp://ftp.ncdc.noaa.gov/pub/data/ghcn/daily/readme.txt
Interestingly, NCDC has deleted it from the web site, but it can be found on the web archive
http://web.archive.org/web/20090713090128/http://cdiac.ornl.gov/ftp/ushcn_monthly/station_history
Figures. And Thanks!
Thanks for the charts, I think they are very convincing. If TOBS requires a half-degree correction, there should be at least a half-degree discontinuity between the afternoon stations and the morning stations, and there should be a big drop in the 1980’s for afternoon stations when much of the USHCN network migrated over to a morning standard.
This is true science, making a hypothesis, making predictions, testing those predictions against real-world measurements. The TOBS crowd left out the most important part, which was to go back and do this sort of sanity-check against the temperature record. No discontinuity means that making systematic TOBS adjustments is unwarranted, and is adding a huge phony warming trend.
Excellent work Steven! Informative discussion.
Thanks I don’t think a lot of people have figured out yet that this discovery is fatal to the NCDC methodology.
Has Anthony Watts figured it out?
No.
I don’t care how or when the thermometer was set in 1936. There were SO many widespread 100 and even 110 degree temperatures, even in places that seldom get them. If you take one look at the data, you don’t have to look twice. Anyone who says it is hotter now than during 1934 or 1936 is a fool and a liar.
Steven,
Would you consider setting up in your header section your posts on TOBS? This is a topic that has been IMHO ceded largely to the lukewarmers who seem at least to me to be a bit gullible. It is important because from even my limited examination of specific temperature stations here in Southern Arizona, TOBS doesn’t look to even do much, except to provide yet more wiggle room for the current crop of gvt_clones to game the numbers to produce artificial warming as requested by their paymasters.
I could just bookmark these posts, but it would be great if there was a directory of them.
Incidentally, one of your greatest graphics of all time; droughts happen, I have also bookmarked and really, that one needs to be in your header section as well.
https://www.youtube.com/watch?feature=player_embedded&v=UqAtal-cn1I
I showed this to a warmista at work. He was stunned. Later, he lectured me that while I might be “smart” there were a lot of people smarter than me. A simple argument from authority. Weak. But I had breached his first line of defense. His subconscious has been perturbed. My work is finished. He is a smart guy. A crack has opened.
Anyhow, this blog simply rocks.
Sure, I will do something like that. I’m just getting started on time of bull shit adjustment..