Roy Spencer wrote a post announcing
UAH Global Temperature Update for September, 2012: +?.?? deg. C,
My betting script is not capable of figuring out who won based on a September Anomaly of +?.?? deg. C. I will wait for further information. I think betting will have to switch to … NOAA? GISS? RSS? Early has been the basis for using UAH. I can pick whichever.
Suggestions?
Maybe you could set up a voting thread. Everyone can vote on whichever dataset they prefer… and we switch to the one that gets the most votes.
Personally, I don’t care all that much which one we switch to… but since we’ve been betting for so long on satellite based anomalies, I’d probably vote for RSS.
Skeptikal–
I’ll probably go for that because it comes out earlier. Is there anything to suggest it might stop reporting? That would be the only concern.
I think that RSS might be easier, but probably of less entertainment value!
I think we should stick with UAH, and also once V6 is published, go back over the last two years betting and adjust who really won!
Aqua 5 shows a strange upward trend over the last week, now at record high. Instrumenet drift?
lucia (Comment #104574)
I’ll probably go for that because it comes out earlier. Is there anything to suggest it might stop reporting? That would be the only concern.
They’ve been reporting late for the past few months, possibly due to dealing with AQUA issues (?)
This graphic suggests RSS dropped AQUA from their TLT analysis last month.
Any idea what’s up with UAH?
Presumably Roy’s post is auto-generated and can’t pick up the figures.
Also, look at Aqua channel 5 over the last few days. It’s rising fast, which is unnatural at this time of year. Perhaps something has gone a little awry somewhere?
NCAR reanalysis also show rising temperature anomalies.
RSS was very late last month (21 Sep).
If you switch to GISS I might participate. More likely if you do an annual GISS predictions bet since I’ve been doing that for several years now. GISS seems to report pretty reliably around half-way through the next month these days.
Use 0.43. Everyone bases their guesses on the daily values on the site, so if you use the average site data vis-a-vis the 2011 values, you get 0.43. Not saying it’s right.
Nick–
Being very late is a problem. I know HadCrut is tending to be late recently too. So.. maybe GISS? Such a conundrum!!
Owen– Not everyone uses Channel 5 to guess. But part of the fun of betting was that you couldn’t get the exact answer based on Channel 5.
For a nominal amount of quatloos, I will undertake to deliver a temperature estimate each month, taking into due consideration all of the major published indices and private communications from Blackboard denizens concerning sharing of quatloo winnings. 😉
I’d stay with UAH. A value will issued by Dr. Roy.. And don’t forget that the new AMSR2 satellite data will going live at any time.
YFNWG:
But is “any time” in October? That’s the issue for October. In the larger scheme of things I don’t mind if UAH has a gap in October. But if we are to bet on October, I need UAH to announce a value! 🙂
Looks like AMSR2 is live… Perhaps Dr. Roy and Christie haven’t incorporated it yet?? Dunno, just guessing. Perhaps AMSR2 isn’t even providing atmospheric brightness temps..
http://suzaku.eorc.jaxa.jp/GCOM_W/data/data_w_product-2.html
YFNWG– I can’t know what Roy is using until after he tells us. Right now, I’m uncertain about whether he will report in October. So….
As to GISS timeliness, Nick Stokes has been regularly putting up a post the day GISS reports in a given month – for this year we had:
* January temps posted February 15 – GISS Temp for Jan 2012 and TempLS
* February on March 11
* March on April 16
* April on May 12
* May on June 15
* June on July 13
* July on August 14
* August on September 15 – August GISS Temp up 0.09°C – ice news (links to all the others at the bottom)
the best I can figure out is it seems to always be either on the second Friday of the month or over the following 6 days. So September’s GISS number should be available between 7 and 13 days from now.
We appear to have a figure of 0.34c for September and revisions to past figures, which I haven’t had a good look at yet.
http://www.drroyspencer.com/
Ray–
Where do you see that the 0.34c figure?
“Do NOT expect the long term warming trend during 1979-2012 to decrease, though, because there are other changes to the long-term time series which cancels out the recent spurious warming”
I guess we now know that satellites are as untrustworthy as thermometers.
Lucia, you should see it if you click on the “Home/Blog” link,i.e.the top link in the list at the top of the page, to the right of the Earth.
It should also be on the page which loads when you click on my link above.
Confusingly, the Latest Global Temp. Anomaly link for August hasn’t been changed to September and it is the same figure.
The actual figure in the list to 3 decimals is 0.338c and August is now 0.208c
Most of the figures since 2010 have changed, most down, but some increased.
Oh wow. So it seems AQUA Channel 5 is on the fritz. That is really too bad. Now I have neither policlimate nor Ch 5 to use! Interesting that removing Ch 5 and looking back at the other two satellites brings UAH more in line with RSS in absolute terms, but the trend is not itself changed.
Ged, do you mean that the trend is still positive?
The trend over the entire UAH series (assuming no changes prior to 2010), has gone from 0.136c/decade to 0.131c/decade.
The trend over the last 30 years has gone from 0.164c/decade to 0.157c/decade.
Personally I think that the results of all of the UAH bets, going back to January 2010 should be re-calculated and the Quatloos re-distributed.
(ONLY JOKING!!!!)
I’m sure all the winners went out and blew their wad on liquor and declared bankruptcy in the jurisdiction of Triskelion. I’ll never get those Quatloos back!
I’d vote against GISS because there never is a final GISS number. You’d have Ray’s issue multiplied a thousandfold, i.e. recalculate all winnings over and over, much like Sisyphus and the boulder.
Somehow I don’t think you’re such a sinner to deserve a fate like that! 😉
Ray,
Oh, no, I mean the trend was not changed due to the loss of AQUA Ch 5, according to Dr. Spencer, “Note that the new v5.5 dataset brings our monthly anomalies over the last few years somewhat more in line with those from RSS, which have been running significantly cooler than ours. The trend change from v5.4 to v5.5, however, only decreases by 0.001 deg. C/decade.”
I think I’m going to lose again in Sept., because after being unusually low in Aug., I added a positive bias to my guess this go round. Oh well …. 🙁
I’ve seen the reported value. But I’m waiting for my montecarlo stuff to run. I don’t want to make my mac cry by asking it to do too much!
Lucia,
“So.. maybe GISS? Such a conundrum!!”
Ah well, TempLS comes out earlier 🙂
I might even bet. The quatloos would be rolling in 🙂
Now I’ll have to reload all the data in my spreadsheet when they finally get around to updating the archive. PITA. JAXA needs to hurry up and get the AMSR2 data from the new satellite launched in May 2012 validated.
Arthur Smith, what is your guess for GISTEMP for the year?
MikeN – back in February (after only 1 month’s GISS results for the year) I guessed 0.65 for the 2012 average. That’s looking rather high – so far we’re at 0.49 for the first 8 months. The La Nina continued deeper and longer than the early forecasts were expecting, but we’re coming out of that now and it should be warm the last few months of the year. So I’m now thinking the annual average will end up around 0.55. We’ll see of course.
Based on nino 3.4, we are already heading into a neutral to la nina phase. The GISS temperature trails nino 3.4 by ~4 months, so (prediction!) global average temperatures should peak in mid December about 0.1 C warmer than August, and then fall in the early part of 2013.
Revising the raw data. That can only mean that nefarious forces are at work to tell us lies. I look forward to a picture of this satellite. Perhaps it is parked too close to the space shuttle.
fun with small town UHI
http://stevemosher.wordpress.com/2012/10/02/city-size-and-suhi/
http://stevemosher.wordpress.com/2012/10/08/more-fun-with-modis/
@Bugs,
There’s a physical reason to adjust — we have a dying satellite channel messing up the results, and we already completely lost one just a bit ago. The whole Aqua sat seems to be showing its age. Thankfully, we can use other satellites to address the issue. Contrast that to GISS which has… what reason to adjust historical figures almost every month?
None the less, the thing to notice is that the changes in the UAH numbers don’t affect its trend more than 0.001 C/decade. It’s more about absolute temperature measurements over the life of an instrument. Again, contrast that to GISS which has significantly changed its trend through historical adjustments without always having a stated reasoning for them. Not nefarious, I would argue, but sloppy or at least not transparent enough to trust the methods.
Ged:
Not all stations report instantaneously (some have taken a decade to update), and as new stations are added the output from the algorithm gets changed (especially if station data from the baseline period changes).
I’m not sure what you mean by not transparent enough. Do you want Gavin to wear a tutu and see-through blouse for you???
😮
They already provide complete code, data, and separately document the algorithms, via published peer-review articles to allow you to completely replicate everything they do. As has been done here.
Complete rewrite of their code in python.
What more could you ask for, other then personal tutelage from them on how their code works? (Maybe they should have a GISS Temperature Code Summer Camp. 😛 )
RSS published a 0.38 anomaly for September.
Lucia,
Project much? 😉
I think it is important to realize that the compilations and adjustments of the temperature data sets is a work in progress. I have found that when my analyses leads to some questions about the data sets a quick email to the proper people will almost always lead to a timely reply. These responders do not always have an answer, or better perhaps, a complete answer but it provides me with an idea of what is “nailed” down and what is not.
I have recently been corresponding with the GHCN people concerning exactly from where and how the unadjusted temperature data are derived and whether it could can change over time. My concerns arose when I found that the differences between adjusted and unadjusted data could be zero going back in time and in fact going way back in time. This concerned me in that it could indicate that the unadjusted sources might have been changed. I think it is very important that the unadjusted original data be kept for all time as it becomes important when the adjusting process changes. I know from my own calculations that if I run the Menne-Williams adjustment algorithm even on data sets that have already been adjusted more adjustments can result and thus having zero differences between unadjusted and adjusted to any great extent is unexpected.
I was informed that I should redo my analyses when the newest version of GHCN comes out in Sept 2012 and then I can repeat my query. The responders to mu queries also noted that the temperatures are not always adjusted when insufficient nearest neighbors exist to use the adjustment algorithm.
Reading this thread reminds me that is time to look for the newest GHCN version. And by the way as I recall GISS is moving or has moved to use the GHCN data set as their own. NCDC of which GHCN is a part provides most of the data for the 3 major data sets so I tend to track GHCN exclusive of GISS and CRU.
It is also of interest that a close reading of the GHCN read.me files indicates that GHCN makes no claims about the form in which they receive the unadjusted data, i.e. it might already have been adjusted by some other algorithm or the one GHCN applies. I assume, however, that all data received is adjusted by GHCN – given sufficient nearest neighbors.
In the link below is the explanation of the changes that occurred on going to the new GHCN version GHCN-M v3.2.0.
It notes that the unadjusted data does not change. This means that my questions about the provenance of the original data probably continues to hold – I’ll have to check. The global mean land temperature trend for the new version changes from 0.94°C/Century to 1.07°C/Century for the period 1901-2011. That is not a small or I would think an insignificant change.
http://www1.ncdc.noaa.gov/pub/data/ghcn/v3/GHCNM-v3.2.0-FAQ.pdf
Paul S (Comment #104716)
October 8th, 2012 at 2:08 pm
“RSS published a 0.38 anomaly for September.”
—————————–
The offset of RSS relative to the UAH baseline is -0.10 C, making the RSS comparison value +0.28 versus the UAH value of 0.34. UAH has been running substantially higher than RSS for the past year or so, as discussed by Dr. Spencer. The decision this month by UAH to retroactively replace the NASA-AQUA measurements with an average of two NOAA satellites has resulted in a substantial downward correction for 2012 (August 2012, for example, went from +.342 to +.208 !!! ). Even with that substantial correction, the September UAH value is still running +.06 degrees warmer than the baseline-corrected RSS value. Puzzling.
Ged,
Please show us an example of a “significant change to its trend through historical adjustments” not documented here:
http://data.giss.nasa.gov/gistemp/updates_v3/
http://data.giss.nasa.gov/gistemp/updates/
Certainly nothing in the last 5 years, which is how long GISS has provided realtime updates on their methodology. Nor have changes to the GISTEMP algorithm and underlying data caused changes in the trend nearly as large as some of the revisions to the satellite analyses over equivalent time periods. A single error in the UAH (or RSS) analysis, or a failure of a single satellite (because there are are so few) throws the results out of whack.
Kenneth,
“It is also of interest that a close reading of the GHCN read.me files indicates that GHCN makes no claims about the form in which they receive the unadjusted data,”
I think there’s no mystery there – the unadjusted data is transferred directly from CLIMAT forms submitted by NMO’s. They are posted here.Since most of them arrive within a few days of end of month, I doubt that there is time to adjust. Certainly in Australia, the BoM posts the temps in real time, half-hourly, and then a daily summary on the current month.At the end of month, that’s what goes on the CLIMAT form.
The unadjusted data was originally issued (V1) on CD. No way of adjusting that later. AFAIK most of that data is still there on the unadjusted file. I have files going back a while and I haven’t seen changes.
Nick Stokes, why would you “doubt that there is time to adjust” when adjustments are (as far as I know) always done automatically by programs? I can’t see any reason time would be an issue.
For example, I’d wager the data GHCN gets isn’t truly “raw” data because it has undergone quality control checking. If I’m right, the data has been “adjusted” before GHCN gets it, and that was done in an automated manner.
GHCN Monthly.
if you are talking about older records GHCN “raw” isnt raw at all. A nice big portion of it comes from adjusted WWR. ‘raw’ means that they apply no adjustments. They flag stuff of course.
monthly data products sent to GHCN would not be adjusted by NWS before sending them out. If you look at how folks do adjustments ( homogenization ) they would do long studies. That gives rise to the problem of having one record submitted to GHCN and then another homogenized record posted on their web site.. for example. Canada has this issue. They submit to GHCN and then the write a paper doing homogenization and publish that record.. which is for example the record that Jones uses.
The only effective way to unravel it is to go to daily records.
Kenneth. you need to dig a little deeper to find the actual sources for GHCN M. its there in the documentation.
Owen (Comment #104720),
RSS dropped AQUA in August, though don’t seem to have made any retroactive changes, probably because their analysis wasn’t as dependent on that one satellite.
Another thing you can see on that chart is that RSS have incorporated observations from Metop-A since 2007. This satellite has not been mentioned at all by UAH so presumably has not been used.
It would be useful to see individual trend contributions from each satellite when making these comparisons, along the lines of this AVISO tool for sea level altimeter data.
Brandon,
“Nick Stokes, why would you “doubt that there is time to adjust†when adjustments are (as far as I know) always done automatically by programs?”
Because most adjustments are done because of readings elsewhere. You have to wait till you get them, and I imagine GHCN gets them first.
Anyway there’s no reason why NMO’s would adjust data. The purpose of adjustment is to get a better global index. That’s not their job.
If you want to see how it’s done here, here is our half-hourly data in real time. Here is the running monthly summary. And here is the CLIMAT form as submitted.
You can check the numbers. For Sept, mean max was 17.7, min 7.3; mean was 12.5. Just as it says here.
Interesting to see GHCN have come up with a new version of their adjustment code v3.2.0. I wonder if it will be any better than the erroneous previous version. At first glance it looks like the fake warming is still there.
Kenneth’s experience does not match mine. I asked GHCN back in January about the large past-cooling adjustments introduced in v3.1.0 and they have still not replied except for a short acknowledgement saying ‘stay tuned for updates’.
Arthur Smith, thanks. I was evaluating Thomas Fuller’s chances of winning his bet with Joe Romm, which you thought Romm would win about 70-80% chance earlier this year. How much would you say Fuller’s chances improved based on what you’ve seen since then?
I think your.55 is high estimate. .49 for 8 months means you are 48 below, ared you expecting a .67 average the rest of the year?
MikeN, it’s early days, but I grow more hopeful that my bet–a politically motivated attempt to blindly catch the top of the sawtooth waveform–might be a lucky win… Maybe I’ll start buying lottery tickets as well.
Nick Stokes (Comment #104722)
Steven Mosher (Comment #104725)
NS/SM, unfortunately your comments are of no help to me in attempting to determine the state of the unadjusted data as reported by GHCN. I am well aware of the sources used and the other details of collection. I will continue to pursue answers from GHCN.
What is puzzling to me is the following:
In most of the major populated nations of the world, I find on plotting the absolute difference between the GHCN unadjusted and adjusted monthly mean temperatures averaged over all reporting stations, a monotonic decrease in the difference with time, i.e. the adjustments have become smaller over time. In fact in recent years the differences are close to zero for many these nations. This situation could result from the nation’s unadjusted data being adjusted in a manner similar to that used by GHCN or less likely the unadjusted data requires little or no adjustments.
It is important to remember that GHCN applies its adjustment algorithm to all incoming station data with sufficient nearest neighbors. It is also important to note that the algorithm used by GHCN will result in further corrections in goodly portion of cases for already adjusted data when applied a second time. What I see in the difference plots is not only the change to smaller differences coming forward in time, but, when going back decades in time, straight line plateaus over extended periods of time that show that all the stations for that nation received an adjustment of equal magnitude. Sometimes the plateau line is at zero and sometimes different than zero. The exceptions to these nations stations averages showing plateaus are the US and Australia and until recent times Canada and Japan.
If we consider that GHCN should have maintained all the unadjusted data they received from their sources in the original form and only the adjusted data will change when the version/methods of adjustment change, I cannot readily explain those plateaus I see resulting from the same difference over all the stations of a nation. It is though a cart blanche nation adjustment was made. When GHCN applies a new algorithm that process would have to be applied to all the unadjusted data going back in time. It becomes very problematic to understand under those conditions a plateau period back decades ago where the difference between all of a nation’s stations would be zero, because the current algorithm has only been in use for a couple of years.
Another explanation for these plateaus is that somehow the unadjusted data has changed over time – and that is not a good thing.
cce (Comment #104721)
GHCN has recently updated a version of an adjustment algoritm – and not changed the algorithm – and obtained a trend change for 1901-2001 from 0.94 degrees C per century to 1.07 for the global mean land temperatures.
I would say that temperature adjustments are a work in progress – satellite and surface.
GISS uses GHCN adjusted data almost exclusively, so I would assume that trend change will appear in their data set also.
Kenneth,
“What is puzzling to me is the following:
In most of the major populated nations of the world, I find on plotting the absolute difference between the GHCN unadjusted and adjusted monthly mean temperatures averaged over all reporting stations, a monotonic decrease in the difference with time, i.e. the adjustments have become smaller over time.”
This is just because the adjustment algorithm starts from the present and works backwards. The adjustment is always zero at the present time. Since it is only the trend that is of interest, it does not matter where you set the zero of the adjustment.
Kenneth,
“What is puzzling to me is the following:
In most of the major populated nations of the world, I find on plotting the absolute difference between the GHCN unadjusted and adjusted monthly mean temperatures averaged over all reporting stations, a monotonic decrease in the difference with time, i.e. the adjustments have become smaller over time. ”
This is just because the adjustment algorithm starts from the present and works backwards. The adjustment is always zero at the present time. Since it is only the trend that is of interest, it does not matter where you set the zero of the adjustment.
“This is just because the adjustment algorithm starts from the present and works backwards. The adjustment is always zero at the present time. Since it is only the trend that is of interest, it does not matter where you set the zero of the adjustment.”
The adjustments are made to the station that the algorithm shows a need for an adjustment. As noted in the link I gave above for the GHCN changes with the new data set version, the changes/adjustments were mostly to the period before 1970. You would have to show me where it is stated that the adjustments are made to give zero adjustment at present time. My concerns had nothing to do with how the trend is calculated or affected. They have to do with a potential change in the unadjusted values.
I would suppose that the algorithm needs a certain amount of time series data in order to make an adjustment and that there can be the algorithm’s capability of handling “end” effects to be considered here, but I also know that an adjustment required at a time in the recent past could make an adjustment that covered into current time.
>catch the top of the sawtooth waveform
The start year does change things quite a bit.
2000 has a 36 while 2010 was 63 for a +27. Then your next lowest are 44 and two 49s.
However, 1999 was a 34, so 1999-2008 would have made your chances tougher, especially with 2009’s 57 added to Romm’s tally instead of yours.
Now if you had done 2001-2010 you get the 63 and drop the 36, and perhaps cruise to victory.
Nick Stokes:
I guess the problem is you and I are using “adjust” and “adjustments” differently. You say the purpose of them “is to get a better global index,” meaning you’re only talking about adjustments done for creating a global index. I, on the other hand, think there are many reasons adjustments might happen, and not all are for that purpose. As I mentioned above, quality checking is a form of adjustment, and it happens. That can be confirmed as a source you provided says:
Of course, a more likely source of questions comes from the fact the process you describe is a recent one, and it isn’t true for older data.
Kenneth,
GISS documented the switch to GHCN 3.2 in its September 26th update, and they linked to a description of NOAA’s changes. These facts are in stark contrast to the beliefs of some posting here and elsewhere.
Kenneth, that’s a good question. I don’t know of anywhere in the GHCN papers or documentation where they say that the adjustment is defined to be zero at the present time. My statement of this as a fact was just based on downloading their data and looking at the adjusted and unadjusted numbers, which always agree for the most recent data for all countries except USA which is handled a bit differently.
cce, the description you refer to is the same file Kenneth himself linked to in #104719 🙂
“Because most adjustments are done because of readings elsewhere. You have to wait till you get them, and I imagine GHCN gets them first.
Anyway there’s no reason why NMO’s would adjust data. The purpose of adjustment is to get a better global index. That’s not their job.”
Nick Stokes point here about why it would be difficult for the unadjusted data to be adjusted before arriving at GHCN for adjustment is well taken and only deepens the mystery I have about the plateaus I see in the differences between adjusted and unadjusted data by country.
I am in the process of locating and doing the same analysis on the new version of GHCN.
“Anyway there’s no reason why NMO’s would adjust data”.
This is not correct. NMOs do adjust data, to take into account station moves and changes in measurement procedures. This makes sense – they are the people best placed to do such corrections. For the case of Iceland, this was discussed by the iceland weather blog in a series of posts earlier this year, see for example
http://icelandweather.blog.is/blog/icelandweather/entry/1228460/
For Iceland at least, and probably many other countries, we have the following absurd process for obtaining GISTEMP:
1. Iceland Met office adjust data
2. GHCN then add adjustments using their algorithm
3. GISS add their own homogeneity adjustment.
Paul Matthews (Comment #104760)
If you think about how these adjustment are made, i.e. with nearest neighbor stations comparisons , I think the catch is that a station that produces a temperature record and sends it to GHCN in a timely manner would not have the information (from other stations) to adjust it before sending the data. If all the data were first collected by an intermediary and that collector immediately adjusted it before sending it on to GHCN you would have a point.
This doesn’t appear to be the case. Following your Iceland MO link I found both unadjusted and IMO-adjusted datasets for ReykjavÃk. Comparing these the most recent data (the adjusted set only goes up to 2000) are identical but, going back a few decades reveals discrepancies, obviously where ajustments have been made.
I then opened up the Reykjavik station entry at GISS (pre-GISS adjustment version) and picked a suitable year for cross comparison (1970). Comparing with both IMO-hosted datasets I found that the GHCN-sourced data at GISS was a perfect match for the IMO unadjusted data, and different from the IMO-adjusted.
Tried to make an edit to my previous post but it doesn’t seem to have worked, so here it is:
I should clarify that I was checking the monthly data, not just the annual. That is to say every monthly entry for 1970 is identical in GHCNv3 and the unadjusted data, but different from the IMO-adjusted dataset. I’ve just checked and the same is true for 1969. Pretty low odds on this being coincidence.
Paul S, it’s documented that GISTEMP uses adjusted GHCN data.
More detail here:
GHCN = Global Historical Climate Network (NOAA/NCDC) version 3SCAR = Scientific Committee on Arctic Research
Basic data set: GHCN - ftp://ftp.ncdc.noaa.gov/pub/data/ghcn/v3
ghcnm.latest.qca.tar.gz (adjusted data)
If you go to the README file for GHCN
V3 contains two different dataset files per each of the three elements."QCU" files represent the quality controlled unadjusted data, and
"QCA" files represent the quality controlled adjusted data. The unadjusted data are often referred to as the "raw" data. It is important to note that the term "unadjusted" means that the developers of GHCNM have not made any adjustments to these received and/or collected data, but it is entirely possible that the source of these data (generally National Meteorological Services) may have made adjustments to these data prior to their inclusion within the GHCNM. Often it is difficult or impossible to know for sure, if these original sources have made adjustments, so users who desire truly "raw" data would need to directly contact the data source.
Paul S, thanks, I think you are right, my step 1 does not occur.
The ‘raw’ input data used by GHCN, the qcu files, come from the icelandic data ‘as originally published’ (which can also be found here), not from the version adjusted by the iceland met office.
Carrick (Comment #104794),
Yes, I’m aware of that: The data I checked on the GISS site was clearly marked ‘GHCNv3 (adj)’. My point was that GHCN appear to be using the unadjusted met station data, not the IMO’s adjusted set.
Paul S, I do agree that the data from Iceland for that period were not adjusted. What the GHCN people are trying to warn us about though is “Often it is difficult or impossible to know for sure, if these original sources have made adjustments, so users who desire truly “raw” data would need to directly contact the data source.”
It didn’t occur with this particular period of this data set, but the data that gets handed over to the GHCN often happens without them having a lot of say about the data feed. It would be much easier if all countries adopted the position that the raw and adjusted sets for their country be made available, for all periods where there is data, free of charge.
That said, it is my impression that CRU uses the local adjustments, when available, and then makes few other changes to the data. In a way, that is superior to adjustments that are make by computer.
See the discussion of automatic versus manual processing here.
I was curious awhile back about the differences in the amount of temperature adjustment required on going from country to country in the world – and probably because of issues of TOB adjustments that did not appeared to be used globally. I had done a country by country analysis on the GHCN Monthly v 3.1.0 station mean temperatures for 24 of the larger countries of the world with reasonably extended records. I saw some what I thought were curious results that I could not resolve as to possible causes. I put my questions to the GHCN people and they replied that I should wait until the v 3.2.0 version was available and then do my analysis again and then query them on any remaining questions. I have completed the same analysis on the 3.2.0 version and my questions remain the same.
My analysis involved taking the absolute difference between the Adjusted and Unadjusted station data for GHCN monthly mean temperatures on a country by country basis and calculating an average absolute difference by year for each country for the time period 1900-2011. My temperature data was extracted from the links below.
ftp://ftp.ncdc.noaa.gov/pub/data/ghcn/v3/ghcnm.tavg.latest.qca.tar.gz
ftp://ftp.ncdc.noaa.gov/pub/data/ghcn/v3/ghcnm.tavg.latest.qcu.tar.gz
The results of my analysis are presented in the links below with each of 4 links showing the results for 6 countries.
http://imageshack.us/a/img6/9219/adjunadjghcn3201.png
http://imageshack.us/a/img855/9057/adjunadjghcn3202.png
http://imageshack.us/a/img829/59/adjunadjghcn3203.png
http://imageshack.us/a/img838/43/adjunadjghcn3204.png
I can summarize what I observe in the following list:
1. All 24 countries average absolute temperature adjustments decrease monotonically towards zero with time.
2. The amount of adjustment over time varies significantly by country.
3. The US uniquely has not reached zero difference between Adjusted and Unadjusted temperatures in the past few years.
4. Some countries show plateaus over extended years in the average station absolute differences between Adjusted and Unadjusted tempertures going back in time and some of these difference are 0.
I can conjecture a couple of potential explanations for the tendency noted in 1 above, with one being that the measured temperatures are performed more scrupulously than in the past and require less adjustment. Page 18 of the link immediately below shows a graph with the temperature adjustments made over the period 1895-2010. It can be seen that while there is a fall-off in adjustments in recent times, the adjustments in these graphs do not indicate the monotonic changes in average station adjustments we see in the table I produced and linked.
ftp://ftp.ncdc.noaa.gov/pub/data/ghcn/v3/techreports/Technical%20Report%20NCDC%20No12-02-3.2.0-29Aug12.pdf
The other conjecture on my part is that somehow the unadjusted temperatures are not properly reported here or these temperatures are being adjusted and being sent to GHCN as unadjusted temperatures. Knowing that even adjusted temperatures (and even with the GHCN algorithm) could require further adjustments with the GHCN algorithm on a second iteration makes me curious on how no adjustments are required for many countries in recent years and some going back many decades in time.
In the following link GHCN notes the uncertainty of the condition of the unadjusted/raw data that they receive and I will not excerpt it here as it has already been in a previous post by Carrick.
ftp://ftp.ncdc.noaa.gov/pub/data/ghcn/v3/README
Notice that in Denmark there is a period of approximately 30 years starting in 1915 that has zero adjustments over all the stations in the country. While there only 10 stations for the Denmark over the 1900-2011 period, the period in question had 6 stations operating. I found that result at face value to be difficult to accept other than assuming those stations over that time period were an example of pristine temperature measurements.
Now there may be very simple and logical reasons for these observations that I find curious and for which I do not have an easy explanation. I plan to email GHCN with my latest analysis and observations and ask for their comments. I would appreciate any that might also come from readers of this post.
Paul,
The accusation (by Ged) was that GISS does not “state a reason” for “large historical adjustments” and that is “sloppy” and “not transparent.” This, of course, is false. GISS documents all significant changes, both in the GISTEMP software and the data used. This includes the recent switch to GHCN 3.2.