The complete UAH v6.0 data for July were released on Friday. I present all the graphs for various regions, and as well summaries for easier comparison. The Pause still refuses to go away, despite all expectations.
These graphs show the furthest back one can go to show a zero or negative trend (less than 0.1 +/-0.1C per 100 years) in lower tropospheric temperatures. I calculate 12 month running means to remove the small possibility of seasonal autocorrelation in the monthly anomalies. Note: The satellite record commences in December 1978- now 37 years and 8 months long- 452 months. 12 month running means commence in November 1979. The y-axes in the graphs below are at December 1978, so the vertical gridlines denote Decembers. The final plotted points are July 2016.
[CLICK ON IMAGES TO ENLARGE]
Globe:
The Pause is 3 months shorter.
And, for the special benefit of those who think that I am deliberately fudging data by using 12 month running means, here is the plot of monthly anomalies, which shows that The Pause is over by my rather strict criterion:
+0.33C/100 years since December 1997- not exactly alarming. The Pause will return sooner with monthly anomalies than 12 month means of course.
Northern Hemisphere:
The Northern Hemisphere Pause has ended as expected. Note the not very alarming warming of 0.28 +/- 0.1C per 100 years for half the record compared with 1.38C for the whole period.
Southern Hemisphere:
The Pause has shortened by another 4 months, but still, for well over half the record, the Southern Hemisphere has zero trend.
Tropics:
The Pause has shortened by another 2 months with the El Nino influence, but is still over half the record.
Tropical Oceans:
The Pause has shortened by another 3 months- the El Nino now having a strong effect on the 12 month means.
Northern Extra Tropics:
The Pause by this criterion has ended in this region, however note that the slope since 1998 is +0.34 +/- 0.1C per 100 years compared with +1.6C for the whole period. That’s still embarassingly slow warming.
Southern Extra Tropics:
The Pause has lengthened again by another month.
Northern Polar:
The Pause has decreased by 1 month.
Southern Polar:
The South Polar region has been cooling for the entire record- 36 years 9 months.
USA 49 States:
The Pause is 2 months shorter.
Australia:
The Australian Pause is one month longer.
The next graphs summarise the above plots. First, a graph of the relative length of The Pause in the various regions:
Note that the Pause has ended by my criteria in the Northern Extra Tropics and the Northern Hemisphere, but apart from the North Polar region, all other regions have a Pause for over half the record, including the South Polar region which has been cooling for the whole record.
The variation in the linear trend for the whole record, 1978 to the present:
Note the decrease in trends from North Polar to South Polar.
And the variation in the linear trend since June 1998, which is about halfway between the global low point of December 1997 and the peak in December 1998:
The only region to show strong warming for this period (18 years 2 months) is the North Polar region: the Northern Extra Tropics, Tropics, the Northern Hemisphere, and the Globe have very mild warming but all other regions (including all of the Southern Hemisphere) are Paused or cooling. The imbalance between the two hemispheres is obvious. The lower troposphere over Australia has been strongly cooling for more than 18 years- just shy of half the record.
And finally, here is a plot of Global UAH versus CO2 concentration at Cape Grim from January 1996 to June 2016:
Now that’s a Pause!
Tags: climate, global warming, Pause, UAH
kenskingdom 
August 7, 2016 at 8:52 pm
[…] The Pause Update: July 2016 […]
August 7, 2016 at 9:27 pm
Very nice summary again. I’ll mark this as something to be distributed to my warmist acquaintances – some of them are friends! – and it should be in the in tray of numerous politicians too.
From my viewpoint it would be very time-saving if you could give the URL for the numbers behind the plots. I could go exploring the sites myself of course, but can’t spend much time on this just now. A direct reference to the numbers would be wonderful.
Robin
August 8, 2016 at 8:11 am
Point taken: I overlooked the links. I’ll update now.
KS
August 8, 2016 at 6:03 pm
Thanks, Ken.
August 8, 2016 at 5:38 am
obama and his regime are working overtime to manufacture temperature data that comport to his radical agenda.
August 9, 2016 at 7:49 am
how do you convert from the line equation(in each plot) to your stated T/100yr?
August 9, 2016 at 7:56 am
The equation gives the trend for change in temperature per month. Multiply this by 1200 months to convert to trend per 100 years.
August 11, 2016 at 11:13 am
Many thanks…good work!!
August 9, 2016 at 5:34 pm
Ken, I still find your continued use of 12 month rolling averages to massage your data very perplexing. You stated when you first used this technique back in January 2016 ‘ there can still be a weak seasonal signal, so from now on I will use 12 month running means of monthly anomalies ’.
Using the rolling average technique does indeed reduce the seasonal signal but only partially. For instance the raw data produced a seasonal signal of 0.01 (standard deviation of the averages for the 12 months). The smoothing, by rolling averages means, reduces the standard deviation to 0.0049. As I have pointed out previously this has the adverse effect of reducing the trend values in the vast majority of cases Additionally the statistical significance of a trend fit to a time series is seriously compromised by the additional serial correlation introduced by the rolling average procedure. I will leave the issue regarding statistical significance of pauses for another time.
A much better method that totally eliminates the seasonal signal is the standard method of generating anomalies by subtracting, from the data, the average for that month. This, by the way, is how anomalies in the UAH data set are already calculated ( using the 1981 to 2010 UAH averages as baselines) so you just have to extend this approach by using the average baseline for the full period from December 1978 until the present. This totally obliterate the seasonal signal while having minimal effect upon the trends.
For instance for the global data the trend over the entire interval is 1.215 degrees per century for the raw UAH data (using 1981 to 2010 as the base line) while for the seasonally corrected data (using the base line from Dec 1978 until July 2016) has a trend value of 1.214 degrees per century (i.e less than 0.1% smaller).In contrast the averaged data has a trend of 1.141 degrees per century (about 6% smaller) .
So my advice, which I hope Ken will not continue to ignore, is to stick to the raw data. It may kill off a few of the rolling average induced pauses with the current UAH data but most of these pauses will continue to expire naturally as time goes on. This trend has been evident for the past year.
Regarding RobinEdwards comment, I strongly endorse his approach. I would love it, if Ken could provide a downloadable file or files with his calculations. In the absence of this, the calculations can be done relatively easily with some Excel knowledge.
Personally I think it is important for people to make up their own minds by critically examining the data themselves to see “if they are being sold a pup’.
For those who might need some assistance in conducting this exercise I will point out that the link above that Ken provides above is to a text file. This can be saved as an Excel readable CSV file by right clicking on the file (assuming you are using Windows) and saving with the extension .csv. Then you can read the data into Excel (saving as .xls file if needed) and using the’ text to columns’ menu item in the ‘data’ tab after highlighting column A. Use the fixed width option and ignore the message about overwriting data.
You will then able to plot the data (best to use a scatter plot ) converting the data from year and month in the first two columns and inserting a new column that contains the date calculated by the formula = year+(month-1)/12 i.e. December 1978 should be 1978.917 . This column can be used as the X-axis of the scatter-plot. The trend value will be the trend in degrees C per annum and obviously can be converted to a trend value per century by simply knocking off the two leading zeros. Personally I think is preferable to using a line plot where you have multiply the trend by 1200 to get the trend per century. This requires some mental gymnastics (beyond the capability of the feeble minded like myself, despite my familiarity with base 12 arithmetic from the pre-decimal currency days) or a calculator/spreadsheet.
I have linked to an image of example of my spreadsheet calculations for the raw data for the global data at -https://s9.postimg.org/rowgyvain/data.jpg . The trends have been calculated using the slope function and the trends that are negative are shown with pale red background and trends less than 0.001 degrees C per annum are shown with pale yellow backgrounds.
A Matlab plot of the same data is shown here – https://s9.postimg.org/fo5bpofkv/Matlab_Capture.jpg .
Ken obtained a result of 0.33 degrees per century above for the same raw data from December 1997 but as you can see from my links it probably should be closer to 0.35 degrees per century. This appears to be a minor quibble but his calculations seem to diverge from mine for many other data sets and Ken appears to be making decisions regarding pauses at this level of accuracy.
I am the first to admit that I have been known to be wrong on many occasions and will probably continue to be. If I am wrong, it would be reassuring to know that Ken is calculating things correctly, otherwise we have to consider that Ken’s prolific contributions to the field of ‘pausology’ are, in many cases, misleading.
Despite my misgivings, as a result of auditing Ken’s data, I still believe Ken is providing a useful forum to debate issues such as the pause(s).
It is so much better than the ‘echo chamber’ sites (on both sides of the debate) that typically moderate out opposing viewpoints.
Thanks Ken.
August 10, 2016 at 11:49 am
I think you take the prize for all time longest comment. Perhaps you need to start your own blog?
August 10, 2016 at 11:42 pm
I’ve looked at your plot, which I have to presume, is standard stuff that can be done in MATLAB. I’ve never used Matlab. All my climate analyses are done in my own stats software package. Thus I always choose generate confidence intervals for the least squares line – in simple regression this is easy enough – but I seldom if ever see such things in material posted in blogs. It would be sobering for climatologists to be able to see the confidence intervals (at 95%, say) for the best trend line /and/ for further examples from the same population.
As a general comment on line fitting to climate data, I have severe reservations about generating (and publishing) linear models applied to climate time series without prior visual inspection of the data to see if common-sense would indicate that it is a reasonable thing to do.
Many climate series (of widely varying types) can easily be shown to include very abrupt (step) changes, which in my considerable experience seem to occur at effectively randomly distributed times. Traditional “smoothing” is I think designed to disguise abrupt change. Thus the ultimate smoothing, which is taking the average over the period of interest, totally ignores potentially vital information. A fit, such as a linear fit, is the next most violent smoothing procedure, and by its nature also ignores the wealth of information contained in the /original/ observations. All smoothing procedures discard data that /appear/ to go against local perceived trends, and they all have a problem with the start and end data.
Our friends(?) in NASA and its offshoots not only manipulate the results from analysis of observations but /actively and arbitrarily/ adjust the observations themselves! Is this science? Clearly not.
Work with the original unsmoothed data until it is clear that it is reasonable to fit a mathematical model, linear or otherwise, to a significant portion of the observational data is sensible. Only then make pronouncements about some summary of what the data can tell us about historical climate.
As for “projecting” climate on a regional basis, I firmly believe that on a long term basis it is impossible.
Robin (Bromsgrove, UK)
August 12, 2016 at 12:00 pm
Hi Robin,
With regards to confidence intervals your comment ‘I seldom if ever see such things’ is another puzzle.
As they say ‘seek and thee shall find’. Here are some blogs where this material is presented very thoroughly including uncertainties– Nick Stokes at https://moyhu.blogspot.com.au/p/temperature-trend-viewer.html, Kevin Cowtan at http://www.ysbl.york.ac.uk/~cowtan/applets/trend/trend.html and lucia at http://rankexploits.com/musings/2008/correcting-for-serial-autocorrelation-cochrane-orcutt/ . The latter has a link to a downloadable spreadsheet where uncertainties are calculated in great detail.
These sites all include the effects of serial correlation in the data which needs to be included to get a statistical significant estimate of the uncertainties. By the way, I did include in my Excel spreadsheet download of some tabs that evaluate unbiased (not including serial correlation) of the UAH global data set.
As for your point regarding fitting a straight trend line to the data. Yes it obviously does not provide the level of information, with only limited parameters (the slope and intercept and uncertainties ), compared to the information rich 450 data points of the original data set.
It can only provide a statistical valid (in the right circumstance) measure of the long term trend. It is important in this context as the debate is about the existence or non-existence of pauses. If you have been following Kens blog for the past few years, this is the area which Ken specializes in.
However if you are interested in the shorter term variations due or ENSO signals and volcanic activity etc. then Kens previous and “interesting” post regarding the influence of MEI upon temperatures is relevant.
Finally for a good basic presentation of trend limes and uncertainties due to short term variations see http://climatica.org.uk/climate-science-information/uncertainty . For a more detailed examination of uncertainties and trends see http://tinyurl.com/zoq29p3.
August 13, 2016 at 7:28 am
Many thanks for the excellent links to places that I did know about. I’ve already done some exploring and find that as I hoped here is stuff referring to the Quenouille correction for serial correlation of residuals, as well as the textbook (and necessary) calculation of standard stuff like SEs of estimates, which do not always appear.
I’ve not yet found graphical displays of confidence intervals for trend lines, or for future individual observations, but no doubt they are there somewhere, just waiting to be found.
Nick Stokes’ site is very impressive.
Thanks again,
Robin
August 9, 2016 at 6:19 pm
Ken
Well done on these graphs in particular the breakdown by region, continent and hemisphere. I live in the southern hemisphere after being born in the northern hemisphere. I spend a lot of time travelling between both hemispheres to China, America and UK. For years I have been highlighting the fact that the southern hemisphere is not warming which even the IPCC acknowledge in their biased reports. I also keep reminding folk in Australia that the weather in Australia is not following the northern trends, now thanks to you I have some greats graphs to reinforce that message – well done and regards from David
August 10, 2016 at 4:19 pm
Thanks Ken for posting my comment despite its inordinate length. As I am not a great fan of blog wars that permeate the blogosphere , I will try and restrain myself to more digestible bite sized chunks. Maybe less than 140 characters?
Have you managed to track down the source of our discrepancy in trend values? To reduce the likelihood of again producing another of my epic sagas it could be a good idea to provide a link to your calculations as suggested by RobinEdwards.
Maybe something along the lines of the author of http://www.kiwithinker.com/climate-bet/ who provides a link to download his well laid out spreadsheet.
In a similar spirit of openness here I have a downloadable link to my Excel spreadsheet at
https://drive.google.com/open?id=0B1sQ-dHZoXtSWkU4dC1oZW9YTGs .
Your readers can download and fiddle with it at their own leisure. They can update it monthly and copy and paste data into the appropriate columns for any regions from the UAH data .
As I indicated above it is best for someone who is not sure about pauses etc. to do the calculations themselves rather than being spoon-fed by others.
Ken thanks again for the opportunity to critique some of the material in your blog. I think you would agree It is important that climate science data is audited whether it is produced by professional scientists or by others.
.
August 13, 2016 at 3:54 pm
After looking at temp data-sets it’s very easy to be sceptical of the accuracy of claims made since 1979. If UAH V 6 is showing about 0.11 C a decade and HAD 4 is showing about 0.16 C it still seems to be an exercise in who to believe.
For all we know the HAD 4 trend may have been higher between 1850 to 1950. It’s hard to believe that there really was an accurate measure of global temp all those years ago.
HAD 4 shows about 0.8 C increase since 1850 ( 0.5 C / century) and GISS LOTI shows about + 0.95 C since 1880 ( 0.7 C / century). So who is correct? Yet the Concordia Uni study found just 0.7 C increase globally since 1800 and attributed our OZ AGW responsibility since then to be just 0.006 C.
Lomborg has a team of stats, maths and economic experts who claim that we will waste 100 trillion $ over the next 84 years to deliver a reduction in temp by 2100 between 0.05 to 0.17 C.
IOW no measurable difference at all and yet they danced in the streets of Paris when the COP 21 deal was delivered.
No wonder even Dr Hansen ( the father of CAGW) called it just BS and fra-d and said if you believe in solar and wind energy you must also believe in fairy tales.
August 14, 2016 at 11:15 am
At WUWT Werner Brozek showed the number of years for the satellite temp data where there is no statistically significant warming. UAH V 6 showed no stat sig for 23 yrs 1 mth. RSS showed no stat sig for 22 yrs and 8 months. I thought the Lower trop temps was supposed to increase faster than the surface, according to AGW theory. Yet here we have no stat significant warming in the L trop for well over half the satellite record. BTW he uses Nick Stoke’s data to derive his stat sig.
August 15, 2016 at 1:13 am
Hello Ken and MikeR
It is nice to get a bit involved with some people who know what they are talking about in respect of the climate scene, and of the data that underlies it, and of suitable analytical techniques for addressing it.
Have either of you used the Quenouille Correction on real data sets? I am looking for a data set that has had a linear “model” fitted, explicitly giving the degrees of freedom for the residuals, and the first order autocorrelation for the residuals actually quoted, together with the Quen… statistic and adjusted degrees of freedom.
When I do this I also compute the ratio of the t statistics (normally 95%) for the nominal DF and the Q… DF. With the long series that I generally seem to work with this ratio is generally as close to 1.00 as makes no practical difference whatsoever to the confidence intervals hyperbolae. Am I doing something wrong, I ask myself, since there’s often a lot of fuss about the “correction”?
With much shorter time series, say 50 observations, I can get ratios that can be 1.1 or more, so with most people’s viewpoint this might affect ones conclusions about the series. For me, with a practical background of advising industrial scientists on experimental designs, and doing the analyses of their data, I still would not be worried by this ratio. In the context of the really doubtful climate “observations” that we are served up with by “The Establishment” “scientists”, like GISS I feel that it is of little practical significance too.
Just asking!
I’ll look out for more comments, but am going on holiday very soon.
Cheers, Robin
August 15, 2016 at 8:16 am
I’m not sure why your comment went into moderation! Sorry, I’m not familiar with Quenouille correction so can’t help you. Enjoy your holidays.
August 15, 2016 at 9:19 pm
Hi Robin,
Like yourself, I am learning new statistical techniques as I go. I came across the jack-knife technique pioneered by Quenoille several months ago when reading the Berkeley Earth paper regarding uncertainties in global temperature measurements for the GISS, NOAA and HadCrut data sets . The paper can be downloaded at http://static.berkeleyearth.org/papers/Methods-GIGS-1-103.pdf.
The jack-knife procedure described in the paper involved successively removing 1/8th of the stations at a time and determining how the distribution of average temperatures of the remaining 7/8th were affected by this process. This then allowed estimates of the uncertainties in the temperature measurements.
Robin, I know you have an interest in that area of the reliability of global temperature measurements so you should find the article interesting reading.
Quenoille also introduced a correction for autocorrelation to estimates of the standard error of the trend for AR1 serially correlated data by appropriately reducing the number of degrees of freedom. I am busy porting my Matlab code (my trial is about to run out!) that analyses the trends and uncertainties for all 27 UAH regions, RSS, GISS, NOAA and HadCrut data to C++ and have almost completed the task. The software uses the number of degrees of freedom (n) and the Quenoille correction n= (1-r)/(1+r) for an AR1 process, where r is the autocorrelation coefficient. This is the same method as described in the Appendix ( see the equations A1, A2 and A3 of the paper (http://tinyurl.com/zoq29p3) I referred to above.
Robin, I will take on board your suggestion to provide the values of r and the number of degrees of freedom.
So far my software indicates there is only one statistically significant pause for a 95% confidence limit (for South Pole, ocean data and only for a limited time – for a start date between 1984 and 1992). For the remaining 30 data sets there are no statistically significant pauses for any starting date.
I will report, in more detail, if Ken does not mind, later.
Robin. I would like to join Ken in wishing you an enjoyable break.
August 16, 2016 at 9:01 am
Robin and Mike,
I encourage you to continue your conversation but it might take up a bit of room here.
Can I suggest that you exchange email addresses here if you don’t mind them being displayed.
Ken
August 16, 2016 at 10:12 am
Hi Ken,
Due to my privacy concerns ( I refrained from filling out my census data online) , i would definitely prefer not to have my working Email address displayed.
I do have an email address (miker45107@gmail.com) for sites that require an Email address for signups etc.. This normally contains only junk mail. I rarely check the Email that arrives at this address but I will now periodically check it for any correspondence.
If Robin, or indeed anyone, would like to correspond with me with regard to matters relevant to this blog I would be happy to answer within a reasonable time frame. Of course any unsavoury Emails will be binned.
August 17, 2016 at 4:26 am
[…] Perhaps the showing in Antarctica is connected to the fact that the Earth hasn’t warmed since February of 2008. […]
August 17, 2016 at 6:43 pm
Hi Ken ,
Totally OT, when was the last time you were able to get graphs from
http://tidesandcurrents.noaa.gov/sltrends/sltrends.html
Do you have many graphs saved?
Thanks.
August 17, 2016 at 7:39 pm
I don’t think I’ve ever used this source, sorry.
KS
August 17, 2016 at 9:42 pm
Thanks Ken,
Maybe it will come back with some data later.
Cheers.
August 21, 2016 at 3:53 pm
Andy you can see the SL graphs at this NOAA site. If that is what you want?
http://tidesandcurrents.noaa.gov/sltrends/sltrends_global.htm
September 6, 2016 at 1:14 am
[…] The Pause Update: July 2016 […]
September 11, 2016 at 3:40 pm
Ken, all the other data files have been updated for August at Roy Spencer’s blog but not the Lower Troposphere. Is the LT more difficult to generate or process?
September 11, 2016 at 5:12 pm
Yes I’ve been wondering what’s going on, but it has taken up to the 14th before.
September 11, 2016 at 9:01 pm
Hi Ken,
All the August results, except for TMT, have been available since earlier today and I’m now looking forward to your usual update.