4) NH temperature ensemble anomalies:
fig 49. Click to enlarge
Yearly NH temperature data (mostly land) illustrated together as “best fit”, click to enlarge.
In this writing have mostly chosen mostly to use 5 year averages to reduce noise and make trends more clearly visible to the reader:
“CSST” : A closeup of the Land and Ocean temperature declines 1960-1974, this time with the 4 SST 5 yr avg. graphs from the CSST:
4.1) Land-Air Temperature minimum and Ocean-Water Temperature minimum
A Land-Air Temperature Minimum - hereafter referred to as “LATM” - occurs 1964-74 whereas an Ocean-Water Temperature Minimum - hereafter referred to as “OWTM” - occurs a few years later 1972-78.
Both the LATM and the OWTM has a magnitude of around 0,25 K cooling compared to 1960 temperatures. (I used several smoothed datasets for the CSST earlier in this writing, so the OWTM dip is slightly underestimated by the CSST graph).
Air temperatures over the oceans are obviously affected by temperatures from air over land as well as temperatures of the sea surface. And what we see is, that Ocean Air temperatures to a high degree shows both the LATM and the OWTM.
The fast 0,25K cooling 1961-64 of land air is accompanied by a slightly smaller ocean air cooling and a significant smaller ocean water (SST) cooling:
In the period 1964-72 when land air temperatures appears mostly constant (interrupted by a short warming around 1968), the ocean air and surface water temperatures cools down to the land air temperature level.
From 1976 to 1980 we then see the opposite scenario: A fast 0,25K warming of land air is accompanied by a slightly smaller ocean air warming and a significant smaller ocean water (SST) warming:
These observations seem to support that data are not completely random, but might very well be of useful quality. For example, it does not look like a freak error that SST ends up 0,1 K lower in 1980 than in 1960. In general it appears evident that Land temperatures has a faster variability than SSTand it seems that land vs. ocean temperatures are near equilibrium around 1960-62 and again in 1972-75. That is, even though land temperatures has fast variability (at least in the above examples) SST and land temperatures reaches equilibrium temperature on the longer term. The above illustration is obviously just a scratch in the surface of the science involved. Later we will compare temperatures with Solar activity and ENSO variations.
Can we use a 800-1200 km radius from coastal temperature stations, and then claim that huge ocean areas are covered in the land air series?
If you focus on the years 1974-78 it appears that Ocean air temperatures are quite equal to Ocean water temperatures. In this period it seems that the land air series does not very well include this ocean air trend, and episodes like this questions how well ocean air is represented from coastal cities. So maybe the 1200 km radius over ocean from coastal temperature stations is overestimated.
I believe that a closer study of comparing these datasets might shed more light on this.
4.2) Natural influences on NH temperatures
In the above graphic, I have made some vertical blue and red lines. The illustration shows solar activity (indicated by sunspot number) and then the ENSO index (El nino warming vs La Nina cooling effect).
I made a blue vertical line mostly when solar activity AND ENSO index suggests cooling. I made a red vertical line mostly when solar activity AND ENSO index suggests warming.
Do we see compliance between temperatures and the natural forcings? Blue lines often accompanied by cooling and red lines often accompanied by warming? Yes, to a satisfactory degree.
(In addition, We have not included factors like PDO, AMO, volcanoes etc.)
In this context I have to show the historical and truly amazing illustration from IPCC showing the combined natural forcings extremely inferior to almost anything we humans do:
This illustration will appear in history books hundreds of years from now, just like we today see “odd science” illustrations in history books.
5) Other temperature series
5.1) NCEP and ERA-40 temperatures:
Unlike the smoothed SST series, I just used 5 yr average for NCEP and ERA-40, and these series cover both land and ocean:
The trends 1960-80 seen in data so far are matched by the NCEP and ERA-40 temperature series rather well. Both series are land + ocean, and it is expected, that the land-air temperature minimum in is less represented than the ocean water temperature minimum since the land area is smaller than the ocean area. The NCEP illustrates just this, a smaller 1960´ies temperature minimum than 1970´ies minimum.
However, for the ERA-40 series the land-air temperature minimum is hardly visible. At least to some degree, the missing land-air minimum is likely to be due to adjustments. This I suggest since NCEP data is partly used as basis for the ERA-40.
We see a temperature dip of 0,25 K after 1960, we see 1980 temperatures not too far from 1960. The consistency of data in general appears acceptable indeed between all the original land series, SST series, NCEP and ERA-40.
It is beyond the scope of this writing to comment on all temperature series, but I find the ERA-40 project impressive. ERA-40 is carried out by the European Centre for Medium-Range Weather Forecasts - ECMWF. In general I recommend all to get acquainted with it, since obviously my comments in the following are just a scratch in the surface and not in any way “the truth”. Check it out for the full story: http://www.ecmwf.int/research/era/do/get/era-40
The ERA-40 project started around year 2000, when an impressive portion of raw data was collected from 15-20 sources, and also the huge pile of SST data from ships. ERA-40 cover temperature data for the years 1958-2001.
Here, the coverage in 1966 and 67 – the coverage changed in that period as shown:
All in all it would appear that we have a much larger database for world temperatures than GHCN today.
Here the change in cover for GHCN (by NOAA, used by GISS and CRU today) over time:
Illustration by Joanne Nova. Fig 59.
In 2004, I guess the ERA-40 project was ready with their temperature data series for 1958-2001, but I haven’t seen it widely published. In around 30 writings from ECMWF on te ERA-40 project I have only noticed one person known from the climate gate mails or known from GISS, NOAA, Hadley or CRU. In 2004 it seems that the ERA-40 project has called in Phil Jones to explain something odd. The ERA-40 data did not match the CRU data. In ERA-40, the years 1958-64 appears globally largely as hot as the years 1980-94. (Somewhat like we see in the Raobcore early version). CRU has the 1958-64 around 0,2K colder than ERA-40 compared to modern years. :
Phil Jones in 2004 makes a writing where he seeks to explain the discrepancy together with a team:
My impression from this writing is, that any discrepancy from CRU temperatures in Jones world equals an error, and he seeks for these errors in ERA-40 data only. From this writing the above graphic was taken, and below the info from their table 1:
However these numbers shows a banal error it seems: When ERA-40 SH trend is 0,04K and NH trend is 0,13K, the global trend for ERA-40 is 0,085K and NOT 0,1K (!) . So the difference is 25% bigger than Jones writes here.
To better illustrate the differences, I have recalculated to trend differences for the whole 43 year period in question. Further more I show the difference to CRU directly:
Now its easier to see how big the differences are.
And again, if SH difference is – 0,39K and NH difference is – 0,26K, the overall global difference is –0,325 K and not just – 0,26K.
But its a stunning error in these central numbers: difference ERA-40 vs CRU NH: -0,26K , difference globally: -0,26K and then difference SH… : -0,39K.!!
(If NH and global is -0,26K, obviously SH should be too. Was this paper written without too much review and scrutinizing?
Anyway, from this writing:
” The discrepancy between the ERA-40 and CRU … is particularly marked before 1967. This appears to be related to limited availability of surface air temperature observations for ERA-40 combined with a net warm bias in the model background forecasts of two-metre temperature over this period.
At the time of data supply, NCAR’s holdings for the early years of ERA-40 had some serious deficiencies..”
So the Phil Jones team appears to partly blame the shift of coverage above between 1966 and 67. The differences in world coverage is around 4% - but of course, I these 4% of new area in 1967 really shows a spectacular warming trend, who knows? Lets calculate:
If these approx. 4% of the globe results in 0,32 K too cold a trend for ERA-40, then these areas should have had a 8 K warming trend, erroneously not included in ERA-40.
Am I wrong again and again? Maybe so! As I said, I’m just scratching the surface - so, checkout the original – a voluminous work – and judge for yourselves:
The Jones team is also highlighting, that discrepancy issues mostly originates from the Southern hemisphere. Their fig 10 shows, that Australia adds too much cold to ERA-40 compared to CRU, but it certainly also shows that Antarctica adds too much heat in the ERA-40 data. Australia and Antarctica together does not explain any SH discrepancy then, and Jones needs to find an explanation for the missing heat in SH elsewhere, it seems.
From the fig 61 - 62 it appears that SH vs CRU differences are ven larger than the NH differences we are analysing in the present writing.
Finally: I have not seen a later adjusted version of ERA-40 and thus not a later version that is more in line with CRU data (like Raobcore 1.4 etc.).
5.3) Temperature Proxies
In an article on WUWT 2009 i made a summary of temperature proxies of the Holocene period. I made a summary of some of the most quoted temperature proxies available for the 20´th century, and below is a closeup. Data series are mostly for NH but SH is also represented to a minor degree. Tree ring series dominates these proxy data, but methods are mixed:s
At first glance, the general picture of cooling after 1940, and also after 1960 matches what we have seen in the original temperature data as well as anyone could demand. However, the magnitude of the decline is reduced since multi proxies are 30, 40 and 50 years averaged etc. The number of data series especially after 1980 are reduced and the last years thus has lower quality of data. But despite Gaussian filters and averaging, even the 1960 peak is actually quite visible. So why talk about a “divergence problem”? 1980 is not so much warmer than 1975, but since 1970-75 are shown too warm (due to heavy averaging) it’s a little tough to evaluate such details. But in general Land proxies does not include the Urban Heat Island warming errors like normal Land air temperatures, the one should expect Land proxies in general showing colder trend than Land air temperatures from conventional temperature stations mostly from cities. And this is what we see, without UHI, the warming in recent years is not as strong as we normally see it.
In general my purpose with this writing is to examine temperature trends of data without too much adjusting of data. Non the less, I came across a RATPAC NH surface temperature set, so I will give it a few words.
RATPAC worldwide comprises just around 85 temperature stations – so the NH temperature set has a number of stations similar to the old Angel and Korshover temperature measurements. As we saw with Angel and Korshover original temperature data, even with such a small number of stations, they got results very similar to Chen, Jones, Hansen, Vinikov etc. at the time. But… why launch a new project as late as year 2005 with such a low number of stations as we see for the RATPAC data? In the light of the endless row of reanalysis projects etc. etc. - why does NOAA in 2005 initiate this little RATPAC project?
Original temperature data has been adjusted by Lanzante, Klein and Seidel to form a dataset “LKS”, and on top off these adjustments NOAA adjusts and update the LKS data further in 2005. So, honestly I was… sceptic… when a reader sent the NH RATPAC data to me.
Here´s how the double-adjusted NOAA RATPAC data ends up showing the NH temperatures, the red graph:
Even though the Angel-Korshover 75 data (black) source is quite similar to RATPAC (red), the LKS + RATPAC adjustments results in a far warmer trend than the original temperatures of Angel Korshover 75. A divergence between Angel and Korshover and RATPAC of around 0,18 K appears over just few years in the early 1960´ies.
So far in this writing every single temperature series of any kind shows an approximately 0,25 K temperature drop between 1960 and 1980 . The only real difference is when the minimum occurs. But for RATPAC, even the cooling of ocean temperatures 1960-1975 is reduced with approx 75%.
Bottom line: If the adjustments leading to the RATPAC results were correct, then all the other data I have shown is faulty. Not only faulty, but faulty in a common way so that errors of all kinds just happens to create a consistent “error-trend” of cooling 1960-80 that also happens to match solar activity, ENSO and the most quoted temperature proxies fairly well.
6) Estimate of NH Land temperatures
First for practical reasons, A weighted NH land temperature series was created from the temperature data published 1974-84:
In fact, the different land-air temperature series published 1974-84 has so similar trends that the specific weight of each series would not change much.
On this basis we can show to what degree today’s Land temperature series differs from the original Land temperature data:
As expected, the original Land temperature data published 1974-84 has a bigger dive after 1939, but a nice surprise it is, that Vinnikov 2005 data for the decline 1939-67 is almost identical to the original temperature data. And to some degree this also goes for GISS land temperature data.
CRUTEM3 and NCDC Land data shows the decline after 1939 in land temperatures somewhat reduced.
While CRUTEM3 data for 1925-81 has a temperature trend 0,18K warmer than the Original temperature data, Vinnikovs temperatures trend in the period is just 0,045 K warmer than the original temperature data. For Vinnikovs data and GISS data, greatest difference starts in the mid 1960´ies while this occurs in the mid 1950´ies for NCDC and CRUTEM3 data.
Last changed: 13th July, 2010 at 15:35:31