T
PART2: The perplexing temperature data published 1974-84 and recent temperature data. |
| Posted by Frank Lansner (frank) on 13th July, 2010 |
PART 2
3) Presentation of some of the 1974-84 temperature data sets used.
Each data source will be compared to a “NH temperature ensemble” of land temperatures, all 5 year averaged.
3.1) Angel and Korshover 1975
Angel and Korshover made high quality temperature recordings from balloons. Below is shown A) CRU/Philip Jones temperature trends by 1982 compared with B) Angel and Korshover surface temperature trends by 1982, both Northern Hemisphere. Both datasets are from above sea surface, and mostly from land:
Fig 8.
The 2 temperature trends overall have a fine resemblance. But Angel and Korshover used just 43 radiosonde stations in the Northern Hemisphere while Jones 1982 used all stations available meeting his criteria’s, generally around 1000 stations.
Just like the IPCC graphic in the introduction, we see that consistency is fair despite very different methods and very different number of stations.
Fig 9.
Above: Angel and Korshovers radiosonde stations. Using Modern GISS std. Coverage radius of 1200 km, it becomes evident that most of the NH land area is in fact covered to some degree by the Angel and Korshover stations. These have been spread out to cover the Earth best possible.
The black areas – the areas not covered at all – are distributed very randomly over the whole Northern hemisphere, and thus, should these black areas have a temperature trend significant different from the green covered areas, then Angel and Korshovers data would be wrong. But it is likely, that black areas scattered and spread out on the Northern hemisphere just by coincidence should have a common trend significantly different from the rest of the Northern hemisphere? No, of course not, and therefore it is in fact not really a surprise that we see a good agreement between Jones 1982 and Angel and Korshover 1982.
The number of stations in Angel and korshover is just 43, but they are spread out evenly over the Northern hemisphere – and therefore useful.
Here, from Angel and Korshover, a compare with other studies of that period, all supporting the significant temperature decline after 1958:
Fig 10-11.
The raobcore tropics added by me for compare. Its generally accepted, that the decline 1940-75 was larger in the Northern hemisphere than the southern hemisphere and the tropics, so the tropic data shown is most likely to be and understatement of the decline 1958-1976.
Angel and Korshover 75 compared with the “NH ensemble” click to enlarge:
fig 12.
3.2) Chen 1982
This work is remarkable since the results where published approximately the same time Jones made his 1982 land NH publication, but this work of Chen´s covers the period 1949-72 for many different areas: land NH, ocean NH, combined NH, land SH, ocean SH, combined SH, land global, ocean global and global combined. This work started in 1976… ended 1982. Chen used NCAR data for land and ocean with the following coverage, in his 5x5 grids:
fig13.
And in addition he used Metoffice ship journals covering for January in average 1949-72 (Antarctic has less cover in southern winter from ships):
Fig 13b.
Chens decline 1949-72 appears similar to Jones 1981:
fig 14.
Chens data are available in written numbers month for month 1949-1972 for all the large areas of the globe.
Chen 82 compared with the “NH ensemble” click to enlarge:
fig 15.
3.3) Folland 1984 / Ocean temperatures
For ocean temperature data we normally use either SST (Sea Surface Temperature) or MAT (Marine Air Temperature). SST is temperature 2m beneath the water surface, and MAT is air temperature over the ocean surface.
Folland showed in 1984 what appear to be errors in the ocean temperature data sets. The most significant issue was a major warm peak around 1941-46. Folland (84) adjusted for these problems in 1984:
fig 16.
I have seen 2 explanations for adjusting the 1940-46 SST peak down:
1) during war, all over the world an night peoble dared not stand on their boats with a flash light when collecting the bucket of sea water. So all over the world SST collecters have had the same scare? Maybe a German bomber plane would appear in the south pacific? The Hudson Bay? Near the great barrier rief? Near Brasil? Alaska? Sure thing. As a consequence water was taken in by the machine inlet under water surface.
2) The material of the buckets where shifted simultaneously and the new material led to a different temperature inside the bucket.
In both cases – probably misquoted by me - the problem is supposed to be connected to the buckets used to collect the water:
fig17.
However, the 1940´ies strong warm peak in ocean temperature data was not just a water temperature problem (SST). In fact, the warm peak of the marine air temperature (MAT) was even stronger than the sea surface water (SST) warm peak.
So the very idea, to focus on the water buckets when explaining the 1940´ies warm ocean temperature peak appears surprisingly wrong. Or perhaps people at sea started using buckets for air just around 1940?
Further discussed here: http://climateaudit.org/2005/06/19/19th-century-sst-adjustments/
None the less, the ocean temperature peak around 1940 does appear “too large”, but I find it more likely that the problem is related to a much reduced area of the globe where temperature data was collected at sea during war time.
Below how the Folland 84 version of ocean data of Northern hemisphere data is presented in Jones et al 86:
fig18.
Below a compare with other Marine Air Temperature sets:
fig 19.
The Folland 84 “global” is a simple average of Folland SH and NH. These data held together with Chen 82 Marine Air Temperature shows a decline after 1940 rather similar to other Northern Hemisphere original temperature data.
The COADS series appears a bit more like the unadjusted MAT data. (Series named “Tisdale” are taken from: Bob Tisdales “Climate Observations”: http://bobtisdale.blogspot.com/2009/03/part-2-of-large-sst-discontinuity-also.html )
fig 20
The HadMAT1 also shows the decline 1940-75 but slightly reduced in magnitude.
MAT vs. SST:
When reading Chens 1982 publication, the surface temperatures for both land and ocean are (obviously) air temperatures, MAT. In fact, I have not noticed any publication of the Earths temperatures published before 1985 where ocean water temperatures are used in stead of oceanic air temperatures when illustrating the Earths temperatures.
But for decades now, it is in fact SST and not MAT that is combined with land temperatures to illustrate world temperatures.
Why did SST take over the role of MAT when illustrating World temperatures?
fig 21.
In general SST appears to show less variation than MAT, which creates some debate today if its really a good idea to use SST for global temperature trends (!). None the less, at some point after 1984 the decision to use SST in stead of MAT in global temperatures definitely reduced the size of the temperature decline after 1940 significantly.
Below, a similar comparison, this time more adjusted data from the 1990´ies, but the same general difference in the decline trend after 1940 between SST and MAT:
fig 22.
Northern Hemisphere; Corrected NMAT, this analysis- Follans 95 (dotted); corrected NMAT (Bottomley et al., 1990) (dash-dot); corrected SST, MOHSST6 (solid).
( But still, all datasets support a significant decline after 1960 – which does not support that IPCC cut tree proxies after 1960 due to falling trend).
Bob Tisdale on climate observations describes problem for the SST temperature corrections too. We see that HADISST has been adjusted down so that the 1940 peak is practically gone (!) whereas the COADS SST versions has a larger temperature peak::
fig 23.
Compared with Folland 84 SST, some SST series are fundamentally different around 1940-60, for example ERSST.v3b. But many things can happen to data from the oceans…
fig 24.
Today it seems there is an endless row of different SST datasets. MOHSST6 (adjusted by Folland in 1995) appears to resemble Folland 84 much:
fig 25. click to enlarge
SST-graphs after 1960 shows quite similar trends, however, the SST before 1960 shows an extremely noisy picture.
“CSST” = Combined SST.
For practical use in this writing I have created an average SST curve from 1960 and forwards. Before 1960 the differences in SST have a chaotic character and I find pre-1960 SST hard to use for now. On the other hand, the post-1960 SST temperatures actually show good agreement even between the old Folland 84 graph and later material. Therefore the “CSST” average graph appears solid to use (at least until 1990 where some inconsistencies appears again).
The CSST curve is a rather smoothed graph calculated as an average of the above series of smoothed and 5 yr averaged data.
Folland 84 MAT and SST compared with the “NH ensemble” click to enlarge:
fig 26.
3.4) Hansen 81
Hansen published temperature data from land stations placed on continents and islands thus covering all Land area + around half the Ocean area. The approximate extend of the coverage is given in his 87-publication:
Fig 27.
So, according to Hansen, roughly, half the Earths oceans are covered – around two thirds of the NH ocean:
fig 28.
In Hansen 81, their results appear not to be regarded as “land only” temperatures.
Ocean-temperatures measured from islands or coastal cities are likely to include some degree of urban heat. Therefore it would not be surprising if these land-ocean temperatures had warmer trends than Sea Surface Temperatures.
Hansen 81 temperature result has been published in 81 and again in 86 slightly differently:
Fig 29.
A) The original Hansen 1981 graph of the Northern Hemisphere showing a large decline 1937-72 of around 0,45 K.
B) Jones et al 86 shows “Hansen 1981 NH” now cut down to just 0,25 K.
A) and B) “Hansen 81 NH” are different because A) covers only 23N-90N, the Extratropics of the Northern Hemisphere whereas the 1986 edition – given to Jones via personal communication – is the entire NH.
In fact, the B) graph resembles the data from Hansen and Lebedeff 87.
fig 30.
Original Hansen 81 temperature graphs. Fortunately we have tropic temperatures from the original Hansen 81 in addition to the extra tropics NH Hansen 81, so we can roughly generate the real Hansen full NH 81 and compare it with the Jones 86 version of Hansen 81:
fig 31.
We thus have a compare NH 1981 land stations vs. NH land stations 1986. I find the resemblance of the 81 and the 86 version convincing all the way up to 1963. - A superb match definitely showing my original Hansen 81 calculation to be useful (even though I had to use the temperatures from full tropic for the NH tropics).
The interesting lesson from this graph is, that suddenly around 1963 the newer 86 version is adjusted up compared to the 81 version, see the grey curve in the bottom of the graph.
What type of adjustment can explain this sudden significant divergence only starting in 1963?
Again, using Hansen-Lebedeff 87 in stead of the Jones 86 version of Hansen 81:
fig 32.
Decline trend for the period, Hansen 81 -0,34K Hansen Lebedeff 87: -0,25K.
The argument to justify the 1938-72 decline in Hansen 86-87 compared to Hansen 81 goes, that the change from NH extratropics to full NH.
This we will test using Chen 82 data 1949-72. Using Chens data, the cooling trend is reduced 0,08K when expanding the area from NH extra-tropics to full NH.
But the Hansen data looses 0,25 K of the cooling trend when expanding from NH extra-tropics to full NH:
fig 33.
So according to Chen, we are still missing an explanation of how Hansen can loose 0,25K of the decline trend when only approximately 0,08K can be explained by the bigger area of the full NH. We are missing 70% of the explanation for the missing cooling trend 1949-72 in the 1986 version of Hansen data.
We cant tell exactly what Hansen did to the temperature decline and why in 1986-1987, but take a good long look at the graphic from Hansen Lebedeff 1987below, it illustrates the temperature development from 1940-65:
fig 34.
At first glance it would appear that most areas has warmed slightly in the very period of temperature decline after 1940??
Note that Hansen has chosen to make the yellowish colours start already at – 0,5 K… In my opinion, you cannot possibly chose yellowish colours starting at – 0,5 K not knowing that this will affect the overall message in the graphic. What ever the purpose was, this little trick does hide the temperature decline after 1940 to some degree.
This brings us to a brief look at the NASA´s web site NH temperatures today:
http://data.giss.nasa.gov/gistemp/graphs/.
After the Hansen Lebedeff 87 publication, I see no significant changes of the GISS NH Land temperature for the temperature data up to 1980. However on the NASA website today, Northern Hemisphere temperatures appears like this:
fig 35.
On the NH temperatures from NASA today, the decline after the 1940´ies appears reduced again. But! The “Northern Hemisphere” on the NASA site is not the original GISS-product from land stations on continents and islands as described earlier.
sources: GHCN 1880-06/2010 + SST: 1880-11/1981 HadISST1
12/1981-06/2010 Reynolds v2
- Resulting in a reduced appearance decline after 1940 (but later, the SST use more and more is a limit on the warming trends - we will get back to this). In pother words, the rather chaotic problems and adjustments of SST mentioned under "Folland 84.."has been incorporated in the GISS NH illustrations today.
Hansen 81 compared with the “NH ensemble” click to enlarge:
fig 36.
3.5) Jones 82
Phil Jones, 1982:
“There is, as yet, no statistical reason to associate recent warming with atmospheric CO2 increases”.
With this fresh opening, let’s look at some open minded science from Jones 82. As for Hansen 81, it is often said that Jones 1982 is only NH-land data, but again, take a closer look at the coverage area below – Jones 1982 is also covering some ocean areas, especially in the Atlantic. Full circles indicate data from the 1950´ies (– which is just what I focus on!):
Fig 37.
(In fact, as Jones writes, it’s only the grids with an “x” that are totally without grid data even for Jones “Maximum grid”.)
So, as I read Jones 1982, he’s trying to produce the best possible NH temperature data set – and thus obviously to some extend including not only land but also some ocean areas. Jones 1982 temperatures NH:
fig 38.
At first glance, we just see the well known peak around 1938 and the decline until around 1972. (In the present writing, I use the extention of Jones 1982 with data to 1983 from: http://www.icsu-scope.org/downloadpubs/scope35/chapter15.html )
Heres Jones words in 1986: "The results shows little trend in the nineteenth century, marked warming to 1940, relatively steady conditions to the mid-1970s and a subsequent rapid warming. The warmest 3 years have all occured in the 1980s."
So the “perplexing” cooling after 1940 is now “relatively steady conditions”. How could Jones over just 4 years totally change his viewpoint? Here we see the Jones 86 NH temperatures:
Fig 39.
Like Hansen/Lebedef 87, Jones also changed data in 1985-86 leading to a reduction of the 1938-72 temperature decline. Hansen 87 showed a smaller temperature decline after 1940 still using land data, while Jones reduced the temperature decline after 1940 when introducing ocean data.
Like we did for Hansen 1981, let’s compare with Chen 1982, who listed all data for us to use for any combination of the large areas. When comparing land+sea with sea data, the differences between Jones 1982 and 1986 should not be as big as for Chen 1982 data sets because, Jones 1982 includes large sea areas as well:
Fig 40.
So for 1949-72, when Chen adds the whole NH sea area to his data, this reduces his NH temperatures with a tiny 0,02 - 0,03 K .
But when Jones goes from land + partly sea in 1982 to land + sea in 1986, the temperature decline is reduced 5 times as much: 0,14 K. So in this light, it seems that sea area in fact only explains a fraction of why Jones decline 1949-72 was reduced 0,14K.
Jones 86 uses SST (for the first time in the world) and thus he has included the SST data from the period with high uncertainties around 1940 as described under “Folland 84”. Only later, other scientists started to use SST and thus reduce the post-1940 temperature decline this way.
Like Hansen temperature data, the Jones / CRU decline 1930´ies – 1970´ies has been reduced further in later data.
fig 41.
CRUTEM3 temperatures is the land temperatures while Brohan/Hadcrut3v is land + SST. So SST or not, temperature data was changed.
fig 42. Click for further examples.
Jones 82 compared with the “NH ensemble” click to enlarge:
fig 43.
3.6) Vinnikov1980
These data, “The Russian Surface temperature set”, is originally based on the work of Budyko 69, and later updated by Vinnikov in 1980. It was the first continuous temperature series from 1891 to 1980 for the Northern hemisphere, so, Vinnikov actually preceded Hansen 81 and Jones 81-82 long NH series. Vinnikov 1980 has a minor problem since its southern limit is 17,5N and not 0N – so a fraction of NH is missing.
Data was made in 5x10 grids, the number of stations in 1940 was around 1000 and in 1960 2000 (from the “Sharova maps” 1960-67” used by Bodyko). According to Alan Robock, these maps where analysed very carefully, and they where meant to create a basis for being able to predict the weather – a high priority, obviously.
Just like Jones, Hansen etc. the records where treated with respect to station locations, breaks in records and other errors in data.
Allan Robock made a description of “The Russian Surface temperature set” in 1982.
fig44a
The Russian data set today (2005) - now with a full NH 0-90N coverage - shows almost the full 1940-75 decline it did in 1980. So, while Jones, Hansen and all important players on the “temperature data market” today claims that they have found errors of all kinds, Vinnikov and colleagues, has not found reason to change temperature estimates much back in time:
Fig 44b
But.. buttom line is: The Vinnikov data is NOT in compliance with Jones and Hansen data after 1984, the corrections for the decline 1938-72 made for Hansen and Jones data is largely still not agreed with by the Russian team. Just a very limitied reduction of the post-1940 decline in the 2005 data – and remember, the 1980 Vinnikov was actually 17,5N-90N while Vinnikov 2005 is full NH.
Vinnikov 80 compared with the “NH ensemble” click to enlarge:
fig 45.
3.7) Yamamoto 1975
The Japanese scientist operates with a good coverage of NH and in fact also SH:
fig 46.
fig 47.
According to Yamamoto, also on global scale the Earth has witnessed quite a decline1958-65, but the data of SH appears slightly too short (?) to say this with confidence.The Yamamoto paper is in Japanese, so I will not give more details for now.
Yamamoto 78 compared with the “NH ensemble” click to enlarge:
fig 48.
Last changed: 16th July, 2010 at 10:08:00
BackComments
| Murder | By Unknown on 4th January, 2011 at 19:37:14 |
| This is not data massaging, it's outright massacre. | |
Add Comment