Frank Lansner sends word of this paper in progress, relevant to the current discussions on the U.S. surface temperature record, and writes:
In the attached article of mine I make an estimate of Hadcrut’s adjustments to us temperature data over the years.
They do not really resemble NOAA´s adjustments:
– As if Hadcru finds reason to step-change in 1946 while NOAA don’t. And NOAAs (mostly TOBS) issues are supposed to take place around 1970-90, but this is not confirmed by Hadcrut. – Frank
Here’s the full paper in progress:
USA TEMPERATURE DEVELOPMENT 1880-2010 FROM UNADJUSTED GHCN V2 DATA
By Frank Lansner, Civil engineer Chemistry
Software engineer (SAP), Novo Nordic IT, Bagsvaerd, Denmark
This paper is part of the RUTI project, described further: http://hidethedecline.eu/pages/ruti.php
ABSTRACT
Temperature records worldwide are used to estimate a warming signal due to increase of CO2 and are thus key parameters when deciding an appropriate climate policy. Despite the fact that the contiguous USA has the best availability of temperature data in the world, there is a large difference between recent [2] and earlier [1] published temperature data from GISS. This raises questions about the robustness of data.
The USA temperature trend calculated from unadjusted Global Historical Climate Network (GHCN) data shows the 1930s as the warmest decade, around 0,2 K warmer than 2000-2009. The calculated temperature trend is based on data from 826 stations, and is virtually identical to that of Hansen et al. 1999 [1].
The calculated temperature trend 1930-2010 is around 0,5 K colder than the updated 2011 GISS US temperature trend [2], and around 0,4 K Colder than temperature trend calculated from Hadcrut stations in USA.
The increased warming trend found in the Hadcrut data appears to originate more due to the choice of temperature stations used than due to adjustments of station data.
Keywords: temperature adjustments, temperature stations, anthropogenic climate impacts
INTRODUCTION
The role of CO2 as a warming agent in the atmosphere is the suspected driver of basically all unfortunate developments in the climate as described in the recent IPCC paper: “Special Report on Managing the Risks of Extreme Events and Disasters to Advance Climate Change Adaptation”.
The “extreme events and disasters” in connection with changes in ice cover, sea levels, precipitation and frequency of large tropical storms and more mentioned in the above IPCC paper [9-10] is based on the magnitude of the recent warming supposed to be caused mostly by CO2 in the Atmosphere. Thus, the “extreme events and disasters” mentioned in the above IPCC writing is dependent on the correctness of temperature data.
This is why credible analyses of temperature data are vital for any estimation of the potential hazards of CO2 and thus the policy to be carried out.
In this analysis, estimates of USA temperature trend 1880-2010 were made and the results compared to USA data published by NASA in 1999 [1] and 2008 [2]. Results were also compared to USA temperature stations used by Hadcrut.
For the Hadcrut data set, other issues are investigated. For example, the selection of temperature stations in the Hadcrut subset of those available suggests a bias towards stations from large urban areas. The temperature data from stations used by Hadcrut was compared to the bulk of the stations from the same data source, unadjusted GHCN. This was done to evaluate how much of the warming trend in Hadcrut data originated simply from the choice of temperature stations done by Hadcrut. The temperature trends for Hadcrut stations should resemble the general temperature trend using unadjusted GHCN v2 data.
To learn further about the increased heat trend in Hadcrut data, data from the unadjusted GHCN dataset – as used by Hadcrut were compared with the of temperature data in Hadcrut adjusted version. This way an estimate of the actual adjustments was made.
The nature of the adjustments in Hadcrut data was then compared to the adjustments for the USA temperature data described by NCDC for the USHCN dataset [3-4]. If adjustments used by Hadcrut and NCDC to US station data are not similar, this will not support the robustness in the adjustments of US temperature data. USHCN is a collection of temperature stations from the USA with longer periods of data available. USHCN was developed by Thomas Karl et al. 1990 [4].
Due to the impact of geography on temperature trends, a method “two-zones-averaging method” was used to improve the estimate of what land area is best covered by what stations. For example coastal temperature stations often show a significantly different trend than nearby non-coastal stations [5]. This information was used in determining sub-regional classifications of temperature stations. Furthermore, to avoid the introduction of possible non-climatic factors, data from large urban areas were not included. Additionally, stations with lack of data before 1940 or after 1980 were not used. In total, 826 stations were used from unadjusted GHCN.
METHODS
USA temperature data used were unadjusted monthly GHCN v2 and it was obtained from Appinsys.com. At the time of data retrieval, GHCN data was latest updated March 2011, years 1880-2010 are used. The Hadcrut data were obtained also from Appinsys.com, latest updated December 2009. Years 1880-2008 are used. All graphs shown have a base period of 1961-90 unless otherwise noted.
Temperature trend was calculated for 5×5 grids, and grid results where averaged by weighting for the area of US land territory for each grid using a “two-zone-averaging” method.
“Two-zone-averaging method”:
Figure 1. For each grid, the area of two zones where estimated: One with temperature stations showing colder temperatures 1998-2008 than 1930-1940 (blue areas on fig 1), and another zone showing colder temperatures 1998-2008 than 1930-40 (red areas on fig 1). (Black stars shown on fig 1 are temperature stations included in the Hadcrut dataset. Temperature trends on the graph are 5 year averaged).
On the graph, The legend text “Colder 13 st 48%” explained:
1) temperature trend within the areas with colder trend
2) 13 such temperature stations found for the 5×5 grid
3) the area covered by cold trended temperature stations is estimated to cover 48% of the US land area within the 5×5 grid.
Temperature stations are unevenly distributed and thus a raw average of all temperature stations belonging to one grid may be misleading. Using 2 zones reduces this source of error.
Figure 2. As discussed in “RUTI: Coastal temperature stations” [5-6], land areas zone 1 and 3 are most likely to be affected by marine air temperatures. On the other hand low land non-coastal areas, zone 2, and higher eleva
ted areas not facing marine air, zone 4, both seem to share a non-coastal temperature trend.
Figure 3. Example from NE US from “RUTI: Coastal temperature stations [5]. Although figure 2 shows 4 zones, often the zones influenced by marine air (zone 1 and 3) show similar temperature trends, just as non-coastal zones (2 and 4) show similar trends. Thus, it makes sense to work with just 2 zones: Cold and warm trended, shown as blue and red respectively hereafter.
There are other ways that geography locally can affect temperature – changing snow/ice cover, rivers etc. – and such locations will be part of either the cold or warm trended zones too. Graphs on fig 3 are 5 year averages.
Also, a Hadcrut dataset calculated on basis of 87 temperature stations located in the USA was made. (the Station Key west located far from the mainland was not used). Hadcrut use numerous strongly urban located stations, but all was included when calculating data from the Hadcrut stations. Dataset for Hadcrut stations where calculated for each 5×5 grid, and each grid was weighted with respect to area of US land territory – just like the RUTI USA data.
Finally a dataset of “Unadjusted GHCN dataset for stations used by Hadcrut was calculated just like the Hadcrut dataset. Of the 87 Hadcrut stations used for USA, 77 has data publicly available from Unadjusted GHCN, and thus it is possible to analyse temperature trend of these unadjusted stations used by Hadcrut.
RESULTS
Figure 4. The distribution of GHCN stations in the USA . Blue: 1998-2008 colder than 1930-1940. Red: 1998-2008 colder than 1930-40 based on unadjusted GHCN data. The coastal and mountain locations of warm trended stations are visible.
In many areas, the cold and warm trended stations can co-exist within short distances, and thus in such areas adjustments of temperature stations based on results from other nearby temperature stations should normally be avoided. This also questions the still lower number of temperature stations used by GHCN, Hadcrut and GISS, especially after 1990. This approach of low number of temperature stations lowers the accuracy of the resulting overall temperature trend.
In the Eastern USA, one can spot a number of warm trended temperature stations located on the Atlantic side of the Appalachian Highlands, but for the Rocky Mountains the terrain is more nuanced and so is the distribution of warm trended and cold trended stations. For more details on this, see [6].
Figure 5. Average temperature trends of the 2 zones shown (red and blue graphs) for each grid. The unit on the graphs is shown in fig 1. Data is 5 year averaged. For some grids, the difference between temperature trend of the 2 zones approaches 1 K when comparing recent decade with temperatures before 1940.
Figure 6. On this illustration, the cold and warm averages has been calculated into one temperature trend for each grid weighted with respect to estimated area coverage by warm and cold trended areas. The red/blue numbers shown on the graphs are the temperature average 1998-2008 compared to 1930-40.
Figure 7. Temperature trends for all grids (48 contiguous states of USA) where then combined, producing a single temperature anomaly dataset here after referred to as “RUTI USA Contiguous 48 states”. This resulting graph shows a remarkable agreement with that published by Hansen 1999 [1]:
Figure 8. The similarity RUTI vs. Hansen 1999 is obvious. Visible differences in 5 year averaged temperatures between Hansen et al. 1999 and RUTI USA, occurs mostly before 1895.
Figure 9. RUTI USA, warmest decade found in unadjusted GHCN data for USA was the 1930´ies, around 0,2 K warmer than the decade 2000-09. Warmest year found was 1934, followed by 1921 and 2006.
Figure 10. Comparison of US temperature datasets: Temperature trends from Hansen et al. [1] and RUTI USA datasets are shown, and now in addition, the resulting adjusted temperature dataset from Hadcrut US stations (red) and adjusted temperature dataset from GISS updated US temperatures [2] (green) are shown. On this graphic, base period of datasets where set equal with RUTI USA for the period 1930-39.
The GISS updated US dataset shows the last 10-15 years to be around 0,5 K warmer than the RUTI USA dataset (and thus the original Hansen et al. 1999 [1] dataset ). Hadcrut stations shows recent decade around 0,4 K warmer than RUTI USA (and thus Hansen et al. 1999 [1] ).
This difference is shown again in fig 11 now with base period 1895-1905 where it amounts to around 0,42 K:
Figure 11. Most of the temperature stations used by Hadcrut have unadjusted temperature data publicly available from GHCN (black graph)
The similarity between the red and black graphs (stations chosen by Hadcrut, adjusted vs. unadjusted) suggests, that adjustments to Hadcrut US temperature data have a relatively small impact: 0,12 K of warming of trend.
This indicates, that most of the difference between the 826 stations from Unadjusted GHCN and then the adjusted Hadcrut temperature trends, approx. 0,42K, is due to the Hadcrut choice of temperature stations, approx. 0,3K.
From unadjusted GHCN, the 77 stations used by Hadcrut has around 0,3K more heat trend 1900-2010 than th
e temperature trend calculated in this writing using 826 stations.
Figure 12. Locations and populations in thousands for the Hadcrut US temperature stations [€€€€]. The raw average population around Hadcrut temperature stations is 1.3 mio people. The typical Hadcrut station is placed in a town with more than hundred thousand inhabitants.
Figure 13. The trend of Hadcrut adjustments of the US temperature data can be estimated by comparing unadjusted GHCN temperature data with Hadcrut adjusted temperature data for the US temperature stations chosen by Hadcrut (black graph fig 13)
The trend of Hadcrut adjustments shows a larger step change around 1946, of 0.12 K.
The larger step change in 1946 for Hadcrut stations was not expected, since specifically for the USA, adjustments announced by NCDC for USHCN temperature trends [3-4] shows a gradual adjustment with the bulk of adjustments done for mostly the period 1970-90, not at all a step change mostly around 1946 (see red graph fig 13).
It seems that NCDC does not agree with Hadcrut on when adjustments should take place. And vice versa: The Hadcrut adjustments for 1970-1990 do not support the adjustments done by NCDC.
CONCLUSION
Calculation of USA temperature trend (“RUTI USA”) from Unadjusted GHCN confirmed temperature trend published by Hansen et al. 1999 [1]. Fig 10 shows that the updated GISS temperature trend for USA [2] since the 1930´ies has around 0.5 K more heat trend after recent adjustments. Considering that the USA temperature data has the highest quality in the world, the large adjustments of 0.5 K may question the quality of temperature data world wide.
The extra heat found for Hadcrut US stations in comparison with both Hansen et al. 1999 [1] and the results from the present paper amounts to approx. 0.42 K since the 1930´ies. Far most of the Hadcrut temperature stations (77 out of 87 stations) are available unadjusted from GHCN, and thus it was possible to estimate that only approx. 0.12 K of the extra roughly 0.42 K heat trend in Hadcrut US temperature trend originates from adjustments. The remaining roughly 0.3 K of extra heat trend for Hadcrut US temperature stations seems to originate from the choice of temperature stations from GHCN included in the Hadcrut USA subset.
Thus the unadjusted temperature stations used by Hadcrut generally has more heat trend than neighbouring unadjusted temperature stations, but even so, Hadcrut add heat trend when they adjust their temperature data.
In addition, the trend of adjustments done to Hadcrut US temperature stations shows one larger step change in 1946, other wise flat trend. This does not at all resemble the NCDC adjustments to USHCN temperature data as one would expect (see fig 13). This mismatch in the nature of adjustments between Hadcrut and NCDC USA temperature data lowers confidence to temperature adjustments done by both.
Hadcrut temperature stations are often located in medium to large urban areas. Hadcrut only use 12 rural temperature stations, but there are many more than 12 useful rural temperature stations in the USA available. A bias towards using more warm trended temperature stations was also questioned by the Russian IEA [7-8] for the Russian land temperatures estimated by Hadcrut.
Does temperature data turn out to be credible and solid as basis for policy making and solid input to models that predicts the future development of temperatures and consequences?
USA temperature data has the highest quality in the world, so if there is an area where temperature data should be robust it is USA. None the less, for each stone turned, hardly any support of robust modern versions of adjusted temperature data turns up.
The most robust result was found when analysing USA temperature data is the good match between RUTI (Unadjusted GHCN) and Hansen 99. The conflict between RUTI (Unadjusted GHCN) and Hansen 99 on one side and Hadcrut + Hansen 2008 temperature trends on the other side speaks against credibility of temperature data to be used for policy making.
ACKNOWLEDGEMENTS
Thank you: Verity Jones, Joanne Nova, Alan Cheetham and Oguzhan Tandogac.
REFERENCES
[1] Hansen et al. 1999: “GISS analysis of surface temperature change “
http://pubs.giss.nasa.gov/docs/1999/1999_Hansen_etal.pdf
[2] Data for GISS USA are updated online at http://data.giss.nasa.gov/gistemp/graphs/Fig.D.txt
[3] USCHN online description: http://www.ncdc.noaa.gov/oa/climate/research/ushcn/ushcn.html
[4] Karl, T.R., C.N. Williams, Jr., F.T. Quinlan, and T.A. Boden, 1990: United States Historical Climatology Network (HCN) Serial Temperature and Precipitation Data, Environmental Science Division, Publication No. 3404, Carbon Dioxide Information and Analysis Center, Oak Ridge National Laboratory, Oak Ridge, TN, 389 pp.
[5] RUTI: Coastal temperature stations, online : http://hidethedecline.eu/pages/ruti/coastal-temperature-stations.php or
[6] RUTI: USA, online http://hidethedecline.eu/pages/ruti/north-america/usa-part-1.php
[7] Russian complaint orig pdf in Russian: http://www.iea.ru/article/kioto_order/15.12.2009.pdf
[8] RiaNovosti: http://en.rian.ru/papers/20091216/157260660.html
[9] IPCC Special Report on Managing the Risks of Extreme Events and Disasters to Advance Climate Change Adaptation, http://ipcc-wg2.gov/SREX/
[10] IPCC Special Report on Managing the Risks of Extreme Events and Disasters to Advance Climate Change Adaptation, PDF overview: http://www.ipcc.ch/pdf/press/ipcc_leaflets_2010/ipcc_srex_leaflet.pdf
“Temperature records worldwide are used to estimate a warming signal due to increase of CO2…”
On what basis is this statement made? As far as I can see the temperature records can estimate warming, not a ‘warming signal’. Further, I do not see how they provide any information that would allow you to ascribe a cause to said warming. Hopefully he’s not following the ‘well, temps are increasing, and CO2 is increasing, therefore the temp increase is caused by the CO2 increase’ meme.
Will William Roberts be commenting on this?
I long for the day that somebody recognizes that a 5×5 grid is a non-uniform coverage of a sphere and somebody in climate science learns how to tesselate a sphere properly and scalably and above all, uniformly.
If climate science were open sourced, this is one of the very first things I’d work on — a set of routines to take the Earth as a sphere, tesselate it with a control variable N that dictates the number of tiles and hence granularity of the tesselation, a simple match routine that takes latitude/longitude and returns the index of the tile in which it falls, an integration routine (sum over tiles), a gradient routine (loop over NN tiles, fit e.g. spline curve in the different directions), a tile-vector-statistics routine, all laid out to use “effortlessly” so that a list of e.g. station data associated with lat/long could instantly be transformed into a uniform tesselation at any granularity desired and supported by the tesselation algorithm.
Then perhaps, maybe, we could put the damnable error of treating a sphere as if it is locally flat on continent-sized regions behind us.
But no, even professional earth scientists — people who teach others that the Earth is not flat — still use a mercator projection to do areal statistics. Welcome to Antarctica — a continent 25000 miles wide and 5000 miles high, almost as large as the rest of the land area of the Earth put together — in terms of its 5×5 cell count.
Sigh.
rgb
Frank Lansner, your website includes the following description of your RUTI project:
”RUTI ”Rural Unadjusted Temperature Index”. RUTI is not all rural nor all unadjusted. However, RUTI is a temperature index aiming to use still more rural data (less use of city and airport data), still more unadjusted data when available and reasonable.”
How is your RUTI dataset different than the GHCN/CAMS dataset?
Caption on Fig 1 defines red and blue stations identically. One “colder” should be “warmer”? Same problem occurs later.
rgbatduke says:
August 1, 2012 at 12:06 pm
Map projection problems are not pretty, but there are solutions to most issues involved in mapping spherical surfaces to flat. Cartography after all has been around a long while. There are so-called equal area projections that do preserve areas but not shape or distance which should achieve what you are discussing. Presumably if an equal-area projection like a Mollwieide equal area projection was used, maps might look more reasonable. The problem is that “area preserving” projections don’t preserve distance and “distance preserving ” projections don’t preserve area. As I gather it, “adjustments” and homogenization are based on nearest-neighbor stations. You could simply use a distance preserving projection to handle the nearest-neighbor aspects of the “homogenization” and than re-project to an equal area scheme to get proper land-area estimates. Most GIS systems are quite capable of this. and GRASS is both free and opensource.
Is Frank Lansner looking for or welcoming web based peer-review? If so, maybe that point needs an update at the bottom of the root post.
@rgbatduke,
Assuming anyone on the climate team actually attempted this, it still wouldn’t come out right because they would ignore the reality that the earth is not actually a true sphere.
@rgbatduke: when I first read up on Globalwarmmongering, I realised that many of the people involved were, by the standards of much science, dim.
Bill Marsh says:
August 1, 2012 at 11:45 am
“Temperature records worldwide are used to estimate a warming signal due to increase of CO2 and are thus key parameters when deciding an appropriate climate policy.”
On what basis is this statement made? As far as I can see the temperature records can estimate warming, not a ‘warming signal’. Further, I do not see how they provide any information that would allow you to ascribe a cause to said warming. Hopefully he’s not following the ‘well, temps are increasing, and CO2 is increasing, therefore the temp increase is caused by the CO2 increase’ meme.
===========================================================================
I think that line is intended to say how temperature records have been used by some. I don’t think he’s saying he agrees with that. (But I may have misunderstood.)
The past and the present is made UNFCCC conform?
error on Fig 4 ? both red and blue stations are “1998-2008 colder than 1930-40”
or is it that blue is ref adjusted data, and red is as stated unadjusted data and just needs making a little clearer
I think pretty much all attempts to make adjustments irreparably damages the data, and makes drawing a definitive trend impossible. Measurements, are the data. Not some homogenized mish mash of made up values.
Frank,
I hope you get some valuable and constructive feedback from this posting. I did think about your RUTI analysis when Anthony published his draft paper on Sunday,
One of the take home message for me in all this is that the people (you, Anthony et al., Christy, Roger Pielke Sr, McNider, Roy Spencer) looking at these issues (siting – macro-, micro-, and changes) are tending to look at them from different angles and each getting a very different result from the orthodox NCDC/CRU/GISS result.
Some time ago I wrote a post entitled “Climate Parallax” http://diggingintheclay.wordpress.com/2011/03/06/climate-parallax/ in which I said:
and
I stand by that.
rgbatduke says:
August 1, 2012 at 12:06 pm
“I long for the day that somebody recognizes that a 5×5 grid is a non-uniform coverage of a sphere and somebody in climate science learns how to tesselate a sphere properly and scalably and above all, uniformly.”
Areas of 5×5 grids are obviously calculated as should be.
I have all relevant data, calculations, US maps and 5×5 grid maps showing the cold vs. warm zones as I estimate them. Im not sure in what form and to what extend these informations shold be available, im very open for suggestions.
This method has the advantage, that if someone suggests i missed a relevant station, then further stations will normally be located within the warm trended or cold trended areas respectively. And since i weigt the warm and cold trends with respect to size of these areas, even stations not used are not likely to change things. K.R. Frank
Jono1066, you are correct. the best peer review is a blog discussion like this!
Under fig 4 it should be: ” Red: 1998-2008 warmer than 1930-40 ”
To ALL:
Its very important for me if all notices what Fig 11 tells:
The black curve shows the stations chosen by Hadcrut, but RAW values, that is unadjusted GHCN V2.
So, both the black and the blue graphs are same data source: unadjusted GHCN.
The point: The unadjusted GHCN US trend for the stations also used by Hadcrut has a far warmer trend than the bulk of unadjusted GHCN US stations.
So, First Hadcrut choses more warm trended stations to work with and then they add a little extra heat trend (resulting in the red graph).
Hi Bob, I think you put a very relevant question:
“How is your RUTI dataset different than the GHCN/CAMS dataset?”
As shown Hadcrut has a preference for urban stations, and they also tend to warm adjust as shown here for USA.
GHCN on the other hand has the unadjusted V2 set which is often (not always) identical to original temperature series, GHCN V2 unadjusted is thus one of the best data sources we have, and therefore RUTI is to a large extend building on Unadjusted GHCN. (Also Nordklim, NACD and several other sources are used including original papers I ask peoble to send me.).
When it comes to USA, RUTI is only based on GHCN unadjusted as source.
So whats the difference?I have to chop my answer up since my browser has a problem with this text window.. to be continued.
Continued. GHCN unadjusted has a HUGE issue that I try to deal with in RUTI. Data does seem mostly unadjusted, but very often datasets are cut, limited, periods are missing etcetc.
Notice this example, turkey, an extreme example of GHCN issues just to show a point:
http://hidethedecline.eu/pages/ruti/asia/turkey.php
In this case, yes, data are probably correct, but only the very largest citiest are used (and thus public available). This to show that GHCNs issue is mostly data-selectiveness rather than ajdustments.
There are all kinds of selectiveness in GHCN, the above promoted Urban trends but many other types of GHCN selectiveness can be found, often geographic “cherry-picking”.
Coast lines and hills/mountains that faces oceans quite often has a trend 1930-2010 warmer than non-coastal areas, and lower areas that then do not face ocean air.
Not rarely, coasts and coast-facinf mountains represents a not too large fraction of the land area, and there fore should not count more than the areas they represent.
But identifying areas of same heat trend and then using the size of this area then suddenly it does not matter how many stations are placed where. Its the size of the areas of same temperature trend that counts, not the number of stations that GHCN or anyone else wants to use.
Also, by identifying areas of similar trends, it also becomes more reliable to stich temperature series within a region of similar trend.
I will give you a sad example of how GHCN deselcting can omit cold trended data, East central China:
http://hidethedecline.eu/media/ARUTI/Asia/China/fig22.jpg
Some stations that on Google maps appear rural seems to show a cold trend. but notice how ALL datasets has been chopped up in early years and thus only shows a few warm years each. But the trend is clear.
However, if you do a “blind” mathematical averaging of all East central china GHCN data, then the urban series dominates completely:
http://hidethedecline.eu/media/ARUTI/Asia/China/fig21.jpg
And when GHCN shows only 5-6 warm early years for each rural dataset, then when averaging, the rural signal is thinned to almost nothing. And Best thinks they have something useful….
Here is an axample from Australia where the importance of limiting temperature trends STRICTLY to the areas relevant:
http://hidethedecline.eu/media/ARUTI/Australia/fig3.jpg
Notice that several coasts has their own thin coastal areas for temperature trends. This is important if you want a result where coastal warm trends do not dominate wrongly.
taken from
http://hidethedecline.eu/pages/ruti/australia.php
Using a dataset like GHCN “just like that” with math gives you exactly the result you where supposed to get. And peoples (some on the BEST crew) actually thinks they did something right…
Hi Verity, you good woman 🙂 allways a plessure to hear from you!
I will check out your link, very interesting. And then I have to tell that im in the middle of an approx 4 month long project that are in many ways unique. I will present a long row of longer raw temperature datasets no one has known about before and more. hehe…
Nice work, Frank. I always like to read your posts.
Small typo: Past tense plural of ‘is’ is ‘were,’ not ‘where.’
” the contiguous USA has the best availability of temperature data in the world”
Who says?
(This is a genuine question.)
typo in the article “Figure 1. For each grid, the area of two zones where estimated:” should be “Figure 1. For each grid, the area of two zones were estimated:”
I can’t trust my eyes ! After trashing Muller (and really everybody else) for dashing forward with non-peer-reviewed papers, here we now have one that hasn’t even been finished. How come the change of mind?
Did you mean to say:
“The black curve shows NOT the stations chosen by Hadcrut, but RAW values,….”
TO rogerknights regarding the fig 11:
I wrote:
Its very important for me if all notices what Fig 11 tells:
The black curve shows the stations chosen by Hadcrut, but RAW values, that is unadjusted GHCN V2.
An you ask, i I meant: “The black curve shows NOT the stations chosen by Hadcrut, but RAW values,….”
No, the black curve IS stations used by hadrcrut, but I have shown how UNADJUSTED GHCN looks for these stations chosen by hadcrut.
This way we can compare the bulk of GHCN Uadj. (Blue) with the GHCN uadj for stations cherry picked by hadcrut (black).
The result is, that even RAW data from GHCN shows warmer trend when only looking at the stations used by hadcrut.
Is this clear (i hope..)
And finally the RED graph shows how Hadcruts versions of the hadcrut stations appears.
So now we can see, that most extra trend in Hadcrut compared with raw GHCN is due to cherry picking more warm trended stations, not adjustments.
This is important because many tries to defend adjustments. But can they defend cherry picking of warm trended stations?
>>
rgbatduke says:
August 1, 2012 at 12:06 pm
I long for the day that somebody recognizes that a 5×5 grid is a non-uniform coverage of a sphere and somebody in climate science learns how to tesselate a sphere properly and scalably and above all, uniformly.
<<
For all the complaints lodged against GCM modelers, they have already figured this part out. They use quasi-homogeneous resolution grid models. Two examples are icosahedral grid and conformal cubic grid. I happen to like the icosahedral grid, but then you have deal with small spherical triangles. I guess you could combine two of them to form a spherical parallelogram shape–something with four sides.
Jim
It seems to me that in the absence of a decent universal, statistically unadjusted, statistically unmolested and free of (TOBS) raw temperature data set then this paper will suffer the same fate as Anthony’s. As an observer, this fate is what I call the venomous Mosher effect (vMe).
After more than a lifetime or two of using crap raw temperature data with embedded TOB’s acting as a frivolous fundamentalist sideshow, then in my view it’s time to throw out all the raw temperature data sets and simply start over again, this time using universally accepted, uniform siting’s and uniform digital recording equipment.
Frank;
No, your Germanic English often not is clear. Speaker natively co-author get should you.
Duster says: August 1, 2012 at 12:51 pm
rgbatduke says: August 1, 2012 at 12:06 pm
“There are so-called equal area projections that do preserve areas but not shape or distance which should achieve what you are discussing.”
You don’t need an equal area projection to get an equal area tessellation. This map shows an example of what TempLS now uses.
One can imagine that coastal statons have less variable or even lower temperatures than comparable sites inland. However, I cannot understnd the meaning of the observation that inland stations have steeper trends than coastal stations. The reason can easily be seen if you extrapolate forwards or backwards. Before too long, you reach unlikely temperatures, where the divergence between coastal and inland is impossibly large. It is possible that the trends can differ if there is a transient event in the study period, but they must become more parallel when the transient effect ends.
What would be your explanation for the transient effect that produces trends that are different for coastal and inland sites?
Bob T,
Here are shorter version of my answer:
GHCN may be cheery-limiting data creating a bias in data.
The effects of cherry-limiting available years in data can be limited by resolving the situation area for are around the globe.
This has a huge impact on the resulting trend, and this is a huge bulky tough work, but you cant just skip this step and jump to do math on a dataset like GHCN.
So RUTI use GHCN but analyses bit for bit to encounter typically “missing heat” in the past.
The more i study “BEST” the poorer it seems. For example theyhave soooooo many datasets..
But for Denmark for example they take the most UHI poluted station – Copenhagen – from 5 different sources, and then this is supposed to help the quality….
If BEST actually was aiming for the truth, they would have used 50 DIFFERENT datasets from Denmark. BEST in one big show with an agenda.
hi Geoff Sherrington, you ask:
“What would be your explanation for the transient effect that produces trends that are different for coastal and inland sites?”
First, takt a look just to see how beautiful cold and warm trended stations in Eastern USA are distributed: Warmer trend at the coast and the inland mountain sides facing Atlantic winds. kind of a “mountain coast line” in the apalachees:
http://hidethedecline.eu/media/ARUTI/USA/fig13.jpg
And the why this distribution?
There are 3 main reasons for different temperature profiles coastal vs. non-coastal:
1) water acts as a temperature buffer. This means that a coastal station may show both the 1940 wamr and the recent warm peak less.
This is a problem when GHCN and others uses a coastal stations up to for example 1990, and then continues data from an inland or even urban station with much larger heat peaks.
2) Oceans take much longer time to warm up. Thus, ocean warming in later years may to some degree be a delayed reposnse from the oceans.
Thus, even though warming effect on Earth is constant, then a coastal station can still show (delayed) ocean warming.
Therefore non-coastal stations are to be preferred if you want to examine present heating effect on Earth.
3) Winds: In Denmark and hollans for example, warmer preiods normally comes with more wind from West. Thus the Atlantic coastal stations have more influence from oceans in the periods of warming.
This extra ocean dominans in warm periods keeps warm peaks even further down.
There is a station on the danish island “Fyn” i am examining at the moment. It turns out that the 8 o´clock trend in the last 100 years show strong warming while the 14.00 and 21.00 series are pretty constant.
Air masses from ocean can over night travel to central Fyn and thus affect the morning temperatures. While the sun and cluds etc fast dominates when the day starts.
K.R. Frank
Frank Lansner: In your August 1, 2012 at 4:25 pm rely to my question How is your RUTI dataset different than the GHCN/CAMS dataset? you did not mention GHCN/CAMS.
In other words, you did not answer my question. I’m not playing this game with you again, Frank.
Goodbye
Meanwhile in New Zealand
http://climaterealists.com/index.php?id=10030
“In other words, the juries will be directed to rule that climatologists refused to disclose the evidence because to do so would prove they intentionally falsified the climate records to get a predetermined outcome.”
Mwhite: Stunning!!!!!!!!!!!!
Niwa MUST know that a refuse of access to their “science” (that created heat trend out of a flat trend) will be very dangerous for them.
This means that their “science” is so weak that they prefer legal problems.
From the article;
”
NIWA’s decision renders an almighty self-inflicted wound to the government agency’s already dire credibility. But worse, the move will be regarded as contempt of court and thus permits the court to grant the plaintiff’s motions for punitive sanctions, including summary judgment. As such, this would bring a swift victory for skeptics with profound legal ramifications around the world. In the sparsely-measured southern hemisphere the New Zealand climate data is critical to claims about a verified global temperature record.
“
NIWA uses a set of 7 devices for the whole of New Zealand. One of which, used to cover most of the north island, is bsed at the airport in Masterton, in the Wairarapa. This thermometer represents NZ north of Wellington and south of Auckland. That’s a large chunk of NZ represented by a single thermometer.
Climatology seems to have taken full advantage of the classic “enough rope to hang themselves”. It’s not the fall that kills you when the hangman’s trap door opens, it’s the sudden stop …
METHODS Figure 1: should be “zone showing WARMER temperatures 1998-2008 than 1930-40 (red areas on fig 1).”
Oliver, thank you for mentioning, another reader also noted this and its correct.
K.R. Frank