All posts tagged Arctic Temperature Reconstruction

Reykjavik Temperature Adjustments – a comparison

Summary

On 20th February, Paul Homewood made some allegations that the temperature adjustments for Reykjavík were not supported by any known reasons. The analysis was somewhat vague. I have looked into the adjustments by both the GHCN v3 and NASA GISS. The major findings, which support Homewood’s view, are:-

  • The GHCN v3 adjustments appear entirely arbitrary. They do not correspond to the frequent temperature relocations. Much of the period from 1901-1965 is cooled by a full one degree centigrade.
  • Even more arbitrary was the adjustments for the period 1939-1942. In years where there was no anomalous spike in the data, a large cool period was created.
  • Also, despite there being complete raw data, the GHCN adjusters decided to dismiss the data from 1926 and 1946.
  • The NASA GISS homogenisation adjustments were much smaller in magnitude, and to some extent partly offset the GHCN adjustments. The greatest warming was of the 1929-51 period.

The combined impact of the adjustments is to change the storyline from the data, suppressing the early twentieth century warming and massively reducing the mid-century cooling. As a result an impression is created that the significant warming since the 1980s is unprecedented.

 

Analysis of the adjustments

There are a number of data sets to consider. There is the raw data available from 1901 to 2011 at NASA GISS. Nick Stokes has confirmed that this is the same raw data issued by the Iceland Met Office, baring a few roundings. The adjustments made by the Iceland Met Office are unfortunately only available from 1948. Quite separate, is the Global Historical Climatology Network dataset (GHCN v3) from the US National Oceanic and Atmospheric Administration (NOAA) I accessed from NASA GISS, along with the GISS’s own homogenised data used to compile the GISTEMP global temperature anomaly.

The impact of the adjustments from the raw data is as follows

The adjustments by the Icelandic Met Office professionals with a detailed knowledge of the instruments and the local conditions, is quite varied from year-to-year and appears to impose no trend in the data. The impact of GCHN is to massively cool the data prior to 1965. Most years are by about a degree, more than the 0.7oC total twentieth century global average surface temperature increase. The pattern of adjustments has long periods of adjustments that are the same. The major reason could be relocations. Trausti Jonsson, Senior Meteorologist with the Iceland Met Office, has looked at the relocations. He has summarized in the graphic below, along with gaps in the data.

I have matched these relocations with the adjustments.

The relocation dates appear to have no impact on the adjustments. If it does affect the data, the wrong data must be used.

Maybe the adjustments reflect the methods of calculation? Trausti Jonsson says:-

I would again like to make the point that there are two distinct types of adjustments:

1. An absolutely necessary recalculation of the mean because of changes in the observing hours or new information regarding the diurnal cycle of the temperature. For Reykjavík this mainly applies to the period before 1924.

2. Adjustments for relocations. In this case these are mainly based on comparative measurements made before the last relocation in 1973 and supported by comparisons with stations in the vicinity. Most of these are really cosmetic (only 0.1 or 0.2 deg C). There is a rather large adjustment during the 1931 to 1945 period (- 0.4 deg C, see my blog on the matter – you should read it again:http://icelandweather.blog.is/blog/icelandweather/entry/1230185/). 
I am not very comfortable with this large adjustment – it is supposed to be constant throughout the year, but it should probably be seasonally dependent. The location of the station was very bad (on a balcony/rooftop).

So maybe there can be some adjustment prior to 1924, but nothing major after. There is also nothing in the this account, or in the more detailed history, that indicates a reason for the reduction in adjustments in 1917-1925, or the massive increase in negative adjustments in the period 1939-1942.

Further, there is nothing in the local conditions that I can see to then justify GISS imposing an artificial early twentieth century warming period. There are two possible non-data reasons. The first is due to software which homogenizes to the global pattern. The second is human intervention. The adjusters at GISS realised the folks at NOAA had been conspicuously over-zealous in their adjustments, so were trying to restore a bit of credibility to the data story.

 

The change in the Reykjavík data story

When we compare graphs of raw data to adjusted data, it is difficult to see the impact of adjustments on the trends. The average temperatures vary widely from year to year, masking the underlying patterns. As a rough indication I have therefore taken the average temperature anomaly per decade. The decades are as in common usage, so the 1970s is from 1970-1979. The first decade is actually 1901-1909, and for the adjusted data there are some years missing. The decade of 2000-2009 had no adjustments. The average temperature of 5.35oC was set to zero, to become the anomaly.

The warmest decade was the last decade of 2000-2009. Further, both the raw data (black) and the GISS Homogenised data (orange) show the 1930s to be the second warmest decade. However, whilst the raw data shows the 1930s to be just 0.05oC cooler than the 2000s, GISS estimates it to be 0.75oC cooler. The coolest decades are also different. The raw data shows the 1980s to be the coolest decade, whilst GISS shows the 1900s and the 1910s to be about 0.40oC cooler. The GHCN adjustments (green) virtually eliminate the mid-century cooling.

But adjustments still need to be made. Trausti Jonsson believes that the data prior to 1924 needs to be adjusted downwards to allow for biases in the time of day when readings were taken. This would bring the 1900s and the 1910s more into line with the 1980s, along with lowering the 1920s. The leap in temperatures from the 1910s to the 1930s becomes very similar to that from 1980s to the 2000s, instead of half the magnitude in the GHCNv3 data and two-thirds the magnitude in the GISS Homogenised data.

The raw data tell us there were two similar-sized fluctuations in temperature since 1900 of 1920s-1940s and from 1980s-2010s. In between there was a period cooling that almost entirely cancelled out the earlier warming period. The massive warming since the 1980s is not exceptional, though there might be some minor human influence if patterns are replicated elsewhere.

The adjusted data reduces the earlier warming period and the subsequent cooling that bottomed out in the 1980s. Using the GISS Homogenised data we get the impression of unprecedented warming closely aligned to the rise in greenhouse gas levels. As there is no reason for the adjustments from relocations, or from changes to the method of calculation, the adjustments would appear to be made to fit reality to the adjuster’s beliefs about the world.

Kevin Marshall

 

by manicbeancounter on 23/02/2015  •  Permalink
Posted in Arctic, Climate Change
Tagged , , ,

Posted by manicbeancounter on 23/02/2015

https://manicbeancounter.com/2015/02/23/reykjavik-temperature-adjustments-a-comparison/

Pages2K Revised Arctic Reconstructions

Climateaudit reports

Kaufman and the PAGES2K Arctic2K group recently published a series of major corrections to their database, some of which directly respond to Climate Audit criticism. The resulting reconstruction has been substantially revised with substantially increased medieval warmth. His correction of the contaminated Igaliku series is unfortunately incomplete and other defects remain.

This post is on comparing the revised reconstruction with other data. In the comments Jean S provides a graph that compares the revised graph in red with the previous version in black. I have added some comparative time periods.

  1. The Maunder minimum of 1645-1715 corresponds to a very cold period in the Arctic. The end of the minimum was associated with a rebound in temperatures.
  2. The Dalton minimum of 1790-1820 corresponds to a period of sharply declining temperatures, with the end of the period being the coldest in 2,000 years. The end of the minimum was associated with a rebound in temperatures.
  3. The early twentieth century shows about 1.1oC of warming from trough to peak in a time period that corresponds to the 1911-1944 trough-to-peak warming of the global temperature series. It is about twice the size of that calculated globally by HADCRUT4 and GISTEMPa, consistent with there being greater fluctuations in average temperatures at the poles than in the tropics.
  4. The late twentieth century shows about 0.5oC of warming from trough to peak in a time period that corresponds to the 1976-1998 trough-to-peak warming of the global temperature series. This is broadly in line with that calculated globally by HADCRUT4 and GISTEMPa. This possibly corroborates data of individual weather stations having a warming adjustment bias (e.g. Reykjavik and Rutherglen) along with the national data sets of USA (Steve Goddard) and Australia (Jennifer Marohasy and Joanne Nova). Most of all, Paul Homewood has documented adjustment biases in the Arctic data sets.
  5. The proxy data shows a drop in average temperatures from the 1950s to 1970s. The late twentieth century warming appears to be a mirrored rebound of this cooling. Could the measured reductions in Arctic sea ice cover since 1979 partly be due to a similar rebound?

In conclusion, the Pages2K Arctic reconstruction raises some interesting questions, whilst corroborating some things we already know. It demonstrates the utility of these temperature reconstructions. As Steve McIntyre notes, the improvements partly came about through recognizing the issues in the past data set. Hopefully the work will continue, along with trying to collect new proxy data and refine existing techniques of analysis.

UPDATE 23.00

In the above, it is evident that the early twentieth century (c.1911-1944) Arctic warming in the revised reconstruction was twice the size of late twentieth century (c.1976-1978) warming, when global temperature anomalies show the later period as being greater in size. Steve McIntyre’s latest post shows that at least part of the answer may lie in the inclusion of the Okshola, Norway speleothem O18 and Renland, Greenland O18 series. These proxies both show a downturn at the end of the twentieth century. This might conceivably be a much greater influence on the discrepancy than either adjustment biases in temperature data, or differences between actual, not fully known, temperature anomalies between the Arctic region and the World. However, we will get a better understanding by eliminating the obvious outliers in the proxies and by continuing to positively seeking to eliminate bias in the global surface temperature anomalies.

Kevin Marshall

Notes

  1. Earlier this year I calculated the early twentieth century warming rates for the HADCRUT and GISTEMP series. They are


  2. From the same posting the 1976-1998 warming rates are



     

by manicbeancounter on 04/10/2014  •  Permalink
Posted in Arctic, Global Warming
Tagged , , ,

Posted by manicbeancounter on 04/10/2014

https://manicbeancounter.com/2014/10/04/pages2k-revised-arctic-reconstruction/