Removing UHI distortion - the elephant in the sitting room

Climategate 2.0 shows climate apologist Phil Jones referring to climate scientist Goodrich as a "jerk" for producing this beautiful demonstration of the magnitude of the UHI effect:

Getting UHI accurately removed from our trends is essential for three reasons. First, it completely removes the alarming recent spike in "global" warming and shows that variation is well within natural bounds. Second, the record then shows solar correlation again, showing that the cause is well within natural explanations. Third, UHI has been falsely quantified in IPCC 2007 AR4 (see below). I've gone down the route of examining clean data separately from infected data: inspired also by Anthony Watts' essay showing UHI alive and well; and Steve McIntyre's comments on the disinformation about UHI at RealClimate (and in official sources generally). My Primer notes the wide range of data distortion issues.

The seasonal temperature record for Salehard, Siberia, shows remarkably a huge upsurge in recent years for winter temperatures. The only reasonable conclusion is the Urban Heat Island effect, which in chilly Siberia is most noticeably increasing in winter months, as district heating improves. Click on the picture to enlarge.

Ira Glickstein at WUWT suggested 0.8°C/century global warming due to: (1) Data Bias 0.3ºC, (2) Natural Cycles 0.4ºC, and (3) AGW 0.1ºC. But I think there is much more artificial biassing. Steve Mosher's comment points to IPCC AR4's sleight of hand by Trenberth and Jones, in sweeping McKitrick and Michaels' 2004 paper on Urban Heat Island effects under the carpet, thus leaving Jones' faulty 1990 paper to continue to allow ridiculously small UHI corrections to the global temperature records (see below). The Abstract of McKitrick and Michaels' latest (2007) paper says that by using their method, "...nonclimatic effects reduces the estimated 1980-2002 global average temperature trend over land by about half." Unlike Jones, Trenberth, and the rest of "the Team", Ross uses appropriate statistics properly.

Let's continue with a favourite of mine, the presentation of Dr Andrei Illarionov at the 2010 Heartland Institute conference (U-tube here and here). Andrei shows locations for the 476 Russian records actually available, then cuts it to Phil Jones' 121, then cuts it further to 37, 12, and finally 4. Using Phil Jones' methods, he shows that just four records suffice to show the whole graph with 1.29°C temperature rise over the last 100 years - the four oldest records, which of course are urban. Arkhangel'sk (1813), Yeniseysk (1871), St Petersburg (1733), and Astrahan (1837).

Then Andrei divides the data into groups of different population size. He takes the average temperatures for the period 1921-1945, and compares them with the averages for 1961-1990, then with the averages for 1991-2006 (see two insets). Where there are no local residents the trends are, in both cases, a huge 0.6°C less than those for urban areas. Unfortunately it looks as if there are not enough stations from unpopulated places to show trends for the full century. The records for zero / very low population appear to be significantly lower than many records classed as "rural", as Roy Spencer's work below bears out. Even without this lowest group, a trend difference of nearly 0.5°C between the 10 most rural and the 10 most urban stations emerges. And since the urban stations' trend approximates to Jones' official Russian trend, Illarionov's evidence shows that Jones' trend is completely overrun by a 0.5°C per century urban heat island effect.

If that is true in Russia, what about elsewhere? Striking evidence emerges regarding the LACK of significant temperature increase in a number of significant places. First, there is the average of 24 US rural records, as picked out by a 6th grader schoolchild in order to compare with "twinned" urban records* right across the States. These show no overall temperature rise in a century of records. The urban records, on the other hand, show an overall rise of ~1.5°C. Second, look at New Zealand temperatures, when they are relieved of the fraudulent weighting they were given. No significant increase whatsoever. Third, look at a rural record from Pretoria, South Africa, chart adapted from Henry P. Temperature records for May to September were used as the Pretoria climate is sunny during this time and therefore provides a sensitive record without scattering from cloudiness, so it should show any overall global warming during this period. No increase here either. Fourth, another look at US rural and urban records, from a paper by Dr Edward Long. Here we see the familiar up-down-up, with a significant difference: rural temperatures recently are LOWER than they were in the 1930's; the overall trend is up, but only slightly, and visibly well within natural variations. Urban data show the familiar AGW picture. The difference between urban and rural records in 2000 AD looks like about 0.6°C.

Phil Jones of UEA famously said to climatologist Warwick Hughes, "Why should I make the data available to you, when your aim is to try and find something wrong with it?" Warwick Hughes wrote a critique of the Australian data used by East Anglia to arrive at their 0.8°C/century, saying

Our conclusion is that Australian temperatures have exhibited no upward trend over the past century. This conclusion is in contrast to the continental warming trend implicit in the East Anglia study from the stations chosen. The main reasons why our conclusions differ are that the East Anglia study:
• included a number of heat island affected city records;
• excluded other long term records from rural Australia.

The East Anglia study drew heavily upon data from major urban centres. One reason for this is the likelihood that such centres' temperature records have been kept more professionally; in addition, it is easier to trace... adjustments due to station shifts in major centres... It may well be that the records from more remote stations are inferior to those of the city sites. However, an examination of peaks and troughs for city and remote sites show a consistent pattern after adjustment for the trend. This indicates that one possible source of error - human error - is unlikely to be systematically present... Although the East Anglia study claims to have taken [UHI] into account in deriving its trends, we consider it to have done so inadequately. Our own findings... are consistent with those others have produced, especially in the U.S.

Hughes' graphs compare the average temperature trend for the 25 regional and remote Australian stations, for which data was available over the years since 1882, with the average temperatures for the six Australian capital cities. The East Anglia study used five Australian capital cities out of its 13 long term stations. These 25 records are: Geraldton (1), Narrabri (2), Hay (3), Albany (4), Rottnest Island Lighthouse (5), Walgett (6), Deniliquin (7), Bourke (8), Cape Naturaliste Lighthouse (9), Coonabarabran (10), Echuca (11), Cooma (12), Darwin (13), Moruya Heads Pilot Station (14), Omeo (15), Dubbo (16), Gabo Island Lighthouse (17), Bathurst (18), Strathalbyn (19), Mt. Gambier (20), Yamba (21), Wilsons Promontory Lighthouse (22), Newcastle Signal Station (23), Cape Otway Lighthouse (24), Alice Springs (25).

In rsearching this article, I viewed every one of the 250-odd graphs obtained by John Daly from the GISS records: "What the Stations Say". These will be less tampered-with than more recent GISS records, nevertheless they are recent enough for their pattern to be - what is that famous word - "unequivocal" - in showing little, if any, overall rise. Every record chosen has a message. Most are rural records which, while varying considerably from year to year, show little if any significant overall rise. This is particularly evident with the US. It is John Daly about whose death Phil Jones remarked to Michael Mann "In an odd way this is cheering news!" and the comparison here reveals why Jones might say this. The comparison is startling. The time periods are actually compatible since both stop slightly short of now.

Jones published a study in the Journal of Geophysical Research in 2008, that incorporated "improved" data for the stations that had been used in the 1990 study. Figure 6 (above, right).

"shows the comparisons (as anomalies from the 1954-1983 period) between the averages of the 42 rural and 42 urban sites used in 1990 compared with averages from the same stations from the CMA network. The dashed lines are the averages for the rural and urban sites in eastern China from the 1990 paper. The solid lines are the averages from the same stations from the CMA network. It is clear from the graph that the trends of the CMA data for both the rural and urban networks agree almost exactly with the results from the 1990 paper."

And still, the divergence between rural and urban records that all other studies show, is not there.

"The 2008 study undertook additional analyses and did conclude that there was a likely urbanization trend in China of 0.1 °C per decade for the period 1951-2004. The study maintains that, allowing for this, there was still a large-scale climatic warming of 0.15 °C per decade over the period 1951-2004 and 0.47 °C per decade over the period 1981-2004."

OR WAS THE "WARMING" IN THIS STUDY, ACTUALLY UHI?

Finally I'd like to recall the work of Roy Spencer (here and here) in attempting to quantify UHI corrections needed. While I've bypassed this route, and simply gone for trustworthy individual rural records and/or trustworthy statistics to sieve out the rural records, his work is important, in making a case for a significant presence of UHI, and UHI increases, that correlate to population, and population increases, in tiny populations that would generally be deemed rural and "beneath the radar" of UHI. Much of this may also apply to Jones & Wang's Chinese studies. Could it correlate to land use changes perhaps? Spencer's figures do seem to bear out the work of Illarionov above, that suggest perhaps adding an extra 0.1°C/century to the 0.6°C of Long or the 0.5°C of Illarionov. That however is speculative and needs further quantifying.

Now we can return to Ira Glickstein's figures distributing causes of the 0.8°C agreed global temperature rise of the last century. I'm going to plump for Natural Change 0.3°C (with fluctuations far exceeding this overall figure) and Data Bias 0.5°C. As to genuine human effects,there is still room for land use change. But for CO2 there is no room, in my opinion. The continuing steady CO2 rise can be explained in the centennial time scale we know exists in ice-core records for CO2 following temperature: the time between vast quantities of CO2-rich water sinking in polar waters during the Little Ice Age, and emerging in the tropics to out-gas, centuries later. Check here to compare our contributions with the natural CO2 cycle.

*Pairs of NASA GISS records used by "Peter": Gardiner to Portland, West Point to Central Park, Maryland to Albany, Lowville to Syracuse, Hemlock to Rochester, Angelica to Buffalo, Smithfield to Raleigh, Santuck to Charlotte, Arcadia to Fort Myers, Inverness to Tampa, Newnan to Atlanta, Philo to Columbus, Vevay to Cincinnati, Whitestown to Indianapolis, Brinkley to Little Rock, Amite to Baton Rouge, Albia to Des Moines, Saint Peter to Minneapolis, Kingfisher to Oklahoma City, Boerne to San Antonio, Los Lunas to Albuquerque, Tombstone to Tucson, Roosevelt to Phoenix, Morgan Como to Salt Lake City, Cuyamaca to San Diego, Lemon Cove to Fresno, Colfax to Sacramento, Hood River to Portland.

Page updated 1st December 2011