One of the early climate-skeptic arguments against the consensus view that the world is warming to a dangerous extent, was that the instrumental record of temperature was corrupted by the Urban Heat Island (UHI) effect. Wikpedia has a good summary and even RealClimate refers to their summary.

The argument was that a failure to correct for the (well-known) effect where urbanisation increases local temperatures led to exaggeration of the temperature trend measured by local weather stations and this, in turn, led to an overestimation of the average degree of warming across the world.

The climate-skeptic argument has been widely criticised. SkepticalScience dismisses it like this (and the link has further detail):

Scientists have been very careful to ensure that UHI is not influencing the temperature trends. To address this concern, they have compared the data from remote stations (sites that are nowhere near human activity) to more urban sites. Likewise, investigators have also looked at sites across rural and urban China, which has experienced rapid growth in urbanisation over the past 30 years and is therefore very likely to show UHI. The difference between ideal rural sites compared to urban sites in temperature trends has been very small...

But a recent paper by the climate-skeptic Anthony Watts and others (press release here, full draft paper pdf here) argues that, when properly classified by site-quality, there is a significant difference in the degree of warming reported between good and bad sites which leads to a significant exaggeration of the reported US temperature trend.

The results are summarised well by this picture showing the apparent differences between the reported warming of good versus bad sites: figure from watts et. al. paper

The paper argues:

A reanalysis of U.S. surface station temperatures has been performed using the recently WMO-approved Siting Classification System devised by METEO-France’s Michel Leroy. The new siting classification more accurately characterizes the quality of the location in terms of monitoring long-term spatially representative surface temperature trends. The new analysis demonstrates that reported 1979-2008 U.S. temperature trends are spuriously doubled, with 92% of that over-estimation resulting from erroneous NOAA adjustments of well-sited stations upward.

So the question is: does the new paper from Watts et. al. convincingly suggest that, when sites recording actual temperatures are properly classified, the warming trend is lower in sites where the UHI is less significant? Is Watts' observation-based argument credible? Does it undermine previous rebuttals of the influence of UHI on the temperature record?

NB. Watts doesn't claim there is no warming, just that the extent of warming is lower when only good-quality temperature-recording sites are used. So the question is about the quality of the instrumental record not about whether warming is happening.

  • 1
    This question is premature as Watts's paper is being reconsidered to think about "time of observation" adjustments. Briefly, if people who read maximum-minimum thermometers once a day change the time of the reading, this can introduce a systematic change to the readings, and Watts and his co-authors are considering whether they are satisfied with how they have treated this.
    – Henry
    Aug 3, 2012 at 21:25
  • Is there a "consensus view that the world is warming to a dangerous extent"? Would be nice if things were so clear cut.
    – going
    Aug 6, 2012 at 6:18

1 Answer 1


No, it should not be convincing to anyone who cares about using data meaningfully and scientifically.

Raw ambient temperature data is seldom useful in and of itself. Particularly when:

  1. temperature measurements get taken at different times of day, because temperature varies in known patterns with time of day; this requires Time of Observations (TOBs) corrections. One of the paper's authors, McIntyre, has even publicly acknowledged that this is a significant flaw in the paper (ref 6). Despite McIntyre's own less-than-distinguished history with statistics in climate science. (ref 7)

  2. temperature sensors move from one location to another, because that causes systematic changes in recorded temperature:enter image description here (figure from ref 1)

  3. the instruments themselves are sometimes changed, causing systematic changes in recorded temperature. (ref 3,4,5)

By ignoring all of those changes, by stripping out the data corrections as Watts et al do, information is stripped from the data. So whatever is claimed to be shown by the uncorrected data, is no longer meaningful. No conclusions can be taken from Watts et al. Not about underlying trends, nor about urban heat islands, nor about differing trends at high-quality and low-quality stations. The only thing the paper shows is that its authors have failed to do rigorous scientific analysis.


  1. Skeptical Science analysis of the paper
  2. Initial thoughts on the Watts et al draft
  3. A cooling bias due to MMTs?
  4. Menne et al 2009
  5. Quayle et al 1991
  6. "When I had done my own initial assessment of this a few years ago, I had used TOBS versions and am annoyed with myself for not properly considering this factor. I should have noticed it immediately. That will teach me to keep to my practices of not rushing. Anyway, now that I’m drawn into this, I’ll have carry out the TOBS analysis, which I’ll do in the next few days " from McIntyre's blog
  7. McIntyre claimed to have reproduced a hockey stick from red noise. However, what he did was generate 10,000 sets of red noise, and then filter down to just the 100 of these that most resembled a hockey-stick, for his paper. 7a 7b 7c 7d
  • 1
    I'm not expert enough to know whether these criticisms are right, but statistically speaking surely the issue is whether the complex corrections are correlated with the site quality? If they are not, then systematic differences between the trends based purely on site quality would still be significant. Watts, I thought after a quick read, was adding new data to the debate because he introduced a better metric for site quality based on direct site observation, not some statistically derived theoretical metric.
    – matt_black
    Aug 4, 2012 at 13:05
  • Might be worth adding the BEST study, as this was the reason Muller disagreed in the first place. He found that the models had accounted for the urban heat island effect very well, that the measures were accurate and that the statistics could be done a number of ways and give the same answer. Nov 28, 2012 at 2:01

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .