I’m going golfing in Pennsylvania in December and will be going in January also. So before you educated readers tell me there is a difference in weather and climate and before you go back eating your chinese and toasting to the betterment of the human condition, let me say the land masses are heating up and will be the demise of most land dwelling creatures including people. The human condition is that people are reactive, as the state of the world shows today. In a very short while, perhaps 300 years, land masses will be in the 110 degree range, and there is nary one thing we can do about it. Go back and observe, analyze, and make predictions, but do it as a Don Quixote, because we are not getting out of this one.
Economic activity is rooted in the now global capitalist push for both the desire for energy and desire for profit, deadly bedfellows for the planet, and mankind lacks the collective proactive skills to do anything different than buy and burn fossil fuels until we all die. Are temperature trends affected by economic activity? decidedly, yes. The question should be Is economic activity causing temperature trends, and we all know the answer to that.
Well, if by “econmic activity” they mean consumption of fossil fuels and forests, economic activity has a conspicuous effect. And the economic slowdown expected by many would be a relief. Moreover, or so it seems to me, if they divorce the economy from consumption of fossil fuels and forests, they risk irrelevance on economic or ecological grounds.
Economy:
Economic activity centres on coal mining, supplemented by fishing and trapping. In the final decades of the 20th century, tourism, research, higher education, and some high-tech enterprises like satellite relay-stations grew significantly. A 200 nautical mile (370 km) Fisheries Protection Zone around Svalbard was established in 1977 pursuant to the Act of 17 December 1976 relating to the Economic Zone of Norway. Despite recent discussions, Russia and Norway dispute their maritime limits in the Barents Sea and Russia’s fishing rights beyond Svalbard’s territorial limits within the Svalbard Treaty zone.
The Svalbard Undersea Cable System which started operation in January 2004 provides dual 1440 km fiber optic lines from Svalbard to Harstad via Andøy, needed for communicating with polar orbiting satellite stations on Svalbard, some owned by the National Aeronautics and Space Administration (NASA) and the National Oceanic and Atmospheric Administration (NOAA), both United States government agencies.
The Norwegian state-owned coal company employs nearly 60% of the Norwegian population on the island, runs many of the local services, and provides most of the local infrastructure. Coal production has increased significantly over the past 10 years, rising from less than 500,000 tons in 1994 to over 2,500,000 tons in 2004.[6]
Exploration for oil and natural gas is underway.
Coal mining in Svalbard:
The Ny Ålesund mine was closed down in 1963 after an explosion in 1962 when 21 lives were lost, and has since been converted to a scientific post.
As of 2006, there are three operational coal mines in Svalbard. There are large mines in Sveagruva (production 2 million tonnes per year,[13] and Barentsburg, while the small mine in Longyearbyen is used mainly to supply the town’s own power plant.
Demographics:
Svalbard has a population of approximately 2,400 people as of 2005. Approximately 70% of the people are Norwegian; the remaining 30% are Russian, Ukrainian and Polish.[citation needed] The official language of Svalbard is Norwegian. Russian is used in the Russian settlements, but formerly, Russenorsk was the lingua franca of the entire Barents Sea region. The annual population growth is -0.02%
Economic activity obviously has no direct effect on climate. Some things related to economic activity, such as urbanization, energy use, and pollution, do have effects on climate, both local and global, but economic activity in and of itself has no effect and it’s just silly to state that it does.
Economic activity means money changing hands (if we’re talking about GDP), and a few electrons, piles of paper, or sacks of gold changing hands are not going to change the climate.
At best, it’s sloppy use of language. At worst, it makes the results meaningless by not distinguishing between agriculture and coal plants (high climate effect/GDP) and a financial district (minimal climate effect/GDP).
PS: It appears that this paper is wrong on plenty of other counts, as mentioned in the main article.
Re 4. Indeed. This site has been losing its innovative edge (and some of its contributors), and has deteriorated into a column commenting other people’s work. This is not what the headline promises: Climate science by climate scientists. Furthermore, the choice of topics gives the impression that the so-called climate contrarians are the only people finding out anything interesting (true or not) about climate and its change.
You said: “I think it is difficult to argue that factors such as the urban heat island effect plays an important role here”
Consider the fact that many of the weather stations are located within the ever expanding “Heat Islands” mentioned. It would seem logical that the temperatures taken over time would reflect a corresponding rise with the growth of the urban heat islands.
What’s with the RealClimate bashing in 4 and 6? This post exemplifies what’s best about RC: it’s educational and directly on-topic. This caviling borders on trolling.
Don’t blame RC for sidetracking the discussion from real climate science to bogus pseudo-science. Place the blame squarely where it belongs: on McKitrick, Michaels, and Loehle.
The non-stop stream of sloppy research in order to discredit genuine climate science makes it necessary for RC to set the record straight. Keep up the good work!
Rasmus : They find that the greatest differences between measured and adjusted trends at Svalbard and other places in the Arctic and Antarctic (See marked sites in Figure below). This is not convincing. Thus, the results themselves provide examples of spurious values obtained by their analysis. Even if they were identified as ‘outliers’ (Svalbard was apparently not one), the fact that their analysis produced highest corrections for economic activity at these places suggest that their analysis is not very reliable.
I’m still reading M&M2007, but it seems they study two kinds of “artifact”: one due to local anthropic effect on measured data, other due to inhomogeneities. Could the Arctic and peri-Arctic biases be mainly related to observationnal difficulties (inhomogeneity) rather than economic activity as you suggest here ? There are probably few meteo. stations north to 60°N, so a lot of interpolation.
For example, I empirically observed that on Nasa Giss (Gistemp), when you smooth at 1200 km, 2007 is bit warmer than 1998, but when you smooth at 250 km (no data when no covergage), that’s the contrary. The main differences between the two estimates arise from Africa and… peri-Arctic, precisely (70-90°N). http://data.giss.nasa.gov/gistemp/
[Response: Good question. If you want to look at trends over a longer period, inhomogeneouity is definately a problem, but is has not been a problem since 1979. THe very strong recent warming is also supported by bore holes on Spitzbergen. -rasmus]
I think it’s also worth pointing out that their analysis assumes a priori that any correlation with economic activity must perforce be related to a contamination of the surface temperatures by urban heat effects. This is certainly not the the only possibility, and the fact that they have apparently discovered urban heating in the satellite trends as well should have alerted them to this fact.
For instance, tropospheric ozone and black carbon emissions have local forcing components that are closely related to local emissions. Large scale land use similarly. The test of whether these factors produce correlations like those reported by M&M is available in the IPCC AR4 model archive. The fact that in 3 years since they first started this analysis, they didn’t once take a model simulation over the same period and do the same calculation is telling. This is of course in addition to the rather poor statistical significance alluded to by Rasmus. Checking their conclusions with random 25 year sequences from a model control run would have been a good test of the robustness to climate variability. Again something they didn’t apparently think of.
Another point : this paper of Hinkel et al. showed that UHI effect can be huge in peri-Arctc areas too : at Barrow (Alaska), 71,3°N, during the winter period, the urban area is 2.2 K warmer than the rural area.
Hinkel K.M. et al. (2003), The urban heat island in winter at Barrow, Alaska, International Journal of Climatology, 23, 1889-1905.
The paper under discussion is outside my areas of expertise, but I get the impression that the author primarily focused on data from the physical world. Getting a handle on the quality of that data seems to be an issue.
For any given piece of research there are basically an uncountable number of things that were not done; some of more importance, some of less. Let’s make a list of what apparently has not been thought of with respect to, say, the NASA/GISS ModelE computer code. Many would say that because of what apparently has not been thought of, the ‘data’ from the code has no quality and is useless.
[Response: Your argument can be applied to anything and therefore nothing has quality or usefulness. Including your argument. Which is then self-contradicting. Think of it this way instead. All analysis is incomplete in some way. Thus conclusions should always be preliminary. As more analysis is done, the preliminary conclusion becomes stronger or falls. Writing an op-ed declaring that M&M proves that the surface record (and satellite record!) are contaminated by UHI effects is rather putting the cart before the horse, don’t you think? Especially, since most of the ideas for further analysis were already put to them in Rasmus’s original comment in the journal, and are pretty obvious in any case. If you have some obvious things to test with ModelE or new analysis you want to do, go ahead. That’s why the code is public. - gavin]
It is important that RC stay current with those such as McKitrick who believe that there are aspects to the science which are overlooked. RC represents the mainstream and as such is under continual attack. I often come here looking for specific refutations of contrarian views, especially when (as often) those views undermine orthodoxy.
RC should present current work and latest understandings as well, of course, but it is simply necessary to, sometimes, comment on other work, especially when that work gains a bit of traction.
We don’t see RC wasting much time disputing obvious denier rhetoric, but when a paper represents a seeming legitimate effort to do science, I (and I’m sure others) want to know what RC thinks.
At the center of this discussion is the explicit role of the GHGs in the observed warming. The terrestrial data likely remain contaminated to some degree by factors related to land use change, including vegetation change, urban heat island, changes in wetland distribution, etc. I have no doubt that due diligence has been exercised to remove the effects of the urban heat island from the terrestrial record (this has been discussed at length elsewhere on RC and other forums), but it is reasonable to conclude that artifacts related to economic activity likely remain. In light of this probability, which data should we use to validate the forcing effect of the GHGs? What are the implications of using only the ocean record?
I am an academic ecologist, and I speak and write regularly about land use change and climate change. While I have a better understanding than most on issues related to climate sensitivity, I would benefit from a direct, clear discussion of the relative importance of the forcing agents as they relate to the temperature record. Has the IPCC AR4 derived an incorrect estimate of climate forcings and feedbacks related to the GHGs? While it is unlikely that we can obtain complete certainty on this issue, it is important for mainstream climate scientists to continue to debate this issue in a manner that can be interpreted by rest of the academic community.
There is an ongoing tug of war between the global climatologists and those academics that seem to hail more from the meteorology groups. This debate is often contaminated by pejorative, and could benefit from more light and less heat. Within the academy there is a small, but influential and vocal group of GHG skeptics, including Pielke, Sr., M&M, Christy and others who variously argue that we are placing too much emphasis on mitigating the GHGs, to the detriment of the economy and human well being. While it is apparent that a low carbon economy could be a thriving economy, climate scientists must be as clear as possible about the explicit role of the GHGs. Pielke, Sr. has repeatedly claimed that the IPCC estimate of the forcing effect of the GHGs is in error. I urge the scientists at RC and elsewhere to develop a detailed analysis of this and deal with this issue as best they can. Please don’t refer me to an earlier post – instead let us see the most recent thinking on this issue explicitly addressing the most recent claims.
This issue is especially crucial to the construction of sound policy. I have recently returned from DC where I spoke with policy makers about the energy bill and cap-and-trade. My personal view is that a risk assessment approach is far preferable to a cost/benefit approach, and thus we should aggressively mitigate GHG emissions. Unfortunately, this is not sufficiently compelling to policy makers given that there are some credible scientists who argue that our money is better spent on other measures. It is arguable that we should also develop a portfolio of mitigation options related to compensating for changes in land use.
I look forward to seeing more discussion of this issue at RC and in the peer-reviewed literature. I agree with 8 and 9. It is essential for RC to do its best to clarify and set the record straight. Although I am skeptical of its conclusions, M&M2007 is being published in a top tier journal, and thus needs to be addressed as serious science.
I find it a bit ironic when people use satellite data measurements to argue that GHG is unimportant. They rely on the fact that these measurements are derived using the very same type of physical laws as those predicting an enhanced greenhouse effect due to increased GHG levels (neglecting feedback processes).
What makes this especially ironic is the fact that we measure the levels of greenhouse gases at various altitudes by means of the increased opacity of the atmosphere to various wavelengths (i.e., “channels” in satellite lingo). This is the mechanism by means of which greenhouse gases have an enhanced greenhouse effect as their concentrations increase over time.
The probability that RC will comment on a paper being actively discussed elsewhere is approaching unity…
and…
the choice of topics gives the impression that the so-called climate contrarians are the only people finding out anything interesting (true or not) about climate and its change.
Complaints of this sort are right out of the creationist/IDist playbook. When scientists and mathematicians attack claims and arguments made by [ID]creationists, the response is … “look, our work is so important ‘darwinists’ have to attack it to support the ‘darwinist conspiracy’. So obviously, we’re right!”.
I suspect it’s common among science denialists of all flavors.
Paul, your golfing experience notwithstanding, the east coast of the U.S. has actually experienced cooling. This site explains that in the FAQ section on the Michael Crichton book “State of Fear.”
As for this article, I think the author missed the point. The HADcrut dataset and analysis presumes a randomness in the data for the purposes of assessing error. Even if the dataset is smooth and continuous, if there is a non-random component (i.e. a correlation), the correlation will need to be modeled in the error bars. IF that didn’t happen, this new paper is pretty significant. I suspect the reviewers at JGR-Atmos have representation in IPCC and must believe this new paper is significant.
Economic impact on temperature? Depends on where the thermometer is.
[Response: Thanks for the nice picture. The thermometer is located near the airport (with about two landings/departures a day), to the far left of the picture. Other measurements from nearby sites, such as Ny Ålesund, Sveagruva, Hopen & Bjørnøya, show similar warming as Longyear byen. There is no economic activity near these sites, except for at Sveagruva. -rasmus]
Every action within an economic system has associated with it an ensemble of ghg emissions. Those emissions can be traced and accounted for in the carbon accounts of the economic unit, and specific patterns and relationships can be identified amongst the myriad of self-organising economic units and subsystems. Thus, provided one accepts the relationship between ghg and temperature, it would be fine to assume that temperature trends are affected by economic activity. Economic activity is in effect an order parameter. Interestingly, the inverse problem can be examined. Does temperature trend affect economic activity? It certainly does.
The utilization of natural resources is the very basis of economic activity, therefore it IS economic activity with a direct and measurable impact on many facets of the environment, climate included,.
Money does not change hands just so the “exchangers” can have fun!
Every time I start my internal combustion engine I am taking part in econimic activity. Why? Because I have to work to buy gasoline, and in a cascade of “economic activity” a fossil fuel company has to employ and pay people to make the fuel available.
When I start my automobile, just because money is not changing hands at that instant, does preclude that activity from being economic.
We all know what is meant by the term, is it then not a bit, in your word, “silly” to argue that economic activity has no direct effect on climate?
Statistics is such a difficult and counter-intuitive subject that we and the public need the best possible assistance in countering the claims of liars. Statistics makes life too easy for liars otherwise. Since very smart and honest professors have been known to make mistakes in statistics, checking by other professors is needed. Thank you, rasmus, for making us aware of a few of the pitfalls the unwary may fall into, and the level of care required to write a good paper on the subject of climate.
Here is a planned ‘economic activity’ which ought to have a positive impact, not only on climate, but also on the well-being of the peoples of the Sahel:
Could someone please tell me how this paper got through peer-review at JGR-Atmosphere? I am stunned and disheartened that something this horribly flawed is coming out of an AGU publication.
If urban heat islands are to be discounted when looking for trends - surely they must be included when looking at the big picture.
If one looks at those classic images of The Earth at Night one can see the twinkling evidence of mankind’s presence most places. And those lights are mostly identifying the location and extent of UHIs.
Is not the heat from UHIs entitled to sit at the same table as heat from bare earth when we ask about the present global temerpature? When does the data from the UHI’s get considered - when their sweltering streets, shimmering towers and streaming chimneys cover more than 50% of the earth, or somewhat sooner?
The UHI effect IS accounted for, and its (rather negligible) effect is mentioned in the IPCC 2007 and other documents. It is not whether UHI is discounted or not; the question is what effect that has on the the Global mean temp, and regardless of what McKitrick says, it is not a practical amount. It is also not influencing increased ocean heat content, melting ice caps and glaciers, satellites showing tropospheric warming or strato cooling, etc
Is not the heat from UHIs entitled to sit at the same table as heat from bare earth when we ask about the present global temerpature? When does the data from the UHI’s get considered - when their sweltering streets, shimmering towers and streaming chimneys cover more than 50% of the earth, or somewhat sooner?
Love the poetry, but the answer is still no.
Not when you are trying to determine global or latitudinal trends in temperature. Need to keep things weighted according to area - and cities are pretty tiny in comparison to the countryside.
*
Fortunately we don’t seem to have to worry all that much regarding urban heat islands — most of the time. As is well-known, there are various corrections made to eliminate any urban heat island effect, and they appear to have been quite successful:
Assessment of Urban Versus Rural In Situ Surface Temperatures in the Contiguous United States: No Difference Found
Thomas C. Peterson
Journal of Climate, VOL. 16, NO. 18, 15 September 2003 http://www.ncdc.noaa.gov/oa/wmo/ccl/rural-urban.pdf
*
Likewise, according to Peterson and Vose (1997) analysis cited by IPCC 2001, the long-term (1880 to 1998) rural (0.70 C/century) and full set of station temperature trends (0.65 C/century) showed that rural stations were trending slightly higher. More recently, a 1998 analysis for the long-term trends (1951-1989) rural (0.80 C/century) and full set of station temperature trends (0.92 C/century) showed urban stations trending slightly higher.
However, in both cases, the difference between urban and rural trends were not statistically significant. As such, it would appear that the Urban Heat Island effect is either negligible or corrected for when climatologists derive their trends for temperature.
What I would be interested in is whether GISS corrects for the Urban Heat Island effect in Barrow. I would presume they do. That is standard operating procedure, I believe. And we have known that it would be a problem if left uncorrected since 1983. For the winter, not summer, since for the latter it is weak to non-existent.
Please see:
Conversely, summer demand for lighting and interior temperature maintenance is minimal compared with mid-latitude or tropical settings, where air conditioning is required. The presence and strength of the UHI in Arctic regions, therefore, has a strong seasonal component with maximum development and intensity in winter, and only weak or nonexistent expression in summer (Benson et al., 1983).
Hinkel K.M. et al. (2003), The urban heat island in winter at Barrow, Alaska, International Journal of Climatology, 23, 1889-1905.
Firstly, I’ve only had a 15 minute look at this paper, but the quality of the analysis concerns me.
In the abstract M&M say, “…we test the null hypothesis that the spatial pattern of temperature trends in a widely-used gridded climate data set is independent of socioeconomic determinants of surface processes and data inhomogeneities. The hypothesis is strongly rejected (P= 7.1×10−14 )…”
Rasmus is completely correct. They have not considered correlation between data points. If they’d doubled the number of grid points the p value would be infinitesimal.
I noticed they left out outliers - such as Arctic and Antartic “hot-spots” when fitting the model. Outliers can be a result of measurement errors or even chance results, but high lattitude “hot-spots” are clearly neither of these. They are regions of stronger than average warming trend. These outliers are most likely a sign of model failure.
I’m still coming to terms with their model, but the inclusion of many non-significant parameter estimates in Table 1 is a worry.
I suspect another big problem with the model is omitted variable bias. The rate of southern hemisphere (and tropical) warming is slower than northern hemisphere warming and that includes landmasses. But economic activity is strongly associated with the extra-tropical nothern hemisphere. As far as I can tell there is no variable in the model to account for this hemisphere effect. This effect will then be incorporated into other variables associated with hemisphere - i.e. GDP, coal production etc - resulting in a biased estimate of these effects, the so called “omitted variable bias”.
Using the regression model to filter the
extraneous, nonclimatic effects reduces the
estimated 1980-2002 global average
temperature trend over land by about half.
So I’m confused - is there a global average temperature or isn’t there?
All due apologies for going OT.
On CNN this morning (around 6:30AM 10 Dec), their science reporter Miles O’Brien did a story on Al Gore at the Nobel Prize ceremony. This is a SCIENCE reporter! Instead of illuminating us on the SCIENCE (and Gore’s efforts to publicize the science) behind the Prize, Miles lead off with Chris Horner (of “Politically Incorrect Guide to GW”) denouncing Gore. I won’t repeat Horner’s trash talk, only that it was vicious ad hominem.
Why they gave this luddite corporate hack a platform in the middle of this story is puzzling, to say the least!
The piece may show up on CNN.com. This style of reporting is despicable and must be protested.
Maybe OT, but does anyone have any idea of what kind of money it would cost in the way of funding to research and produce a paper like this? McKitrick has been recieving what looks to be rather substantial funding from the Social Sciences and Humanities Research Council to do this work (into the $100,000s).
What I’m trying to understand might be in the post but I haven’t been able to dig it out. Sorry.
I take it that a simple summary of the paper’s contention is that climatic terrestrial temperature measurements have been and are overstated because of increases in economic activity. My question: do they contend that economic activity per se is the culprit, or that heat islands are the culprit — stemming from economic activity which tends to move things from rural to urban areas. It’s not obvious how economic activity per se can add any heat without extreme stretching of the meaning of “economic activity”, other than maybe its increase in GHG output, which would form a ludicrous circular logic. In other words does their argument boil down to simply the heat island effect? Or would they claim something else?
I’m posting this before reading the comments which might have the answer….
[[Economic activity is rooted in the now global capitalist push for both the desire for energy and desire for profit]]
And yet the Soviet Union and Peoples’ Republic of China were/are the most egregious polluters in the world.”
Not really, at least so far as GHGs are concerned: neither approached/approaches the per capita emissions of the USA, Canada, Australia, or even western Europe - though not for want of trying. With regard to the PRC, a considerable proportion of current emissions result from manufacture for export to the rich world; and more fundamentally, it is arguable that China is now an integral part of the capitalist world-system (although the USSR was not). It is the very success of capitalism in increasing economic activity that now threatens to bring about its own collapse.
Hello Rasmus. Thank you for your comments on my new paper. Here are some responses.
Spatial autocorrelation is an issue, in principle, with any cross-sectional study. I agree with you. You should have mentioned, though, that we applied a GLS estimator with White’s HCCME terms and clustering structure built in. Adding in local spatial AC coefficients would, for many of the regions, be redundant on top of the exiting off-diagonal elements. My conjecture is it would not affect things. However, that’s no more than a conjecture. Perhaps a reader who is interested, and better than me at programming, will figure out the math to put spatial AC controls in the GLS estimator while also controlling for heteroskedasticity and clustering.
I accept your concerns about whether we used the most updated data possible. It was a large data base to put together. It’s available at my web site. If someone wants to swap in columns with newer series (making sure the definitions are consistent) then the code can easily be re-run.
I don’t agree with your concerns about over-fitting. Over-fitting becomes a detectable problem when you have a high r2 and very low t-stats. We don’t have that, and the variance inflation factors indicate that our covariates are contributing unique explanatory power.
Your paragraph beginning “I have not examined the economic data…” seems to rule out using socioeconomic covariates under any circumstances. Yes, they change abruptly at national borders. Yes, an ideal data set would have them change continuously, but discrete changes doesn’t mean a variable can’t be used in a regression model. You’ve set up a criterion where it’s Heads-disqualified, Tails-disallowed. Can you state what circumstances you would permit socioeconomic covariates for this type of test?
You raise concerns about spurious results, but we have a few tests for this, including the endogeneity discussion in Sct 4.4. Can you be more specific? Your point was rather vague.
The oceans are obviously not at issue here. Perhaps there is an issue whether data collected from ship intakes will prove to be comparable to data collected by the argo network, but that’s for others to examine.
It is not true that our discussion of the effect of urbanization and land-use change rested only on our 2004 paper. In the on-line preprint (http://www.uoguelph.ca/~rmckitri/research/jgr07/jgr07.html)
pages 4-11 discuss anthropogenic surface processes and inhomogeneities, and there are many references therein.
Yes, we used the UAH data. I will eat my toque if that choice matters greatly, but, again, the data base is on-line and others can easily check.
Is 24 years too short to extract a trend? Well, 30 years would be better. If we had 30 years and the same results emerged, would your position change? I doubt it. So maybe the point is at most a secondary one.
Re the use of MSU data. Nothing in my paper disputes the idea that GHG are infrared-absorbing, or that oxygen emits microwaves. I’m not in a position to say anything about these things either way. But your qualifier is key: “…neglecting feedback processes…” Feedback processes are pretty much what is at issue.
Your conclusion says that there may very well be some contamination of the data. I suppose this admission represents progress. But considering the importance of the data at issue, is this an adequate response on your part? I have made the case that there is substantial contamination of the data. You don’t accept my results, which is your prerogative, but if you want to argue that there is only a small contamination problem, taking into account both surface processes and inhomogeneities (i.e. not just UHI effects), you should make the case with clear empirical methods.
Oh, and bcl, this kind of research doesn’t cost much at all. I used a bit of time of one of my research assistants for part of the data assembly. Otherwise the data are free and I used software I already own. And there are page charges for JGR. I am funded by SSHRCC for a range of research projects.
[Response: Thanks for your response, Ross. I think that your model is over-fit because I think that you have not eliminated the dependence and include too many inputs without any clear/understood connection. A regression analysis will always find a combination of weights giving the ‘best’ fit. You find greatest ‘biases’ in locations far away from places such as the Arctic and Antarctic. I don’t find that convincing. -rasmus]
If the oceans are warming, and we know that heat is going in and not going out or just being redistributed, then we know an external agent is acting on the climate system and this is unequivocal across the globe from ice sheet/glacier responses to SST/atmosphere/surface temperatures, snow cover decline, etc. I am not sure why this wouldn’t be an issue in a study going over UHI impact of global land temperatures, since one would imagine this external agent is also acting on land. One couldn’t say that the CO2+feedbacks are just acting on oceans and polar and rural regions, but not in urban areas, and that UHI makes up for this in the instrumental record.
As for the UAH data, it was not right, so shouldn’t have been used but I have no reasonable insight into how that would have effected your study. — C
Using the regression model to filter the
extraneous, nonclimatic effects reduces the
estimated 1980-02 global average
temperature trend over land by about half.
So I’m confused - is there a global average temperature or isn’t there?
The following might help:
Inter-annually, the 18-year Pathfinder data in this study showed global average temperature increases of 0.43 Celsius (C) (0.77 Fahrenheit (F)) per decade.
By comparison, ground station data (2 meter surface air temperatures) showed a rise of 0.34°C (0.61°F) per decade, and a National Center for Environmental Prediction reanalysis of land surface skin temperature showed a similar increasing trend in global and land surface temperature, in this case 0.28°C (0.5°F) per decade. Skin temperatures from TOVS also prove an increasing trend in global land surface temperatures. Regional trends show more temperature variations.
According to the Pathfinder data, it would appear we have slightly underestimated the trend in the global temperature (by 0.09 C) — using ground-based measurements from 1981-1998. NCEP reanalysis of the data had shaved another 0.06 C, so it underestimated the trend by 0.15 C, according to the Pathfinder data. Not sure how statistically significant that is though.
The real question here is whether measured global warming has been exaggerated by changing human activity in the vicinity of the measuring stations.
There is no question that the measurements themselves are affected by local human activity. We attempt to adjust for this “contamination” (to use McKitrick’s term) algorithmically.
Are these adjustments correct? For purely algorithmic reasons, I have long been skeptical.
The way to test this is to do precisely the sort of study that McKitrick has done. The correlation that McKitrick has shown between socioeconomic activity and the temperature anomaly is truly startling. I’m surprised that so many here are willing to dismiss it so quickly.
I agree with Rasmus that spatial correlations would tend to reduce the confidence with which McKitrick makes his conclusions (and I simply don’t have the background to understand McKitrick’s response), but even if the conclusions are overstated, the correlation is real, and difficult to question.
Isn’t it incumbent on purveyors of global temperature data to prove that their data is NOT “contaminated”?
Doesn’t McKitrick’s analysis STRONGLY suggest that this is not the case?
Does anyone know of a site with graphs of cloud cover and temperature anomalies for Longyearbyen? I’m particularly looking for records that cover the period from about 1935 to present. Anomalous high temps at west-facing near-ocean locations are my interest, and this looks like a prime candidate. All the papers I’ve googled have been behind a pay-to-view wall.
I would expect it to exhibit a rising temperature spike starting in 1939.
re #34/Rasmus: I think overfitting due to dependence and too many inputs without clearly understood connection might be a problem also in other areas like multiproxy temperature reconstructions (many uncalibrated proxies regressed on, say, instrumental temperature PCs). Maybe you could help Ross by explaining how the problem has been avoided, e.g., in the landmark paper by Mann et al. (1998)?
[Response: Step-wise regression is often preferred, but I’d recommend a proper cross-validation. -rasmus]
[Response: Cross-validation (and objective selection rules) are the key here, as indeed emphasized by Mann et al (1998), and driven home fairly convincingly by the followup papers by Wahl and Ammann we’ve referenced above. You should read these (and obviously, re-read Mann et al 1998) to understand the issues better. -mike]
Need to keep things weighted according to area - and cities are pretty tiny in comparison to the countryside.
I would have thought you needed to keep ‘things’ weighted in terms of energy content per unit area not just area alone? Thus won’t the energy of the UHIs figure rather more strongly in the overall picture?
Otherwise it just seems to me that we are looking for trends among a sample we are ever-reducing as old stations are knobbled by the UHI effect. Isn’t this obscuring how ‘inconvenient’ the truth really is about warming globally?
Yet would not the UHIs put more (in fact most) anthropo-energy into circulation, capture more insolation and (because of effects like inversions holding higher concentrations of GHGs) retain more energy locally at lower altitudes?
That there is still an upward trend to be found in the remaining non-UHI stations is useful info, but what do we see if we use all stations on say a 10km square grid (or 10″ grid or whatever) with stations within a square averaged and look at the past and present energy content (Joules) of the total atmosphere globally over time?
Isn’t that total energy content including UHIs the figure that matters because that energy is the meat in the sandwich between the energy additions to the system and the energy loss to space which in turn defines the global temperature?
Re: #11 Response by Gavin
[……..If you have some obvious things to test with ModelE or new analysis you want to do, go ahead. That’s why the code is public. - gavin]
Is this a valid challenge? Many (Most ?) of us don’t have the expertise or the capacity( by a long shot) either in our gray matter or on our hard drives to run alternative tests on Model E or any equivalent. Also Nasa/ Giss works as a team and on company time. Do you really expect a poster working alone in his spare time to take you up on the above statement. I don’t much like the way the war in Iraq has been handled, but if somebody were to say “We’ll put you in charge - go ahead and solve the debacle”, I’m sure I couldn’t do any better.
You shouldn’t have to have the knowledge of say a four star general or an expert climate modeler in order to make good faith criticisms. Also this works both ways. Because you’re unhappy with the M&M paper, you shouldn’t have to develop your own paper on the same topic, using their method of analysis or another method of analysis with their own or your own data.
[Response: I agree. Good faith criticisms should be welcome. But McKitrick’s op-ed and declarative statements seem to have done before any of those criticisms were dealt with. (PS. you could run ModelE on your home linux box if you want, and we have a port for windows in the works- you don’t need to do that to make good faith criticisms, but if you wanted to see how it worked you could in fact see for yourself). - gavin]
In principle, global average temperature over a given period would be defined as the integral over the surface area, then integrated over the time period under consideration, then divided by the product of the surface area times the time.
As stations measure temperature over a limited number of points, one will need to interpolate temperatures between stations over a given area, but in no small part this will involve removing the effects of Urban Heat Islands. Technically such an average will be refered to as being weighted by area and will make reference only to temperatures, not heat content.
Satellite measurements should be more accurate, both in arriving at global average temperatures and in identifying trends over a given period as they are able to take readings from a far larger number of points. As such when Pathfinder satellite gives a larger trend in global temperature of 0.43 C/decade but groundbased measurements give 0.34 C/decade, I suspect that the higher satellite-based trend is more accurate.
But yes, urban temperatures are important for their own sake — but they are relatively insignificant when compared to the effects of greenhouse gases, and without corrections for UHI would give a distorted picture of the rise in global average temperature. However, given the results of Pathfinder, it would appear that we have overcompensated for the effects of UHI.
#41 Timothy, what is Pathfinder Satellite ? (The link leads to another discussion without a precise reference). If you look at UAH data for lower troposphere trends on land and NH (used by M&M 2007), you get 0,24 K/dec (0,33 K/dec CRU, 0,34 K/dec NCDC, 0,29 K/dec Giss, cf AR4 tab. 3.2 p. 243). Anyway, even if the trends were the same, I guess it’s not the M&M2007 purpose to look at global correlation, rather to assess local (grid by grid) correlations between Ts, Ttropo, economic activity.
Other point (more general) : UHI is detectable even for small towns (see Torok 2001), not just big cities, and anthropic effects on surface energy budget is not limited to urbanization.
Torok S.J. et al. (2001), Urban heat island features of southeast Australian towns, Aust. Met. Mag., 50 1-13.
Over-fitting: Re. Rasmus, Ross McKitrick @34, Jean @39,
Although I have reservations and criticisms of the model generally, I think the concerns about over-fitting are over-done. Over-fitting usually results in models which follow data noise well but have large errors on parameter estimates and poor predictive precision. It will not however have much influence on the global test of the null hypotheis, which is stated in the abstract as:
“…the null hypothesis that the spatial pattern of temperature trends in a widely-used gridded climate data set is independent of socioeconomic determinants of surface processes and data inhomogeneities.”
This hypothesis is rejected with a p=7.1*10^-14.
My bigger concerns are with the model itself. Two of which Rasmus has stated:
1) Failure to appropriately incorporate spatial correlation of temperature measurements. This will result in exaggerated p-values.
2) Poor model fit for land areas at extreme latitudes. This is evidence of outright model failure - the model does not fit the data as it is inappropriate.
And an additional issue:
3 Omitted variable bias. http://en.wikipedia.org/wiki/Omitted_variable_bias
This is a bias that appears in parameter estimates - such as the effect of economic activity on local temperature measurement in Ross’s model - that is due to the omission of a variable that is better able to explain the relationship between location and temperature. Such a variable might be hemispheric effects, or oceanic effects that extend beyond coastlines.
Essentially omitted variable bias is a type of poor model specification that results in biased estimates of model parameters even when the model fits the data.
Re. Gavin @11 and my earlier comment about omitted variable bias.
Gavin’s concerns about tropospheric ozone and black carbon emissions having local forcing components is another example of omitted variable bias. A fitted model that excludes these will attribute these forcing effects to other variables in the model which have large values at the same locations, such as GDP.
Anyone care to hazard an estimate of the percentage of human induced CO2 of the 100 ppmv that finds its way into the +.7 observed trend in the mean global anomaly?
[Response: That’s not a well-posed question. See here for a discussion. - gavin]
If you look at UAH data for lower troposphere trends on land and NH (used by M&M 2007), you get 0,24 K/dec (0,33 K/dec CRU, 0,34 K/dec NCDC, 0,29 K/dec Giss, cf AR4 tab. 3.2 p. 243). Anyway, even if the trends were the same, I guess it’s not the M&M2007 purpose to look at global correlation, rather to assess local (grid by grid) correlations between Ts, Ttropo, economic activity.
Pathfinder was using skin temperature — which is actually closer to the surface than land-based.
They state:
Furthermore, satellite skin temperatures have global coverage at high resolutions, and are not limited by political boundaries. The study uses Advanced Very High Resolution Radiometer Land Pathfinder data, jointly created by NASA and the National Oceanic and Atmospheric Administration (NOAA) through NASA’s Earth Observing System Program Office. It also uses recently available NASA Moderate Resolution Imaging Spectroradiometer skin temperature measurements, as well as NOAA TIROS Operational Vertical Sounder (TOVS) data for validation purposes. All these data are archived at NASA’s Distributed Active Archive Center.
Inter-annually, the 18-year Pathfinder data in this study showed global average temperature increases of 0.43 Celsius (C) (0.77 Fahrenheit (F)) per decade.
Anyway, it wouldn’t surprise me if lower trop and ground-based were actually closer to one-another. In fact, I seem to remember they are almost the same — at least for the continental US — but I will have to check. However, I do remember there is somewhat greater variability with lower trop — the hot years tend to be higher.
Other point (more general) : UHI is detectable even for small towns (see Torok 2001), not just big cities, and anthropic effects on surface energy budget is not limited to urbanization.
Torok S.J. et al. (2001), Urban heat island features of southeast Australian towns, Aust. Met. Mag., 50 1-13.
Certainly it is detectable. But it is also in very large part accounted and corrected for.
Fortunately we don’t seem to have to worry all that much regarding urban heat islands — most of the time. As is well-known, there are various corrections made to eliminate any urban heat island effect, and they appear to have been quite successful:
Assessment of Urban Versus Rural In Situ Surface Temperatures in the Contiguous United States: No Difference Found
Thomas C. Peterson
Journal of Climate, VOL. 16, NO. 18, 15 September 2003 http://www.ncdc.noaa.gov/oa/wmo/ccl/rural-urban.pdf
… and I gave additional sources, as well as pointed out that we knew about Barrow in 1983. Somehow I suspect the UHI at Barrow has been corrected for. In any case, there are a fair number of corrections made depending upon location, altitude, etc. There is the law of large numbers which cancels out much of the local variation. Then there is also the Park Cool Island effect. All of this is covered in a fair amount of detail in the paper by Peterson. I believe it has become a classic.
In any case, pointing out literature which shows and details an Urban Heat Island effect isn’t sufficient for criticizing the estimated trends in temperature. One needs to show that the features of UHI have not been accounted for, that they are not cancelled out by various corrections made to the data in combination with the law of large numbers, etc.. But oddly enough, climatologists at NASA GISS seem to keep up with the literature and know what needs to be accounted for, typically. I suppose this might be because it is their job.
Re # 37 Hank Roberts: “Can we assume the journal’s peer reviewers approved the online preprint version? I am not sure why they differ.”
I don’t know about J of Geophysical Research, but here is Science magazine’s policy on online rapid publication (Science Express):
“Each week, Science selects several papers for rapid online publication in advance of the scheduled print publication date. These papers are published essentially as supplied by the authors, with minimal copyediting by Science; a fully copyedited version appears later In print.” http://www.sciencemag.org/about/authors/prep/gen_info.dtl#express
Over-fitting: Re. Rasmus, Ross McKitrick @34, Jean @39,
Although I have reservations and criticisms of the model generally, I think the concerns about over-fitting are over-done. Over-fitting usually results in models which follow data noise well but have large errors on parameter estimates and poor predictive precision. It will not however have much influence on the global test of the null hypotheis, which is stated in the abstract as:
“…the null hypothesis that the spatial pattern of temperature trends in a widely-used gridded climate data set is independent of socioeconomic determinants of surface processes and data inhomogeneities.”
I would think that if you wanted to check for the accuracy of global temperature trends calculated by means of ground-based observations using satellites, you would calculate the the global average trend using satellite measurements, e.g., Pathfinder, then compare. The same would hold for latitudinal averages.
Bringing in additional factors such as economic activity, etc. simply overcomplicates the math and introduces more opportunities for losing or distorting a signal that can be obtained in a fairly straightforward fashion. But I suppose this is what is meant by “over-fitting.”
Would it be not easier to disprove the effects of UHI’s by simply using Antarctica, Arctic data as well as every remote location on Earth? There is no argument in these stations about economics, unless they have been massively paved for no reason. As a pure form of temperature trends there are none so potent. Being in a very remote location, I already know that there is a strong warming, Most people need not be knowledgeable about the real finer climate details which get marred in mud on purpose, the more compelling arguments are simple and straightforward.
Like November 2007 NASA GISS Latitude analysis, North of 60 North was the strongest warming
How many big cities contributed to that result? BTW With Nov 07 in , 2007 is the warmest year in history for the Northern Hemisphere, a lot of this warming was well away from mega cities.
In the past few weeks, RC has offered criticism of Scafetta, Loehle and McKitrick. As the comments poured in, I found my eyes rolling at the cheerleading of the RC groupies. The most valuble responses came from the authors themselves, defending their work and responding the the criticism.
So my thought is this, when doing such a post, wouldn’t it be most valuble to invite the author to respond, and close the comments to all others at least for a little while. I would love to see a little back and forth between author and critic. Then the comments could be opened to the RC admiration society for some typical back slapping and piling on.
Perhaps you can make useful contributions in showing where the RC team went wrong, instead of the typical “RC groupy” attacks. The 3 works were bad science, and in my opinion show a breakdown of peer review. To say the least, they were not defensible; to say the worst, I think people out there are concerned only with forwarding certain notions and will do what it takes to get the madvanced (like up the solar contribution, up the UHI contamination, up the MWP) I would almost like to see an article on the peer review process and what is going on now.
RealClimate is doing a great job here. Other blogs are dedicated exclusively to bashing other peoples work and generally by people who do not sit on the mainstream of that work. Scientific-sounding but content-free material going around the internet is probably not a good thing, and I thank RC for going over the material.
In the past few weeks, RC has offered criticism of Scafetta, Loehle and McKitrick. As the comments poured in, I found my eyes rolling at the cheerleading of the RC groupies. The most valuble responses came from the authors themselves, defending their work and responding the the criticism.
So my thought is this, when doing such a post, wouldn’t it be most valuble to invite the author to respond, and close the comments to all others at least for a little while. I would love to see a little back and forth between author and critic. Then the comments could be opened to the RC admiration society for some typical back slapping and piling on.
As someone who you would probably regard as one of the groupies, while I may not like the way you put things all that much, you have brought up an interesting idea. A discussion between the authors and the contributors could form the backbone of later discussion, giving it more structure and making it less likely to veer off into unrelated topics once the “groupies” and “anti-groupies” come in.
On-topic there is a better chance for a process of discovery. Offtopic things can easily turn into bullsessions. It would provide a better opportunity to learn — and I think the caliber of discussion that would be possible may be something that the contributors would enjoy.
But obviously this will be up to the contributors and the authors. Perhaps letting the authors know that their papers will be discussed either way will mean they will be more likely to attend.
1. Got any nice pictures of the thermometer clear of the tarmac?
2. re
“Other measurements from nearby sites, such as Ny Ålesund, Sveagruva, Hopen & Bjørnøya, show similar warming as Longyear byen.”
- Is this measurement data available online?
3.re
“There is no economic activity near these sites, except for at Sveagruva.”
- Got any nice pictures?
[Response: The measurements are being done very carefully. High-quality stuff. -rasmus]
“Cheerleading?” “Groupies?” “Admiration society?” “Backslapping?” “Piling on?” I think Real Climate is being confused with another site with “climate” in the title.
he most valuble responses came from the authors themselves, defending their work and responding the the criticism.
You assume a couple of things that have not been proven to be accurate.
1. Those commenting aren’t competent to do so. There are several physicists and at least one professional statistician here who are clearly competent to comment, and whose comments aren’t simply “cheerleading”.
2. You assume the authors of these denialist pieces are honest, and therefore are open to honest discourse. [edit]
There was a great thread over in Tamino’s blog … [edit]
“Cheerleading?” “Groupies?” “Admiration society?” “Backslapping?” “Piling on?” I think Real Climate is being confused with another site with “climate” in the title.
I agree it was put in a rather derogatory fashion, and likewise it isn’t very descriptive of what goes on here. But the author may have had a good idea nevertheless — despite himself. Different objectives require different methods and different approaches. I believe being more systematic and more methodical — perhaps even going section by section — might serve ours. One slight modification might be in order, though: limited, non-leading questions from ordinary participants. Just a thought.
Re 51
I didn’t say they went wrong, I just think it would be best for the authors to be the first ones to address the criticism.
Re 52
Backbone and structure would be great. BTW groupie isn’t necessarily a bad thing, way back when I was playing Rugby, I loved the groupies.
Re 53
If they don’t respond, open it up.
Re 55
I didn’t assume anything. Yes, my eyes often roll when reading CA. They have groupies and cheerleaders as well. You resent the groupie characterization, but you freely to throw out denialist and liar. We are all entitled to our own opinion.
I know the RC scientists can handle a one on one, I am sure the authors in question can handle it, and I think that a brief closed discusion would be very enlightening. Then open comment.
I have found a very informative site with hundreds of articles on every phase of climate change in — sciencedaily.com, a great adjunct to real climate and reports stretching back for many years. Your contributers ought to check this one out.
As it is the correlation between the surface and the lower troposphere (TLT) records are ~90%, but even though the areas compared (5o x 5o) are quite large, I’ll bet that a more sophisticated analysis, for example, using some sort of weighted averaging for neighboring boxes would be better.
One’s first thought on the surface ground correlation would be that it should be high, because of a near radiative equilibrium between the ground and the lower troposphere (they are exchanging energy rapidly by radiation), which leads to the further thought that this equilibrium would be perturbed by clouds, and that unless the effects of cloud cover is controlled for the whole thing is spinach.
The Svalbard Luft record is plotted here as monthly anomalies, alongside another older, but rather distant, record. Read the y-axis scale…
I guess you are right. I had hit one, and silly me, I thought that was the only one. But in an extreme climate you would undoubtedly have more than a few. And the airport would be a great place for recording temperatures, far removed from the rest of town.
Having now perused the article and digested its contents and methodology, I had the following very general thoughts:
1)The very complexity of the model seems to make it almost inevitable that spurious correlations will develop in the data. Science is replete with admonitions to avoid unnecessarily complicated models-from Occam’s “I will not multiply causes…” to von Neumann’s “Give me 4 parameters and I will fit an elephant; five and I will make him wiggle his trunk.” It is crucial to ensure that the added complexity actually adds information and doesn’t just become an exercise in “curve-fitting”. Climate models do a very good job of constraining their forcings and parameters independently of trends they are trying to model. I don’t see much “theory” guiding this model and the types of correlations it is seeking.
2)An example of the type of spurious correlation that concerns me is this: We know that the industrialized countries are disproportionately in the north of the Northern Hemisphere. This is precisely where we expect to see the most economic activity and where climate models also predict the most warming (due to geographic features–e.g. greater land mass, polar amplification…). Could it be that the article merely rediscovers this well known fact?
3)It is a mistake to lump all “economic activity” together. Deforestation and Reforestation are both economic activities, but presumably have opposite effects. Economic growth fueled by manufacturing would presumably have a different signature than economic growth fueled by e-commerce…
4)To suggest that warming is half of current estimates is actually quite surprising. As pointed out above, this would put it below estimates for the Oceans–physically unreasonable. Also, we know we are starting to see signs of significant feedbacks–saturation of the Ocean’s ability to take up CO2, outgassing from melting permafrost, and these are occurring in areas that are far from economic development.
5)As emphasized repeatedly, GHG forcing is pretty well constrained. If we’ve seen only half the predicted warming, where’s the rest of it. Is physics wrong? Is it being masked by some other factor? If so, will it kick in with a vengeance at some future date? Since I don’t think physics is wrong, I would point out that even if M&M2007 were correct, the implications might be much more alarming than reassuring.
Finally a response to “Lucky”: Like it or not, Lucky, this is how science gets done. You throw your ideas to the wolves of the community. Some of them wind up as “wolf muscle”. Some wind up as wolf crap. The measure is found in the subsequent influence they have. M&M2007 is unlikely to be cited in very much future work, primarily because it is not that useful. There are flaws in the methodology and validation that call the results into question. More importantly, though, it makes the claim that data are contaminated but does little to indicate the nature or origin of those contaminations or what to do about them. Had their goal been to shed light on the issue they purport to treat, I suspect a less ambitious project might have had more influence. Instead, I think their goal is to say the problem doesn’t exist. However, as pointed out above, even ef their contention were right (and this is doubtful), its implications could be grim rather than reassuring.
Re # 37 Hank: “Can we assume the journal’s peer reviewers approved the online preprint version? I am not sure why they differ.”
I tried responding to this yesterday, but the post got lost in the ether (AGW denialists and skeptics take note: Even RC groupies sometimes don’t get their comments posted):
I don’t know about the Journal of Geophysical Research, but here is an excerpt from Science magazine’s policy on rapid online publication (Science Express):
“Each week, Science selects several papers for rapid online publication in advance of the scheduled print publication date. These papers are published essentially as supplied by the authors, with minimal copyediting by Science; a fully copyedited version appears later In print. …”
If the JGR policy is similar, this could explain the discrepancy you noted.
In reply to #37, Yes, that is the approved version.
#50: I suspect the readers would appreciate such an exchange as well. It is essentially what happens when a comment is submitted to a journal, which remains the most appropriate way to address technical challenges.
For those who are convinced that the paper and its results are just wrong, wrong, wrong, you have to put your arguments into the form of a testable hypothesis. My paper takes the hypothesis that local temperature trends are independent of local economic activity and shows that it fails a test. Various speculations have been offered above to the effect that surface data are not contaminated but these test scores could nonetheless be obtained under restricted conditions. Maybe you’re right, but you’re going to need an encompassing statistical model to show it.
Ray (#63) - we control for latitude, not to mention the tropospheric trend at each latitude. Unnecessary complexity of a model does not usually lead to spurious gains in significance, it more typically leads to collinearity and loss of significance. That’s not a problem here, and we do test for spurious correlations. On your 3rd point, we don’t lump all economic activity together, we include a variety of indicators to pick up both cross-sectional and rate-of-change effects. You seem to be objecting that the model is both overspecified and underspecified.
Bruce (#44) - If the problem is omitted variable bias it should be easy to prove. Raising the mere possibility of it doesn’t make for much of a counterargument, since any regression model could suffer from it, and you can’t prove its absence. The IPCC suggested (Chapter 3 page 244) that the correlations are due to naturally-caused coincidence:
McKitrick and Michaels (2004) and De Laat and Maurellis (2006) attempted to demonstrate that geographical patterns of warming trends over land are strongly correlated with geographical patterns of industrial and socioeconomic development, implying that urbanisation and related land surface changes have caused much of the observed warming. However, the locations of greatest socioeconomic development are also those that have been most warmed by atmospheric circulation changes (Sections 3.2.2.7 and 3.6.4), which exhibit large-scale coherence. Hence, the correlation of warming with industrial and socioeconomic development ceases to be statistically significant. In addition, observed warming has been, and transient greenhouse-induced warming is expected to be, greater over land than over the oceans (Chapter 10), owing to the smaller thermal capacity of the land.
So you could add controls for AO, NAO, PDO etc to my statistical model and — if the IPCC is right — the socioeconomic effects will vanish. Or maybe Eli is right (#61) and it’s all due to cloud cover.
But then again, maybe not. And considering what rides on this data set not being contaminated, I hope the practitioners in the RC audience will agree that the issue deserves some serious attention rather than just casual dismissal.
[Response: The seriousness of our attention goes in inverse proportion to how the authors spin their results. In this forum, you are all about the investigation and understanding, yet in the National Post op-ed you instead claim that the surface temperature rise is “an exaggeration” (no ifs, no buts, no caveats about the existence of other possibilities) and that the IPCC “concedes … that … its main data set is contaminated”. This is completely untrue. I would suggest that your hyping of this result is a big disincentive to other researchers taking your hypothesis seriously. - gavin]
Hank, I’m afraid we won’t know until they actually publish. The “date received” and “date published” might provide a quick indication, as a long lag may indicate that the article went through significant revision. However, I’d be surprised if there were major substantive differences. The suggestion that the new references support the contention of the 2004 paper struck me as a little bit stretched.
#65 in addition to Gavin’s comments. What about the far North? Are +10 C monthly anomalies an exaggeration or an error? And the over all not so small Polar temperature trends a mistake? The push to claim GT temperature trends as a UHi mistake completely falls apart with data from remote stations.
Why not look at remote stations data alone? The case will be closed if it was so.
I myself wouldn’t be at all surprised if there is a correlation between local economic activity and local temperature trends. My question is, “How would you determine the direction of causation?” Does economic activity affect apparent trends in temperature, or do actual trends in temperature affect economic activity?
We are dealing with climates that are barely habitable in the latitudes under consideration. Furthermore, if all we were concerned with was a static economic activity, this would not result in a higher apparent trend in temperature. To have a higher apparent trend in temperature as the result of some Urban Heat Island effect, one has to have economic growth. But since we are dealing with arctic regions, any small increase in temperature will make possible considerably more economic growth, e.g., growing broccoli and strawberries in Greenland.
Consider the following…
Theory: in subarctic regions, the rate of increase in economic activity will show a strong positive correlation with the rate of increase in temperatures as higher temperatures decrease costs and make available more resources, e.g. days in the growth season. Null hypothesis: no such correlation exists. Test: check for correlation. Result: a strong correlation exists.
Question for Ross McKitrick: how does one distinguish between the theory that actual higher temperatures result in increased economic activity vs. increased economic activity resulting in spuriously high temperature readings?
I hadn’t read McKitricks Op-Ed before I made my comment #41 and see now that it contains some loaded, or non-objective,words and phrases, such as the word manipulations in the first paragraph, referring to the temperature graph at the end of the op-ed. Later on he refers to the “biases of their lead authors” of the IPCC. Hardly an objective or good faith criticism.
I know that both sides refer to the word contamination when referring to unadjusted data, which I think is too strong and gives a wrong impression. Contaminated connotes impurity or infectation. When economists make statements like ‘the price of gas in 1995, adjusted for inflation to 2007 $’ no one infers that the 1995 data is infected. When correcting fathometer soundings,many years ago, for the purpose of making coastal charts, by taking nansen bottle casts to compensate for temperature and salinity adjustments for the speed of sound in water, we didn’t consider the raw data as infected. Astronomers correct for lens and atmospheric effects all the time.
Almost all raw data require known corrections and adjustments. Adjustments for UHI are no exception.
Partisanship is a truism in the discipline of climate science. There seems no getting away from it even from the most authoritative voices. Gavin is correct that Ross’s National Post op ed clearly represents strong views which of course makes for interesting reading, whether it is right, wrong, exaggerated or incomplete. Let the reader beware. Gavin, you are guilty of the same partisanship editorializing. Your defense of Al Gore’s movie is case and point.
There are facets in the M&M2007 paper that are worthy of further analysis and research regardless of partisan views. What has not been mentioned yet but warrants a passing acknowledgment, is the fact that Ross has made his data available for others to scrutinize and reproduce. If his analysis is wrong, what better way to prove it?
Since data quality is pretty important to science — especially experimental science, why not just scrap all these thermometers in cities and even suburbs for that matter and evenly disperse them across the Earth’s uninhabited land mass. Measurements could be sent by solar powered satellite uplink bursts so that they never need to be visited or their observations corrupted by any kind of vehicle traffic.
I suppose the answer to “why not” is $$$. Still when your main data source has to be “adjusted” for UHI effect, me thinks any final result can be obtained depending on who is doing the adjusting…
Perhaps urban/suburban ground based temperature measurement should just be thrown out until better data is available.
The ANALYSIS - which is what it is - is more likely wrong, and almost, but not quite, completely wrong, and the arguments made in this post ESTABLISH that. It’s not “his data,” either. It’s data that’s out in the public domain, and other, better studies have been done. This approach is not fraud, nor is it pseudoscience, but it’s a very low bar. Particularly pernicious is that this paper is not very different from the authors’ first, unsupported paper.
If Gavin’s defense of most of the presentation in AIT is partisan, then given that it squares with the picture of a sampling of scientists around the world, it must be a big party he’s in, and its agenda must be actual science, vs. the politicized and economic interests model of reality presented in Ian McLeod’s post.
Moreover, this represents Orwell’s “duckspeak,” unfortunately - the reflexive posting of “the data is available” where it doesn’t apply and the reflexive demand that people must replicate the analysis in order to criticize it, even though the paper itself shows signs of serious flaws.
Ian McLeod, It is not partisanship to insist on good science. M&M2007 really does little to elucidate the problem it purports to consider. It does not shed light on the nature of the contamination and the very complexity of the model makes it nearly inevitable that it would find some correlation–spurious or not. As to Gore, he is a layman who actually got most of the science right. That is to be commended. He is also alone among politicians on the global stage in his unrelenting efforts to get people to pay attention to this threat. I find the howls over Gore’s Oscar and Nobel from the political right amusing, as all a rightwing politican would have had to do to deprive him of it would be share the stage with him in calling attention to threat. And despite the fact that many on the right have acknowledged the threat, none had the courage or foresight to do so.
Ross, #63, thanks for your response. I was wondering if you could respond to my other point–the fact that if your estimates of surface warming are correct and we are starting to see positive feedbacks already, this would be cause for serious concern rather than complacency.
Also, your discussion of trying to include multiple indices for economic growth sort of illustrates my point. How do you know you are using the right indices? Without a real theoretical framework to guide you as to the types of contamination you are looking for, adding multiple indices may unnecessarily complicate the model while still not capturing important differences in regional growth patterns.
Finally, there is the question of what you expect to be done with your research. Even if we were to take at face value your conclusions, they give little indication how to reliably estimate and correct for the biases you say you see. Moreover, I don’t think it is advisable to adopt the “Don’t Worry, Be Happy” approach you seem to have taken in you Post editorial, given the unmistakable indications of significant change we are seeing independent of any global temperature estimate. The real concern here is when do we reach the point where natural ghg feedbacks overtake our own contribution, since at that point mitigation becomes pointless. It is not alarmist to be alarmed by significant perturbations to a physical system with known but ill characterized positive feedbacks.
Re David Goebel @ 71: “Perhaps urban/suburban ground based temperature measurement should just be thrown out until better data is available.”
As if that would make the warming in the Arctic and the acceleration of sea ice and glacial melting in Greenland go away (discussed in raypierre’s most most recent report from the AGU meeting).
Gavin has repeatedly addressed this very proposal in past topic comments, btw. The existing urban stations are kept precisely because they have a long, unbroken temperature record and any deviations due to UHI can be corrected for. Newly created stations would have exactly zero temperature record, which, I suppose, is pretty much what those who advocate this course desire. No record, no increase, at least not for some time off into the future.
Gavin, you seem to be making a lot about Ross’s language in an OP ed. An Op ed is like a movie. Nobody expects it to live up to the standards of peer review.
An Op ed is like the posts that Hansen makes on his personal site. [edit - quote’s wrong please check your source]
Do you want your science judged by those remarks? Obviously not. Anymore than Hansen wants his science judged by his personal remarks, and any more than Ross deserves to have his science judged by his Op-ed.
[Response: Which planet are you living on? Hansen gets judged on his personal remarks all the time. McKitrick’s brand of ’science by op-ed’ undermines every supposedly scientific statement he makes. You don’t get a pass on making stuff up just because it’s not a journal article. - gavin]
Gavin, you are guilty of the same partisanship editorializing. Your defense of Al Gore’s movie is case and point.
How is saying “The science in Al Gore’s movie is largely consistent with the consensus view of climate scientists” an example of “partisanship editorializing”?
Since data quality is pretty important to science — especially experimental science, why not just scrap all these thermometers in cities and even suburbs for that matter and evenly disperse them across the Earth’s uninhabited land mass. Measurements could be sent by solar powered satellite uplink bursts so that they never need to be visited or their observations corrupted by any kind of vehicle traffic.
I suppose the answer to “why not” is $$$. Still when your main data source has to be “adjusted” for UHI effect, me thinks any final result can be obtained depending on who is doing the adjusting…
Perhaps urban/suburban ground based temperature measurement should just be thrown out until better data is available.
You write, “Since data quality is pretty important to science — especially experimental science, why not just scrap all these thermometers in cities and even suburbs for that matter and evenly disperse them across the Earth’s uninhabited land mass.”
As Jim Eager pointed out in 73, without the urban and subrural, one wouldn’t have any real trends to speak of — beyond what is provided by means of satellite measurements. Thus even if one were to create an entirely new network, the trends produced by such a network would be lacking in statistical significance for some time. However, it is also worth pointing out that a great many corrections are made which reduce and according to more prominent analyses, efectively eliminate the Urban Heat Island effect.
There are also reasons for thinking that it is not particularly significant, e.g., Park Cool Islands. But more significantly, we do have means for determining the trend in temperatures which are largely independent of ground-based networks. In the lower troposphere, we have UAH and RSS trends, both of which are based off of Microwave Sounder data. Likewise we have Pathfinder — but this measures skin-temperatures, making it closer to the surface than so-called ground-based measurements, and while I know that the Pathfinder trend for 1980-2002 was 0.43 C/decade, I do not know how this should be compared to ground-based measurements.
Independently of the models, it would be natural to assume that the trend in the lower troposphere would be roughly the same as surface measurements. However, models would project a lower troposphere trend which is 1.3 times higher that that of the surface. Given ground-based measurements with a trend of 0.187 C/decade for January 1982 to December 2004 and model projections that the trend in lower troposphere temperatures should be 1.3 times the trend in surface measurements, one would expect a tend in the lower troposphere of 0.2431 C/decade. This compares quite favorably with the RSS trend of 0.239 C/decade, differing by 1.7%, which I would assume is well within the range of expected statistical error.
However, it does not compare quite so well with UAH with its trend of 0.163 C/decade for the same period — assuming that models are correct. In fact, it suggests that ground-based measurements are inflated by roughly 49%, or else that models are wrong and that surface measurements are roughly 15% above lower troposphere measurements, or that somehow both the models and surface temperatures are wrong.
It is worth noting that UAH has had a troubled history. For example, in 2005, it was discovered that John Christy’s algorithm used for processing the Microwave Sounder data was incorrectly adjusting for the difference between night and day. Nevertheless, in some way that still is unclear to me, Ross’ study employs UAH. But I do not know whether their analysis includes the correction for the difference between night and day.
I suppose the answer to “why not” is $$$. Still when your main data source has to be “adjusted” for UHI effect, me thinks any final result can be obtained depending on who is doing the adjusting…
Multiple organizations act as a double-check, for example, there are the people at NASA GISS and those at Hadley MET, as well as a great deal of literature devoted to the analysis of data. If there are discrepencies, one looks for the source of these discrepancies.
In contrast I have noticed that Patrick Michaels, Ross McKitrick, John Christy and Stephen McIntyre all belong to the Exxon-funded George C. Marshall Institute, so you might find it preferable to leave out motives and instead begin with the assumption that work is being done in good faith.
One point that concerns me is whether — when considering trends in economic development — you are considering them at the local or national levels.
It would seem that any cause of a spurious trend in the readings of a given ground-based station would be the result of local economic development, not national. Particularly if the cause of the spurious trend were the Urban Heat Island effect. So when considering economic development, it would seem appropriate to compare economic development at the local level rather than the national level.
Furthermore, it would seem that an increasing Urban Heat Island effect would be required to produce a spurious trend in temperatures rather than a one-time distortion. This too would be a function of local economic development, not economic development at the national level.
Then I saw the socioeconomic variables which you claim are correlated with trends in apparent local temperature:
g = GDP/ million square kilometers
e = education
p = percent population growth
m = percent growth in real average income
y = percent growth in real national gross domestic product
c = percent growth in coal consumption
pg 47
Quantifying the influence of anthropogenic surface processes and inhomogeneities on gridded global climate data
Ross R. McKitrick & Patrick J. Michaels http://www. uoguelph.ca/~rmckitri/research/jgr07/M&M.JGRDec07.pdf
Furthermore, you state:
Our data set has a low resolution for strictly local measures of economic density within countries. Since we detect significant effects on temperature trends even with low spatial resolution we conjecture that if future studies are able to examine the issues at the subnational level, even more significant and detailed results will emerge.
pg. 38
These would appear to be at the national level. As such I find it difficult to see how this would tend to explain trends in temperature by way of an Urban Heat Island effect — which is necessarily limited in nature.
Likewise, given the fact that the former Soviet Union has an economy largely in disarray, whereas Europe, Canada and the United States are doing comparatively well over the period from 1979 to 1999 with the former Soviet Union experiencing weaker warming would produce a substantial amount of the correlation which you see. And as such, if one were able to explain why the trend in temperature over Siberia is weaker than the trend throughout much of the rest of the subarctic, this would be in essence an alternative explanation of the very same “correlations” uncovered by your analysis.
Furthermore, given the wide variety of economic measures and a limited number of countries in the upper northern latitudes, it would seem fairly easy to select a combination of economic measures for which there would exist spurious correlations. Particularly if no causal explanation of the relationship between these economic measures and the trend in temperatures is given or required.
So at this point I have to ask:
Do you have any sort of causal explanation of the correlation relationship between the economic measures you’ve selected and the trends in temperature?
#70, equivalency is the lowest form of argumentation. McKitrick’s paper arrives at some pretty heady conclusions that if valid would more or less upend the state of the mainstream science, (something I gather would not upset him). Despite this, and despite the truism that the making of groundbreaking discoveries is more often an indication of mistake than virtuosity, he doesn’t see fit to investigate the former possibility through the normal channels. Quite the contrary, he promptly shouts his results from a mountaintop to the public at large (and the PR machine eager for these types of results), as the man says, without ifs, buts, or caveats about the existence of other possibilities.
Meanwhile, Gavin- who has, at least informally, consulted Al Gore as to the state of the science if I have my facts right- defends AIC’s treatment of the then mainstream science. Just fyi, even disregarding the difference in context and character of these instances, and even allowing that each equivalently reveals an opinion about the veracity of the mainstream science, they are hardly equivalent. ‘Partisanship’ and ‘bias’ are not a binary tests of validity. To illustrate a biased journalist, which is to say any and all journalists, can write an article that reflects a genuine effort to cover the story and wherever that takes him or her, (plenty such instances exist), or a work of hackery bent on distortion, (and, certainly, everything in between). It goes without saying, any two such articles are not equivalent despite the existence of an opinion that shapes the story to a greater or lesser extent in both cases. Anyone attempting to short circuit a judgment about the relative validity of work produced by ‘partisan’ or ‘biased’ authors by inane citation of the corresponding existence of a point of view is doing everyone involved a disservice, not to mention playing their part in the stunting of human discourse.
Speaking of which, I find it interesting that in the comments here McKitrick has not seen fit to explain why we might expect to see UHI effects in the satellite record, what accounts for, given his titanic findings, large high latitude temperature anomalies, or the disconnect that his manuscript implies in the terrestrial and oceanic temperature records, to name a few. Of course, were interested in those questions, they probably would’ve occurred to him before he published his paper, not least before that paper became the foundation of a message meant to influence public opinion.
Steven Mosher, A scientist must be especially circumspect in his statements to a lay audience about scientific matters. Looking at the Post editorial, it is hard to reach any other conclusion but that the results are being vastly oversold–and to a gullible audience. This is improper to say the least. James Hansen’s occasional allegedly intemperate remarks concern politics and policy, not science. I presume you would not deny him a right to his political opinions just because he is a scientist.
) 
9 December 2007 at 10:39
I’m going golfing in Pennsylvania in December and will be going in January also. So before you educated readers tell me there is a difference in weather and climate and before you go back eating your chinese and toasting to the betterment of the human condition, let me say the land masses are heating up and will be the demise of most land dwelling creatures including people. The human condition is that people are reactive, as the state of the world shows today. In a very short while, perhaps 300 years, land masses will be in the 110 degree range, and there is nary one thing we can do about it. Go back and observe, analyze, and make predictions, but do it as a Don Quixote, because we are not getting out of this one.
9 December 2007 at 10:57
Economic activity is rooted in the now global capitalist push for both the desire for energy and desire for profit, deadly bedfellows for the planet, and mankind lacks the collective proactive skills to do anything different than buy and burn fossil fuels until we all die. Are temperature trends affected by economic activity? decidedly, yes. The question should be Is economic activity causing temperature trends, and we all know the answer to that.
9 December 2007 at 11:10
Well, if by “econmic activity” they mean consumption of fossil fuels and forests, economic activity has a conspicuous effect. And the economic slowdown expected by many would be a relief. Moreover, or so it seems to me, if they divorce the economy from consumption of fossil fuels and forests, they risk irrelevance on economic or ecological grounds.
9 December 2007 at 11:32
The probability that RC will comment on a paper being actively discussed elsewhere is approaching unity.
9 December 2007 at 12:07
re Svalbard:
Economy:
Economic activity centres on coal mining, supplemented by fishing and trapping. In the final decades of the 20th century, tourism, research, higher education, and some high-tech enterprises like satellite relay-stations grew significantly. A 200 nautical mile (370 km) Fisheries Protection Zone around Svalbard was established in 1977 pursuant to the Act of 17 December 1976 relating to the Economic Zone of Norway. Despite recent discussions, Russia and Norway dispute their maritime limits in the Barents Sea and Russia’s fishing rights beyond Svalbard’s territorial limits within the Svalbard Treaty zone.
The Svalbard Undersea Cable System which started operation in January 2004 provides dual 1440 km fiber optic lines from Svalbard to Harstad via Andøy, needed for communicating with polar orbiting satellite stations on Svalbard, some owned by the National Aeronautics and Space Administration (NASA) and the National Oceanic and Atmospheric Administration (NOAA), both United States government agencies.
The Norwegian state-owned coal company employs nearly 60% of the Norwegian population on the island, runs many of the local services, and provides most of the local infrastructure. Coal production has increased significantly over the past 10 years, rising from less than 500,000 tons in 1994 to over 2,500,000 tons in 2004.[6]
Exploration for oil and natural gas is underway.
Coal mining in Svalbard:
The Ny Ålesund mine was closed down in 1963 after an explosion in 1962 when 21 lives were lost, and has since been converted to a scientific post.
As of 2006, there are three operational coal mines in Svalbard. There are large mines in Sveagruva (production 2 million tonnes per year,[13] and Barentsburg, while the small mine in Longyearbyen is used mainly to supply the town’s own power plant.
Demographics:
Svalbard has a population of approximately 2,400 people as of 2005. Approximately 70% of the people are Norwegian; the remaining 30% are Russian, Ukrainian and Polish.[citation needed] The official language of Svalbard is Norwegian. Russian is used in the Russian settlements, but formerly, Russenorsk was the lingua franca of the entire Barents Sea region. The annual population growth is -0.02%
9 December 2007 at 12:32
Economic activity obviously has no direct effect on climate. Some things related to economic activity, such as urbanization, energy use, and pollution, do have effects on climate, both local and global, but economic activity in and of itself has no effect and it’s just silly to state that it does.
Economic activity means money changing hands (if we’re talking about GDP), and a few electrons, piles of paper, or sacks of gold changing hands are not going to change the climate.
At best, it’s sloppy use of language. At worst, it makes the results meaningless by not distinguishing between agriculture and coal plants (high climate effect/GDP) and a financial district (minimal climate effect/GDP).
PS: It appears that this paper is wrong on plenty of other counts, as mentioned in the main article.
9 December 2007 at 12:35
Re 4. Indeed. This site has been losing its innovative edge (and some of its contributors), and has deteriorated into a column commenting other people’s work. This is not what the headline promises: Climate science by climate scientists. Furthermore, the choice of topics gives the impression that the so-called climate contrarians are the only people finding out anything interesting (true or not) about climate and its change.
9 December 2007 at 12:37
You said: “I think it is difficult to argue that factors such as the urban heat island effect plays an important role here”
Consider the fact that many of the weather stations are located within the ever expanding “Heat Islands” mentioned. It would seem logical that the temperatures taken over time would reflect a corresponding rise with the growth of the urban heat islands.
Respectfully,
Joe Alderman
9 December 2007 at 13:53
What’s with the RealClimate bashing in 4 and 6? This post exemplifies what’s best about RC: it’s educational and directly on-topic. This caviling borders on trolling.
9 December 2007 at 13:55
Re: #4 (Steven Mosher), #6 (Dodo)
Don’t blame RC for sidetracking the discussion from real climate science to bogus pseudo-science. Place the blame squarely where it belongs: on McKitrick, Michaels, and Loehle.
The non-stop stream of sloppy research in order to discredit genuine climate science makes it necessary for RC to set the record straight. Keep up the good work!
9 December 2007 at 14:03
Rasmus : They find that the greatest differences between measured and adjusted trends at Svalbard and other places in the Arctic and Antarctic (See marked sites in Figure below). This is not convincing. Thus, the results themselves provide examples of spurious values obtained by their analysis. Even if they were identified as ‘outliers’ (Svalbard was apparently not one), the fact that their analysis produced highest corrections for economic activity at these places suggest that their analysis is not very reliable.
I’m still reading M&M2007, but it seems they study two kinds of “artifact”: one due to local anthropic effect on measured data, other due to inhomogeneities. Could the Arctic and peri-Arctic biases be mainly related to observationnal difficulties (inhomogeneity) rather than economic activity as you suggest here ? There are probably few meteo. stations north to 60°N, so a lot of interpolation.
For example, I empirically observed that on Nasa Giss (Gistemp), when you smooth at 1200 km, 2007 is bit warmer than 1998, but when you smooth at 250 km (no data when no covergage), that’s the contrary. The main differences between the two estimates arise from Africa and… peri-Arctic, precisely (70-90°N).
http://data.giss.nasa.gov/gistemp/
[Response: Good question. If you want to look at trends over a longer period, inhomogeneouity is definately a problem, but is has not been a problem since 1979. THe very strong recent warming is also supported by bore holes on Spitzbergen. -rasmus]
9 December 2007 at 14:17
I think it’s also worth pointing out that their analysis assumes a priori that any correlation with economic activity must perforce be related to a contamination of the surface temperatures by urban heat effects. This is certainly not the the only possibility, and the fact that they have apparently discovered urban heating in the satellite trends as well should have alerted them to this fact.
For instance, tropospheric ozone and black carbon emissions have local forcing components that are closely related to local emissions. Large scale land use similarly. The test of whether these factors produce correlations like those reported by M&M is available in the IPCC AR4 model archive. The fact that in 3 years since they first started this analysis, they didn’t once take a model simulation over the same period and do the same calculation is telling. This is of course in addition to the rather poor statistical significance alluded to by Rasmus. Checking their conclusions with random 25 year sequences from a model control run would have been a good test of the robustness to climate variability. Again something they didn’t apparently think of.
9 December 2007 at 14:17
Another point : this paper of Hinkel et al. showed that UHI effect can be huge in peri-Arctc areas too : at Barrow (Alaska), 71,3°N, during the winter period, the urban area is 2.2 K warmer than the rural area.
Hinkel K.M. et al. (2003), The urban heat island in winter at Barrow, Alaska, International Journal of Climatology, 23, 1889-1905.
9 December 2007 at 15:34
re: #11
The paper under discussion is outside my areas of expertise, but I get the impression that the author primarily focused on data from the physical world. Getting a handle on the quality of that data seems to be an issue.
For any given piece of research there are basically an uncountable number of things that were not done; some of more importance, some of less. Let’s make a list of what apparently has not been thought of with respect to, say, the NASA/GISS ModelE computer code. Many would say that because of what apparently has not been thought of, the ‘data’ from the code has no quality and is useless.
[Response: Your argument can be applied to anything and therefore nothing has quality or usefulness. Including your argument. Which is then self-contradicting. Think of it this way instead. All analysis is incomplete in some way. Thus conclusions should always be preliminary. As more analysis is done, the preliminary conclusion becomes stronger or falls. Writing an op-ed declaring that M&M proves that the surface record (and satellite record!) are contaminated by UHI effects is rather putting the cart before the horse, don’t you think? Especially, since most of the ideas for further analysis were already put to them in Rasmus’s original comment in the journal, and are pretty obvious in any case. If you have some obvious things to test with ModelE or new analysis you want to do, go ahead. That’s why the code is public. - gavin]
9 December 2007 at 15:35
It is important that RC stay current with those such as McKitrick who believe that there are aspects to the science which are overlooked. RC represents the mainstream and as such is under continual attack. I often come here looking for specific refutations of contrarian views, especially when (as often) those views undermine orthodoxy.
RC should present current work and latest understandings as well, of course, but it is simply necessary to, sometimes, comment on other work, especially when that work gains a bit of traction.
We don’t see RC wasting much time disputing obvious denier rhetoric, but when a paper represents a seeming legitimate effort to do science, I (and I’m sure others) want to know what RC thinks.
9 December 2007 at 15:48
I am unable to pull up the original papers (subscription required, etc.)
Is the “Michaels” in the M&M2007 paper Patrick J. Michaels at the University of Virginia?
Thanks!
(And, FWIW, I also thought the analysis by Rasmus was interesting and useful — Thanks!)
[Response: Yes. FYI there is a preprint on McKitricks’ site. - gavin]
9 December 2007 at 15:54
At the center of this discussion is the explicit role of the GHGs in the observed warming. The terrestrial data likely remain contaminated to some degree by factors related to land use change, including vegetation change, urban heat island, changes in wetland distribution, etc. I have no doubt that due diligence has been exercised to remove the effects of the urban heat island from the terrestrial record (this has been discussed at length elsewhere on RC and other forums), but it is reasonable to conclude that artifacts related to economic activity likely remain. In light of this probability, which data should we use to validate the forcing effect of the GHGs? What are the implications of using only the ocean record?
I am an academic ecologist, and I speak and write regularly about land use change and climate change. While I have a better understanding than most on issues related to climate sensitivity, I would benefit from a direct, clear discussion of the relative importance of the forcing agents as they relate to the temperature record. Has the IPCC AR4 derived an incorrect estimate of climate forcings and feedbacks related to the GHGs? While it is unlikely that we can obtain complete certainty on this issue, it is important for mainstream climate scientists to continue to debate this issue in a manner that can be interpreted by rest of the academic community.
There is an ongoing tug of war between the global climatologists and those academics that seem to hail more from the meteorology groups. This debate is often contaminated by pejorative, and could benefit from more light and less heat. Within the academy there is a small, but influential and vocal group of GHG skeptics, including Pielke, Sr., M&M, Christy and others who variously argue that we are placing too much emphasis on mitigating the GHGs, to the detriment of the economy and human well being. While it is apparent that a low carbon economy could be a thriving economy, climate scientists must be as clear as possible about the explicit role of the GHGs. Pielke, Sr. has repeatedly claimed that the IPCC estimate of the forcing effect of the GHGs is in error. I urge the scientists at RC and elsewhere to develop a detailed analysis of this and deal with this issue as best they can. Please don’t refer me to an earlier post – instead let us see the most recent thinking on this issue explicitly addressing the most recent claims.
This issue is especially crucial to the construction of sound policy. I have recently returned from DC where I spoke with policy makers about the energy bill and cap-and-trade. My personal view is that a risk assessment approach is far preferable to a cost/benefit approach, and thus we should aggressively mitigate GHG emissions. Unfortunately, this is not sufficiently compelling to policy makers given that there are some credible scientists who argue that our money is better spent on other measures. It is arguable that we should also develop a portfolio of mitigation options related to compensating for changes in land use.
I look forward to seeing more discussion of this issue at RC and in the peer-reviewed literature. I agree with 8 and 9. It is essential for RC to do its best to clarify and set the record straight. Although I am skeptical of its conclusions, M&M2007 is being published in a top tier journal, and thus needs to be addressed as serious science.
Sincerely,
Stephen
9 December 2007 at 16:10
rasmus wrote in the essay:
What makes this especially ironic is the fact that we measure the levels of greenhouse gases at various altitudes by means of the increased opacity of the atmosphere to various wavelengths (i.e., “channels” in satellite lingo). This is the mechanism by means of which greenhouse gases have an enhanced greenhouse effect as their concentrations increase over time.
9 December 2007 at 16:20
and…
Complaints of this sort are right out of the creationist/IDist playbook. When scientists and mathematicians attack claims and arguments made by [ID]creationists, the response is … “look, our work is so important ‘darwinists’ have to attack it to support the ‘darwinist conspiracy’. So obviously, we’re right!”.
I suspect it’s common among science denialists of all flavors.
9 December 2007 at 17:31
Paul, your golfing experience notwithstanding, the east coast of the U.S. has actually experienced cooling. This site explains that in the FAQ section on the Michael Crichton book “State of Fear.”
As for this article, I think the author missed the point. The HADcrut dataset and analysis presumes a randomness in the data for the purposes of assessing error. Even if the dataset is smooth and continuous, if there is a non-random component (i.e. a correlation), the correlation will need to be modeled in the error bars. IF that didn’t happen, this new paper is pretty significant. I suspect the reviewers at JGR-Atmos have representation in IPCC and must believe this new paper is significant.
9 December 2007 at 17:38
This view of Longyearbyen give a slightly different perspective than the Google Earth view you provided.
http://static.panoramio.com/photos/original/2129791.jpg
Economic impact on temperature? Depends on where the thermometer is.
[Response: Thanks for the nice picture. The thermometer is located near the airport (with about two landings/departures a day), to the far left of the picture. Other measurements from nearby sites, such as Ny Ålesund, Sveagruva, Hopen & Bjørnøya, show similar warming as Longyear byen. There is no economic activity near these sites, except for at Sveagruva. -rasmus]
9 December 2007 at 17:51
Every action within an economic system has associated with it an ensemble of ghg emissions. Those emissions can be traced and accounted for in the carbon accounts of the economic unit, and specific patterns and relationships can be identified amongst the myriad of self-organising economic units and subsystems. Thus, provided one accepts the relationship between ghg and temperature, it would be fine to assume that temperature trends are affected by economic activity. Economic activity is in effect an order parameter. Interestingly, the inverse problem can be examined. Does temperature trend affect economic activity? It certainly does.
9 December 2007 at 17:59
Robert Edele #5:
The utilization of natural resources is the very basis of economic activity, therefore it IS economic activity with a direct and measurable impact on many facets of the environment, climate included,.
Money does not change hands just so the “exchangers” can have fun!
Every time I start my internal combustion engine I am taking part in econimic activity. Why? Because I have to work to buy gasoline, and in a cascade of “economic activity” a fossil fuel company has to employ and pay people to make the fuel available.
When I start my automobile, just because money is not changing hands at that instant, does preclude that activity from being economic.
We all know what is meant by the term, is it then not a bit, in your word, “silly” to argue that economic activity has no direct effect on climate?
Steve Horstmeyer
9 December 2007 at 18:18
Statistics is such a difficult and counter-intuitive subject that we and the public need the best possible assistance in countering the claims of liars. Statistics makes life too easy for liars otherwise. Since very smart and honest professors have been known to make mistakes in statistics, checking by other professors is needed. Thank you, rasmus, for making us aware of a few of the pitfalls the unwary may fall into, and the level of care required to write a good paper on the subject of climate.
9 December 2007 at 19:25
Here is a planned ‘economic activity’ which ought to have a positive impact, not only on climate, but also on the well-being of the peoples of the Sahel:
http://biopact.com/2007/12/eu-and-africa-to-build-green-wall.html
9 December 2007 at 20:36
John Norris (#19) wrote:
You can get windspeed here:
http://www.unis.no/research/geology/Geo_research/Ole/TemperatureGruvefjell.htm
Several meters per second looks pretty standard.
This has links to the information on the various climate monitoring equipment in the area:
http://www.unis.no/research/geology/Geo_research/Ole/ClickableLongyearbyenSurroundings.htm
… including the location of the thermometer.
9 December 2007 at 21:02
And satellites are SUCH hotbeds of economic activity …
And your panorama makes it clear that Longyearbyen is a small place.
Was that photo created by the surface stations photo documentation project, by any chance?
(end sarcasm)
9 December 2007 at 21:22
Could someone please tell me how this paper got through peer-review at JGR-Atmosphere? I am stunned and disheartened that something this horribly flawed is coming out of an AGU publication.
9 December 2007 at 22:30
If urban heat islands are to be discounted when looking for trends - surely they must be included when looking at the big picture.
If one looks at those classic images of The Earth at Night one can see the twinkling evidence of mankind’s presence most places. And those lights are mostly identifying the location and extent of UHIs.
Is not the heat from UHIs entitled to sit at the same table as heat from bare earth when we ask about the present global temerpature? When does the data from the UHI’s get considered - when their sweltering streets, shimmering towers and streaming chimneys cover more than 50% of the earth, or somewhat sooner?
9 December 2007 at 23:01
Re 25
The UHI effect IS accounted for, and its (rather negligible) effect is mentioned in the IPCC 2007 and other documents. It is not whether UHI is discounted or not; the question is what effect that has on the the Global mean temp, and regardless of what McKitrick says, it is not a practical amount. It is also not influencing increased ocean heat content, melting ice caps and glaciers, satellites showing tropospheric warming or strato cooling, etc
9 December 2007 at 23:19
Nigel Williams (#25) wrote:
Love the poetry, but the answer is still no.
Not when you are trying to determine global or latitudinal trends in temperature. Need to keep things weighted according to area - and cities are pretty tiny in comparison to the countryside.
*
Fortunately we don’t seem to have to worry all that much regarding urban heat islands — most of the time. As is well-known, there are various corrections made to eliminate any urban heat island effect, and they appear to have been quite successful:
Assessment of Urban Versus Rural In Situ Surface Temperatures in the Contiguous United States: No Difference Found
Thomas C. Peterson
Journal of Climate, VOL. 16, NO. 18, 15 September 2003
http://www.ncdc.noaa.gov/oa/wmo/ccl/rural-urban.pdf
*
Likewise, according to Peterson and Vose (1997) analysis cited by IPCC 2001, the long-term (1880 to 1998) rural (0.70 C/century) and full set of station temperature trends (0.65 C/century) showed that rural stations were trending slightly higher. More recently, a 1998 analysis for the long-term trends (1951-1989) rural (0.80 C/century) and full set of station temperature trends (0.92 C/century) showed urban stations trending slightly higher.
However, in both cases, the difference between urban and rural trends were not statistically significant. As such, it would appear that the Urban Heat Island effect is either negligible or corrected for when climatologists derive their trends for temperature.
Please see:
2.2.2.1 Land-surface air temperature
http://www.grida.no/climate/ipcc_tar/wg1/052.htm#2221
*
What I would be interested in is whether GISS corrects for the Urban Heat Island effect in Barrow. I would presume they do. That is standard operating procedure, I believe. And we have known that it would be a problem if left uncorrected since 1983. For the winter, not summer, since for the latter it is weak to non-existent.
Please see:
10 December 2007 at 0:30
Am I the only one who noticed that reducing the land warming trend by 50% would make it lower than the ocean warming trend??
land =http://www.cru.uea.ac.uk/cru/data/temperature/crutem3vgl.txt
ocean =http://www.cru.uea.ac.uk/cru/data/temperature/hadsst2gl.txt
given that land responds faster to solar warming “seasons” it would make some sense that land should also warm faster
also there are some cool spots that surprised me like the east coast of the U.S.A and Australia
personally I think their trying to make a case for urban heat continents..
[Response: You get the prize for being the first to mention it. Well spotted! -rasmus]
10 December 2007 at 0:53
Firstly, I’ve only had a 15 minute look at this paper, but the quality of the analysis concerns me.
In the abstract M&M say, “…we test the null hypothesis that the spatial pattern of temperature trends in a widely-used gridded climate data set is independent of socioeconomic determinants of surface processes and data inhomogeneities. The hypothesis is strongly rejected (P= 7.1×10−14 )…”
Rasmus is completely correct. They have not considered correlation between data points. If they’d doubled the number of grid points the p value would be infinitesimal.
I noticed they left out outliers - such as Arctic and Antartic “hot-spots” when fitting the model. Outliers can be a result of measurement errors or even chance results, but high lattitude “hot-spots” are clearly neither of these. They are regions of stronger than average warming trend. These outliers are most likely a sign of model failure.
I’m still coming to terms with their model, but the inclusion of many non-significant parameter estimates in Table 1 is a worry.
I suspect another big problem with the model is omitted variable bias. The rate of southern hemisphere (and tropical) warming is slower than northern hemisphere warming and that includes landmasses. But economic activity is strongly associated with the extra-tropical nothern hemisphere. As far as I can tell there is no variable in the model to account for this hemisphere effect. This effect will then be incorporated into other variables associated with hemisphere - i.e. GDP, coal production etc - resulting in a biased estimate of these effects, the so called “omitted variable bias”.
10 December 2007 at 1:43
From the abstract of the paper
Using the regression model to filter the
extraneous, nonclimatic effects reduces the
estimated 1980-2002 global average
temperature trend over land by about half.
So I’m confused - is there a global average temperature or isn’t there?
10 December 2007 at 6:40
PaulM posts:
[[Economic activity is rooted in the now global capitalist push for both the desire for energy and desire for profit]]
And yet the Soviet Union and Peoples’ Republic of China were/are the most egregious polluters in the world. Funny how that works.
10 December 2007 at 9:51
All due apologies for going OT.
On CNN this morning (around 6:30AM 10 Dec), their science reporter Miles O’Brien did a story on Al Gore at the Nobel Prize ceremony. This is a SCIENCE reporter! Instead of illuminating us on the SCIENCE (and Gore’s efforts to publicize the science) behind the Prize, Miles lead off with Chris Horner (of “Politically Incorrect Guide to GW”) denouncing Gore. I won’t repeat Horner’s trash talk, only that it was vicious ad hominem.
Why they gave this luddite corporate hack a platform in the middle of this story is puzzling, to say the least!
The piece may show up on CNN.com. This style of reporting is despicable and must be protested.
10 December 2007 at 10:08
Maybe OT, but does anyone have any idea of what kind of money it would cost in the way of funding to research and produce a paper like this? McKitrick has been recieving what looks to be rather substantial funding from the Social Sciences and Humanities Research Council to do this work (into the $100,000s).
10 December 2007 at 10:51
What I’m trying to understand might be in the post but I haven’t been able to dig it out. Sorry.
I take it that a simple summary of the paper’s contention is that climatic terrestrial temperature measurements have been and are overstated because of increases in economic activity. My question: do they contend that economic activity per se is the culprit, or that heat islands are the culprit — stemming from economic activity which tends to move things from rural to urban areas. It’s not obvious how economic activity per se can add any heat without extreme stretching of the meaning of “economic activity”, other than maybe its increase in GHG output, which would form a ludicrous circular logic. In other words does their argument boil down to simply the heat island effect? Or would they claim something else?
I’m posting this before reading the comments which might have the answer….
10 December 2007 at 11:19
Re #30 (Barton Paul Levenson)
“PaulM posts:
[[Economic activity is rooted in the now global capitalist push for both the desire for energy and desire for profit]]
And yet the Soviet Union and Peoples’ Republic of China were/are the most egregious polluters in the world.”
Not really, at least so far as GHGs are concerned: neither approached/approaches the per capita emissions of the USA, Canada, Australia, or even western Europe - though not for want of trying. With regard to the PRC, a considerable proportion of current emissions result from manufacture for export to the rich world; and more fundamentally, it is arguable that China is now an integral part of the capitalist world-system (although the USSR was not). It is the very success of capitalism in increasing economic activity that now threatens to bring about its own collapse.
10 December 2007 at 11:39
Hello Rasmus. Thank you for your comments on my new paper. Here are some responses.
Spatial autocorrelation is an issue, in principle, with any cross-sectional study. I agree with you. You should have mentioned, though, that we applied a GLS estimator with White’s HCCME terms and clustering structure built in. Adding in local spatial AC coefficients would, for many of the regions, be redundant on top of the exiting off-diagonal elements. My conjecture is it would not affect things. However, that’s no more than a conjecture. Perhaps a reader who is interested, and better than me at programming, will figure out the math to put spatial AC controls in the GLS estimator while also controlling for heteroskedasticity and clustering.
I accept your concerns about whether we used the most updated data possible. It was a large data base to put together. It’s available at my web site. If someone wants to swap in columns with newer series (making sure the definitions are consistent) then the code can easily be re-run.
I don’t agree with your concerns about over-fitting. Over-fitting becomes a detectable problem when you have a high r2 and very low t-stats. We don’t have that, and the variance inflation factors indicate that our covariates are contributing unique explanatory power.
Your paragraph beginning “I have not examined the economic data…” seems to rule out using socioeconomic covariates under any circumstances. Yes, they change abruptly at national borders. Yes, an ideal data set would have them change continuously, but discrete changes doesn’t mean a variable can’t be used in a regression model. You’ve set up a criterion where it’s Heads-disqualified, Tails-disallowed. Can you state what circumstances you would permit socioeconomic covariates for this type of test?
You raise concerns about spurious results, but we have a few tests for this, including the endogeneity discussion in Sct 4.4. Can you be more specific? Your point was rather vague.
The oceans are obviously not at issue here. Perhaps there is an issue whether data collected from ship intakes will prove to be comparable to data collected by the argo network, but that’s for others to examine.
It is not true that our discussion of the effect of urbanization and land-use change rested only on our 2004 paper. In the on-line preprint (http://www.uoguelph.ca/~rmckitri/research/jgr07/jgr07.html)
pages 4-11 discuss anthropogenic surface processes and inhomogeneities, and there are many references therein.
Yes, we used the UAH data. I will eat my toque if that choice matters greatly, but, again, the data base is on-line and others can easily check.
Is 24 years too short to extract a trend? Well, 30 years would be better. If we had 30 years and the same results emerged, would your position change? I doubt it. So maybe the point is at most a secondary one.
Re the use of MSU data. Nothing in my paper disputes the idea that GHG are infrared-absorbing, or that oxygen emits microwaves. I’m not in a position to say anything about these things either way. But your qualifier is key: “…neglecting feedback processes…” Feedback processes are pretty much what is at issue.
Your conclusion says that there may very well be some contamination of the data. I suppose this admission represents progress. But considering the importance of the data at issue, is this an adequate response on your part? I have made the case that there is substantial contamination of the data. You don’t accept my results, which is your prerogative, but if you want to argue that there is only a small contamination problem, taking into account both surface processes and inhomogeneities (i.e. not just UHI effects), you should make the case with clear empirical methods.
Oh, and bcl, this kind of research doesn’t cost much at all. I used a bit of time of one of my research assistants for part of the data assembly. Otherwise the data are free and I used software I already own. And there are page charges for JGR. I am funded by SSHRCC for a range of research projects.
[Response: Thanks for your response, Ross. I think that your model is over-fit because I think that you have not eliminated the dependence and include too many inputs without any clear/understood connection. A regression analysis will always find a combination of weights giving the ‘best’ fit. You find greatest ‘biases’ in locations far away from places such as the Arctic and Antarctic. I don’t find that convincing. -rasmus]
10 December 2007 at 12:36
Re 34 McKitrick
“the oceans are not an issue here”
If the oceans are warming, and we know that heat is going in and not going out or just being redistributed, then we know an external agent is acting on the climate system and this is unequivocal across the globe from ice sheet/glacier responses to SST/atmosphere/surface temperatures, snow cover decline, etc. I am not sure why this wouldn’t be an issue in a study going over UHI impact of global land temperatures, since one would imagine this external agent is also acting on land. One couldn’t say that the CO2+feedbacks are just acting on oceans and polar and rural regions, but not in urban areas, and that UHI makes up for this in the instrumental record.
As for the UAH data, it was not right, so shouldn’t have been used but I have no reasonable insight into how that would have effected your study. — C
10 December 2007 at 13:16
John Cross (#29) wrote:
The following might help:
According to the Pathfinder data, it would appear we have slightly underestimated the trend in the global temperature (by 0.09 C) — using ground-based measurements from 1981-1998. NCEP reanalysis of the data had shaved another 0.06 C, so it underestimated the trend by 0.15 C, according to the Pathfinder data. Not sure how statistically significant that is though.
10 December 2007 at 13:36
> It is not true that our discussion of the effect
> of urbanization and land-use change rested only
> on our 2004 paper. In the on-line preprint …
Can we assume the journal’s peer reviewers approved the online preprint version? I am not sure why they differ.
10 December 2007 at 14:00
The real question here is whether measured global warming has been exaggerated by changing human activity in the vicinity of the measuring stations.
There is no question that the measurements themselves are affected by local human activity. We attempt to adjust for this “contamination” (to use McKitrick’s term) algorithmically.
Are these adjustments correct? For purely algorithmic reasons, I have long been skeptical.
The way to test this is to do precisely the sort of study that McKitrick has done. The correlation that McKitrick has shown between socioeconomic activity and the temperature anomaly is truly startling. I’m surprised that so many here are willing to dismiss it so quickly.
I agree with Rasmus that spatial correlations would tend to reduce the confidence with which McKitrick makes his conclusions (and I simply don’t have the background to understand McKitrick’s response), but even if the conclusions are overstated, the correlation is real, and difficult to question.
Isn’t it incumbent on purveyors of global temperature data to prove that their data is NOT “contaminated”?
Doesn’t McKitrick’s analysis STRONGLY suggest that this is not the case?
10 December 2007 at 14:18
RE Longyearbyen
Does anyone know of a site with graphs of cloud cover and temperature anomalies for Longyearbyen? I’m particularly looking for records that cover the period from about 1935 to present. Anomalous high temps at west-facing near-ocean locations are my interest, and this looks like a prime candidate. All the papers I’ve googled have been behind a pay-to-view wall.
I would expect it to exhibit a rising temperature spike starting in 1939.
TIA.
JF
10 December 2007 at 15:08
re #34/Rasmus: I think overfitting due to dependence and too many inputs without clearly understood connection might be a problem also in other areas like multiproxy temperature reconstructions (many uncalibrated proxies regressed on, say, instrumental temperature PCs). Maybe you could help Ross by explaining how the problem has been avoided, e.g., in the landmark paper by Mann et al. (1998)?
[Response: Step-wise regression is often preferred, but I’d recommend a proper cross-validation. -rasmus]
[Response: Cross-validation (and objective selection rules) are the key here, as indeed emphasized by Mann et al (1998), and driven home fairly convincingly by the followup papers by Wahl and Ammann we’ve referenced above. You should read these (and obviously, re-read Mann et al 1998) to understand the issues better. -mike]
10 December 2007 at 15:46
Thanks Timothy 26.
I would have thought you needed to keep ‘things’ weighted in terms of energy content per unit area not just area alone? Thus won’t the energy of the UHIs figure rather more strongly in the overall picture?
Otherwise it just seems to me that we are looking for trends among a sample we are ever-reducing as old stations are knobbled by the UHI effect. Isn’t this obscuring how ‘inconvenient’ the truth really is about warming globally?
Yet would not the UHIs put more (in fact most) anthropo-energy into circulation, capture more insolation and (because of effects like inversions holding higher concentrations of GHGs) retain more energy locally at lower altitudes?
That there is still an upward trend to be found in the remaining non-UHI stations is useful info, but what do we see if we use all stations on say a 10km square grid (or 10″ grid or whatever) with stations within a square averaged and look at the past and present energy content (Joules) of the total atmosphere globally over time?
Isn’t that total energy content including UHIs the figure that matters because that energy is the meat in the sandwich between the energy additions to the system and the energy loss to space which in turn defines the global temperature?
10 December 2007 at 17:23
Re: #11 Response by Gavin
[……..If you have some obvious things to test with ModelE or new analysis you want to do, go ahead. That’s why the code is public. - gavin]
Is this a valid challenge? Many (Most ?) of us don’t have the expertise or the capacity( by a long shot) either in our gray matter or on our hard drives to run alternative tests on Model E or any equivalent. Also Nasa/ Giss works as a team and on company time. Do you really expect a poster working alone in his spare time to take you up on the above statement. I don’t much like the way the war in Iraq has been handled, but if somebody were to say “We’ll put you in charge - go ahead and solve the debacle”, I’m sure I couldn’t do any better.
You shouldn’t have to have the knowledge of say a four star general or an expert climate modeler in order to make good faith criticisms. Also this works both ways. Because you’re unhappy with the M&M paper, you shouldn’t have to develop your own paper on the same topic, using their method of analysis or another method of analysis with their own or your own data.
[Response: I agree. Good faith criticisms should be welcome. But McKitrick’s op-ed and declarative statements seem to have done before any of those criticisms were dealt with. (PS. you could run ModelE on your home linux box if you want, and we have a port for windows in the works- you don’t need to do that to make good faith criticisms, but if you wanted to see how it worked you could in fact see for yourself). - gavin]
10 December 2007 at 17:24
RE Nigel Williams (#40)
In principle, global average temperature over a given period would be defined as the integral over the surface area, then integrated over the time period under consideration, then divided by the product of the surface area times the time.
As stations measure temperature over a limited number of points, one will need to interpolate temperatures between stations over a given area, but in no small part this will involve removing the effects of Urban Heat Islands. Technically such an average will be refered to as being weighted by area and will make reference only to temperatures, not heat content.
Satellite measurements should be more accurate, both in arriving at global average temperatures and in identifying trends over a given period as they are able to take readings from a far larger number of points. As such when Pathfinder satellite gives a larger trend in global temperature of 0.43 C/decade but groundbased measurements give 0.34 C/decade, I suspect that the higher satellite-based trend is more accurate.
But yes, urban temperatures are important for their own sake — but they are relatively insignificant when compared to the effects of greenhouse gases, and without corrections for UHI would give a distorted picture of the rise in global average temperature. However, given the results of Pathfinder, it would appear that we have overcompensated for the effects of UHI.
10 December 2007 at 18:04
#41 Timothy, what is Pathfinder Satellite ? (The link leads to another discussion without a precise reference). If you look at UAH data for lower troposphere trends on land and NH (used by M&M 2007), you get 0,24 K/dec (0,33 K/dec CRU, 0,34 K/dec NCDC, 0,29 K/dec Giss, cf AR4 tab. 3.2 p. 243). Anyway, even if the trends were the same, I guess it’s not the M&M2007 purpose to look at global correlation, rather to assess local (grid by grid) correlations between Ts, Ttropo, economic activity.
Other point (more general) : UHI is detectable even for small towns (see Torok 2001), not just big cities, and anthropic effects on surface energy budget is not limited to urbanization.
Torok S.J. et al. (2001), Urban heat island features of southeast Australian towns, Aust. Met. Mag., 50 1-13.
10 December 2007 at 19:07
Over-fitting: Re. Rasmus, Ross McKitrick @34, Jean @39,
Although I have reservations and criticisms of the model generally, I think the concerns about over-fitting are over-done. Over-fitting usually results in models which follow data noise well but have large errors on parameter estimates and poor predictive precision. It will not however have much influence on the global test of the null hypotheis, which is stated in the abstract as:
“…the null hypothesis that the spatial pattern of temperature trends in a widely-used gridded climate data set is independent of socioeconomic determinants of surface processes and data inhomogeneities.”
This hypothesis is rejected with a p=7.1*10^-14.
My bigger concerns are with the model itself. Two of which Rasmus has stated:
1) Failure to appropriately incorporate spatial correlation of temperature measurements. This will result in exaggerated p-values.
2) Poor model fit for land areas at extreme latitudes. This is evidence of outright model failure - the model does not fit the data as it is inappropriate.
And an additional issue:
3 Omitted variable bias.
http://en.wikipedia.org/wiki/Omitted_variable_bias
This is a bias that appears in parameter estimates - such as the effect of economic activity on local temperature measurement in Ross’s model - that is due to the omission of a variable that is better able to explain the relationship between location and temperature. Such a variable might be hemispheric effects, or oceanic effects that extend beyond coastlines.
Essentially omitted variable bias is a type of poor model specification that results in biased estimates of model parameters even when the model fits the data.
10 December 2007 at 19:15
Re. Gavin @11 and my earlier comment about omitted variable bias.
Gavin’s concerns about tropospheric ozone and black carbon emissions having local forcing components is another example of omitted variable bias. A fitted model that excludes these will attribute these forcing effects to other variables in the model which have large values at the same locations, such as GDP.
10 December 2007 at 19:39
Anyone care to hazard an estimate of the percentage of human induced CO2 of the 100 ppmv that finds its way into the +.7 observed trend in the mean global anomaly?
[Response: That’s not a well-posed question. See here for a discussion. - gavin]
10 December 2007 at 20:01
Charles Muller wrote
> what is Pathfinder Satellite? (the link leads to
> another discussion without a precise reference …
the discussion includes this reference:
http://earthobservatory.nasa.gov/Newsroom/NasaNews/2004/2004042116878.html
and at the bottom of that page
> For more information and images … visit:
http://www.gsfc.nasa.gov/topstory/2004/0315skintemp.html
10 December 2007 at 20:08
Charles Muller (#43) wrote:
The link was to my earlier comment that included a relevant quote and the link to my source:
April 21, 2004
Earth Observatory: NASA News Archive
Satellites Act as Thermometers in Space, Show Earth has a Fever
http://earthobservatory.nasa.gov/Newsroom/NasaNews/2004/2004042116878.html
I simply didn’t want to duplicate the quote.
Charles Muller (#43) wrote:
Pathfinder was using skin temperature — which is actually closer to the surface than land-based.
They state:
Anyway, it wouldn’t surprise me if lower trop and ground-based were actually closer to one-another. In fact, I seem to remember they are almost the same — at least for the continental US — but I will have to check. However, I do remember there is somewhat greater variability with lower trop — the hot years tend to be higher.
Charles Muller (#43) wrote:
Certainly it is detectable. But it is also in very large part accounted and corrected for.
In 26, I had stated:
… and I gave additional sources, as well as pointed out that we knew about Barrow in 1983. Somehow I suspect the UHI at Barrow has been corrected for. In any case, there are a fair number of corrections made depending upon location, altitude, etc. There is the law of large numbers which cancels out much of the local variation. Then there is also the Park Cool Island effect. All of this is covered in a fair amount of detail in the paper by Peterson. I believe it has become a classic.
In any case, pointing out literature which shows and details an Urban Heat Island effect isn’t sufficient for criticizing the estimated trends in temperature. One needs to show that the features of UHI have not been accounted for, that they are not cancelled out by various corrections made to the data in combination with the law of large numbers, etc.. But oddly enough, climatologists at NASA GISS seem to keep up with the literature and know what needs to be accounted for, typically. I suppose this might be because it is their job.
10 December 2007 at 20:10
Re # 37 Hank Roberts: “Can we assume the journal’s peer reviewers approved the online preprint version? I am not sure why they differ.”
I don’t know about J of Geophysical Research, but here is Science magazine’s policy on online rapid publication (Science Express):
“Each week, Science selects several papers for rapid online publication in advance of the scheduled print publication date. These papers are published essentially as supplied by the authors, with minimal copyediting by Science; a fully copyedited version appears later In print.”
http://www.sciencemag.org/about/authors/prep/gen_info.dtl#express
10 December 2007 at 20:41
Bruce Tabor (#44) wrote:
I would think that if you wanted to check for the accuracy of global temperature trends calculated by means of ground-based observations using satellites, you would calculate the the global average trend using satellite measurements, e.g., Pathfinder, then compare. The same would hold for latitudinal averages.
Bringing in additional factors such as economic activity, etc. simply overcomplicates the math and introduces more opportunities for losing or distorting a signal that can be obtained in a fairly straightforward fashion. But I suppose this is what is meant by “over-fitting.”
10 December 2007 at 21:22
Re 34 McKitrick
“the oceans are not an issue here”
Being less polite than Chris, oceans provide a simple bull(you know what) test of the hypothesis. It fails.
It seems obvious to me also that MSU data, in so far as it is affected by surface temperature has to average over very large areas.
10 December 2007 at 22:20
Would it be not easier to disprove the effects of UHI’s by simply using Antarctica, Arctic data as well as every remote location on Earth? There is no argument in these stations about economics, unless they have been massively paved for no reason. As a pure form of temperature trends there are none so potent. Being in a very remote location, I already know that there is a strong warming, Most people need not be knowledgeable about the real finer climate details which get marred in mud on purpose, the more compelling arguments are simple and straightforward.
Like November 2007 NASA GISS Latitude analysis, North of 60 North was the strongest warming
http://data.giss.nasa.gov/cgi-bin/gistemp/do_nmap.py?year_last=2007&month_last=11&sat=4&sst=0&type=anoms&mean_gen=11&year1=2007&year2=2007&base1=1951&base2=1980&radius=1200&pol=reg
How many big cities contributed to that result? BTW With Nov 07 in , 2007 is the warmest year in history for the Northern Hemisphere, a lot of this warming was well away from mega cities.
10 December 2007 at 22:28
In the past few weeks, RC has offered criticism of Scafetta, Loehle and McKitrick. As the comments poured in, I found my eyes rolling at the cheerleading of the RC groupies. The most valuble responses came from the authors themselves, defending their work and responding the the criticism.
So my thought is this, when doing such a post, wouldn’t it be most valuble to invite the author to respond, and close the comments to all others at least for a little while. I would love to see a little back and forth between author and critic. Then the comments could be opened to the RC admiration society for some typical back slapping and piling on.
10 December 2007 at 23:17
Re 50
Perhaps you can make useful contributions in showing where the RC team went wrong, instead of the typical “RC groupy” attacks. The 3 works were bad science, and in my opinion show a breakdown of peer review. To say the least, they were not defensible; to say the worst, I think people out there are concerned only with forwarding certain notions and will do what it takes to get the madvanced (like up the solar contribution, up the UHI contamination, up the MWP) I would almost like to see an article on the peer review process and what is going on now.
RealClimate is doing a great job here. Other blogs are dedicated exclusively to bashing other peoples work and generally by people who do not sit on the mainstream of that work. Scientific-sounding but content-free material going around the internet is probably not a good thing, and I thank RC for going over the material.
10 December 2007 at 23:26
Lucky (#50) wrote:
As someone who you would probably regard as one of the groupies, while I may not like the way you put things all that much, you have brought up an interesting idea. A discussion between the authors and the contributors could form the backbone of later discussion, giving it more structure and making it less likely to veer off into unrelated topics once the “groupies” and “anti-groupies” come in.
On-topic there is a better chance for a process of discovery. Offtopic things can easily turn into bullsessions. It would provide a better opportunity to learn — and I think the caliber of discussion that would be possible may be something that the contributors would enjoy.
But obviously this will be up to the contributors and the authors. Perhaps letting the authors know that their papers will be discussed either way will mean they will be more likely to attend.
10 December 2007 at 23:28
re rasmus response to #18
1. Got any nice pictures of the thermometer clear of the tarmac?
2. re
“Other measurements from nearby sites, such as Ny Ålesund, Sveagruva, Hopen & Bjørnøya, show similar warming as Longyear byen.”
- Is this measurement data available online?
3.re
“There is no economic activity near these sites, except for at Sveagruva.”
- Got any nice pictures?
[Response: The measurements are being done very carefully. High-quality stuff. -rasmus]
10 December 2007 at 23:30
# 50 Lucky
What if the author in question doesn’t respond?
11 December 2007 at 0:02
“Cheerleading?” “Groupies?” “Admiration society?” “Backslapping?” “Piling on?” I think Real Climate is being confused with another site with “climate” in the title.
11 December 2007 at 0:38
You assume a couple of things that have not been proven to be accurate.
1. Those commenting aren’t competent to do so. There are several physicists and at least one professional statistician here who are clearly competent to comment, and whose comments aren’t simply “cheerleading”.
2. You assume the authors of these denialist pieces are honest, and therefore are open to honest discourse. [edit]
There was a great thread over in Tamino’s blog … [edit]
[Response: No more on this please. -gavin]
11 December 2007 at 1:52
cce (#54) wrote:
I agree it was put in a rather derogatory fashion, and likewise it isn’t very descriptive of what goes on here. But the author may have had a good idea nevertheless — despite himself. Different objectives require different methods and different approaches. I believe being more systematic and more methodical — perhaps even going section by section — might serve ours. One slight modification might be in order, though: limited, non-leading questions from ordinary participants. Just a thought.
Anyway, it certainly isn’t my decision.
11 December 2007 at 2:18
Re 51
I didn’t say they went wrong, I just think it would be best for the authors to be the first ones to address the criticism.
Re 52
Backbone and structure would be great. BTW groupie isn’t necessarily a bad thing, way back when I was playing Rugby, I loved the groupies.
Re 53
If they don’t respond, open it up.
Re 55
I didn’t assume anything. Yes, my eyes often roll when reading CA. They have groupies and cheerleaders as well. You resent the groupie characterization, but you freely to throw out denialist and liar. We are all entitled to our own opinion.
I know the RC scientists can handle a one on one, I am sure the authors in question can handle it, and I think that a brief closed discusion would be very enlightening. Then open comment.
11 December 2007 at 3:46
I have found a very informative site with hundreds of articles on every phase of climate change in — sciencedaily.com, a great adjunct to real climate and reports stretching back for many years. Your contributers ought to check this one out.
11 December 2007 at 5:18
More Svalbard:
#18, from Rasmus:
…umm, far right
#22:
No, it’s way out of shot, to the right.
The Svalbard Luft record is plotted here as monthly anomalies, alongside another older, but rather distant, record. Read the y-axis scale…
11 December 2007 at 9:20
As it is the correlation between the surface and the lower troposphere (TLT) records are ~90%, but even though the areas compared (5o x 5o) are quite large, I’ll bet that a more sophisticated analysis, for example, using some sort of weighted averaging for neighboring boxes would be better.
One’s first thought on the surface ground correlation would be that it should be high, because of a near radiative equilibrium between the ground and the lower troposphere (they are exchanging energy rapidly by radiation), which leads to the further thought that this equilibrium would be perturbed by clouds, and that unless the effects of cloud cover is controlled for the whole thing is spinach.
11 December 2007 at 9:21
GlenFergus (#57) wrote:
I guess you are right. I had hit one, and silly me, I thought that was the only one. But in an extreme climate you would undoubtedly have more than a few. And the airport would be a great place for recording temperatures, far removed from the rest of town.
Thanks!
11 December 2007 at 9:45
Having now perused the article and digested its contents and methodology, I had the following very general thoughts:
1)The very complexity of the model seems to make it almost inevitable that spurious correlations will develop in the data. Science is replete with admonitions to avoid unnecessarily complicated models-from Occam’s “I will not multiply causes…” to von Neumann’s “Give me 4 parameters and I will fit an elephant; five and I will make him wiggle his trunk.” It is crucial to ensure that the added complexity actually adds information and doesn’t just become an exercise in “curve-fitting”. Climate models do a very good job of constraining their forcings and parameters independently of trends they are trying to model. I don’t see much “theory” guiding this model and the types of correlations it is seeking.
2)An example of the type of spurious correlation that concerns me is this: We know that the industrialized countries are disproportionately in the north of the Northern Hemisphere. This is precisely where we expect to see the most economic activity and where climate models also predict the most warming (due to geographic features–e.g. greater land mass, polar amplification…). Could it be that the article merely rediscovers this well known fact?
3)It is a mistake to lump all “economic activity” together. Deforestation and Reforestation are both economic activities, but presumably have opposite effects. Economic growth fueled by manufacturing would presumably have a different signature than economic growth fueled by e-commerce…
4)To suggest that warming is half of current estimates is actually quite surprising. As pointed out above, this would put it below estimates for the Oceans–physically unreasonable. Also, we know we are starting to see signs of significant feedbacks–saturation of the Ocean’s ability to take up CO2, outgassing from melting permafrost, and these are occurring in areas that are far from economic development.
5)As emphasized repeatedly, GHG forcing is pretty well constrained. If we’ve seen only half the predicted warming, where’s the rest of it. Is physics wrong? Is it being masked by some other factor? If so, will it kick in with a vengeance at some future date? Since I don’t think physics is wrong, I would point out that even if M&M2007 were correct, the implications might be much more alarming than reassuring.
Finally a response to “Lucky”: Like it or not, Lucky, this is how science gets done. You throw your ideas to the wolves of the community. Some of them wind up as “wolf muscle”. Some wind up as wolf crap. The measure is found in the subsequent influence they have. M&M2007 is unlikely to be cited in very much future work, primarily because it is not that useful. There are flaws in the methodology and validation that call the results into question. More importantly, though, it makes the claim that data are contaminated but does little to indicate the nature or origin of those contaminations or what to do about them. Had their goal been to shed light on the issue they purport to treat, I suspect a less ambitious project might have had more influence. Instead, I think their goal is to say the problem doesn’t exist. However, as pointed out above, even ef their contention were right (and this is doubtful), its implications could be grim rather than reassuring.
11 December 2007 at 10:01
Re # 37 Hank: “Can we assume the journal’s peer reviewers approved the online preprint version? I am not sure why they differ.”
I tried responding to this yesterday, but the post got lost in the ether (AGW denialists and skeptics take note: Even RC groupies sometimes don’t get their comments posted):
I don’t know about the Journal of Geophysical Research, but here is an excerpt from Science magazine’s policy on rapid online publication (Science Express):
“Each week, Science selects several papers for rapid online publication in advance of the scheduled print publication date. These papers are published essentially as supplied by the authors, with minimal copyediting by Science; a fully copyedited version appears later In print. …”
If the JGR policy is similar, this could explain the discrepancy you noted.
11 December 2007 at 11:31
Ray, you read the full article, can you answer the question I asked way back at #37?
> It is not true that our discussion of the effect
> of urbanization and land-use change rested only
> on our 2004 paper. In the on-line preprint …
Can we assume the journal’s peer reviewers approved the online preprint version?
11 December 2007 at 11:55
In reply to #37, Yes, that is the approved version.
#50: I suspect the readers would appreciate such an exchange as well. It is essentially what happens when a comment is submitted to a journal, which remains the most appropriate way to address technical challenges.
For those who are convinced that the paper and its results are just wrong, wrong, wrong, you have to put your arguments into the form of a testable hypothesis. My paper takes the hypothesis that local temperature trends are independent of local economic activity and shows that it fails a test. Various speculations have been offered above to the effect that surface data are not contaminated but these test scores could nonetheless be obtained under restricted conditions. Maybe you’re right, but you’re going to need an encompassing statistical model to show it.
Ray (#63) - we control for latitude, not to mention the tropospheric trend at each latitude. Unnecessary complexity of a model does not usually lead to spurious gains in significance, it more typically leads to collinearity and loss of significance. That’s not a problem here, and we do test for spurious correlations. On your 3rd point, we don’t lump all economic activity together, we include a variety of indicators to pick up both cross-sectional and rate-of-change effects. You seem to be objecting that the model is both overspecified and underspecified.
Bruce (#44) - If the problem is omitted variable bias it should be easy to prove. Raising the mere possibility of it doesn’t make for much of a counterargument, since any regression model could suffer from it, and you can’t prove its absence. The IPCC suggested (Chapter 3 page 244) that the correlations are due to naturally-caused coincidence:
So you could add controls for AO, NAO, PDO etc to my statistical model and — if the IPCC is right — the socioeconomic effects will vanish. Or maybe Eli is right (#61) and it’s all due to cloud cover.
But then again, maybe not. And considering what rides on this data set not being contaminated, I hope the practitioners in the RC audience will agree that the issue deserves some serious attention rather than just casual dismissal.
[Response: The seriousness of our attention goes in inverse proportion to how the authors spin their results. In this forum, you are all about the investigation and understanding, yet in the National Post op-ed you instead claim that the surface temperature rise is “an exaggeration” (no ifs, no buts, no caveats about the existence of other possibilities) and that the IPCC “concedes … that … its main data set is contaminated”. This is completely untrue. I would suggest that your hyping of this result is a big disincentive to other researchers taking your hypothesis seriously. - gavin]
11 December 2007 at 12:16
Hank, I’m afraid we won’t know until they actually publish. The “date received” and “date published” might provide a quick indication, as a long lag may indicate that the article went through significant revision. However, I’d be surprised if there were major substantive differences. The suggestion that the new references support the contention of the 2004 paper struck me as a little bit stretched.
11 December 2007 at 12:51
#65 in addition to Gavin’s comments. What about the far North? Are +10 C monthly anomalies an exaggeration or an error? And the over all not so small Polar temperature trends a mistake? The push to claim GT temperature trends as a UHi mistake completely falls apart with data from remote stations.
Why not look at remote stations data alone? The case will be closed if it was so.
11 December 2007 at 13:27
I don’t (amateur reader here) see how this approach avoids just equating fuel use and economic activity.
This chart for example — if you didn’t have the label, what would you think it described?
http://web.whittier.edu/academic/math/jmiller/United%20States%20National%20Debt_files/usdebt1.gif
11 December 2007 at 13:52
I myself wouldn’t be at all surprised if there is a correlation between local economic activity and local temperature trends. My question is, “How would you determine the direction of causation?” Does economic activity affect apparent trends in temperature, or do actual trends in temperature affect economic activity?
We are dealing with climates that are barely habitable in the latitudes under consideration. Furthermore, if all we were concerned with was a static economic activity, this would not result in a higher apparent trend in temperature. To have a higher apparent trend in temperature as the result of some Urban Heat Island effect, one has to have economic growth. But since we are dealing with arctic regions, any small increase in temperature will make possible considerably more economic growth, e.g., growing broccoli and strawberries in Greenland.
Consider the following…
Theory: in subarctic regions, the rate of increase in economic activity will show a strong positive correlation with the rate of increase in temperatures as higher temperatures decrease costs and make available more resources, e.g. days in the growth season. Null hypothesis: no such correlation exists. Test: check for correlation. Result: a strong correlation exists.
Question for Ross McKitrick: how does one distinguish between the theory that actual higher temperatures result in increased economic activity vs. increased economic activity resulting in spuriously high temperature readings?
11 December 2007 at 17:10
I hadn’t read McKitricks Op-Ed before I made my comment #41 and see now that it contains some loaded, or non-objective,words and phrases, such as the word manipulations in the first paragraph, referring to the temperature graph at the end of the op-ed. Later on he refers to the “biases of their lead authors” of the IPCC. Hardly an objective or good faith criticism.
I know that both sides refer to the word contamination when referring to unadjusted data, which I think is too strong and gives a wrong impression. Contaminated connotes impurity or infectation. When economists make statements like ‘the price of gas in 1995, adjusted for inflation to 2007 $’ no one infers that the 1995 data is infected. When correcting fathometer soundings,many years ago, for the purpose of making coastal charts, by taking nansen bottle casts to compensate for temperature and salinity adjustments for the speed of sound in water, we didn’t consider the raw data as infected. Astronomers correct for lens and atmospheric effects all the time.
Almost all raw data require known corrections and adjustments. Adjustments for UHI are no exception.
11 December 2007 at 19:37
#65 Ross/Gavin
Partisanship is a truism in the discipline of climate science. There seems no getting away from it even from the most authoritative voices. Gavin is correct that Ross’s National Post op ed clearly represents strong views which of course makes for interesting reading, whether it is right, wrong, exaggerated or incomplete. Let the reader beware. Gavin, you are guilty of the same partisanship editorializing. Your defense of Al Gore’s movie is case and point.
There are facets in the M&M2007 paper that are worthy of further analysis and research regardless of partisan views. What has not been mentioned yet but warrants a passing acknowledgment, is the fact that Ross has made his data available for others to scrutinize and reproduce. If his analysis is wrong, what better way to prove it?
12 December 2007 at 1:00
Since data quality is pretty important to science — especially experimental science, why not just scrap all these thermometers in cities and even suburbs for that matter and evenly disperse them across the Earth’s uninhabited land mass. Measurements could be sent by solar powered satellite uplink bursts so that they never need to be visited or their observations corrupted by any kind of vehicle traffic.
I suppose the answer to “why not” is $$$. Still when your main data source has to be “adjusted” for UHI effect, me thinks any final result can be obtained depending on who is doing the adjusting…
Perhaps urban/suburban ground based temperature measurement should just be thrown out until better data is available.
12 December 2007 at 6:09
The ANALYSIS - which is what it is - is more likely wrong, and almost, but not quite, completely wrong, and the arguments made in this post ESTABLISH that. It’s not “his data,” either. It’s data that’s out in the public domain, and other, better studies have been done. This approach is not fraud, nor is it pseudoscience, but it’s a very low bar. Particularly pernicious is that this paper is not very different from the authors’ first, unsupported paper.
If Gavin’s defense of most of the presentation in AIT is partisan, then given that it squares with the picture of a sampling of scientists around the world, it must be a big party he’s in, and its agenda must be actual science, vs. the politicized and economic interests model of reality presented in Ian McLeod’s post.
Moreover, this represents Orwell’s “duckspeak,” unfortunately - the reflexive posting of “the data is available” where it doesn’t apply and the reflexive demand that people must replicate the analysis in order to criticize it, even though the paper itself shows signs of serious flaws.
12 December 2007 at 8:15
Ian McLeod, It is not partisanship to insist on good science. M&M2007 really does little to elucidate the problem it purports to consider. It does not shed light on the nature of the contamination and the very complexity of the model makes it nearly inevitable that it would find some correlation–spurious or not. As to Gore, he is a layman who actually got most of the science right. That is to be commended. He is also alone among politicians on the global stage in his unrelenting efforts to get people to pay attention to this threat. I find the howls over Gore’s Oscar and Nobel from the political right amusing, as all a rightwing politican would have had to do to deprive him of it would be share the stage with him in calling attention to threat. And despite the fact that many on the right have acknowledged the threat, none had the courage or foresight to do so.
12 December 2007 at 9:20
Ross, #63, thanks for your response. I was wondering if you could respond to my other point–the fact that if your estimates of surface warming are correct and we are starting to see positive feedbacks already, this would be cause for serious concern rather than complacency.
Also, your discussion of trying to include multiple indices for economic growth sort of illustrates my point. How do you know you are using the right indices? Without a real theoretical framework to guide you as to the types of contamination you are looking for, adding multiple indices may unnecessarily complicate the model while still not capturing important differences in regional growth patterns.
Finally, there is the question of what you expect to be done with your research. Even if we were to take at face value your conclusions, they give little indication how to reliably estimate and correct for the biases you say you see. Moreover, I don’t think it is advisable to adopt the “Don’t Worry, Be Happy” approach you seem to have taken in you Post editorial, given the unmistakable indications of significant change we are seeing independent of any global temperature estimate. The real concern here is when do we reach the point where natural ghg feedbacks overtake our own contribution, since at that point mitigation becomes pointless. It is not alarmist to be alarmed by significant perturbations to a physical system with known but ill characterized positive feedbacks.
12 December 2007 at 12:21
Re David Goebel @ 71: “Perhaps urban/suburban ground based temperature measurement should just be thrown out until better data is available.”
As if that would make the warming in the Arctic and the acceleration of sea ice and glacial melting in Greenland go away (discussed in raypierre’s most most recent report from the AGU meeting).
Gavin has repeatedly addressed this very proposal in past topic comments, btw. The existing urban stations are kept precisely because they have a long, unbroken temperature record and any deviations due to UHI can be corrected for. Newly created stations would have exactly zero temperature record, which, I suppose, is pretty much what those who advocate this course desire. No record, no increase, at least not for some time off into the future.
12 December 2007 at 13:37
Gavin, you seem to be making a lot about Ross’s language in an OP ed. An Op ed is like a movie. Nobody expects it to live up to the standards of peer review.
An Op ed is like the posts that Hansen makes on his personal site. [edit - quote’s wrong please check your source]
Do you want your science judged by those remarks? Obviously not. Anymore than Hansen wants his science judged by his personal remarks, and any more than Ross deserves to have his science judged by his Op-ed.
[Response: Which planet are you living on? Hansen gets judged on his personal remarks all the time. McKitrick’s brand of ’science by op-ed’ undermines every supposedly scientific statement he makes. You don’t get a pass on making stuff up just because it’s not a journal article. - gavin]
12 December 2007 at 15:56
How is saying “The science in Al Gore’s movie is largely consistent with the consensus view of climate scientists” an example of “partisanship editorializing”?
Statements like this make my head spin.
12 December 2007 at 16:40
David Goebel (#71) wrote:
You write, “Since data quality is pretty important to science — especially experimental science, why not just scrap all these thermometers in cities and even suburbs for that matter and evenly disperse them across the Earth’s uninhabited land mass.”
As Jim Eager pointed out in 73, without the urban and subrural, one wouldn’t have any real trends to speak of — beyond what is provided by means of satellite measurements. Thus even if one were to create an entirely new network, the trends produced by such a network would be lacking in statistical significance for some time. However, it is also worth pointing out that a great many corrections are made which reduce and according to more prominent analyses, efectively eliminate the Urban Heat Island effect.
There are also reasons for thinking that it is not particularly significant, e.g., Park Cool Islands. But more significantly, we do have means for determining the trend in temperatures which are largely independent of ground-based networks. In the lower troposphere, we have UAH and RSS trends, both of which are based off of Microwave Sounder data. Likewise we have Pathfinder — but this measures skin-temperatures, making it closer to the surface than so-called ground-based measurements, and while I know that the Pathfinder trend for 1980-2002 was 0.43 C/decade, I do not know how this should be compared to ground-based measurements.
Independently of the models, it would be natural to assume that the trend in the lower troposphere would be roughly the same as surface measurements. However, models would project a lower troposphere trend which is 1.3 times higher that that of the surface. Given ground-based measurements with a trend of 0.187 C/decade for January 1982 to December 2004 and model projections that the trend in lower troposphere temperatures should be 1.3 times the trend in surface measurements, one would expect a tend in the lower troposphere of 0.2431 C/decade. This compares quite favorably with the RSS trend of 0.239 C/decade, differing by 1.7%, which I would assume is well within the range of expected statistical error.
However, it does not compare quite so well with UAH with its trend of 0.163 C/decade for the same period — assuming that models are correct. In fact, it suggests that ground-based measurements are inflated by roughly 49%, or else that models are wrong and that surface measurements are roughly 15% above lower troposphere measurements, or that somehow both the models and surface temperatures are wrong.
It is worth noting that UAH has had a troubled history. For example, in 2005, it was discovered that John Christy’s algorithm used for processing the Microwave Sounder data was incorrectly adjusting for the difference between night and day. Nevertheless, in some way that still is unclear to me, Ross’ study employs UAH. But I do not know whether their analysis includes the correction for the difference between night and day.
David Goebel (#71) wrote:
Multiple organizations act as a double-check, for example, there are the people at NASA GISS and those at Hadley MET, as well as a great deal of literature devoted to the analysis of data. If there are discrepencies, one looks for the source of these discrepancies.
In contrast I have noticed that Patrick Michaels, Ross McKitrick, John Christy and Stephen McIntyre all belong to the Exxon-funded George C. Marshall Institute, so you might find it preferable to leave out motives and instead begin with the assumption that work is being done in good faith.
12 December 2007 at 19:09
Ross,
One point that concerns me is whether — when considering trends in economic development — you are considering them at the local or national levels.
It would seem that any cause of a spurious trend in the readings of a given ground-based station would be the result of local economic development, not national. Particularly if the cause of the spurious trend were the Urban Heat Island effect. So when considering economic development, it would seem appropriate to compare economic development at the local level rather than the national level.
Furthermore, it would seem that an increasing Urban Heat Island effect would be required to produce a spurious trend in temperatures rather than a one-time distortion. This too would be a function of local economic development, not economic development at the national level.
Then I saw the socioeconomic variables which you claim are correlated with trends in apparent local temperature:
Furthermore, you state:
These would appear to be at the national level. As such I find it difficult to see how this would tend to explain trends in temperature by way of an Urban Heat Island effect — which is necessarily limited in nature.
Likewise, given the fact that the former Soviet Union has an economy largely in disarray, whereas Europe, Canada and the United States are doing comparatively well over the period from 1979 to 1999 with the former Soviet Union experiencing weaker warming would produce a substantial amount of the correlation which you see. And as such, if one were able to explain why the trend in temperature over Siberia is weaker than the trend throughout much of the rest of the subarctic, this would be in essence an alternative explanation of the very same “correlations” uncovered by your analysis.
Furthermore, given the wide variety of economic measures and a limited number of countries in the upper northern latitudes, it would seem fairly easy to select a combination of economic measures for which there would exist spurious correlations. Particularly if no causal explanation of the relationship between these economic measures and the trend in temperatures is given or required.
So at this point I have to ask:
Do you have any sort of causal explanation of the correlation relationship between the economic measures you’ve selected and the trends in temperature?
12 December 2007 at 19:36
#70, equivalency is the lowest form of argumentation. McKitrick’s paper arrives at some pretty heady conclusions that if valid would more or less upend the state of the mainstream science, (something I gather would not upset him). Despite this, and despite the truism that the making of groundbreaking discoveries is more often an indication of mistake than virtuosity, he doesn’t see fit to investigate the former possibility through the normal channels. Quite the contrary, he promptly shouts his results from a mountaintop to the public at large (and the PR machine eager for these types of results), as the man says, without ifs, buts, or caveats about the existence of other possibilities.
Meanwhile, Gavin- who has, at least informally, consulted Al Gore as to the state of the science if I have my facts right- defends AIC’s treatment of the then mainstream science. Just fyi, even disregarding the difference in context and character of these instances, and even allowing that each equivalently reveals an opinion about the veracity of the mainstream science, they are hardly equivalent. ‘Partisanship’ and ‘bias’ are not a binary tests of validity. To illustrate a biased journalist, which is to say any and all journalists, can write an article that reflects a genuine effort to cover the story and wherever that takes him or her, (plenty such instances exist), or a work of hackery bent on distortion, (and, certainly, everything in between). It goes without saying, any two such articles are not equivalent despite the existence of an opinion that shapes the story to a greater or lesser extent in both cases. Anyone attempting to short circuit a judgment about the relative validity of work produced by ‘partisan’ or ‘biased’ authors by inane citation of the corresponding existence of a point of view is doing everyone involved a disservice, not to mention playing their part in the stunting of human discourse.
Speaking of which, I find it interesting that in the comments here McKitrick has not seen fit to explain why we might expect to see UHI effects in the satellite record, what accounts for, given his titanic findings, large high latitude temperature anomalies, or the disconnect that his manuscript implies in the terrestrial and oceanic temperature records, to name a few. Of course, were interested in those questions, they probably would’ve occurred to him before he published his paper, not least before that paper became the foundation of a message meant to influence public opinion.
12 December 2007 at 21:00
Steven Mosher, A scientist must be especially circumspect in his statements to a lay audience about scientific matters. Looking at the Post editorial, it is hard to reach any other conclusion but that the results are being vastly oversold–and to a gullible audience. This is improper to say the least. James Hansen’s occasional allegedly intemperate remarks concern politics and policy, not science. I presume you would not deny him a right to his political opinions just because he is a scientist.