The answer to this question is NO.
An October 9. 2005 article in the Seattle Times included the following,
“American researchers examined the possibility that urban heat was masquerading as global warming in 1997, by comparing data from all over the globe with measurements made only in rural areas. The warming was the same. Last year, David Parker, of Britain’s Hadley Centre for Climate Prediction and Research, settled the question emphatically by comparing measurements taken on calm and windy nights (Parker, D.E. 2004: Large-scale warming is not urban. Nature, November 18, 2004). If urbanization was making the planet look hotter than it really is, the effect should be more pronounced when there’s no wind to dissipate the heat from sweltering cities. But rates of warming were the same whether the wind was blowing or not.”
Our new paper in press at Geophysical Research Letters (GRL) “Should light wind and windy nights have the same temperature trends at individual levels even if the boundary layer averaged heat content change is the same?” has shown that the answer to this question must be no, and that the conclusions reported in the Seattle Times are incorrect. This GRL paper shows that the Parker Nature conclusions in which the temperature trends were the same on windy and light wind nights actually means that the heat content changes in the boundary layer were different!
We find that
“…if the nocturnal boundary layer heat fluxes change over time, the trends of temperature under light winds in the surface layer will be a function of height, and that the same trends of temperature will not occur in the surface layer on windy and light wind nights.”
The abstract of this new GRL paper states
“Long-term climate trends of surface air temperature should not be expected to have the same trends for light wind and stronger wind nights, even if the trends in the boundary layer heat fluxes were the same. Parker [2004] segmented observed surface temperature data into lighter and stronger wind terciles in order to assess whether the reported large-scale global-averaged temperature increases are attributable to urban warming. We conclude, however, that trends at an individual height depend on wind speed, thermodynamic stability, aerodynamic roughness, and the vertical gradient of absolute humidity. We present an analysis to illustrate why temperature values at specific levels will depend on wind speed, and with the same boundary layer heat content change, trends in temperature should be expected to be different at every height near the surface when the winds are light, as well as different between light wind and stronger wind nights. This introduces a complexity into the assessment of long-term surface temperature trends that has not been previously recognized.”
This research raises further questions as to the value of using surface air temperature data to assess global warming as well as the conclusions of the Parker Nature paper. See the weblog of July 11, 2005 entitled The Globally-Averaged Surface Temperature Trend - Incompletely Assessed? Is It Even Relevant? for other serious problems with using surface temperature data for this purpose.
Roger, aren’t we talking about data from literally thousands of stations spread around the world? Having read your paper, it’s not clear to me that the differences discussed would do much more than perhaps change the global mean surface temperature confidence interval slightly, *unless* there’s some reason to think that a disproportionate number of stations were biased in a particular manner. Is there? Even if that were shown to be true, would it really mean much since with GMST we are concerned with long-term anomalies? I’m reminded again of the NSIDC/UI Arctic sea ice extent discussion on this blog, where it turned out that despite the apparent selection of rather different metrics the anomalies were in very close agreement (although still different enough to have an interesting but scientifically meaningless argument about whether a new record low had been set this year).
That said, your paper seems likely to have important implications for work on regional climate.
Comment by Steve Bloom — October 31, 2005 @ 2:31 pm
P.S.: Have you asked Parker for a response?
Comment by Steve Bloom — October 31, 2005 @ 2:32 pm
Steve- Thanks for the quick feedback. For the current blog, the issue we have identified is very much relevant for quantitative temperature trends as light wind nights in all of the data would have the same problem. Indeed, a variety of effects (surface roughness, change of instruement height, etc) can result in significant influences on the temperature trends that are evaluated. This brings into question what is meant by a “surface temperature”.
The origin of this note was a Comment to Nature on the Parker paper earlier this year. David Parker did provide a response. However, Nature rejected my Comment (which is a blatant example of cherrypicking which I will document on this weblog within the next few weeks).
Comment by Roger Pielke Sr. — October 31, 2005 @ 4:17 pm
And thank you! But I want to focus for a moment on the issue of the data versus the anomalies. Going back again the NSIDC/UI business, it turned out that NSIDC uses a standard of 15% ice-covered as their standard and UI uses a smaller percentage. (It’s not clear exactly what that percentage is, but it must be both different and smaller than the NSIDC metric since the UI data shows a consistently greater extent, and we know that the raw data source is the same for both.) In any case, the two sets of data look quite different, and presumably the difference comes from perfectly defensible choices on the part of both. Does this throw the existence of a metric for sea ice extent into question? In a sense it does, but fortunately the anomalies tend to resolve all of the important differences between the two approaches.
The situation with global mean surface temperature seems similar. All of the effects you discuss may be quite real and of undoubted scientific importance, but can they be shown to have a significant effect on the temperature *anomalies*?
Finally, can you post the Parker response (maybe with your original so as to make the context clear)?
Comment by Steve Bloom — October 31, 2005 @ 7:35 pm
Steve- with respect to the sea ice areal extent, we both agree that this past year it was well below average much of the time, and that over several decades, its coverage has decreased. However, its variations over time show that large changes can occur very quickly. I am not aware of any model that can skillfully predict such changes. See, for example, the current UI plots (October 31st) where over 500,000 square kilometers have frozen over since mid October such that it is not now far below the long term average for this time of the year.
On the surface temperature trends, there are quite a few issues with this data set, as summarized in my July 11 weblog in addition to the new one that was posted today. In the context of global warming, however, as I wrote in my paper http://blue.atmos.colostate.edu/publications/pdf/R-247.pdf, it is ocean heat content changes that we should monitor. A global averaged surface temperature trend is an inadequate metric to accurately characterize global-averaged changes in the heat content of the climate system.
I need David Parker’s permission, of course, to post his comment on my earlier rejected comment to Nature. I plan to request that from him, as well as his comments on the GRL paper, and will present a weblog on the Nature rejection when I have all of the needed material.
Comment by Roger Pielke Sr. — October 31, 2005 @ 8:03 pm
Rpger, your first paragraph addressed a point I didn’t raise. I mentioned the sea ice issue only for purposes of drawing an analogy.
In your second paragraph, you discuss the adequacy of global mean surface temperature as a metric (a point that was discussed at length under a different post on this blog) when my question had to do with whether the factors you analyze in your GRL paper are likely to significantly affect the GMST anomalies. It doesn’t seem to me that they would, but I would appreciate knowing if you think otherwise.
Comment by Steve Bloom — November 1, 2005 @ 2:42 am
Dear Roger,
This is very intere sting and I look forward to see the responses to this.
I assume you are familiar with some of the work by John Christy on the
tamperature trends in one of the valleys in CA. He found that all the stations down in the plains, where farmers had started to develope the land, showed a strong warming, while all the foothills-stations did not show any warming. He conclude that irregation etc.. produces a “human made” warming.
So if one compare a city station with a station outisde the city - where there is farming (ahd I thing most such stations sits in these locations where there is some farming - both the city station and the “farm” station will show similar warming. If this is the case - the corrections for the heat urban island effect must be w rong since both stations are affected..
Good luck with your paper
Comment by Paal Brekke — November 1, 2005 @ 6:21 am
Steve- our GRL paper demonstrates that a significant unrecognized bias is in the global historical climate network assessment of temperature trends. Quantifying the actual effect on the trends is needed, if this data set is going to be continued to be used for policy decisions (which, as I have said previously, is, however, a very poor choice). By example, lumping together minimum temperatures from light wind and strong wind nights produces a trend that will amplify an actual nocturnal boundary layer warming.
We also demonstrate a critical issue with the David Parker’s article in Nature. I have contacted him and he requested that I not post his earlier reply on the version of my Comment that I submitted to Nature. He is considering writing a comment for my weblog in response to the GRL paper.
Comment by Roger Pielke Sr. — November 1, 2005 @ 9:13 am
Paal-thanks for your comment. As you identify, there are real surface air temperature (and surface enthalpy) changes associated with land use/land cover (LULC) changes. The introduction of irrigation, for example, is one type of LULC change that we have studied (e.g. see Stohlgren, T.J., T.N. Chase, R.A. Pielke, T.G.F. Kittel, and J. Baron, 1998: Evidence that local land use practices influence regional climate and vegetation patterns in adjacent natural areas. Global Change Biology, 4, 495-504.http://blue.atmos.colostate.edu/publications/pdf/R-198.pdf.
The new GRL paper is associated with bias in the surface air temperature observational data set. This bias occurs whether or not there were LULC changes. Another bias of this type that we have identified is reported in
Davey, C.A., and R.A. Pielke Sr., 2005: Microclimate exposures of surface-based weather stations - implications for the assessment of long-term temperature trends. Bull. Amer. Meteor. Soc., Vol. 86, No. 4, 497–504.http://blue.atmos.colostate.edu/publications/pdf/R-274.pdf.
Clearly, we need to understand real spatially representative temperature changes due to LULC change, but also need to identify the uncertainty that is introduced into the surface temperature record due to biases such as we have identified in our GRL paper.
Comment by Roger Pielke Sr. — November 1, 2005 @ 9:27 am
Roger, how can you assert that something is necessarily significant if it hasn’t been quantified? “May be significant” seems more like a scientifically supportable phrase to use.
Also, maybe I’m dense, but your response that “lumping together minimum temperatures from light wind and strong wind nights produces a trend that will amplify an actual nocturnal boundary layer warming” doesn’t seem to address the issue I raised as to whether the anomalies could still be reliable.
Comment by Steve Bloom — November 1, 2005 @ 1:24 pm
Steve-we have quantified the effect of a 1 Watt per meter squared reduction in long wave cooling using the Stull model. As shown in Figure 3 and Table 1 of our paper, trend differences, as a function of wind speed and height, are on the order of tenths of a degree to one degree. This is a significant effect. Other models (and long-term observations at multiple heights) should be used to quantify further, but the basic micrometeorological issue we raise in our paper will remain.
As to whether the long-term surface air temperature anomalies are reliable, in terms of their spatial representativeness for global warming assessments, our studies identify a range of unresolved issues with the data that indicates they are not quantitatively robust for this purpose.
Having a large number of surface stations and assuming the errors are random ignores the need to determine if systematic biases remain. Since so many of the land analysis grid points have none or just one observation site, such a bias can result in an erroneous value for a global averaged surface temperature.
I urge the community to move to ocean heat content changes for our assessment of global warming which is a much more robust climate metric to assess climate system heat changes.
Comment by Roger Pielke Sr. — November 1, 2005 @ 2:00 pm
Roger - this is an interesting paper that I shall read carefully. Before I do that
I would like to pick up:
“In our analysis, we assume that the wind speeds in the residual layer (e.g., 200m to 500 m) are the same as in the near surface layer (e.g., 2m to 10 m)”. This seems implausible: sfc winds should be lower.
Also:
“we estimate the potential temperature profile at the end of a 12-hour night”
Why the end? If Parker used Tmin, wouldn’t that be in the middle of the night?
Comment by William Connolley — November 2, 2005 @ 12:28 pm
Oops, sorry, still haven’t read it, but:
even with the various problems with the surface record, proposing to swap to ocean heat content ignores the vast problems with that record: principally space-time sparsity.
Comment by William Connolley — November 2, 2005 @ 4:02 pm
Thanks for the response, Roger. This begins to get to the crux of things.
I had a look at the GISS temp page and some of the linked documents, and it does appear that they are generally aware of the issues you raise. Have you asked them what relevant adjustments they made (or didn’t make), and why? If you haven’t, I could try.
Regarding the possibility of an individual station within a single grid point throwing things off due to varying wind, in order to say that the temp anomalies (again as distinct from the direct temp data) are affected, wouldn’t it be necessary to find a corresponding change in the wind over a period of time, and wouldn’t such wind data be readily available?
Regarding global mean surface temperature itself, I never meant to imply that absolute GMST was especially meaningful as distinct from the anomalies. One could spend a great deal of time and effort adjusting the data to account for the wind effect discussed in your paper, establishing e.g. that the base GMST should be adjusted by half a degree, but unless the anomalies change this won’t be of any great interest.
Finally, of course ocean heat content is the ultimate metric for scientific purposes. In fact, couldn’t Tim Barnett reasonably start to produce some kind of annual ocean heat content index now? But imagine if he did: The first thing reporters would ask is what would this mean for surface temps. Can you suggest a way to get around this, or explain why trying to do so would be helpful?
Comment by Steve Bloom — November 2, 2005 @ 4:49 pm
William- good questions! The assumption of a constant wind is not central to the issue we have raised on the temperature trends. We use this model with this assumption, to illustrate that lumping data together at night, regardless of wind speed, introduces a significant, unrecognized bias in the temperature trends. I agree the winds often vary with height, even near the ground; this would add, however, even more complexity to the assessment of temperature trends!
On Tmin, this usually occurs at sunrise, unless a weather pattern shift (e.g. a cold or warm frontal passage at night) occurs.
Comment by Roger Pielke Sr. — November 2, 2005 @ 7:32 pm
Steve- thanks again for your constructive questions and comments. I would welcome if you could ask GISS how they accommodate the issue that we raise in our GRL paper. A statement, however, that this effect is smoothed out by selecting a large number of stations (which seems to be the response whenever I bring up the diversity of problems with the surface temperature data) will not remove systematic biases such as we present in our paper.
On changes over time, while a change in wind speeds would also result in an effect on the diagnosed temperature trends, even without this effect we show it is an important issue with respect to the anomalies, as well as the actual temperatures.
Steve and William- Finally, I agree on the need for ocean heat data. The sampling does appear to be good enough to track heat content changes over time as I discussed in my September 25th 2005 blog entitled “Is Global Warming Spatially Complex?”. When reporters ask how this translates into temperature, we need to move them to the more important question as to how ocean heat content changes alters our weather and other aspects of the climate system.
Comment by Roger Pielke Sr. — November 2, 2005 @ 7:43 pm
I must admit that this, as a non-scientist, perplexes me. If we attempted, in some way, to back out the “bias” due to variable night winds, would we really still be measuring GMST? I think it comes back to what we are looking for in measuring GMST. If it is meant to be a true measure of surface temp, then why would we care about anything other than a statistically valid and robust distribution of sampling sites - a few of which, in fact, should be placed in urban or suburban settings - because after all, a small portion of the planet’s surface is urban?
If we want, instead, to come up with a measure devised specifically to identify temperature anomaly due to global warming…it seems to me there are arguments for this, but we should give it a new name. It would not be GMST, but adjusted GMST. Or something. The pure number is more relevant for assessing impact of global warming, and the second number (assuming it comes out measurably different) is more relevant for assessing the progress / rate of global warming.
Finally, since many proxies records were driven by the “pure” GMST, I’m not sure it would be accurate to use them in a temperature trend graph that used an adjusted number for the last 30 years.
Well…I’m way out of my field…so if this is off-base…i won’t be offended.
Comment by dan allan — November 4, 2005 @ 11:39 am
Dan-I appreciate your feedback. As I have urged, we need to move to the assessment of heat content to assess climate system heat changes, and discard the use of surface temperature for this purpose. Heat changes in the different components of the climate system (the oceans, atmosphere,land, and continental ice) are a direct metric of global warming. The use of surface temperature for this purpose is fraught with problems as we have shown most recently by our GRL paper. We cannot accurately back-out all of the uncertainties that are in the surface temperature record.
Comment by Roger Pielke Sr. — November 4, 2005 @ 5:58 pm