Arguments
Guest post in Guardian on microsite influences
Posted on 28 January 2010 by John Cook
After the recent post on microsite influences, I was asked to write on the subject for the Guardian Environment Blog. So here it is, Climate sceptics distract us from the scientific realities of global warming (I didn't write that headline, btw, I suggested "On measuring temperature: how data analysis trumps photographs" but apparently headlines with the phrase "data analysis" just aren't sexy enough). It's basically a less technical version of the original blog post along with an introduction to the concept of microsite influences (while studiously not using the term 'microsite influences' once). The one thing the article does do, I believe, is more succinctly explain how poorly sited weather stations produce a cooler trend:
The cause of this cooling bias appears to have been a change in instruments. In the late 1980s, many sites converted from Cotton Region Shelters (CRS, otherwise known as Stevenson Screens) to electronic Maximum/Minimum Temperature Systems (MMTS). This had two effects. Firstly, MMTS sensors record lower daily maximums compared to their CRS counterparts. So the switch from CRS to MMTS sensors caused a cooling bias in certain stations.
Secondly, the MMTS sensors were attached by cable to an indoor readout device. Limited by cable length, the MMTS weather stations were often located closer to buildings and other artificial sources of heat. This meant most of the stations with the newer MMTS sensors also happened to fall under poorly sited categories. The net result is that poor stations show an overall cooler trend compared with good stations.
Anyway, it's weird excerpting my own writing so go to the Guardian blog to read the full article. I will say one thing - I'm glad I'm not moderating comments on that website. If you consider the behaviour in most online climate discussions, Skeptical Science users are well above the bell curve as far as constructive scientific dialogue goes. Pat yourselves on the back, people!
I've being trying to encourage the editor of one of our provincial newspapers to have a regular column for the purpose of educating the public on climate science and meteorology but to no avail.
While this result was initially met with dismay, Watts rallied and criticised the result, saying it was made with only a small percentage of stations being rated. I believe some time after this, Anthony Watts made the data on station ratings unavailable to prevent any other data analyses comparing good and bad weather stations - but I'm not sure of the timing of this.
The next analysis was by NOAA who also published an analysis comparing only the good stations to the total record (NOAA 2009):
Again, the trends are near identical (you expect some discrepancy as both records cover slightly different regions). Watts criticised this result as a result of homogenisation (data adjustment) of both the good data and the full dataset. That's why Menne 2010 is interesting in that it uses unadjusted data - this is where the cooling bias is revealed.
The thought of hundreds of bogus weather stations occupies me after seeing this theme now repeated. It would be nice to know what the measurement accuracy of these systems have on their own. If, for instance, the Earth's temperature was actually rising 0.10 degree per decade, you would need at least +-.05 degrees to even begin to substantial this. And I am not talking about termocouple specs or a thermometer's rating. I am talking about the system as a whole (its total accuracy), and how it is set up. (The graphs of figure 2 in the original article were plotting values to within a tenth of a degree, which implies an accuracy of +-.05. Spectacular, but real?)
If a pristine mountain peak in the middle of the Pacific makes sense for monitoring CO2, why not apply similar rigor for weather stations? Instead of trying to salvage this data, maybe better to start from scratch. Dont we have 100 years or so to work on this problem? Much cheaper in the end too.
Aside from issues of accuracy, another important system parameter is repeatability. Repeatibility that can be guaranteed over many years. If you were to simply take the sum of measurements over time from a set of reliable and repeatable instruments, (even if every sensor was buried in a 1 meter cube of cement), this global checksum would tell you something about global warming, because it is the relative change that matters, not the absolute temperature reading.
My last remark. The continental US doesnt exactly seem like the best location on Earth for monitoring global warming.
, with details including where they are and site requirements.
The site lists exactly the things a trained metrologist needs to consider when seeting up any decent temperature monitoring system:
Resolution, Repeatability, Response time, Drift, Hysteresis, and Linearity.
It is interesting that in all of this that Watts and the WUWT members seem to forget that America is actually not the world. Australia is of pretty comparable surface area wise to America and if the stations here report a similar warming trend...
I wait with keen eyes for Watts full analysis!
This is taken from the site
"Temporal Representativeness
In addition to difficulties with the correct exposure of instruments, thought has to be given to changes in the long-term exposure of the site. Buildings in close proximity to the instrument enclosure will result in the area of representativeness being reduced.
For example, when the instrument enclosure at Sydney was installed in 1788, the instruments were representative of a relatively wide area around Sydney. With subsequent construction of high-rise buildings and freeways, climatic and meteorological conditions only 50m from the site are now significantly different to those at the site.
It is important that the station be inspected regularly and any changes in the siting are properly documented."
1788?
Aside from the site, what about technology? Was there even a universal standard set of measurements in 1788?
I must be going nuts, or is it the CO2?
Also, I don't believe we have 100 years to fix the problem. Satellite & surface temperatures are in agreement-+0.16 degrees per decade since 1979. As we don't know for certain just how had the impacts of future rises are going to be, I think it would be irresponsible to wait 100 years to "double check". If we don't take serious action within the next 20-30 years, I believe it will be too late to avert a truly catastrophic rise in temperatures. The fact is that, had we listened to the scientists 20 years ago, instead of letting the fossil fuel industry have its way, CO2 mitigation would have been a *hell* of a lot cheaper!
RSVP: "The continental US doesnt exactly seem like the best location on Earth for monitoring global warming."
Tell that to the "global warming is a scam" crowd, who have been using relatively cooler U.S. temperatures over the last year or two to argue against global warming. Mr. Watts is a full participant in this line of spin. The U.S. surface record is unreliable, except during the times it shows cooler temperatures.
But instead of trying to stay ground and defend the science, I thing the science should be made easier to follow. The only way, in my opinion, to manage that is to make an offensive, where you make it your goal to make climate science understandable for the general public. We still haven't achieved that goal, or maybe we still haven't set that goal...
One reason for that, is that the climate science are complicated. But I thing that it's an achievable goal with the right means. I'm not an activist, but I think that by simplifying (in words) the science somehow we're half way there. This should not be the scientists job, and in many cases I don't think that it always lies in their skills to make a simple argument ;) They should be able to continue their work.
But in some way I know that this is an achievable goal, with the right means.
Regards,
Svatli
We're not talking about absolute temperatures, but instead they're temperature *anomalies*. If I had a termometer in an oven at 200ºC for the whole century, the anomaly would be zero. So having a termometer in a parking lot, while it yields of course a higher temperature, it does not produce increasingly higher temperatures over the years and decades.
Am I oversimplifying it?
"If a pristine mountain peak in the middle of the Pacific makes sense for monitoring CO2, why not apply similar rigor for weather stations? Instead of trying to salvage this data, maybe better to start from scratch. Dont we have 100 years or so to work on this problem? Much cheaper in the end too."
Well, actually, a new temperature monitoring system, designed from the ground up to meet climatology needs, has not only been designed, but deployed.
It's called the US Climate Reference Network.
In fact, it's the CRN siting criteria that Anthony Watts is using to "prove" that certain stations in the Historical Climate Network are "bad" - using standards set in the last decade to categorize stations that are decades or a hundred years old.
One of the results of the Menne 2010 paper is that the several years of USCRN data we have matches the temperatures derived from the historical stations extremely closely. Two separate sets of stations, one set explicitly designed to meet rigorous siting standards and provide optimal spatial coverage. The other a much larger set of stations placed originally to provide data for weather forecasting.
And they match.
And as years go on and they continue to match, it will only increase the confidence of the accuracy of data from the historical network of stations (though within science it's already sky-high).
The historical data has been subject to dozens of tests, and has always passed with flying colors. The only test it hasn't passed is the "I took a photo but did no analysis" test, which is bogus.
Not only do we not have 100 years to wait, there's no reason in the world to throw out existing, perfectly good, data - except for the politics of delayed action.
John
An outgrowth of a study conducted in the late 1990's, the USCRN went online in 2003 to address some of the concern about site based bias in measurements.
More information is available from NOAA.
http://www.ncdc.noaa.gov/crn/
is Menne 2010 the first paper to compare data from the new network with the old, historical weather station network?
More important, has NOAA photographed all 100+ of the new stations, or not? If so, in color, or black-and-white?
"John - I'm sure you are aware that the counter from Watts is that the reason that he wants a fuller sample than that used by Menne before writing it up is because the earliest returns of photos of sites were naturally urban - close to the neighbourhoods of volunteers."
Researchers have previously compared rural with urban stations and have found no significant difference in trend.
I looked hard, but could come up with nothing.
There is something called data homogenization used in data warehousing, but in the absence of metadata it relies heavily on AI technologies like pattern recognition. It is not meant to be used to produce "objective" datasets, just hints.
In other words it is woodoo magic, not a tool for science.
BTW, Menne at al 2010 do depend on data homogenization.
http://www1.ncdc.noaa.gov/pub/data/ushcn/v2/monthly/menne-etal2010.pdf
http://mapcenter.hamweather.com/records/7day/us.html?c=maxtemp,mintemp,lowmax,highmin&s=20090913&e=20090913
If the results are obtained using many measurements from a each of a large number of instruments then the overall trend can be found to much higher precision than the accuracy of individual instruments (if there's no systematic bias in the errors, of course, which is the whole point of the article). You could easily see a 0.1°C/decade increase using only a few dozen thermometers which only had a resolution of 1°C using one measurement from each for each of the 3652 or so days in a decade.
homogenization of a temperature time serie is not a statistical procedure. If, for example, a sensor is moved downhill you correct for the lapse rate. Homogenization means to take into account all the possible sources of bias one can identify.
She says that water vapour is acting as a negative feedback rather than amplyfiying CO2.
http://www.guardian.co.uk/environment/2010/jan/29/water-vapour-climate-change
says no one understands why its decreasing
It is not on the water vapour feedback. Indeed, it focuses on the water vapour just in the lower stratosphere; it might have contributed to the slow down of the tropospheric temperature (as opposed to global warming) increse, but it's something related more to interannual/decadal variability than to the long term trend.
The reason of the sudden drop in year 2000 is not clear. It might be related to the unusual warming of the western pacific ocean and the drop of the tropopause temperature. If this is linked to global warming or it is part of a natural cycle can not yet be assessed.
As far as i know no one has looked at models behaviour as far as the lower stratospheric humidity is concerned. It's then hard to guess a possible physical mechanism responsible for its reduction and if it's a consequence of the already known physics. Gavin Schmidt at RealClimate promised further analysis, we'd better wait for more insights.
As is standard practice in science, unless there are evident flaws in a paper (which seemes to be not the case here) it should be carefully evaluated taking the required time. No nails in the coffin nor smooking guns around, just one more little piece of evidence to fit in the global picture.