On the reliability of the U.S. Surface Temperature Record
Posted on 22 January 2010 by John Cook
The website surfacestations.org enlisted an army of volunteers, travelling across the U.S. photographing weather stations. The point of this effort was to document cases of microsite influence - weather stations located near car parks, air conditioners and airport tarmacs and anything else that might impose a warming bias. While photos can be compelling, the only way to quantify any microsite influence is through analysis of the data. This has been done in On the reliability of the U.S. Surface Temperature Record (Menne 2010), published in the Journal of Geophysical Research. The trends from poorly sited weather stations are compared to well-sited stations. The results indicate that yes, there is a bias associated with poor exposure sites. However, the bias is not what you expect.
Weather stations are split into two categories: good (rating 1 or 2) and bad (ratings 3, 4 or 5). Each day, the minimum and maximum temperature are recorded. All temperature data goes through a process of homogenisation, removing non-climatic influences such as relocation of the weather station or change in the Time of Observation. In this analysis, both the raw, unadjusted data and homogenised, adjusted data are compared. Figure 1 shows the comparison of unadjusted temperature from the good and bad sites. The top figure (c) is the maximum temperature, the bottom figure (d) is the minimum temperature. The black line represents well sited weather stations with the red line representing poorly sited stations.
Figure 1. Annual average maximum and minimum unadjusted temperature change calculated using (c) maximum and (d) minimum temperatures from good and poor exposure sites (Menne 2010).
Poor sites show a cooler maximum temperature compared to good sites. For minimum temperature, the poor sites are slightly warmer. The net effect is a cool bias in poorly sited stations. Considering all the air-conditioners, BBQs, car parks and tarmacs, this result is somewhat a surprise. Why are poor sites showing a cooler trend than good sites?
The cool bias occurs primarily during the mid and late 1980s. Over this period, about 60% of USHCN sites converted from Cotton Region Shelters (CRS otherwise known as Stevenson Screens) to electronic Maximum/Minimum Temperature Systems (MMTS). MMTS sensors are attached by cable to an indoor readout device. Consequently, limited by cable length, they're often located closer to heated buildings, paved surfaces and other artificial sources of heat.
Investigations into the impact of the MMTS on temperature data have found that on average, MMTS sensors record lower daily maximums than their CRS counterparts, and, conversely, slightly higher daily minimums (Menne 2009). Only about 30% of the good sites currently have the newer MMTS-type sensors compared to about 75% of the poor exposure locations. Thus it's MMTS sensors that are responsible for the cool bias imposed on poor sites.
When the change from CRS to MMTS are taken into account, as well as other biases such as station relocation and Time of Observation, the trend from good sites show close agreement with poor sites.
Figure 2: Comparison of U.S. average annual (a) maximum and (b) minimum temperatures calculated using USHCN version 2 adjusted temperatures. Good and poor site ratings are based on surfacestations.org.
Does this latest analysis mean all the work at surfacestations.org has been a waste of time? On the contrary, the laborious task of rating each individual weather station enabled Menne 2010 to identify a cool bias in poor sites and isolate the cause. The role of surfacestations.org is recognised in the paper's acknowledgements in which they "wish to thank Anthony Watts and the many volunteers at surfacestations.org for their considerable efforts in documenting the current site characteristics of USHCN stations." A net cooling bias was perhaps not the result the surfacestations.org volunteers were hoping for but improving the quality of the surface temperature record is surely a result we should all appreciate.
UPDATE 24/1/2010: There seems to be some confusion in the comments mistaking Urban Heat Island and microsite influences which are two separate phenomenon. Urban Heat Island is the phenomenon where a metropolitan area in general is warmer than surrounding rural areas. This is a real phenomenon (see here for a discussion of how UHI affects warming trends). Microsite influences refer to the configuration of a specific weather station - whether there are any surrounding features that might impose a non-climatic bias.
UPDATE 24/1/2010: There has been no direct response from Anthony Watts re Menne 2010. However, there was one post yesterday featuring a photo of a weather station positioned near an air-conditioner along with the data series from that particular station showing a jump in temperature. The conclusion: "Who says pictures don’t matter?"
So the sequence of events is this. Surfacestations.org publishes photos and anecdotal evidence that microsite influences inflate the warming trend but no data analysis to determine whether there's any actual effect on the overall temperature record. Menne 2010 performs data analysis to determine whether there is a warming bias in poorly position weather stations and finds overall, there is actually a cooling bias. Watts responds with another photo and single piece of anecdotal evidence.
UPDATE 28/1/2010: Anthony Watts has posted a more direct response to Menne 2010 although he admits it's not complete, presumably keeping his powder dry for a more comprehensive peer reviewed response which we all eagerly anticipate. What does this response contain?
More photos, for starters. You can never have enough photos of dodgy weather stations. He then rehashes an old critique of a previous NOAA analysis criticising the use of homogenisation of data. This is curious considering Menne 2010 makes a point of using unadjusted, raw data and in fact, it is this data that reveals the cooling bias. I'm guessing he was so enamoured with the water pollution graphics, he couldn't resist reusing them (the man does recognise the persuasive power of a strong graphic).
"How would a slow urban build-up around a rural temp. station be treated via microsite and UHI adjustments?"
This is a good question and is addressed in Urbanization effects in large-scale temperature records, with an emphasis on China (Jones et al 2008) - I give a summary of the paper's results here.
Re the classification of stations, the NOAA have classified their stations also - Menne 2010 performs their analysis with the surfacestations.org classifications and their own.