2.1.4. THE TIME OF OBSERVATION BIASThe “Mean Daily Temperature" which is obtained by one reading per day of the maximum and minimum temperature for the past 24 hours is taken to be the average of these two figures. However, the actual 24 hours for which it applies is the previous 24 hours of the time of measurement, not the actual daily 24 hours. The measurement of Max and Min is made at different times in different places and It also changes over time and from one place and one country to another. This bias in "mean daily temperature" is called the "Time of Observation Bias" (TOB) by the Americans and together with all the other inaccuracies in their measurements, they make a gallant effort to try and "correct" for it. These efforts are described by Vose et al. (2003). There is some very interesting information in this paper. We learn, for example, that "the majority of the US Cooperative observing Network is staffed by volunteers". I wonder what their qualifications are, or who checks up on them and what situations apply in other countries?. They also say "When the observation day differs from the calendar day a "carry over" bias of up to 2.0ºC is introduced into monthly mean temperatures. Also “Non-calendar day observations also result in a "drift" bias of up to 1.5ºC in monthly mean" because there is a carry over from the previous month. If the day is different, then so are the month and the year. They state that there has been a systematic change in the preferred observation time in the US, requiring a large correction they recorded near sunset before the 1940s and switched to mornings after that, giving a "slight" warm bias to the later readings. A diagram showing the distribution of time of observation now for the USHCN (United States Historical Climatology Network) stations shows a wide level of variability. They make a "correction" for the US, which may not apply elsewhere. It is doubtful whether knowledge of conditions 100 years ago is very reliable.
2.1.5. URBAN HEATING AND LAND USE CHANGEThe unrepresentative meteorological temperatures are often measured in places of increasing population, more buildings, more concrete, growing vegetation, more cars, more heating and therefore subject to a positive bias. The evidence that this is happening is overwhelming. It is the only authenticated “anthropogenic” effect on the climate (Gray 2000, McKitrick and Michaels 2004, 2008).
The IPCC have repeatedly quoted the paper by Jones et al. (1991) as evidence that urban heating is negligible. These authors examined an “extensive” set of rural station temperature data for three regions of the world - European parts of the Soviet Union, Western Australia and Eastern China. When combined with similar analyses for the contiguous United States, the results are claimed to be representative of 20% of the land area of the Northern Hemisphere and 10% of the Southern Hemisphere. They worked out the linear slope of temperature anomalies for the rural series in each case and compared it with the same slope for several gridded series. For the Western USSR, it covered the period 1901-1987 and 1930-1987, for Eastern Australia it was 1930-1988 compared with 1930-1997, for Eastern China it was 1954-1983 and for the contiguous United States it was 1901-1984. The differences between urban and rural slopes were only significant at the 5% level for Eastern Australia and for one set of Eastern China. They concluded “It is unlikely that the remaining unsampled areas of the developing countries in tropical climates, or other highly populated parts of Europe, could significantly increase the overall urban bias above 0.05ºC during the twentieth century”
It is unclear whether this small correction has been made for the most recent version of the Jones et al. global temperature series. There are several things wrong with the Jones et al. (1991) paper.
• The quality of the data is even worse than usual. They admit “It is unfortunate that separate maximum and minimum temperature data are not more widely available.”
• The qualification for a “rural” site is a population below 10,000 for Western Soviet Union, below 35,000 for Eastern Australia, and below 100,000 for Eastern China. There is ample evidence (Gray 2000) that urban effects exist in such places.
• They have chosen countries with a continuous record of effective scientific supervision. These are not representative of the rest of the world, where changes of country and adequate supervision are far less common.
Even these countries raise doubts. Russia had a tyrannical regime where statistics were frequently manipulated for political purposes. China had a major famine from the “Great Leap Forward” between 1958 and 1959 and also a manipulation of statistics.
Two of the countries, the contiguous USA and China have such reliable records that, when corrected, they show no global warming, or residual urban influence (see Figures 3 and 4), but these two well monitored countries cannot be regarded as “typical” of the rest of the world. In the very same year there appeared in Geophysical Research Letters another paper which included two of the authors of the previous paper, Wang and Karl (Wang et al. 1991). The abstract of this paper reads “We used 1954-1983 surface temperature from 42 Chinese urban (average population 1.7 million) and rural (average population 150,000) station pairs to study the urban heat island effects. Despite the fact that the rural stations are not true rural stations, the magnitude of the heat islands was calculated to average 0.23ºC over the thirty year period, with a minimum value (0.19ºC) during the 1964-1973 decade and maximum (0.28ºC) during the most recent decades.”
This study appears to have used the same stations that were claimed to have no urban bias in the first paper and now there is an urban bias even if “rural” now includes places with population as high as 150,000. The early paper (Jones et al. 1991) states, of Eastern China, “The stations were selected on the basis of station history: We chose those with few, if any, changes in instrumentation, location or observation times”.
Wang et al. (1991) says “They were chosen based on station histories. We chose those without any changes in instrumentation, location, or observation times”. Both papers were written at the same time and different conclusions made from the same data. Recently, Keenan (2007) has shown that many of the Chinese stations moved several times over the period in question, in one case 15 km and he accuses Wang of outright fraud, as he must have known this at the time.
Confirmation of continuing urban warming in China has been documented by Ren et al (2008) who, from 282 weateher stations in Northern China from 1960 to 2000, that there was an urban bias of 0.16ºC per decade for cities over 500,000 population, down to 0.07ºC per decade for small cities (100,000 to 300,000). The National bias was estimated t 0,11ºC per decade, However, these were all by comparison with “rural” measurements, which were assumed to be immune from urban heating. Another paper used by the IPCC (Solomon et al. 2007) as evidence that urban warming is negligible is by Peterson (2000) "Assessment of Urban Versus Rural In-Situ Surface Temperatures in the Contiguous United States: No Difference Found". This paper supplies much more information on the observation process and its snags than has appeared before.
The IPCC has chosen to consider the phrase "No Difference Found” as implying that it is evidence that no difference exists. The text shows that this untrue. Peterson merely found that his sample size was insufficient to obtain a statistically significant figure.
He studied only three years of data, 1989-91, so he was unable to study "trends". His excuse is rather startling. "A longer period would increase the problem of missing data". The problem of missing data is not otherwise mentioned, but it must be important if it has an influence after only three years in the USA. The data are not given and the problem must be even worse outside the USA. He chose for study 40 clusters of stations, well distributed over the country; a total of 289 stations, 85 "rural", 191 "urban" and 13 "suburban. It was surprising to learn that in the Unites States there are several different types of instrument and shelter. There were 106.9 maximum and minimum liquid-in-glass thermometers in a Cotton Region Shield (CRS, resembles a Stevenson Screen), 142.8 thermistor based instruments in a MMTS shield, 35 hygro-thermometers in an HO-83 housing and 2.3 hygro-thermographs. (The fractions are from changes during the three years). There are photographs of these three types. If the Americans have several different instruments what kinds are used elsewhere? Corrections had to be made for urban/rural location, elevation, Time of Observation Bias, instrumentation and siting. The total remaining overall urban/rural bias before the others were applied was +0.31ºC. This is half the amount claimed to be caused by greenhouse gases since 1900. However, when the other corrections were applied, together with their inaccuracy levels, the urban/rural bias was reduced to +0.04ºC.
The Time of Observation Bias was the largest, accounting for a correction of -0.17ºC. This was because rural stations had a higher proportion of morning readers. Differences in elevation accounted for a correction of -0.11ºC because rural stations in the USA are usually higher up than the cities. Differences in instrumentation accounted for a bias of 0.05ºC because rural stations had a higher proportion of hygro-thermometers that had a warm bias over the period and latitude changes gave a negative bias, -0.06ºC, as urban stations tended to be a little further north than the rural stations.
The fully adjusted urban/rural bias of +0.04ºC was regarded by Peterson as equivalent to zero because it was not significant at the 90% level. But this does not mean that the bias does not exist, as assumed by the IPCC. It merely means that Peterson’s sample size was not large enough to give a result with a higher level of significance. It is simply not true to claim “No Difference Found.”
In most other countries the complex correction procedures carried out by Peterson are impossible as they do not possess the numbers of sites for comparison, or the supervision or the scientific expertise. Corrections for Time of Observation Bias, Elevation, and Instrument Change may be impossible, so the first, unadjusted result of Peterson's, an urban/rural bias of +0.31 ºC, could be the best estimate. Two recent papers by Parker (2004, 2006) seek to show that urban warming does not happen. He argues that because daily mean, Maximum or Minimum Temperatures are not influenced by windy conditions, therefore urban heating is negligible. But the "day" that gives average wind conditions is usually a different "day" from that used for the daily mean, the Maximum and Minimum. In the second paper he seems to have realised this after he wrote the paper, so he puts the problem in Appendix A, where some "private communications" helped him out, but he does not list the ones which did not.
The idea that urban heating should be influenced only by the strength of the wind and not its direction, and that there are no other factors involved, is simply a gross oversimplification of a complex issue.
The only other country that has attempted a similar correction exercise is China and they also show no evidence of greenhouse warming. Jin et al. (2005) used measurements with a MODIS spectrometer on NASA satellites to measure the urbanisation effect globally and over several selected cities. In July 2001, for night time and daytime temperatures, urban areas between 30 and 60 degrees north are eight degrees Celsius above a surrounding forest by day and two degrees above at night. These are much greater than the "corrections" that are made to the surface record. There were also large differences between urban surfaces and cropland and for selected cities.
They make the following comment, which is relevant to the Peterson paper and to the IPCC approach "Urban areas have generally been determined from satellite data from the night time lighting. Such a measure is more an index of electricity use than of urban features that are significant for climate study”. McKitrick and Michaels (2004, 2008) showed that the surface anomalies for 1979 to 2000 were significantly biased by rises in population, energy usage, sample defects and GDP. Removal of these effects reduced the average temperature trend by about half. Pepin and Lundquist (2008) chose temperature records from high altitude weather stations and plotted the average trend over recent years, which is slightly downwards, as are the general trends. Only a few urban stations show a rise. So the glaciers are unlikely to be receding because of "warming" after all.
2.1.11. TEMPERATURE OSCILLATION
