The observed global temperature record

Since the late nineteenth century, the observed average surface temperature has been increasing, as shown in figure 7.1. The data from land comes from about 4,000 stations distributed widely over the the globe, although naturally enough there are more stations in North America and Europe and fewer stations in such places as Antarctica, Greenland, Siberia, and the Sahara Desert. The actual measurements typically are taken twice daily

Figure 7.1. The instrumental record of global average surface temperatures from 1880 to 2009, relative to the mean temperature from 1951 to 1980.

(more frequently in some locations) and are of the air temperature a few meters above ground. The ocean data mostly come from in situ observations from ships and buoys; the measurements are the temperature of seawa-ter itself, although in fact this is a good surrogate for the air temperatures just above the surface.1

As the temperature data on both the land and the ocean are quite nonuniform, the temperatures are first interpolated onto a regular grid from which a globally averaged temperature can be constructed. (Measurements have in fact been taken on land and on sea since well before 1850, but not with sufficient spatial density to enable the direct construction of a gridded data set.) Obviously, such global warming and the ocean a procedure is not error-free; errors come from the measurement errors at individual stations, from sampling errors caused by the fact that there may be insufficient coverage over the globe, and from bias errors caused by possible systematic changes in measurement methods. Possible changes in the surroundings of a station, caused for example by urbanization (discussed below), are also a potential source of error. The combined effect of all these errors on the global average is generally estimated to be quite small, less than 0.1°C since 1950, possibly up to about 0.2°C before about 1930, certainly considerably smaller than the increase of temperature over the twentieth century.

There is one possible source of error that has received much discussion both in the scientific literature and in the media, and that is the effect of urbanization. It is common experience that the temperature in a large city is a little higher than that in the surrounding countryside because of all the energy that is expended in the city in the buildings and by transport. Furthermore, the paving over of the ground by asphalt and concrete reduces evaporation and tends to increase surface temperatures. If the stations that measure temperature are in locations that have experienced increased urbanization over the past 50 or 100 years, then any increase in temperature that they have recorded may in part be caused by that urbanization effect, rather than reflecting a true increase in global temperature. One way to study this effect is to compare the increases in temperature between urban areas and areas that have remained rural. When this comparison is done, the effects on the global temperature record are found to be small, contributing an error of about 0.005°C per decade, and about 0.05°C over the past century, although it should be said that some critics of the temperature record believe this error estimate to be too small. Satellite measurements, discussed below, provide another check on urbanization errors.

A somewhat different source of error, although still a bias error and one that has certain similarities with the urbanization problem, is that the way that temperatures have been measured has changed over the past century, both over water and on land. On land the early shelters were fairly heterogeneous; they have been slowly replaced with more standardized shelters known as Stevenson screens (after Thomas Stevenson, 1818-1887). These shelters are essentially ventilated white boxes that shield the thermometer against precipitation and direct radiation but that allow air to flow past the thermometer and so give accurate measurements of air temperature. Still more recently, some of these shelters have been replaced with mechanically aspirated shelters to further increase the airflow. The errors due to the different box designs are generally thought to be very small (< 0.1°C) at any given station up to 1950 and negligible after that.

Over water, the methods of taking sea-surface temperature have also varied over time, from taking samples in wooden and then canvas buckets in the early part of the record to measuring the temperatures of the water coming into the engine rooms of ships for cooling. Studies suggest that the errors are small, certainly compared to the changes in temperature over the past century, although changes in measuring techniques have led to some apparently artifactual small jumps in the record.2

0 0

Post a comment