Questions about the temperature graphs used for global warming claims.
Contaminated data
Hot cities, not CO2, cause urban thermometers to rise
Ross McKitrick, Financial Post Published: Wednesday, December 05, 2007
Below is the famous graph of "global average surface temperature," or "global temperature" for short. The data come from thermometers around the world, but between the thermometer readings and the final, famous, warming ramp, a lot of statistical modelling aims at removing known sources of exaggeration in the warming trend. In a new article just published in the Journal of Geophysical Research -- Atmospheres, a co-author and I have concluded that the manipulations for the steep post-1980 period are inadequate, and the above graph is an exaggeration. Along the way, I have also found that the United Nations agency promoting the global temperature graph has made false claims about the quality of its data.
The graph at right comes from data collected in weather stations around the world. Other graphs come from weather satellites and from networks of weather balloons that monitor layers of the atmosphere. These other graphs don't show as much warming as the weather-station data, even though they measure at heights where there is supposed to be even more greenhouse-gas-induced warming than at the surface. The discrepancy is especially clear in the tropics.
The surface-measured data has many well-known problems. Over the post-war era, equipment has changed, station sites have been moved, and the time of day at which the data is collected has changed.
Many long-term weather records come from in or near cities, which have gotten warmer as they grow. Many poor countries have sparse weather-station records and few resources to ensure data quality. Fewer than one-third of the weather stations operating in the 1970s remain in operation.
Scientists readily acknowledged that temperature measurements are contaminated for the purpose of measuring climate change. But they argue that adjustments fix the problem. To deal with a false warming generated by urbanization, they have the "Urbanization Adjustment." To deal with biases due to changing the time of day when temperatures are observed, they have the "Time of Observation Bias Adjustment."
And so forth.
How do we know these adjustments are correct? In most studies, the question is simply not asked. A few studies argue that the adjustments must be adequate since adjacent rural and urban samples give similar results. But closer inspection shows some of these papers don't actually give similar results at all, or when they do they define "rural" so broadly that it includes partly urbanized places. Other studies say the adjustments must be adequate because trends on windy nights look the same as trends on calm nights. But the long list of data problems includes issues just as serious under both windy and calm conditions.
The papers describing the adjustments aim to construct data showing what the temperature would be in a region if nobody had ever lived there. If the adjustments are right, the final output should not be correlated with the extent of industrial development and variations in socioeconomic conditions. But in a 2004 study with climatologist Patrick Michaels, we found that the adjustment models were not removing the contamination patterns as claimed. If the contamination were removed, we estimated the average measured warming rate over land would decline by about half. Dutch meteorologists using different data and a different testing methodology had come to the same conclusions.
