December 22, 2007

Contaminated Temperature Data

Filed under: Surface, Temperature History

It’s that time of year again when we see headlines about 2007 being the mth warmest year on record over the past n years whether we are talking about the United States or the world as a whole. Reporters breathlessly reveal that the trend in temperatures is alarming and completely unprecedented over the eons of earth history. The buildup of greenhouse gases is immediately blamed, and we are all left to believe that the rising temperatures can only be explained by human emissions. Rarely does anyone seem to question the quality of the temperature data, and yet, articles appear regularly in the scientific literature showing that the near-surface air temperature measurements are fraught with errors, gaps, and any number of inhomogeneities.

Climate scientists have been writing about these problems for over a century. Long ago, scientists noticed that temperatures in London were substantially higher than the surrounding rural landscape, and urban climatology has been a subdiscipline in the atmospheric sciences ever since. Once the greenhouse debate got fired-up in the late 1980s, countless articles appeared in the literature on everything from the urban heat island to changes in instrumentation to changes in time of observation. Yet another major article related to the issue of contamination to the temperature record has appeared recently in the Journal of Geophysical Research written by two of the leading greenhouse skeptics walking the planet – Ross McKitrick of the University of Guelph and Patrick Michaels of the Cato Institute. Be aware that the peer reviewers of the manuscript would have been fully aware of who conducted the research and wrote the article, they would certainly have known of the international reputations of McKitrick and Michaels, and accordingly, the research would likely have been held to the highest standards of scrutiny.

The authors present a historical context for their research noting that more than 50 years ago, prominent climatologist Murray Mitchell warned that when using weather records to determine trends in climate “The problem remains one of determining what part of a given temperature trend is climatically real and what part the result of observational difficulties and of artificial modification of the local environment.’’ McKitrick and Michaels note that “two types of bias continue to affect the measurement of climate change. Observational difficulties, or data inhomogeneities (such as station moves and closure, record discontinuities, equipment change and changes to the time of observation) are known to have affected records of mean temperature. Modification of the land surface, including urbanization and other economic activity, has been shown to affect local, regional and possibly global meteorology, and thus locally measured temperature data.”

There are countless potential contaminants to weather records, and generally speaking, the United Nations Intergovernmental Panel on Climate Change (IPCC) and many others make the assumption that these contaminants are relatively inconsequential, well accounted for, and/or their effects generally cancel out in determining long-term trends in temperature. If true, there should be no significant relationship spatially between socioeconomic variables and trends in temperature over land areas. If significant relationships can be identified between socioeconomic variables and temperature trends, then contaminants to the temperature records would be confirmed.

McKitrick and Michaels examined the gridded temperature dataset used by the IPCC and many others – they then gathered for each grid cell information on gross domestic product, literacy, months with missing data, growth in human population, economic growth, and growth in coal consumption. To make the analyses as rigorous as possible, they also added the satellite-based lower-tropospheric temperature trend, sea level pressure, a dryness index, length of coastlines, and latitude. They used a very sophisticated set of calculations to identify any socioeconomic signals in the temperature trend data, and to say the least, the signals were loud and clear.

Almost all of the socioeconomic variables were highly statistically significantly related to the temperature trends. The authors note “Taken together, our findings show that trends in gridded climate data are, in part, driven by the varying socioeconomic characteristics of the regions of origin, implying a residual contamination remains even after adjustment algorithms have been applied. Users of gridded climate data products need to interpret their results accordingly.” Furthermore, they state “These results are also consistent with previous findings showing that nonclimatic factors, such as those related to land use change and variations in data quality, likely add up to a net warming bias in climate data, suggesting an overstatement of the rate of global warming over land.”

They produced a global map of where the raw data warmed too much or too little when adjustments were made for the socioeconomic variables (see Figure 1). McKitrick and Michaels comment “Note the regions where the adjustments are minimal are North America, Eastern Europe and Australia. Widespread positive biases are observed in Western Europe and Southeast Asia. Africa and South America contain many regions with missing data, though the map overstates this, because at the equator, the raster squares are smaller than the grid cells they represent, because of the global projection used.”

Figure 1. Differences between observed and adjusted trends around the world. Raster squares correspond to center of 5° x 5° grid cell, but not to size of grid cell itself. Units are °C/decade. A value of, say, 0.1–0.2 means that the observed trend in that cell was between 0.1 and 0.2°C/decade higher than the adjusted trend (from McKitrick and Michaels, 2007).

So what is the bottom line here? The frequency histogram below (Figure 2) shows the distribution of temperature trends for the gridded dataset, the satellite-based lower-tropospheric data, and the data adjusted to account for the socioeconomic variables. In commenting on this figure, the authors state “our analysis does suggest that nonclimatic effects are present in the gridded temperature data used by the IPCC and that they likely add up to a net warming bias at the global level that may explain as much as half the observed land-based warming trend.”

Cancelling half of the “global” warming of the past few decades is highly noteworthy at World Climate Report, but no worthy of coverage elsewhere? We can only imagine the press coverage had they been able to squeeze even more warming out of the IPCC temperature record.

Figure 2. Distributions of temperature trends 1979–2002. (top) IPCC surface data. (middle) Satellite
(MSU) tropospheric data. (bottom) Adjusted surface data (from McKitrick and Michaels, 2007).


McKitrick, R. R., and P. J. Michaels, 2007. Quantifying the influence of anthropogenic surface processes inhomogeneities on gridded global climate data. Journal of Geophysical Research, 112, D24S09, doi:10.1029/2007JD008465.

No Comments

No comments yet.

RSS feed for comments on this post.

Sorry, the comment form is closed at this time.

Powered by WordPress