The hockey stick later fell on hard times, though. It was largely marginalized in the fourth and latest report of the IPCC, published in 2007, because an independent review by the National Academy of Science of the statistical analysis involved in the creation of the hockey stick led to the determination that it was the result of a flawed methodology.13 This expert review might never have happened had not two Canadians, one an economist and the other a retired statistician, dug into exactly how the hockey stick was created.14 It’s not clear that tree rings can be trusted to reconstruct past temperatures to any useable level of accuracy anyway. It turns out that the most recent tree-ring data do not even show the warming that occurred in the second half of the twentieth century, but appear to indicate a cooling instead. This discrepancy is called the “divergence problem.” Craig Loehle argues that tree-ring data cannot be used as temperature proxies for previous warm events, such as the Medieval Warm Period, because the tree rings have not demonstrated sensitivity to unusual warmth in the late twentieth century.15 In other words, tree-ring data have probably underestimated the magnitude of previous warm events in history –which is exactly what the hockey stick did–since those tree rings do not even show the warming over the last fifty years! Therefore, it could be that temperatures in the Medieval Warm Period were considerably higher than today. We simply do not know, and probably never will know. Thermometer data, available since 1900 or earlier, are clearly better than temperature proxies, but they are still rather limited in their geographic sampling. Until recently there have been very few measurements over two-thirds of the Earth: the oceans. Early measurements of ocean temperatures were taken from buckets dipped in the ocean from the decks of ships. Later, temperatures would be taken well below a ship’s water line, in the intake ports for water that cooled the ship’s engine. Most recently, a global network of over a thousand drifting buoys has been deployed specifically for measuring sea surface temperature, salinity, currents, and weather. The differences between these various observing systems scattered through time mean that our estimates of ocean warming to a fraction of a degree over the last hundred years are, at best, uncertain. Fortunately, the global sampling problem is believed to be minimized for long-term trends. Weather patterns move around the globe, generally from west to east, and this tends to average out localized warm or cold events over time, in effect “smearing” them around the Earth. Because the oceans represent such a huge reservoir of heat energy, land temperature changes tend to follow ocean temperature changes over the long run. Therefore, long-term temperature trends over land are believed to be largely a response to long-term changes in the oceans. In fact, it has recently been demonstrated that if the oceans warm for any reason, global land areas can warm even more.16 This makes the oceans a potential key player in long-term climate change. It also means that what the scientific consensus views as a “fingerprint” of anthropogenic warming–stronger warming over land than over ocean–is also consistent with natural climate change. The same is true of the missing tropical upper tropospheric “hot spot” that was expected to result from manmade warming.17 It, too, would be expected to accompany a natural warming of the oceans. I believe that there is no unique fingerprint of anthropogenic global warming. Warming is warming, and it will look basically the same no matter what causes it. One would think that land-based thermometers should provide accurate estimates of temperature changes over the years, but there are substantial sources of error. The most significant of these is the fact that thermometers are usually placed where people live, and people tend to build things, replacing native vegetation with structures, roads, parking lots, and other manmade sources of heat. While vegetation tends to cool the air by diverting some of the sun’s energy into photosynthesis and evapotranspiration, most manmade surfaces and structures just sit in the sun and bake. When more people and more buildings and roads occupy the environment around a thermometer site, the air is heated in the immediate vicinity. This leads to an “urban heat island” warm bias, which typically increases with time as more buildings and parking lots are added. But the urban heat islands have virtually no direct warming effect on the rest of the Earth, since the coverage of the Earth by cities and towns is at most 1 percent.18 In fact, some of the thermometer sites used for climate monitoring have recently been revealed to be contaminated by heat exhaust from air conditioners and heat coming off the roofs or walls of buildings. All these influences add up to a component of the measured warming trend that is entirely local and that spuriously inflates our estimates of global warming. While the thermometer dataset developers claim they have removed this spurious source of warming, there is increasing evidence that much of it remains in the data. One recent estimate is that as much as 50 percent of the warming measured over land in the last thirty years could be spurious, due to various indirect effects of economic growth contaminating the thermometer data.19 If you have not heard about that important study, you have the news media to thank, since they decided it was not worthy of being reported. Satellite instruments provide our only truly global source of temperature information. The measurements are based on either infrared or microwave emissions given off naturally by the atmosphere, and they require carefully calibrated instrumentation to provide a stable long-term record. The instruments carry their own laboratory-calibrated electronic thermometers to convert their measurements of the atmosphere into temperatures. Probably our biggest headache in trying to monitor climate trends with satellites is not related to calibration but to the kind of Earth orbit the satellites are in. At least until recently, all the satellites had local observation times that slowly changed over the years. This would be like trying to determine climate trends from your backyard thermometer by taking measurements at noon one year, then at 1:00 P.M. the next year, then 2:00 P.M. the next, and so on. The day-night temperature cycle gets mixed in with whatever climate variability there is, and so it must be estimated and removed. This problem has been alleviated only since mid-2002 with the launch of NASA’s Aqua satellite, which carries extra fuel to adjust its orbit periodically and so maintain a constant observation time, year after year. This is the primary source of temperature data that John Christy and I use for monitoring global temperature trends. It is clear that a host of problems are involved in the determination of temperature trends and other statistics. All our measures of temperature variability are imperfect. And even if they were perfect, the huge amount of natural variability in the climate system–on time scales from yearly to millennial or longer–makes the definition of a temperature “trend” very difficult. While we can probably say with high confidence that the climate has warmed in the last 50 to 100 years, it is more difficult to say by exactly how much, still more difficult to say whether it is unprecedented or not, and impossible to say what any of this means for future temperatures. Next, we will examine one mode of natural climate variability that I believe plays a crucial role in what is popularly known as “global warming.”