Most scientific measurements claim the definitive precision of a football score. No one, after all, disputes the length of a centimeter, the mass of a gram, or how many points a touchdown is worth.
A new analysis of climate data over more than 100 years, however, suggests that finding the hottest or coldest day, week or year may be less like reading a football score and more like deciding which college teams play for the national championship: Raw empirical data is important, but it isnít everything.
Using 108 years of weather data gathered by several sources, Alabama's state climatologist, Dr. John Christy, tried to answer what most people might consider a simple question: "When was the hottest summer in North Alabama?"
Christy, a professor of atmospheric science at the University of Alabama in Huntsville (UAH), presented his findings this week at the annual meeting of the American Meteorological Society in Orlando, FL.
As state climatologist, Christy says he frequently is asked for a historical perspective on recent or developing weather events.
"People want to know if it has ever been this hot before, or this cold or wet or dry," Christy said. "Hampering our ability to answer these questions is the lack of consistently-observed weather conditions over the time-scale of a century or more."
To determine the reliability of a climate data test sample, Christy spent about a year reading through the handwritten notes and data entries of volunteer "cooperative" weather watchers in North Alabama as far back as 1893.
Merging that data with information from stations operated by the National Weather Service, the U.S. Army and other agencies, he went looking for north-central Alabamaís hottest summer (June, July and August).
What he found were enough breaks, bends and gaps in the data to change the 108-year temperature trend for the area from cooling by 0.07 degrees C per decade to cooling by 0.13 degrees C per decade. Over the course of a century, that means the average temperature has dropped about one degree Fahrenheit more than previously reported in the unadjusted data (-2.34 degrees F vs -1.26 degrees F).
"In essence, what we're doing is identical to how the BCS computer systems work, using different standards to compare similar data that doesn't overlap," said Christy. "How do you compare temperature segments from different places at different times that have no overlap? How do we relate something from 1925 and 2001? The instruments are different, the exposure is different, and development around the site is different -- just like the football teams from Nebraska and Oregon are different."
Many of the data gaps and breaks were minor, such as when an observer in Madison, Ala., missed 16 days when he was struck with the flu and unable to perform his duties during the great influenza epidemic of 1918. This same observer missed nine days in 1924 when a microburst or a tornado destroyed the thermometer shelter and the thermometers.
Some problems were easily identified. In April 1963, for instance, a weather observing station in Athens, Ala., was moved about 20 feet. This led to an almost instant 0.44 degrees C (about 0.8 degrees F) "warming" at that site, affecting data recorded over the next 31 years.
"In the climate record it would show up as a spurious warming, but it was probably that the earlier temperature readings were too cool," Christy said. "I've seen a map of that site and it looks like it was close to the shade of a tree in the afternoon. I'm guessing it was well shaded. But that's all it takes. Move the thermometer 20 feet and youíve got a new climate trend."
Other apparent breaks in the data had more challenging solutions. In 1964 the U.S. Weather Bureau Office at the Huntsville airport was about 1.2 degrees C (more than 2 degrees F) cooler than six nearby stations; August was about 2 degrees C (more than 3.5 degrees F) cooler than the surrounding area.
"During the first five days of that period about 125 mm (more than 4.75 inches) of rain fell in the area," Christy said. "In 1964 the Huntsville weather bureau office was on the edge of a swampy area southwest of the city. It seems likely that evaporative cooling related to saturated or flooded ground in that immediate area kept daily maximum temperatures at the weather station cooler than it was at sites with better drainage.
"This is a case in which topography appears to create a real cooling, rather than an artificial break in the data."
Other problems remain mysteries.
Between 1994 and 1995 the summer high temperatures recorded in Athens jumped about 1.5 degrees C (more than 2.5 degrees F) for no apparent reason, Christy said. "No information from official records or from interviews with observers revealed a change in the site, the instrument or the time at which the data were gathered."
But Athens didnít see the largest unexplained "climate change."
"June, July and August temperatures in Oneota, Ala., were 2.37 degrees C (more than 4 degrees F) warmer in 2000 than in 1999," Christy said. "Seven other stations reporting for both years indicated that the summer of 2000 was from 0.09 degrees cooler to 0.78 degrees C warmer than summer 1999. No obvious explanation for Oneota's relative warmth could be determined, but the large difference for this co-op station implies that a significant, yet-to-be-explained change did occur."
After all of the quirks and kinks were worked out, Christy found that with an average high temperature of almost 35 degrees C (about 95 degrees F), the summer of 1925 was probably the warmest in the North Alabama climate record. But the summer of 1925 only qualifies as probably the warmest because five other warm summers (1930, 1936, 1943, 1952 and 1954) were all within the plus-or-minus 0.4 degrees C margin of error.
Not exactly the stuff from which compelling headlines are written.
While the bottom line result of Christy's research provides little more than a climatological footnote, the implications of this work reach deep into the global warming debate and the hundreds of billions of dollars that are at stake in the U.S. and abroad.
"When someone talks about the entire globe having its hottest year, the public needs to understand how difficult it is to find scientifically reliable long-term climate data just for one spot in the U.S., where we have extensive networks of weather stations and careful data storage," Christy said. "There are margins of error associated with those statements, but those errors are rarely communicated in the media."
Christy said climate researchers rely on data from rural stations, which were thought to be more reliable than urban stations. Data from urban stations can be influenced by spurious warming due to asphalt and concrete replacing grass and trees.
"But data from rural stations can also have many problems and should not be treated as continuous records for climate work," Christy said. "There are discussions going on among state climatologists about how to make the appropriate adjustments to this data.
"Some of us are advocating a site-by-site analysis and correction based on empirical data from nearby stations, which you would use to compare with data from both sides of a break point. When you do that, usually the shift is negligible.
"Sometimes it is remarkable."
The results of Christy's study have also been accepted for publication in the Bulletin of the American Meteorological Society. - By Phillip Gentry
[Contact: Phillip Gentry]