summer, 2003

Detecting environmental trends

By Susan Bacon

Scientists work to describe the world in numbers. They measure environmental descriptors ranging from temperature and plant growth, to radiation and precipitation, oftentimes bringing together data sets from satellites, field expeditions and monitoring towers. But after all of the numbers are collected, a key question emerges: Is anything changing?

Betsy Weatherhead

Betsy Weatherhead, a scientist with the Cooperative Institute for Research in Environmental Sciences and leader in the field of analyzing environmental trends, sifts through data sets from various projects and uses statistical methods to uncover changes in these physical processes.

As human impacts on the environment become deeper and wider reaching every day, and trends such as increasing average temperature become clearer, Weatherhead's work becomes more interesting to policy makers and the general public.

"People have always asked whether life is changing," Weatherhead said. "But the scientific study of trends has accelerated in the last ten years, and interest in trend studies continues to grow."

Scientists trying to detect environmental trends also find Weatherhead's work useful because it can help them find the most cost-effective research design by describing, for example, whether a large number of sampling sites or high quality measurements would be more appropriate to find a certain trend.

Finding a trend is not as straightforward as it may at first seem: It requires more analysis than simply plotting a set of measurements over time. One challenge is that most environmental trends are often very small compared to natural variability. For instance, a trend of a 1-degree Celsius increase in average temperature over several decades would be hard to see when average annual temperature naturally varies - sometimes by several degrees- - from year to year, and when seasonal temperatures vary by tens of degrees. Also, even the best thermometers and other instruments have some level of associated error, which adds uncertainty to measurements and makes it harder to detect a trend, especially when measurements span decades.

The variety of acceptable definitions for trends presents another challenge. Some researchers say a trend is a change over time, while others insist it is a change over a long period of time. Some researchers consider trends to be only linear changes, so would not define the emergence of the ozone hole as a trend. Some say trends are only changes that are anthropogenic, or caused by people.

Weatherhead said she likes to consider all definitions of trends when she looks at a data set. She also keeps her mind open to trends in related measurements. For instance, if she's looking at a change in temperature, she also considers changes in winds because trends in these two factors often influence each other.

"I try to look at data from an integrated perspective, similar to what modelers do," Weatherhead said. "Modelers try to consider the whole picture, and to describe interactions among all of the parts."

Mauna Loa: Stratospheric Aerosol

Trends in data can be difficult to detect. For aerosol measurements at Mauna Loa, large contributions from the eruption of Chichon (1982) and Pinatubo (1991) mask distinct signal in background stratospheric aerosol levels.

A typical trend study takes Weatherhead two to three years. First she looks over the data set and talks with the researchers who collected it to find out what factors naturally influence the data. For instance, if one temperature sensor was located near a black road that was later painted white, it would give different temperature readings before and after the road was painted. Or if a volcano erupted during the timeframe of the study, it could have effects on the numbers that were recorded.

Although this process of describing the quality and uncertainty of the data takes up the majority of Weatherhead's time, it is the part of her work that Weatherhead said she finds most enjoyable. "What I enjoy most is the opportunity to work with people in different fields, whose outstanding expertise, knowledge and insight illuminate the environmental processes that the data describe," she said.

Next, Weatherhead teams up with other statisticians to develop a statistical model that's supported by the physics of the study and can be used to derive the trends. She said she does not try different approaches while processing the statistics, because that could lead to "shopping around for trends." After she finds a trend or a lack of a trend, she does a post-analysis to examine the phenomena more closely and learn, for example, if measurements from certain years dominated the trend. And finally, she communicates the results back to the scientific community through scientific publications.

In all of her projects, Weatherhead said she doesn't rely solely on statistics; she uses the broader body of scientific knowledge to understand how the systems she's examining are supposed to act.

"The big question is what's natural variability and what's a real signal or trend," Weatherhead said. "Statistics can get us part of the way there but only part of the way there. We can never make a statement without an understanding of the system."

That knowledge lets Weatherhead tease out real trends from natural variability and from inconsistencies or errors in the measurements. This process has led her to trends in a variety of measurements. For instance, in the late 1980s, after nations around the world agreed to stop releasing chemicals, such as chloroflurocarbons, that were destroying the ozone layer, a study came out suggesting that there was a decreasing trend in ultraviolet radiation. Researchers had expected that this radiation, which is blocked by the ozone layer, would be increasing. Weatherhead read the study, then took another look at the numbers with the researchers who wrote the original report. Together they found that the original report was unsubstantiated, and that ultraviolet radiation had not been decreasing.

While examining data sets, Weatherhead is careful not to ignore trends that do not appear to be statistically significant. For most scientists to consider a trend significant, they usually have to be 95 percent confident that the data are showing that trend. But, Weatherhead said, sometimes a trend can be important before the 95 percent threshold is attained. The lack of a statistical trend might just mean there aren't enough years of data to show a trend.

"If we wait until it's at 95 percent, we've waited until it's basically obvious," Weatherhead said. "We need to be careful and thoughtful when we look at trends. Just because a trend isn't at that level of certainty doesn't mean there's nothing there."

Now Weatherhead is evaluating the slight rise in average temperatures -- about 1 degree Celsius over the last century -- that scientists connect to global climate change. By comparing temperature measurements taken from surface instruments and from satellites, she is testing whether patterns in the observed warming match what would be expected if the warming were due to anthropogenic causes. As with all of her studies, Weatherhead looks closely at mainstream theories as well as alternative theories for answers to why these trends may be taking place.

"I take alternative theories very seriously because I think they can only strengthen our knowledge," Weatherhead said. "I have yet to find one that refutes the classical theory of climate change, but sometimes they bring new questions to the table."

Looking through this climate change data as well as the many other data sets she works with, Weatherhead is not just trying to answer a single question about how the environment is changing. Rather, she said she's always finding new questions to ask the data, and new answers that can ultimately lead to more questions.