Fishwrap Archive

Why can't we be friends?

Or how does an observation-based scientist get along in a theoretical world?

Dorothy M. Peteet is knee-deep in the mud in the middle of a bog. She is doing field research in Alaska, collecting sediment cores for pollen analysis. Even though the air is crisp, she is sweating from lugging around her field gear and exhausted from hand-drilling her corer as deep into the muck as it will go. The black flies and mosquitoes are biting enough to drive one mad and her stomach is growling angrily, but she could not be happier.

Out in the bog, Peteet feels connected to nature, to life. Through the tops of the spruces surrounding the bog she can see glacier-capped mountains. She can smell the crispness of the air, the pungent aroma of the bog. She can see and feel a menagerie of plant species surrounding and cushioning every step she takes. In some ways, science for her is an excuse to experience wilderness.

James E. Hansen is a physicist who appreciates fieldwork. But his domain is really numbers. He spends much of his days trying to reduce the messy, chaotic weather data collected from meteorological stations around the world to a series of mathematical formulas that will explain climatic patterns and processes and allow him to predict how those patterns and processes will change.

Hansen's job is a frustrating one, for the data often upsets simple, elegant hypotheses of how the climate system works, and frequently foils any attempt to accurately predict the future. But in these days of oft-predicted environmental catastrophe, he persists, hoping his model will warn of impending doom or the lack thereof so that appropriate precautions can be taken while unnecessary ones won't become a burden to humanity. Regardless of how people respond to what his model predicts, Hansen's main concern is that it works.

Peteet and Hansen work together at NASA's Goddard Institute of Space Studies (GISS) in New York. They are locked in a somewhat unholy alliance among earth scientists of the observational and theoretical, in a mutually beneficial relationship where antagonism formerly ruled.

Traditionally, theory-oriented earth scientists looked down on observation-oriented ones because the latter's work lacked a strong mathematical foundation, because it wasn't predictive enough. Observation-oriented scientists often complained that theorists' work wasn't grounded strongly enough in hard data, that theory often had little to do with the real world and its complexities.

Before computers, the empiricists more or less had the upper hand; most of the great theoretical advances in the sciences were based on a solid observational data base. When computers became widely available, however, the theoreticians had their revenge, and in recent decades observation-based science has often been frowned upon and sometimes actively discouraged. But lately the two camps have seen each other more and more as useful collaborators.

Hansen, for example, is one of the developers of the GISS general circulation model (GCM), a type of model that predicts climatic phenomena over long time and large spatial scales. Hansen and his colleagues at GISS were among the pioneers in using GCMs to model the Earth's climate, along with the developers of GCMs at NOAA's Geophysical Fluid Dynamics Laboratory in Princeton, N.J., NOAA's National Center for Atmospheric Research, in Boulder, Colo., Oregon State University in Corvallis, and at the United Kingdom Meteorological Office in Bracknell, England (although many of its researchers were at the University of East Anglia in Norwich).

Peteet is a palynologist. Her primary interest is in using pollen data to reconstruct vegetation cover by analyzing fossil pollen in lake, pond and bog sediments. Hansen needs the vegetation data from the past to calibrate and demonstrate the accuracy of the GCM. Peteet needs the interest in predicting climate to fund her own less economically important research.

Above a diner

GISS occupies several upper floors of a non-descript building -- officially known as Columbia University's Armstrong Hall -- at 112th and Broadway in New York. The building's only claim to fame is that a diner on the ground floor, Tom's, has become immortalized by the television comedy, "Seinfeld." While more people would know the building for the diner, what goes on upstairs may make it one of the most important scientific centers of the late 20th century.

The institute, a division of NASA's Goddard Space Flight Center, was formed in 1961 to do basic research on space and Earth sciences. Most of the early research at GISS was in support of planetary exploration, but in the 1970s, as government support of the Apollo program dried up, GISS personnel began to investigate other problems.

"In particular the idea came up of using space measurements to improve weather forecasts," Hansen said. "Jules Charney at the Massachusetts Institute of Technology had the idea that satellites should be able to measure temperature well enough so that you could use satellite data to initialize the weather models and get improved weather models."

The use of satellites in weather forecasting was enough to justify NASA's interest, so GISS bought a forecasting model from the University of California-Los Angeles. Eventually, GISS researchers began modifying the model so it could serve as a GCM.

Hansen was one of those researchers. The Iowa native earned a Ph.D. in physics in 1967 at the University of Iowa. For his dissertation, Hansen researched the radiative properties of the atmosphere of Venus, particularly the planet's greenhouse effect, whereby gases in the atmosphere prevent heat radiation from escaping to outer space. This leads to higher and higher temperatures on the surface of the planet.

Immediately following his graduation from Iowa, Hansen went to work at GISS planning experiments for an unmanned mission to the planet. Eventually GISS tapped his expertise for another project, encouraging him to work on the model of Earth's climate.

"I was marginally involved in that they needed radiation in their general circulation model," Hansen said. "I worked on that with a colleague of mine, a fellow student from the University of Iowa whom I hired to come here and develop the radiation for that weather model."

Shortly afterward, the satellite meteorology group moved to Goddard Space Flight Center. But Hansen was already hooked.

"By that time I got interested in the climate problem and decided to try to convert that weather model into a climate model," Hansen said. "Working with several different scientists here, we made changes in the physics of the model to make it appropriate for climate instead of weather, putting in processes that are important for longer time scales."

With that, Hansen's group began applying the GISS model to the study of the greenhouse effect. The group's major research breakthrough began with a paper (Hansen, J., Johnson, D., Lacis, A., Lebedeff, S., Lee, P., Rims, D., and Russell, G. 1981. Climatic impact of increasing atmospheric carbon dioxide. Science 233:957-966.) in 1981 based, ironically, not on the predictions of the group's climate model, but on the study of historical climate data.

"In the 1970s when I first started doing the calculations it was not a very popular topic," Hansen said, "because empirical evidence suggested that the earth had been getting colder the last few decades up until that time. So there were alternative suggestions as to what was going on, what was driving the Earth's temperature, namely that man-made aerosols were causing the planet to get cooler.

"I guess one of the first papers that we did that got some notice was probably published in 1981 in Science in which we pointed out that, first of all, that if you look for temperature data for the whole globe instead of just the northern hemisphere, then the planet actually had started warming up in the middle 1960s and that over the past 100 years up to the late 1970s there actually was a warming of about 0.4 of a degree Celsius. That was roughly consistent with the calculations that we did for the greenhouse effect.

"So the perception that the earth was getting cooler was really based on northern hemisphere measurements and not quite up-to-date data."

What Hansen and his colleagues suspected was that volcanic and man-made aerosols in the northern hemisphere, such as ash, soot and sulfur compounds, had been reflecting the sun's radiation back into space, in effect shading the Earth and keeping it cooler. But increasing carbon dioxide in the atmosphere was keeping the radiation from escaping to outer space, thus slowly driving temperatures upward. The paper was one of the first to demonstrate the possibility of a greenhouse effect using actual temperature data.

Their findings were reported by Walter Sullivan on the front page of The New York Times. Sullivan's article helped get attention and research support and helped spark a furious debate, which still goes on, on whether society should take steps to reduce greenhouse gas emissions before a man-made climatic disaster occurs. Part of that debate focuses on whether the predictions of GCMs are reliable enough to justify actions based on their predictions.

One way to assess the reliability of GCM predictions is to have them predict what the climate was like at some time past and compare those predictions to what is known about the period in question.

Data strike again

Peteet came from Georgia to New York in the 1970s to pursue Master's and Ph.D. degrees in biology at New York University. From her major professor, Calvin Heusser, Peteet learned palynology, the reconstruction of past environments from analysis of fossil and subfossil pollen grains.

While at NYU, Peteet began collaborating with Wallace Broecker and others at Lamont-Doherty Earth Observatory of Columbia University. Broecker, a paleoclimatologist, reconstructs past climates from geological and isotopic evidence. He suggested that Peteet and another researcher, David Rind, compare temperature predictions for tropical regions from the CLIMAP (Climate/Long-Range Investigation, Mapping and Prediction) project to temperature estimates obtained from pollen and alpine snow-line analyses. CLIMAP was a global modeling study of the climate during the height of the last ice age, about 18,000 years before present (B.P.).

Rind and Peteet found that the GCM predicted much higher temperatures than indicated by the paleoenvironmental data (Rind, D., and Peteet, D. 1985. Terrestrial conditions at the last glacial maximum and CLIMAP sea-surface temperature estimates: Are they consistent? Quaternary Research 24:1-22.).

Peteet earned her Ph.D. in 1983. While working on the CLIMAP study, she was hired by Hansen to work at GISS.

"We've always had the perspective that we could understand today's climate better by looking at the earth in larger context," Hansen said, "by comparing the earth to other planets as well as the earth's history over longer time scales. So we like to have some people who are actually doing research on other planets. It also makes sense to have somebody looking at paleoclimate."

Hare vs. tortoise

Even though Hansen and Peteet study the earth's climate, the methods they use are different. To model the Earth's climate, Hansen and other modelers start with fundamental equations for the processes they want to simulate and then try to build a mathematical model of radiation balance, oceanic circulation, the water cycle, and other processes. Where the equations are not known or the processes poorly understood, estimates are used (and can be a major source of uncertainty). For example, if the ability of the soil to soak up rainfall is not known, a modeler may set an arbitrary maximum amount of precipitation that can be taken up by the soil. Any rainfall (or snow melt) above that amount would be lost as runoff or evaporate, leading to further impacts on the climate.

Next, modelers translate these equations into computer code, adding in routines for data input and output and for communication with other components of the model. For complex models like GCMs, numerous submodels are made for various components of the climate system, like oceanic circulation, land surface hydrology, wind and precipitation patterns, atmosphere-vegetation interactions and energy transfers of glaciers and ice fields. The computer code is tested, first to see if it can run without crashing the computer, then to see if the input parameters and results appear reasonable.

After the individual submodels are evaluated, the entire model is assembled and again tested to see if it works as a whole. Once the modelers are satisfied it works, they can proceed to simulate the phenomena of interest.

Building and testing a large, complex model like a GCM can take months, and the model may undergo constant revision to keep up with an improved understanding of the processes studied. But once the model is complete in some form, it can be used for a number of experiments. Even with a large model, several experiments can be run over a short period of time.

Unlike modelers, field-oriented scientists may take years to obtain results from their endeavors. A palynologist like Peteet starts by planning a collecting trip. This consists of library research and consultations with colleagues to determine where best to obtain samples to answer a particular question. Then travel and other logistical arrangements must be made to get to and from the site. This step can be relatively easy if one wants to sample in easily accessible areas such as bogs in the Appalachian Mountains, but gets increasingly difficult for hard-to-reach regions like central Alaska, New Guinea or Easter Island.

A typical field day for a palynologist like Peteet begins with the hike to a site where, year after year, pollen is likely to collect and be deposited in layers of sediment. The best sites are usually bogs or pond or lake bottoms. Once on the site, she collects her samples by drilling into the bottom sediments and obtaining a number of long, narrow cores. Taking care to not disrupt the samples, she transports them back to the lab.

Each sediment core may contain thousands of annual layers. At the lab the palynologist must painstakingly date each layer, often comparing several cores to make sure that no single core is either missing a layer or has more than one layer per year. Within each layer, the pollen grains and plant macrofossils must be identified. The pollen and fossil material is then used to reconstruct what the environment must have been like for a specific period. As an example, if a palynologist found spruce and fir pollen and needles in a layer of sediment, he or she would conclude that the climate was cool at the time the sediment was deposited. But if oak or magnolia pollen was found, the scientist would conclude that the climate was warm.

It can take several years to work up a sample collection, Peteet said. Thus it can take several years to publish one's results. Compared to modelers, they are at a distinct disadvantage in the academic world of "publish or perish," where productivity and scientific worth is more often measured by the quantity, rather than quality, of publications one has.


In many of the earth sciences, observation-oriented scientists and modeling-based researchers are in conflict. Historically, observational-based researchers held sway when description and classification of natural phenomena seemed enough. For example, Walter Alvarez, in his new book "T. rex and the Crater of Doom," wrote that geological mapping was a necessary task, "to measure and describe the rocks of the entire surface of the world and to plot their distribution on detailed maps that would be the basis for understanding Earth history."

Much of the research in biology and ecology was concerned with the classification of organisms and ecological units such as forests, grasslands, tundra and deserts from various parts of the world. Most geographers spent their time dividing the world into regions. Climatologists, while having a stronger theoretical orientation, devoted a significant effort to classifying the climates of the world.

Two developments in the post World War II years brought an end to the dominance of the describers and classifiers. First came "big science," with an influx of massive amounts of government funding for research. In a 1976 article, entitled "Ecology since 1900," Robert P. McIntosh wrote that "Most pre-World War II ecologists were university faculty members working as individuals with a few students and minimal funding. The advent of large-scale federal funding and an era when "Grant Swinger" became an apocryphal figure in science greatly changed the expectations of ecologists, as well as those of other scientists." With "Big Science" came big management and changes in the way scientific productivity was measured.

The other development was the "quantitative revolution," largely brought about by the increasing use of computers. While theory had been a component of all the natural sciences, it was virtually impossible to statistically analyze massive amounts of data or model complex physical phenomena without computers. Arild Holt-Jensen, in the second edition of her book, "Geography: History and Concepts," wrote that "The rapid development of model building and the use of quantitative techniques could not have taken place without computers..."

As a result of the quantitative revolution, many primarily observation-based disciplines became more quantitative. Young scientists in the affected fields began looking to physics as an example of what all sciences should be like. Herman H. Shugart, in the book, "A Theory of Forest Dynamics," described the situation for biologists and ecologists in this way: "Biology and ecology, in particular, are "young" sciences that are wobbly on their still-weak mathematical underpinnings. Because modern ecology has placed a strong emphasis on quantitative methodology ..., one sometimes tends to view mathematical sophistication as an index of scientific evolution much in the way that the elaboration of sutures is an index of biological evolution in the ammonoids." One ecologist, Scott Collins of the University of Oklahoma, satirically defined the perception as "physics envy."

The better way

Like most revolutions, the quantitative revolution began devouring its young. Bitter debates raged between older, classically trained scientists and younger, more quantitative scientists. The old guard argued that the models were not grounded enough in hard data, that they were not realistic enough. The revolutionaries countered with the argument that a true science must test hypotheses, must be predictive, and that observation in the absence of hypothesis testing thus did not constitute a scientific endeavor.

Gradually, throughout the 1960s and 1970s, the quantitative scientists began to dominate. As they did, theoretical or modeling studies rose in esteem while old-fashioned observational studies fell out of favor. A scientist had to be testing some theory or hypothesis, or else forgo funding, publications, and eventually a job.

Peteet, and other observationalists like her, survive because they have adapted to the current scientific environment. She has learned how to address important questions without sacrificing her penchant for observation and her desire to go into the field. Yet, scientists like her who have been able to adapt still have a hard time earning the respect of their peers in the modeling camp. One event stands out in Peteet's mind as an example of this disrespect.

Several years ago, she attended a conference of the American Quaternary Association. A modeler there named Pat Butler was speaking to an audience of about 300 paleoecologists, paleogeologists and other paleoscientists. Butler told the crowd that there were two approaches to understanding the world.

"One is from the bottom up," Peteet said as she recounted the scene. "You map glacial moraines, study the rocks, figure out the pollen record, look at the landscape."

"The other approach is to start with a model and work down," she continued. "And he said that clearly the modeling approach is the best because we could do it globally, we could change the model at any time to put in whatever we want.

"All these people, their mouths dropped because this person could come [to the conference] and say that. Sure you can look globally with a model and sure you can put in anything ... But you're not getting reality."

"Part of me understands what he was trying to say," Peteet said, "but he was totally misunderstanding data people. How does he know his model was right?"

The modeler in Peteet's example was perceived by many of those at his talk as rather arrogant. But in the early 1970s, as the quantitative revolution was beginning to take hold, the observationalists were the arrogant ones.

Shugart, a professor at the University of Virginia, began his career as a teen-aged ornithologist and museum collector, in essence a classic observation-based biologist. After receiving his Ph.D. in ecology from the University of Georgia, Shugart went to work at Oak Ridge National Laboratory where he was a co-developer of one of today's most widely used models of forest succession, or the change in forest composition over time. He has also written several books on the potential effects of a changing global environment on the world's ecosystems. Initially, however, he had to struggle for acceptance by his colleagues.

"People don't realize I was primarily a data/field observation person most of my early life," said Shugart. "I was a museum collector when I was 12, unknown to the federal government when they licensed me. I was running bird surveys when I was 14 and publishing them."

Shugart's first job after finishing his Ph.D. was to help build models for the International Biosphere Project, a large-scale cooperative study of entire ecosystems around the globe. He had to talk to post-doctoral researchers at Oak Ridge about their studies. While doing so he felt that many treated him as if he were beneath contempt.

"I'd meet these guys who were supposedly field/data scientists who, clearly at that time, had quite a bit of chauvinism about modelers and didn't know anything about what it was like to be in the field," Shugart said. "I'd sort of lay in wait for them because, in terms of most of them, I had forgotten more about what it was like being in the field than the bastards ever knew."

Causes of conflict

The use of models is not a recent development in the sciences. In fact, anyone who has calculated an average or placed a bet on a horse after studying a racing form has applied a model to data, in the first case, or used a model to make a prediction in the second.

How, then, do such rifts arise within a scientific discipline? Peteet feels like differences in training among different groups of scientists make it difficult for observationalists and theorists to communicate. She cites the struggles that some ecologists have in learning how to use satellite data such as from the Advanced Very High Resolution Radiometer (AVHRR), a weather satellite that measures radiation in bands that are also useful in the study of vegetation.

"When I was going to school my training didn't include global ecosystems, asking the same questions that are being asked now," Peteet said. "So part of the problem is ecologists trying to catch up in looking at the earth from a satellite instead of on the ground.

"But the modelers, too, don't appreciate the complexity of the ground. They override it and think everything can be simplified to certain AVHRR bands from a satellite. The ecologists go nuts because within the ecosystem there are all those complexities that we don't understand."

Hansen agreed. Throughout his career, he has had other people point out complexities that weren't considered in his models.

"It's probable some people know too much about something, and they see you're not explicitly representing that knowledge or process in that model so they're skeptical," Hansen said. "Of course there's some validity to that but its easy to become so skeptical that you can't do anything."

Models, however, can be complex in their own right. Data-oriented scientists, who often lack much training in advanced mathematics, can be at a loss when trying to understand what a modeler is trying to do.

"Another problem I see between the modeling and data communities is that the models are so complicated that the data people can't understand the nuts and bolts of the model," Peteet said. "So most data people aren't going to go try to figure it all out."

Observationalists and theorists tend to run in different scientific circles. The intellectual separation between the two groups of scientists hinders communication, and thus, cooperation.

"They are often in different organizations, in different places, working amongst their peers rather than with each other," Hansen said of the two groups. "You need to actually have the same people working in both areas in order to make it work. You need to have the same goal for the different disciplines."

Shugart speculated that a difference in scientific philosophy leads to much of the conflict.

"There's a philosophy ... that raw observations will become good in science in time," Shugart said of the observationalists' viewpoint. "It turns out that, with rare exceptions, that there's very little proof that that is actually the case. What's normally happened is that you've had theories and stuff that has structured observations and that, for the occasional people who are cluey enough to observe the right stuff, their data has become very valuable."

Shugart cited Charles Darwin, credited with developing the theory of evolution, as one scientist who happened to observe the "right stuff."

"Darwin was a data hog," Shugart said. "This guy goes on one of the early voyages of exploration essentially, and hauls back tons and tons of crap on this boat that sort of sits in a shed for 40 years.

"But the really important thing he did was the synthesis, not the data collection. I mean, nobody gives a damn about Geospiza [the Galapagos finches] on the Galapagos Islands. They are just more out of eight or nine thousand species of little brown birds that are out there in the world. But the fact that he organized, built a theory around them and a lot of other stuff he observed is what's important."

Shugart feels that observationalists tend to overestimate the importance of data in itself.

"I think that most people who collect data feel like they're contributing in a fundamental way, like data are immutable," Shugart said. "If you've got data, you've built something that no one is going to tear down. It's going to build science. In fact, not a lot of it really builds science very much, which is a point that I think data-collecting people are often kind of nervous about. It's sort of like someone telling you there's no Santa Claus when you're 5 years old."

Coming together

William F. Ruddiman is a paleoclimatologist at the University of Virginia. For years, he has straddled the line between observationalists and modelers, beginning his career onboard Lamont Doherty Earth Observatory research vessels, attempting to reconstruct past climates from sediment cores taken from the ocean floor. He has seen the conflict between observationalists and modelers, but it has not been as pronounced in his discipline as it has in others. He believes the two groups are coming closer together.

"My impression is that there are more people now, in my field, that definitely admit that you need both," Ruddiman said. "You need data and you need modelers."

Ruddiman believes in taking a two-pronged approach to his research, using both data and models to get a better understanding of the climate system. "There's an independence there that gets you testing the assumptions of the data and of the model and of the boundary conditions put into the model," Ruddiman said.

"It's wonderful when it works because it gets you out of the limitations of just the data or the model," he said. "The data-model comparison is testing both against each other."

Hansen feels that Peteet's paper with David Rind in 1985 was an example of a two-pronged approach that worked. Without her data, the GISS GCM would have been predicting unreasonably high temperatures in the tropics during the height of the last ice age.

"Dorothy has made a highly important contribution in this issue about whether low latitudes were really colder," Hansen said. "CLIMAP data didn't have the low latitudes much colder during the ice age than they are today. She pointed out evidence on land that it was a lot colder during the ice age ... Working with the modeling people here, its very hard to understand how it could have been that much colder in the high latitudes without being much colder in the low latitudes."

Ruddiman sees observationalists and modelers working more and more closely together, such as in the case of Peteet and Hanson.

"A good scientist who goes to gather data will always be applying his data to models and vice versa," Ruddiman said. "In the ideal, the whole is greater than the sum of its parts."


I'll keep this short. I interviewed Hansen and Peteet in person, with my infant son, Malcolm, in tow for baby relief. Shugart and Ruddiman were interviewed by phone. A lot of the background information I needed was obtained from the World Wide Web or from my personal research library.

David M. Lawrence