PREDICTING THE PAST: NEW PREDICTIVE MODEL HELPS UA ARCHEOLOGISTS PINPOINT SITES

FAYETTEVILLE, Ark. - Using innovative mapping techniques, University of Arkansas researchers have developed a predictive model that enables archeologists to more accurately surmise where long-buried sites are located on a given landscape.

The basic principle addressed by the model is how fundamental physiological needs for water, heat and light are met, said Marvin Kay, associate professor of anthropology, who devised the modeling technique with doctoral student Richard Allan.

"We’re interested in understanding how people use the land over time. We do that by identifying the unique physical constraints to settlement that exist in any given landscape," Kay explained. "Those constraints remain the same whether you’re looking at 10,000 years ago or 100 years ago."

The model works so well that its predictions prove accurate not only for prehistoric sites but for evidence of human habitation throughout the ages, even as late as the 19^th century. Further, for archeological sites already identified, Kay’s model can tell archeologists the season during which the settlement was likely inhabited, based solely on the site’s location.

Kay began developing the predictive model five years ago while working on a National Forest Service project in Taney County, Mo. His technique uses GIS mapping to blend information about variables such as topography, climate, even human endurance - factors that directly impact settlement decisions. Kay then analyzes the overall picture of the landscape to determine the most likely spot where people would settle at a specific time of year.

The layering of information over a landscape and the attempt to predict areas of human habitation are not new endeavors in the field of archeology. What makes Kay’s approach both innovative and effective are the factors he focuses on - the elements that he believes impact the way people have chosen to use the land throughout time.

As an experienced archeologist, Kay understands that the most significant of these limiting factors is the availability of fresh water. But assessing an area’s water resources is not just a matter of counting lakes and streams, according to Kay. Researchers also must consider the unique climate of a region and account for seasonal fluctuations in temperature and precipitation, which can significantly affect a landscape’s appeal to human occupants.

For example, by studying water records for Taney County, Kay found that this area and the Ozarks in general experience extremely dry conditions during the summer. As a result, people crossing the land in the summer tended to settle in the lowlands between hills, where they were more likely to find moisture. In the winter months, when water was abundant, people gravitated toward the uplands and camped near the headwaters of streams.

It’s well-known in archeology that human encampments stick close to the water’s edge. Kay attributes this proximity to the difficulty of transporting water over any distance, and his predictive model takes into consideration the energy a person exerts to survive in a given landscape - whether carrying water or climbing slopes. Richard Allan, a doctoral student in the UA Environmental Dynamics program, studied physiology texts to estimate the caloric cost of such activities and integrated them into the model.

As a general rule, people didn’t move away from streambeds until the 19^th century, when the use of draft animals, wells and dams made it easier to trap and transport water, Kay said. "But occasionally we still find evidence of encampment in unexpected places - not just at a greater distance from the water but situated on steep slopes," he added.

"Often archeologists dismiss these as isolated finds, but if you want to understand land use, those sites become very significant - they mean that some other factor was attracting people to stay there."

Using the latest version of a software program called R.sun - which measures radiant sunlight values in watts/meter-squared/hour - Kay began to examine sun exposure on different aspects of the landscape at different times of the year. As expected, features of the land experienced varied intensities of heat and light, depending on the season. But Kay found that these variations in sunlight correlated with the placement of prehistoric archeological sites.

Seventy-two percent of 418 prehistoric archeological sites were identified to a season of the year in the Taney County project. Of the "isolated find" sites situated on steep slopes, the vast majority were located in areas that caught maximum sun exposure during the winter season, Kay said. The finding indicates that people made settlement choices not only based on water availability but also on warmth and light.

By integrating information about topographical features with that of water sources and availability, physiological restrictions and radiant energy variations, Kay can create maps that tell archeologists where sites - whether prehistoric or modern - are likely to be located.

"Seasoned archeologists can intuitively guess where sites are most likely to be," Kay said. "What we’re trying to do is provide the empirical evidence to back those hunches up - to show that there is a science and a system to guide this work."

The success of Kay’s predictive maps in Taney County captured the attention of the National Park Service. In 2000, the NPS offered Kay an initial grant of $15,000 and gave additional funds of $35,000 this year to apply his model to two Civil War battlefields: Wilson’s Creek in Missouri and Pea Ridge in Arkansas.

In collaboration with historical archeologist Douglas Scott of the NPS Mid-West Archeological Center, Kay plans to test the accuracy and versatility of his model by mapping the battlefields and attempting to predict the location of artifacts, dating from prehistoric times up to the years of the Civil War.

Wilson’s Creek and Pea Ridge make excellent test sites not only because they’re federally-regulated historic parks but also because these landscapes have been repeatedly used by people for more than 11,000 years.

"These parks contain an archeological record dating back to the Ice Age," Kay said. "It’s common to find sites that contain 10,000-year-old artifacts alongside objects that are less than 1000 years old. That’s because people throughout time have used the same process of reasoning to determine the best place to settle - they consider the same factors, so they settle the same spots over and over again. That’s part of the reason this model works."

Having developed and analyzed the maps for both battlefields, Kay and Scott will dedicate the next four years to controlled excavations of each park. The researchers selected 30 meter-square plots at random, each of which will be examined for evidence of habitation. Those that yield artifacts will be compared to the model to determine if they match Kay’s predictions.

"This is a very different archeology than we used to talk about," Kay said. "New technologies and applications are giving us a whole new way to pursue our field. I think this project has the potential to contribute a lot to Arkansas and Missouri and to the science of archeology as a whole."