By Nick Stockton
An advanced computer technique called Agent-Based Modeling (ABM) is helping researchers study how the intricate dynamics between people and the places they inhabit explain changes in both over time, according to researchers speaking Feb. 18 at the 2011 AAAS meeting in Washington, D.C.
Alex Zvoleff, a doctoral candidate from San Diego State University, explained how ABMs are helping him and his colleagues study the linkages between land use and land cover in Nepal.
An ABM is a dynamic virtual environment populated by virtual people, who are called decision-making “agents.” These agents act out lives that are shaped by their immediate context: their friends, neighbors, social programs, and physical surroundings, including the natural environment.
In developing an ABM, researchers use data from real people—including that collected from surveys and observations—to help determine how their life choices affect the natural environment now and how the environment will affect their choices in the future.
In turn, the model must also account for the dynamic relationships of the environment, and how changes in these affect subsequent decisions by the agents. Just as in real life, fluctuations in animal and plant populations, soil erosion, deforestation, and natural disasters all have an effect on agents' decisions.
“The challenge,” said Zvoleff, “is knowing which factors influencing decision-making are relevant to the model.” His research explores the tendency of people in the agricultural Chitwan Valley in southern Nepal to marry and reproduce at younger ages relative to urban populations. Zvoleff created a model to show how these decisions affected land use and land change in the valley.
Zvoleff, with help from his advisor Li An, uses decisions regarding marriage and childbearing, and their associated feedbacks, to predict trends of land-use and land-cover change. In his model, the population grew around established towns and established transportation lines.
As expected, deforestation increased as the demand for fire and building wood increased, which led to impacts on soil erosion. These changes are then reflected by the reproductive and settlement decisions that future agents make.
Zvoleff uses a comprehensive demographic data set of the Chitwan Valley that has been compiled for more than 10 years by researchers at the University of Michigan to provide information about the people of the valley. His model, which he has named ChitwanABM, also accounts for many other factors: landscape features, such as hills, rivers, and forests, all of which can physically restrict decisions; social relationships and family ties that affect an individual's decision-making; land-cover types that influence how agents use the land; and transportation routes, which show accessibility and migration patterns.
The “agent” in the model can represent an individual, a couple, or an entire household, all of which have different effects on outcomes.
Although agent-based modeling is not a new technique, it is growing in popularity to scientists embracing a holistic approach to research, according to Zvoleff. Since each ABM is programmed uniquely for the problem it addresses, the implications are specific to the culture, policies, and landscape in the study area.
Geographic tools are important for representing the data so researchers can visualize their results, Zvoleff said. These maps show “hot spots” of population growth, trends in land-cover change, and how development affects endangered-species habitat.
Zvoleff emphasized keeping the model simple. As Sarah Wandersee of SDSU elaborated, “When the model gets too complex, its results reach too far beyond the supporting data, and you can draw inappropriate conclusions.”
Despite acknowledged limitations about the ability of the models to handle extreme complexity, the potential applications offer exciting opportunities for the next generation of researchers, Zvoleff said.
“I'm very interested in climate," he said. "I would love to see the bigger impact of climate variation and how it will affect human migration.”
Nick Stockton is a recent graduate of the Geography program at Portland State University. In addition to writing about the practical applications of geography, he publishes a blog about sailing. He can be reached at stockton_nick@hotmail.com.
