ATHENS, Ga.--Let's say you own a hundred acres of forest in the mountains of North Carolina. You want the land to stay beautiful, but you also want to sell some of the timber to put your kids through college. Years ago, your best bet would be to cut and pray, but not any more.
A new decision support system that researchers from the University of Georgia have helped design could make your job vastly easier. The system, whose development is directed by the U. S. Forest Service and called NED, is near completion. Instead of having to make a decision strictly on board footage, a manager will soon get help designing a program that might, for example, maintain a continuous overhead canopy, discourage exotic plant and animal species, enhance habitat for the black bear--and still focus on cutting the proper trees for your college fund.
"The development of techniques to deal with these once-intractable problems has been amazing over the past few years," said Dr. Donald Nute, head of the Artificial Intelligence Center and the department of philosophy at UGA. "But in order for them to work, we first have to have deep knowledge of a problem area."
Ecosystems are complex hierarchies. Managing for a single species or problem has sometimes led to disastrous results, and it is now clear that multiple-use management is essential for the best use of any area. The problems are vast and entangling. How can you manage for timber yield if it threatens an already endangered species? How can you manage for an endangered species if it means you won't have college money for your kids?
The answer often lies in the use of decision support systems (DSS), computer programs that help managers make decisions in situations where human judgment is important but where limitations on the ability of judgment impede decision-making. These systems have been in use, in one form or another, for nearly 20 years, but with increasingly powerful and sophisticated computers, DSS has become a vital component of land management. (At least 24 different DSS's are under development for ecosystem management by the Forest Service alone, and NED is one of those.)
Nute and faculty and graduate students in the Artificial Intelligence Program at UGA have been involved with NED for several years, helping to design individual parts of the overall program. Currently involved are four graduate students, Mingguang Xu, Shanyin Liu, Huasong Yin and Geneho Kim.
"We first got involved in 1990 when Michael Raucher, then of VPI, got in touch with us about working on decision support systems in forestry," said Nute. "These DSS's don't take control of the processes like expert systems. The idea of a DSS is to provide support when the user desires it."
The work began with the Forest Management Advisory System (FMAS), which was designed to manage a single stand of red pines that were all the same age. The UGA researchers developed the DSS tools for the system, designing modules for growth-yield and for harvesting. Also linked were hypertext documents that added information at appropriate times for timber managers to consider. The UGA team also wrote an interface to a so-called "legacy code"--a module that could help FMAS interact with programs already available in other programming languages.
While FMAS was designed to manage a single-species, single-age tree stand, NED is taking the process to a new level. The new system, named NED for Northeast Decision Model, reflecting on the part of the country for which it was originally targeted, provides decision support for multiple stands with multiple species, trees of differing ages, timber yield, wildlife, water and even aesthetics. More than 100 individuals from the Forest Service, universities and non-industrial private firms have been involved in designing NED.
"This is real knowledge engineering," said Nute. "We have to work with experts build a model that reflects what they do. We could start, for instance, with a simple English description of the problems, but building the program must always involve two steps: knowledge acquisition and knowledge representation. The first step is where we get the knowledge from the experts, and the second is where we convert it to a form a computer can use."
In addition to its knowledge base, a decision support system must also have what designers call an inference engine, a program that can take data from the user and from problem-solving knowledge then combine them to produce specific recommendations. At present, UGA is the only collaborator on NED designing inference systems and knowledge bases.
The work of the University of Georgia on the problems continues on a broad front. Researchers here are designing an FMAS-like system for three major Georgia tree species, loblolly, slash and longleaf pine. They are also working with scientists at the Forest Service's Southeast Forest Experiment Station and the Coweeta Hydrologic Research Laboratory in North Carolina on systems for the southern Appalachians to include such aspects as nitrogen use, economics and wildlife abundance.
The Forest Service is testing beta versions of NED, and the first full-scale prototype will probably be released by the end of the summer, Nute said.