Mapping Information in Action
Geographic information systems can show what may not be obvious at first glance
by Judson Mead
The Town of Amherst, New York, where the UB North Campus is located, has a population of 130,000 people and about 1,100 white-tailed deer. Deer can occasionally be seen on the campus; a few years ago a disoriented deer crashed through a large window into a dining hall in the Ellicott Complex and, after a few panicked moments, escaped again, reportedly unhurt, back into the "wilderness."
While such incidents are not hazards typical of life in Amherst, neither are they rare. As new housing developments pushed into forest and agricultural land over the past two decades, people began to complain about deer damage to their gardens. Reports in the local media started to refer to the "Amherst Deer Problem," with road accidents involving deer soon becoming the focus of such stories.
By the early 1990s there was a universally accepted sense that there were a lot of deer in Amherst and that they were responsible for a lot of damage. But no one knew anything much more specific than that. In 1994, the town sought to reduce the problem by thinning its deer population with a "bait-and-shoot" program. This aroused intense opposition from some town residents seeking a more humane solution; they eventually won an injunction against continuing the program without more information about the problem and alternative remedies.
In the Geographic Information and Analysis Laboratory, a large windowless space in another building of the Ellicott Complex (not the one the deer crashed in and out of), faculty and students of UB's Department of Geography pursue the application of computing power to problems that have a geographical component. Their tools are geographic information systems, or GIS.
GIS is a computer system capable of assembling, storing, manipulating and displaying geographically referenced information; i.e., data identified by location, according to a U.S. Geological Survey publication.
UB has been among the leaders in studying the development of GIS since the early 1970s, when the field was in its infancy. Duane Marble, now a professor of geography at Ohio State University, brought with him an interest in computer-assisted transportation modeling when he became a member of the UB faculty in 1973. Hugh Calkins, professor of geography, who joined the department two years later, remembers the state of the art at that time: "There was no commercial GIS. An early study we did was to survey all GIS available. Nothing was standardized, and most of what existed at that time was gone a few years later."
Marble and Calkins's surveys of early GIS work contributed significantly to its subsequent development. With the advent of the microcomputer and the standardization of operating systems, commercial GIS took off. The first major GIS software packages appeared in the early 1980s. Now a $2 billion-a-year industry in the United States, GIS is expanding at an annual rate of 20 percent.
In the early days of GIS, UB geographers tried to do computer mapping projects on the university's mainframe computer, but found it next to impossible to get enough time on the system to handle the amount of data such work requires. It was for this reason that the department established the Geographic Information and Analysis Laboratory-one of the first dedicated GIS labs on a U.S. university campus.
The primary feature of a GIS product is its ability to organize large amounts of data into a visual display in which certain kinds of information (magnitude, frequency, relationship, for example) are easy to see in geographic space. This apparent simplification-something like the display of 911 calls related to a particular criminal activity during a particular period (see chart on page 13)-is actually the expression of a huge amount of underlying data: the digitized street map itself, the data files that represent all addresses as x-y coordinates on that map, the data file of all 911 calls and the subset of those calls that are of interest, and the merger of those calls with the street address files. The illustrations that accompany this article are just snapshots of the GIS displays. The GIS display itself can't exist without its data: It is not a picture, but a window.
The early growth of the GIS industry and the obvious promise of the technology prompted the National Science Foundation to establish a National Center for Geographic Information and Analysis (NCGIA) in 1988 with sites at UB, the University of California/Santa Barbara, and the University of Maine/Orono to conduct basic research in geographic information science. David Mark, professor of geography, is director of UB's NCGIA site.
This past fall, the National Science Foundation awarded a five-year, $2.2 million grant to Buffalo's NCGIA site to support a new multidisciplinary, doctoral-level
concentration in geographic information science at UB.
"Social and environmental processes happen in geographic space," according to Mark. "GIS allows us to manage data and analysis of these social and environmental processes." For example, GIS might be used to develop detailed information about what happens when 1,100 white-tailed deer and 130,000 people share the same geographic space.
Such a use occurred to UB Ph.D. candidate Steven Parkansky while he was taking an independent study course in biogeography. He was aware of the deer problem from news accounts and decided to look into it. He made contact with Amherst town officials, who were happy to share their data. "I didn't intend to take the work beyond this 3-credit course," Parkansky says from Moorehead State College in Kentucky, where he now teaches, "but the more I looked at what I had, the more I realized that the problem was perfect for a GIS application." The perfect problem turned into a doctoral thesis under David Mark's supervision.
IS technology doesn't solve problems; it organizes information that may help planners solve-or at least better understand-problems. One of the ways the technology does this is to present information stored in databases in discrete layers. If, for example, you wanted to keep track of crime data and see if crimes could be associated with certain features in a neighborhood, you could make a two-dimensional paper map showing the features you were interested in (bars, parking lots, pay telephones on the street) and then stick pins in the map for each criminal occurrence.
But after only a few busy nights, your map would be a clutter of pins obscuring the features beneath them, and as the pins accumulated at certain spots, the map would be less and less accurate about the location of incidents. If you were trying to keep track of 911 calls in Buffalo, you would need about 300,000 pins in the course of a year and a very large map to make enough room for all of them.
If the information you are interested in is contained in databases-say, type of property in one, crime reports with address and date in another-you can display it on a digital street. To give a simple example: Putting a layer of data called "all car thefts in the past six months" on top of a layer called "all parking lots" on top of a street map might show that car thefts from parking lots are concentrated in certain parts of the city. Or it might show that in a certain part of the city where incidence of car theft is high, some parking lots are far more likely to be victimized than others.
The information has been there all along: Police reports are by necessity very specific, and property use (Where are the parking lots?) can be derived from city records. What GIS does is to make the information available with a few keystrokes on a computer-a few keystrokes, that is, after a great deal of thought about how to make those databases accessible and then a good deal of programming. That is what GIS workers do.
In the case of deer-related vehicle accidents in Amherst, there was no keystroke sequence to execute. What Steven Parkansky had to work with when he started his project were electronic text files of all Amherst police calls since 1991. He sorted these, by year, for the word "deer," then geocoded the addresses of those calls for display on a digitized map of Amherst. He also acquired deer-warning sign locations from the Amherst Highway Department, deer carcass location data from the contractor responsible for picking up road-killed deer in the town, and deer population survey data from the New York State Department of Environmental Conservation (DEC). At his direction, UB biogeography students prepared a detailed land cover map of Amherst from aerial photographs. Now that all the information had been sorted, it could be displayed in ways that are comprehensible and useful to the interested policymaker or lay person.
"I produced a number of maps of deer-related vehicle accidents and some analysis," Parkansky says, "but I realized that the real question was, 'Where are the deer?' Or, more specifically, 'Where are the deer that aren't being counted?'" The DEC deer population figures come from aerial photographs made during late-winter overflights of the town. With snow on the ground and no leaves on the trees, the deer are relatively easy to see. But there were fewer captured on film than would account for the number encountered on the town's roads. DEC overflights do not cover the entire town and the department reports that they may be missing as much as 30 percent of the population where they do count.
arkansky and Mark constructed a habitat model for the Amherst deer based on the assumption that there will be deer present where they can find food, shelter, and water; and, of course, access to such habitat. Parkansky specified eight land classes-such as agricultural, urban, grass, forest-for the cover map of Amherst. From these, he derived four land classes that would support deer and plugged them back into his GIS. He then looked at what factors might contribute to deer-related vehicle accident hot spots, using the hypothesis that land class is a contributing factor. With the habitat model and other data, such as car counts along certain stretches of road, he says he can account for 40 percent of the accidents, which is a significant improvement over anecdote and impression.
Town of Amherst computer services manager Jerry Galkiewicz (who earned an M.A. in geography from UB in 1984) can call up the results of this work on a laptop computer in his town hall office. "Steven offered us his work, and our GIS program was mature enough to support it," Galkiewicz says as he demonstrates successive years of deer-related vehicle accident data. He keeps the information current by adding police reports and carcass pickups monthly. "One thing we have to do with the police call data is to check how the word 'deer' is used," he notes. "If someone calls in a suspicious person on Deer Ridge, that will show up as a deer-related vehicle accident unless we take it out of the database."
Galkiewicz uses the deer data compiled in GIS to support the information needs of the Amherst Town Supervisor's Deer Management Task Force (on which UB's David Mark serves) and a team of consultants considering remedies for the problem. "We'll also use this to evaluate the outcome of any plans the town implements." As a driver, Galkiewicz says that having studied patterns of deer-related vehicle accidents revealed by the GIS, he can now identify likely crossing points along the roads he travels.
Although Parkansky marshaled the data into information the town can use, he points out that this doesn't make him an expert in what to do about the problem. "I'm as incompetent to suggest remedies as anyone else," he says, and then runs through a number of possibilities that come immediately to mind: Whistling devices on cars that would scare deer off the road? More warning signs? Flashing warning signs? Lower speed limits? Contraception? Bait and shoot? "For that matter," he says, "you could continue to thin the herd by road accident."
But the social cost of continuing to thin the herd by accident is high. Deer-related vehicle accidents in Amherst-in 1994, before the bait-and-shoot program, there were nearly 500; now there are around 300 a year-cost an average of $1,500 per occurrence.
The debate, Jerry Galkiewicz says, is about "how does the town want to live with-or without-its deer?" The role of GIS is to make the subject of the debate as clear as possible to all sides.
David Mark refers to the project as "NIMBY meets Bambi." The not-in-my-backyard (NIMBY) half of this wry formulation suggests a search for simpler answers than GIS is likely to point toward. "People don't seem to realize that location matters," he says. "They want a number for the right population of deer in the town. But that's the wrong question: the problem is not with the number of deer, but where they are."