Utilizing Remote Sensing and GIS in Wild Turkey Research
By Robert G. Wright

INTRODUCTION
Wisconsin's highly successful wild turkey restoration program prompted concerns by farmers that turkeys were causing significant crop damage. In 1987, the Wisconsin Department of Natural Resources (WDNR) initiated a study to investigate these concerns. Remote sensing and GIS were important tools in this effort.
      Wild turkeys are native to Wisconsin but were extirpated in the late 1880s, mainly due to habitat changes and unregulated hunting. Wisconsin's restoration program began in 1976, when 17 wild-trapped turkeys from Missouri were released into Vernon County, in southwestern Wisconsin. (Figure 1) Additional releases of Missouri birds, combined with intensive trapping and relocation of their Wisconsin offspring, resulted in rapid population growth and range expansion. By 1993, nearly all suitable habitat in the state was occupied, and the current population is estimated at 300,000 turkeys.
    Because turkeys are gregarious and easily observed when using crop fields, farmers became increasingly concerned about crop damage as turkey populations increased. Addressing these concerns required a two-pronged approach. First, how often were turkeys using crop fields and second, were they eating unharvested crops when they were out there? The question of food habits was addressed by collecting birds observed feeding in crop fields and examining the contents of their digestive tracts.
      Determining how often wild turkeys used crop fields was more complicated because, in Wisconsin, they spend most of their time in woodlands, where they are difficult to observe. In addition, and like most wildlife, wild turkeys are wary and unapproachable. A study of habitat usage based on visual observation would have been biased because of unequal observability in different habitat types, and because of avoidance behavior caused by the observer's presence. These and other obstacles were overcome with the application of remote sensing and GIS technologies.

APPLICATIONS
Radio Telemetry
Radio telemetry is a common technique for remotely monitoring free-ranging wildlife. Each animal wears a transmitter with a unique frequency, and a biologist collects the data of interest, such as survival, movements, or habitat use, using a receiver and antenna. During the winters of 1987-88 through 1989-90, wild turkey hens were captured with rocket-propelled nets, fitted with backpack-style transmitters, and released at their original capture sites. Transmitters had a battery life of two years and a signal range of from 0.5 to 5 km on the ground (depending upon topography and vegetation), to over 30 km from the air.
      Efforts were focused on hens because they often form large "brood flocks" (composed of several hens and their young) that can be seen feeding in crop fields from mid-June to the end of the growing season. These large flocks represented a worst-case scenario for causing crop damage and were most often mentioned in farmers' complaints. During summer and fall, radioed hens were located throughout the day (turkeys roost in trees at night) by triangulating on their transmitter signals from three or more receiver locations. (Figure 2)

Global Positioning System
Measurements of signal direction (bearing) are subject to error, which is usually caused by a combination of signal bounce (due to topography), dense vegetation, equipment performance, and observer error. Because of this, telemetry locations of free-ranging animals are actually only estimates of their true locations, and the statistical confidence of these estimates is directly related to the magnitude of bearing error. Biologists estimate this error by comparing the measured bearings to test transmitters, to the actual bearings to these transmitters.
      By combining GPS technology with turkey nesting behavior, it was possible to use the radioed hens themselves as "test" transmitters. Biologists measured the signal bearings to several hens while they were known to be incubating (turkeys nest on the ground). After leaf-off, differentially corrected GPS locations of the nests were collected, making it possible to determine the hens' "true" (+6.1 m) locations without disturbing them. The average bearing error of +5.50 was then applied to the habitat-use data, in order to generate statistical estimates of the telemetry locations, called "error ellipses." (Figure 2) Error ellipses can be likened to three-dimensional normal curves, where statistical probabilities are highest in the center and lowest along the edges. The size and shape of an individual ellipse is determined by the statistical confidence of the x and y coordinates, and the correlation between them.

Geographic Information System
To apply the most sophisticated statistical approach to the telemetry data, a vector-based coverage of wild turkey habitat was needed. Crop fields, idle fields, pastures, and woodlands were delineated on 1:20,000 black-and-white unrectified aerial photographs and verified with ground truthing and USDA farm program participation records. A private contractor utilized a zoom transfer scope and MYLAR-based USGS 7.5 minute topographic maps to rectify and transfer the delineations to overlay material at 1:24,000. The staff of WDNR's Bureau of Information Management automated the coverage using ARC/INFO. (Figure 3)
      When using telemetry triangulation data to estimate habitat use, error ellipses that include more than one habitat type present a special problem. For example, if an ellipse includes both idle field and woodland habitats, what is the probability that the estimated habitat type (woodland) is incorrect? (Figure 4) Because the error ellipse represents the statistical confidence of the location, random subsampling from within this ellipse is the most appropriate way to estimate these misclassification probabilities. For this study, 100 random subsample points were generated for each location, which were then developed into an ARC/INFO point coverage. The number of random points falling in each habitat type for each location were then used to calculate unbiased estimates of habitat use. (Figure 4)
      The total availability of each habitat type in the study area (for later comparison with habitat use) was also determined using ARC/INFO.

FINDINGS
Habitat Use By Wild Turkeys
Radioed hens were located 721 times during the summers of 1988-90, and 322 times during the fall of 1989. During the summer, hens with broods used crop fields more often (35%) than hens without broods (20%). This was probably because high insect numbers found in crop fields, especially alfalfa, provided the young poults with the high-protein diet needed for rapid growth. Overall, radioed wild turkey hens used crop fields less and woodlands more than was expected based on habitat availability. (Figure 5) As the growing season progressed, they spent less time in crop fields and more time in woodlands.
      The companion study of food habits showed that turkeys did indeed use crop fields for insects during the summer, and for waste grain and insects during the fall. In addition, investigations of crop damage complaints showed no significant damage attributable to wild turkeys during the growing season.

Contribution of Remote Sensing and GIS
Remote sensing and GIS technologies were instrumental in accommodating several sources of bias and error in this study. Without radio telemetry, estimates of habitat use would have been strongly biased toward open habitats, where turkeys are most easily observed. Indeed, this visibility bias was the principle factor contributing to farmers' perceptions of crop damage by turkeys. In addition, remotely monitoring turkeys eliminated the avoidance- behavior bias caused by an observer's presence.
      With GPS technology, it was possible to estimate bearing error through extensive testing in a realistic situation, without the cost of traditional surveying for the test transmitters.
      Prior to the availability of GIS technology, most estimates of habitat use were derived by comparing raw telemetry locations to maps or aerial photos. GIS technology provided not only a more objective way of estimating habitat use, but also allowed a more sophisticated method of incorporating bearing error into these estimates.
      Although this study suggested that wild turkeys do not cause significant crop damage in southwestern Wisconsin, these misconceptions are hard to dispel. The application of remote sensing and GIS technologies greatly enhanced the validity of these results among researchers, wildlife managers, and the general public.

ACKNOWLEDGEMENTS
The staff at the USGS Biological Services Division, Upper Midwest Environmental Sciences Center offices in La Crosse and Onalaska, Wisc., provided invaluable assistance during this project. Carl Korschgen provided access to workstation ARC/INFO, Doug Olsen provided advice on ARC/INFO programming, and Carol Lowenburg provided GPS equipment and assistance. Special thanks to Kevin Kenow for developing the statistical software and generating the random subsample points used in the habitat use analysis.
      This work was funded almost entirely by sportsmen and women through an excise tax on sporting goods, and through the National and Wisconsin Chapters of the National Wild Turkey Federation.

About the Author:
Bob Wright is a Certified Wildlife Biologist and conducted this research while working for the Wisconsin Department of Natural Resources. He is currently a Research Associate at the Center for Ecological Management of Military Lands at Colorado State University, and is stationed at Fort McCoy, Wisc. He can be reached at: N6685 CTH M, Holmen, WI 54636; Telephone 608-526-2511; E-mail: [email protected].