1 A “GAP” ANALYSIS OF BIRD DISTRIBUTIONS IN SOUTHERN ONTARIO USING SIMULATED ANNEALING Daniel McKenney, Lisa Venier, Ian Ball, Janice McKee, Hugh Possingham, and Brendan Mackey 1 ABSTRACT.—In this paper we demonstrate the use of simulated annealing (an optimization technique) on large- scale field survey data to inform the design of protected area networks. We used Ontario’s Breeding Bird Atlas (BBA) data because it has distribution information for 290 species during the breeding season. In southern Ontario, where landscapes have been largely transformed by agriculture and urban development, the BBA data covers 2,360 Universal Transverse Mercator (UTM) grid squares (approx. 10*10km/square). In this study we identified all species that did not occur in the major parks of the Province. We then identified the minimum number and location of UTM grid squares that would represent 5 percent of the breeding occurrences of these species in southern Ontario. There are a number of possible combinations of grid squares that meet the criteria. Four particular squares were selected in every run of the model. We visualize results in a geographic information system (GIS) to provide some insights on potential gaps in bird representation in the southern Ontario park network. We also describe implications and other possible applications. INTRODUCTION The government of Ontario has recently embarked on a review of its protected area network through the Lands for Life initiative (OMNR 1997). The Natural Heritage Areas Framework and Action Plan is aimed at developing a comprehensive system of parks and protected areas: “This natural heritage areas system will strive to represent the full range of natural diversity in the Province and will provide protection for other special natural heritage values important to society for their intrinsic worth” (OMNR 1996). This plan explicitly identifies one key theme as “[to conduct] the science and [identify the] methodologies needed to identify significant areas” [our interpretation]. In recent years, considerable spatial data on aspects of biodiversity have been developed in Ontario. Insights for designing protected area networks could be gained through the application of operations research techniques to much of this data. Results could be of use not only to the Ontario Ministry of Natural Resources and forest industry but also to groups like the Nature Conservancy, the Federation of Ontario Naturalists, the Canadian Wildlife Service, and other groups that have limited resources and want to more effectively focus their conservation efforts. One purpose of protected areas or nature reserve networks is to represent and hence help to sustain biological diversity in a regional context (Margules et al. 1994; see also Mackey et al. 1989). Thus, an element of a protected area review is an assessment of how well current parks and/or reserves sample enduring landform features and the distribution of plant and animal species. The corollary of this problem is to identify locations of species that are not protected by reserves. This is not a trivial problem. For example, Canada has 3,269 species of native plants, 81 species of reptiles and amphibians, and approximately 426 bird species (Mosquin and Whiting, 1992). Pinning down actual and potential distributions is challenging given the wide range of influences affecting species (e.g., climate, nutrients, species interactions, land use—see Gaston 1990 for a review). 1 Daniel McKenney, Lisa Venier, and Janice McKee, Canadian Forest Service, Sault Ste. Marie P.O. Box 490, Sault Ste. Marie, Ontario P6A 5M7. Ian Ball and Hugh Possingham, Department of Environmental Science and Management, The University of Adelaide, South Australia, Australia 5371. Brendan Mackey, Department of Geography, The Australian National University, Canberra, Australia 2700.