1905 Ecology, 82(7), 2001, pp. 1905–1920  2001 by the Ecological Society of America DENSITY AND LINKAGE ESTIMATORS OF HOME RANGE: NEAREST-NEIGHBOR CLUSTERING DEFINES MULTINUCLEAR CORES ROBERT E. KENWARD, 1,3 RALPH T. CLARKE, 1 KATHY H. HODDER, 1 AND SEAN S. WALLS 2 1 Centre for Ecology and Hydrology, Winfrith Technology Centre, Dorchester DT2 8ZD, UK 2 Biotrack, 52 Furzebrook Road, Wareham, Dorset BH20 5AX, UK Abstract. Methods used to estimate home ranges from point locations are based either on densities of locations or on link distances between locations. The density-based methods estimate ellipses and contours. The other class minimizes sums of link distances, along edges of polygons or to range centers or between locations. We propose a new linkage method, using nearest-neighbor distances first to exclude outlying locations and then to define a multinuclear outlier-exclusive range core (OEC) by cluster analysis. The assumption behind exclusion of outliers, that movements inside and outside range cores involve different activities, was supported by data from radio-tagged Common Buzzards (Buteo buteo). We compared the new method with other techniques by using location data from each of 28 goshawks, 114 buzzards, 138 gray squirrels, and 14 red squirrels. Range structure statistics from OECs showed marked differences between species in numbers and extent of core nuclei. Range analysis displays illustrated relationships of range area with age categories, food supply, population density, and body mass within species. The OECs gave highly significant results in three of five within-species tests, perhaps because animal movements in these cases were affected by coarse-grained habitat boundaries. When movements were likely to have been influenced by diffuse social interactions and foraging for scattered prey, the most significant results were from density-based estimators, especially kernel contours that had been optimized by least-squares cross validation. We recommend use of both density and linkage estimators of home range until a basis for a priori choices has been established. Key words: Accipiter gentiles; Buteo buteo; cluster analysis; density-based estimators; home range estimation; minimum link distances; nearest neighbor; outlier exclusion; Sciurus carolinensis; Sciurus vulgaris. INTRODUCTION During the last three decades, radio-tracking has rev- olutionized studies of the use of space by terrestrial vertebrates. Locations of radio-tagged animals can be collected systematically, free from the restrictions and bias that may be imposed on visual records by elusive behavior or dense vegetation. The resulting sets of lo- cations often provide a basis for estimating home rang- es, which are used in studies of animal density, social behavior, predation, and other foraging activities (re- views in Amlaner and Macdonald 1980, Harris et al. 1990, White and Garrott 1990, Samuel and Fuller 1994). Home range estimators are typically used to provide measures of size, shape, and structure. Estimates of home range size may be needed for management pur- poses, such as planning reserves, or as indices of move- ment differences between or within species (Schoener 1968, Hulbert et al. 1996). Shape may be important for analyzing how home ranges conform to the landscape when meeting resource and security requirements Manuscript received 1 December 1998; revised 29 December 1999; accepted 17 March 2000; final version received 20 June 2000. 3 E-mail: reke@ceh.ac.uk (Covich 1976, Redpath 1995), or how home ranges of conspecifics fit together to indicate territoriality or so- cial cohesion (Macdonald et al. 1980, Ims 1988). In- ternal structure may differ between home ranges (Ad- ams and Davis 1967), and range cores that are used with different degrees of intensity may vary in habitat content, or in extent of overlap with neighboring ani- mals (Poulle et al. 1994). The widely used definition of home range as an ‘‘area traversed by the individual in its normal activities’’ (Burt 1943) implies that high- use cores may be separable from peripheral areas that are seldom visited. However, as noted by White and Garrott (1990), how does one define ‘‘normal’’? As a result of requirements to measure range size, shape, and structure, biologists face a wide choice of methods for estimating home range boundaries and in- ternal cores (Dalke and Sime 1938, Mohr 1947, Hayne 1949, Calhoun and Casby 1958, Stickel 1964, Harvey and Barbour 1965, Siniff and Tester 1965, Jennrich and Turner 1969, Koeppl et al. 1975, Dunn and Gipson 1977, Dixon and Chapman 1980, Voight and Tinline 1980, Schoener 1981, Anderson 1982, Don and Ren- nolls 1983, Geissler and Fuller 1985, Samuel and Gar- ton 1985, Hartigan 1987, Kenward 1987, Worton 1989, 1995a, Loehle 1990, Wray et al. 1992a). The methods differ in ability to (1) distinguish core areas as well as