The spatial requirements of populations ought to be central to developing conservation strategies and ecological theory, but they are difficult to identify from the published collection of reported population estimates. Transforming population estimates into density and log values may have clouded recognition of the discontinuous relationship that exists between abundance estimates and the areas within which they were made. I used abundance estimates of terrestrial mammalian Carnivora to identify scale domains of abundance. In scale domain A, the study areas were too small to include all the individuals of the population, and abundance estimates increased propor-tionally with study area size. Scale domain B included high variation in abundance estimates, but these estimates no longer increased with increasing size of study area as they did in domain A. I refer to the smallest area in the scale domain B as the 'threshold area' (the minimum area known to support a population of the particular species). The abundance in scale domain B averaged 52 individuals amongst all but three outlier species. Scale domains C and D were also identified, and included much larger study areas and substantially larger abundance estimates.

The threshold area is the first empirically-derived representation of the minimum space used by populations of each species. However, nearly half of all carnivore abundance estimates were made in areas smaller than the threshold area, and only 4% could be used to represent transitions between scale domains B, C and D. The remainder varied within domains B, C and D due to variation in environmental conditions and research methods. Most comparisons of density between sites and between species have been made without considering whether the estimates represented the true population or some collection of individuals more or less abundant than the population. Threshold area can be used as a standard area against which to estimate and compare population density, in lieu of more appropriately comparing estimates of region-wide density, which generally do not yet exist. Also, threshold area could be predicted with reasonable error rates based on a typical species' body mass, brain mass and home range size. Therefore, conservation biologists may now have a tool to identify the minimum habitat areas needed to conserve populations of carnivore species, and theoretical ecologists can relate ecological space at the population level to allometric variables.