94 Introduction

CHAPTER 4.2 RANGE SIZES

The range size of species has received considerable attention in macroecology (Maurer 1999, Holt 2003) as well as conservation biology (e.g. Thomas et al. 2000). For instance, range sizes have been associated with species abundance (Brown 1984, Gaston 1996a, Gaston & Lawton 1988) and latitude (e.g. McPherson 2003, Chown & Gaston 2000), they are used as parameters to assess the extinction risk of species (Jones et al. 2003, Keith et al. 2000, Purvis et al. 2000, Kunin & Gaston 1997) and they are important components of theoretical models that investigate dispersal (Gaston & Blackburn 2003, Akcakaya 2000) or biodiversity and biogeography (e.g. Hubbell 2001, McGill & Collins 2003). However, conceptual problems appear once theoretically derived ideas are to be tested on empirical data: 1) Range size might be seen as a fractal (as are most environmental data, e.g. Burrough 1981, Bolliger et al. 2003), so its measurement depends heavily on the scale of measurement. Macroecological studies on the range-abundance relation, for instance, often use very different scales of range measurement, such as occupancies of grid cells (or less systematic entities, such as sample sites) of various size or the complete range area within the most extreme records of the species (the extent of occurrence, Gaston 2003). A general distinction was also drawn between the comprehensive (i.e. global) range and the partial range (Gaston 1996a), which describes any part of the range that is under investigation, e.g. the region where a particular field study was conducted. Studies which employ largely different scales in the measurement of range size might actually investigate different phenomena and mechanisms (Blackburn & Gaston 1998, Gaston 1996a, see also Hartley 1998) and might create artefact results at small scales under certain circumstances (Stoffels et al. 2003, Maurer 1999). 2) Except for very few taxa (e.g. birds) in Western Europe and North America, range sizes cannot be measured directly, but have to be estimated from more or less scattered data. For invertebrates, moreover those from tropical regions, data can be expected to be highly incomplete and biased with regard to sampling region, habitat and taxon (e.g. Graham et al. 2004, Fagan & Kareiva 1997, Sobern et al. 2000). Rough measurements such as latitudinal and longitudinal extent of the most extreme records, or minimum convex polygons have been used as surrogates for range size (e.g. Gaston 1994a, 2003), while for increasing precision of a range assessment assumptions about a species habitat preference have to be made. Techniques to implement such assumptions, such as GIS-based models, are discussed in chapter 1.3 & 4.1. Quinn et al. (1996) addressed these problems and showed in a partial range analysis of British butterflies that several range measures, from the occupancy of 100 km2 grid cells to European ranges as well as latitudinal and longitudinal extents, correlate relatively good with each other. Similarly, Blackburn et al. (2004) found that several GIS-based, spatially detailed range estimates for the worlds parrots did not yield substantially different range rankings than crude measures such as latitudinal extents. Brndle et al. (2002a) also report highly significant correlations of range measurements across different scales for European butterflies. However, Brndle et al. (2002a) as well as Cowley et al. (2001) found an influence of scale on the importance of factors that explain range size in correlation analyses (such as density, body size, dietary niche, etc.). In this study, global distribution maps for all Sphingidae that occur in Southeast-Asia were available from a GIS-supported estimate (see below, Beck & Kitching 2004). Range sizes