Plant species invasions along the latitudinal gradient in the United States: comment.

Few biogeographic patterns have been as well documented as the latitudinal gradient in species richness. For North American plants, the decrease in number of species from the subtropics to the arctic is well established (Stevens 1989, Currie 1991, Qian 1999), but becomes less defined as the spatial grain of species richness decreases (e.g., Monk 1967, Glenn-Lewin 1977). There is considerable interest among biogeographers and community ecologists as to whether clear latitudinal patterns are manifest at relatively fine spatial resolutions, where the signature of local processes such as historical disturbances or fine-scale environmental variation could obscure broad-scale geographic drivers of richness (Rahbek and Graves 2001, Whittaker et al. 2001). Because adequate data have generally been lacking, however, there remained until recently no systematic attempt to address the latitudinal gradient of plant richness for the full extent of the United States using relatively fine-scale data. Stohlgren et al. (2005) examined native and exotic vascular plant species richness at the county level for the 48 conterminous U.S. states in relation to latitude, climate, topographic and biotic factors, and human disturbance. By comparing a variety of multifactor models explaining native and exotic plant richness, Stohlgren et al. (2005) concluded that: (1) there is no relationship between native plant richness and latitude at the county level; (2) the strongest single predictor of exotic species richness is native species richness; and (3) bird species richness, a surrogate for habitat heterogeneity, is the best predictor of native plant richness. If true, these observations would represent a major advance in the study of plant diversity patterns, as they would suggest scale thresholds where broad-scale geographic patterns are no longer manifest due to the increasing importance of other factors such as local habitat heterogeneity. However, we argue that these conclusions are inaccurate, stemming from uncritical evaluation of completeness in the data set and from artifacts deriving from inappropriate data transformation. Furthermore, we analyze a more appropriate data set to the issue of county-level patterns of native and exotic species richness, and obtain different results than that of Stohlgren et al. (2005). A major conclusion of Stohlgren et al. (2005:2301) is that ‘‘regression analyses showed no relationship between latitude and native plant species density (richness).’’ We assert that this result is an artifact of two major deficiencies in their study; namely, that (1) their data set of county richness values underestimates true county richness, and (2) richness values were incorrectly transformed to account for variation in survey area. Plant data used by Stohlgren et al. (2005) come from the Biota of North America Program (BONAP), a collection of taxon occurrences from the literature and from herbarium surveys recorded by county (Stohlgren et al. 2003). These data are widely regarded as the ‘‘standard plant data set for many government and non-government agencies’’ (Stohlgren et al. 2005:2299), and we agree that this is the most comprehensive collection of plant occurrence records for North America. However, records for many counties were not themselves collected with the aim of providing complete floristic lists of counties (see Palmer 1995). Thus, although some counties are well collected, others remain less well described for a variety of reasons (such as low survey effort). In addition to adding error to the assessment of county-level richness patterns (thereby undermining the signal of geographic pattern in the data), variation in floristic completeness among counties could cause spurious correlations between native and exotic richness; as well-surveyed counties are more likely to contain more occurrences of both native and exotic species. The use of such data for the analysis of richness patterns is appropriate if researchers make sufficient attempts to correct for, or are at least be able to quantify, the variation in floristic completeness among counties. Stohlgren et al. (2005) give no assessment of data quality in their analysis. Moreover, the only attempt made to reduce the bias associated with including poorly surveyed counties is the removal of 114 counties with fewer than 100 occurrences. This procedure is both insufficient (114 is ,4% of the data set) and flawed, in that occurrences are a function of both survey effort and Manuscript received 3 February 2006; accepted 6 March 2006. Corresponding Editor: H. Hillebrand. 1 Department of Biology, University of North Carolina, Chapel Hill, North Carolina 27599 USA. 2 Research and Collections Center, Illinois State Museum, 1011 East Ash Street, Springfield, Illinois 62703 USA. 3 Botany Department, Oklahoma State University, Stillwater, Oklahoma 74078 USA. 4 Present address: Department of Biology, Syracuse University, 130 College Place, Syracuse, New York 13244 USA. E-mail: [email protected]