Journal of Vegetation Science 18: 307-312, 2007 © IAVS; Opulus Press Uppsala. FORUM - Is tree diversity different down-under? - 307 FORUM Is tree diversity different in the Southern Hemisphere? Burns, K.C. School of Biological Sciences, Victoria University of Wellington, P.O. Box 600, Wellington, New Zealand; Fax +64 4 463 5331; E-mail [email protected]; Web http://www.vuw.ac.nz/staff/kevin_burns/index.htm Abstract Introduction Questions: Is tree diversity higher in the southern hemisphere? Are latitudinal asymmetries in diversity sensitive to sampling Latitudinal variation in species diversity has intrigued effects? biologists for over two centuries (Hawkins 2001). How- Location: 198 forested locales worldwide. ever, it has only recently been appreciated that declines Methods: I re-analysed the Gentry database, which I augmented in diversity towards the poles may differ substantially with an additional survey from New Zealand. Data were used to test whether latitudinal declines in tree diversity differ between between hemispheres. In a recent review of latitudinal the northern and southern hemispheres. Data were also used to diversity asymmetries, Chown et al. (2004) argue that test whether hemispheric asymmetries in diversity are sensitive “…simple exercises – such as plotting richness values to sampling effects, or geographic variation in tree densities. for different latitudes or latitudinal bands against each Results: Area-based measurements of species diversity are other for the hemispheres and examining the resulting higher in the southern hemisphere. However, southern forests relationship… – rarely appear in the literature. Thus, it is house denser plant populations. After controlling for geographic not yet clear how common or strong hemisphere-related variation in tree densities, diversity patterns reverse, indicating asymmetry is.” (p. 460, Chown et al. 2004; cf. Hillebrand tree diversity is higher in the northern hemisphere. 2004). Conclusions: Latitudinal changes in tree diversity differ between hemispheres. However, the nature of hemispherical In a pioneering study, Gentry (1988) suggested that asymmetries in species diversity hinges on how diversity tree diversity might be higher ʻdown-underʼ in the south- is defined, illustrating how different definitions of diversity ern hemisphere. He graphically illustrated that species can yield strikingly different solutions to common ecological diversity of woody plants follows a bell-shaped distribu- problems. tion with latitude, rising from low diversity levels at the poles to a peak near the equator. But the peak in species diversity appeared to occur in the southern hemisphere. Keywords: Latitudinal diversity gradient; Sampling effect; Diversity also seemed to decline more rapidly with Species diversity. latitude in the northern hemisphere. However, Gentry (1988) based this observation on a limited number of forest inventories and he did not statistically test for Nomenclature: Allan (1961) and Moore & Edgar (1970). differences in diversity between hemispheres. Species diversity is deceptively difficult to character- ize. The most common measure of diversity is on an area- basis (i.e. species density, α-diversity, or the number of species present in a given area). However, area-based estimates of species diversity can be confounded by population density, if the number of individuals sampled varies among sampling points (Bunge & Fitzpatrick 1993; Gotelli 2001). Such ʻsampling effectsʼ can have impor- tant consequences to our understanding of how factors such as insularity and productivity influence taxonomic diversity (Hector et al. 2002; Forbes et al. 2001; Chiarucci et al. 2004; Lawes et al. 2005). However, “standardizing 308 Burns, K.C. FORUM data sets by area… may produce very different results continents (Africa = 18, Australasia = 38, Europe = 5, compared to standardizing by the number of individu- North America = 57, South America = 126), with over als collected, and it is not always clear which measure 80% of sampling points coming from North and South of diversity is appropriate” (p. 379, Gotelli & Colwell America. 2001). In an attempt to increase the representation of poorly Here, I test whether tree diversity is higher in the sampled regions and latitudes, I used Gentryʼs (1982) pro- southern hemisphere. Using the Gentry database, I tocol to sample an additional site in New Zealand. These evaluate whether hemispherical clines in area-based data were collected in Otari-Wiltonʼs Bush (41º14' S, measurements of tree diversity (i.e. species richness 174º45' E), which contains a large, undisturbed stand of per unit area) differ between hemispheres. I then as- conifer-broadleaf forest on the southern tip of the North sess latitudinal trends in tree population density and Island, New Zealand (see Burns & Dawson 2005 for a test whether latitudinal asymmetries in tree diversity detailed site description). Elevation ranges between 70 remain unchanged after controlling for variation in plant m and 280 m above sea-level, mean annual temperature density. is 12.8 ºC and total annual rainfall averages 1249 mm (Anon. 1996). I tested for hemispherical differences in area-based Methods measures of species diversity (i.e. number of species per 0.1 ha) using the general linear model procedure in Analyses were conducted using the Alwyn H. SPSS (Anon. 2002). The absolute value of latitude was Gentry Forest Transect Data Set (Phillips & Miller used as a covariate and hemisphere (north or south) was 2002). Gentry and his colleagues sampled a total of considered a fixed factor. The full model, comprised of 226 forested locales across the globe with a sampling the independent effects of the covariate and the fixed- design comprised of 10 separate transects, each meas- factor and their interaction was assessed. Following uring 2 m × 50 m (Gentry 1982). The first transect Engqvist (2005), the interaction term was used to test typically began from a randomly chosen starting point for differences in the slope of relationships for each in undisturbed forest and was oriented in a random hemisphere. Separate tests were conducted for trees, direction. Each subsequent transect was then oriented lianas (lianas and hemi-epiphytes combined) and total in a random direction within a 180° arch from the end species diversity. of the previous transect. All plants > 2.5 cm diameter Next, I re-assessed hemispherical differences in at breast height (1.37 m) that were rooted within 1 m of species diversity after controlling for latitudinal varia- the transect line was identified to species. Plants were tion in plant density. Least-squares regression was used classified as belonging to morpho-species when their to evaluate the relationship between plant density and taxonomic identity was uncertain. Each plant was also latitude, and the relationship between plant density and categorized according to growth habit, either as a tree, species richness per unit area. Standardized residuals of liana or hemi-epiphyte. As a result, each subsequent 0.1 the relationship between plant density and area-based ha plot provides an estimate of plant population density measures of species diversity were then subject to the and species richness per unit area for each growth form. same general linear model procedure described above, The full dataset is freely available on line (http://www. using the same covariate (absolute value of latitude) and mobot.org/MOBOT/research/gentry/data.shtml) and is fixed-factor (hemisphere). Diversity and density estimates discussed in detail by Phillips & Miller (2002). were log10 transformed to conform to normality assump- Of the 226 plots included in the dataset, 29 differ tions. Separate tests were again conducted for trees, lianas from the protocol described above. Lianas and hemi- and total species diversity. epiphytes were not recorded in 6 sites, and 23 sites were Several additional analyses were conducted to control not sampled over a full 0.1 ha (i.e. < 10 transects were for geographic and latitudinal differences in sampling sampled). To promote unbiased comparisons, these plots intensity. In addition to the analysis of the full dataset (N were omitted from analyses, leading to a sample size of = 198), analyses were repeated after removing all sites 197 standardized inventories. Forest inventories were located above 42° latitude to account for latitudinal differ- also not distributed homogeneously across the globe. ences in sampling intensity (N = 188). Analyses were also First, temperate forests in the northern hemisphere were repeated on North and South American sites exclusively, sampled more intensely than south-temperate forests. to account for differences in sampling intensity between The highest latitude sampled in the southern hemisphere continents. In the separate analysis of American sites, was in south-central Chile (40°43' S, 72°18' W), while ten samples located on Caribbean islands were also omitted sites in North America and Europe were located above to control for insularity effects (N = 148). In all three 42° latitude. Sampling intensity also differed between sets of analyses, the dependent variable and covariate FORUM - Is tree diversity different down-under? - 309 Table 1. F-statistics from general linear models of the effect of latitude and hemisphere on species diversity of lianas (including hemi-epiphytes), trees and both life forms combined. Results from area-based diversity estimates (species richness/0.1 ha) are shown alongside individual-based diversity estimates