Abstract
Aim Studies of species turnover commonly assume that turnover is a critical determinant of species richness patterns. But the concordance in patterns of turnover and species richness along gradients is poorly known. Here we characterize elevational patterns of species turnover and test whether turnover and species richness are strongly related. Location Sixty‐two elevation gradients world‐wide, from 17° S to 43° N . Methods We used elevational range data for six terrestrial vertebrate groups to characterize species turnover between neighbouring elevational bands. We measured turnover as Simpson's dissimilarity, a metric that is unaffected by measured differences in species richness among recorded samples. To assess differences from random patterns, elevational turnover was compared with three null models (hard, soft and no boundaries). Lastly, elevational turnover was compared with the combined species richness of neighbouring elevational bands. Analyses were conducted at three grain sizes (200, 400 and 800 m elevation). Results We found no consistent, repeated patterns in elevational turnover. Variability among gradients was very high, with most datasets displaying multiple but inconsistently located peaks. Concordance between null predictions and empirical turnover was poor (average r 2 for 200, 400 and 800m grains were: hard boundaries 0.06, 0.12 and 0.15; soft boundaries 0.06, 0.11 and 0.14; unbounded 0.03, 0.07 and 0.10; respectively), although many empirical values fell within the confidence intervals of the null model. Correlations of turnover and species richness were generally poor, but increased with analysis grain (average r 2 = 0.19, 0.33 and 0.54, respectively). Main conclusions Turnover cannot serve as a general explanation for richness patterns within elevational gradients. Elevational turnover patterns are highly idiosyncratic, change with scale, and are often indistinguishable from random patterns. Despite the common assertion that the highest species richness occurs where distinct, dominant communities turn over on mountains (e.g. low‐ and high‐elevation communities at a middle ecotone), we found no strong support for such Clementsian‐structured patterns.