What Is the Observed Relationship between Species Richness and Productivity? Comment

The form of the species richness–productivity relationship (SRPR) remains central to our understanding of species diversity controls. Mittelbach et al. (2001) built on Waide et al. (1999) to provide the first largescale, formal meta-analysis of the SRPR, based on 257 data sets from 171 publications. Their aim was to classify each relationship as negative, positive, U shaped, or hump shaped, in the context of recent suggestions that the general form of the SRPR is hump shaped (e.g., Rosenzweig and Sandlin 1997, Huston 2001). The central findings of the meta-analysis are that there is no single general pattern, and that patterns are scale and taxon dependent. We concur with these conclusions (also see Chase and Leibold 2002), and applaud the intent of their analysis. However, we raise concerns over three issues: (1) the treatment of scale; (2) the treatment of surrogate productivity variables; and (3) potential bias in the statistical procedures followed in the meta-analysis. The subset of studies we examine in detail is of tree data sets, classed by Mittelbach et al. (2001) as ‘‘regional’’ or ‘‘continental-global’’ in scale. A priori, we expect a positive SRPR to be evident for trees at the macro-scale (see O’Brien et al. 1998, Whittaker et al. 2001). Such a relationship is consistent with our general understanding of geographical gradients of species richness in trees (Currie and Paquin 1987, O’Brien 1993, 1998) and of estimates of net primary productivity (NPP) variation across biomes (Lieth and Whittaker 1975, Esser 1998). Our reexamination was prompted by the (to us) unexpected degree to which