Case studies: contrasting latitudinal gradients in trees and reef fish

We used our approach to analyze two empirical datasets documenting latitudinal diversity gradients (LDG) in reef fish and trees. The trend of increasing diversity from poles to equator is one of most prominent global biodiversity patterns that occurs in many taxa and at different spatial scales (Fine, 2015; Hillebrand, 2004; Willig et al., 2003). All components, particularly N and the SAD, likely vary along the gradient, though how they combine to form the LDG at a given scale, and whether this varies among taxa, is less well known.
For example, N is expected to vary with energy- or resource availability and, accordingly, the more-individual hypothesis (MIH) is one of the classic explanations for the LDG (Brown, 2014; Srivastava & Lawton, 1998; Wright, 1983). Historically, the MIH has referred to a collection of different mechanisms by which higher total abundance translates to higher species diversity, including effects on extinction and speciation rates (Evans et al., 2005; Scheiner & Willig, 2005; Storch et al., 2018). However, here we use the term more narrowly to only mean passive sampling effects (Coleman et al., 1982), which is the process by which larger communities (e.g. in the tropics) randomly sample a larger portion of a species pool than small ones (e.g. in temperate regions) (Wright, 1983). Abundance-related processes that influence extinction (e.g. demographic stochasticity) and diversification rates over the longer term likely alter the SAD and regional species pool, and therefore would be captured by SAD-effects in our framework. Indeed, there are a large number of ecological and evolutionary mechanisms that shape and maintain latitudinal gradients in regional SADs. These include differences in time for speciation, environmental stability, species interactions, and niche-processes (Fine, 2015). While the LDG is generally strongest at larger spatial grains (Hillebrand, 2004), it is largely unknown how such species pool gradients combine with gradients of total abundance to determine local-scale diversity gradients.
Here, we applied the analytical framework to analyze latitudinal gradients of two publicly available datasets with standardized community surveys: (1) forest trees from the Gentry plot dataset (Gentry, 1988, Phillips & Miller, 2002), and (2) reef fish from the Reef Life Survey (Edgar et al., 2020; Edgar & Stuart-Smith, 2014). Importantly, both datasets use a fixed sampling effort in terms of plot/transect size for their respective sites. Therefore, latitudinal variation in sample diversity reflects changes in the regional species pool (SAD) as well as natural variation in the observed number of individuals (i.e. more-individuals effect).
Because our main focus was to illustrate the analytical framework, rather than an exhaustive analysis of these datasets, we reduced both datasets into one latitudinal “slice” to minimize other well-known confounds, such as biogeographic factors, that influence the magnitude of the gradient. For trees, we focused on the plots located in the Americas, so as to avoid the potential influence of continent on tree diversity (Qian & Ricklefs, 2000). And for the reef fish, we only included surveys from the Indo-Pacific area where diversity is highest, and biogeographic effects (e.g., distance from diversity center) were minimized (Blowes et al., 2017). For both data sets, we excluded sites with fewer than 20 individuals (we also used different cutoff-levels to test the robustness of our results). Supplementary Figure S3 shows the geographical location of samples included in our analyses.
After selecting the sites, we dissected the observed diversity of each sample into the SAD-component and the N-component, assuming a reference sample size of n=20. To do this, we calculated the observed richness and the rarefied richness (Sn) for n=20, and derived the corresponding ENS values using Eqn 3 (i.e. EN and En, respectively). En represents the SAD-component. The difference between EN (total diversity) and En (SAD-component) is the diversity component that results from the changes in N or the more-individuals effect (N-component). We then modeled the two components along the latitudinal gradient using simple linear models with absolute latitude as the independent variable, and the SAD and N components as dependent variables. We used the regression coefficients (or slopes) as the effect sizes for the respective components. Since our partitioning framework is additive and models are linear, the effect sizes (i.e. slopes) of the two components add up to the effect size (i.e. slope) of the total diversity gradient.
Both trees and reef fish showed similar slopes along their respective latitudinal gradient for the overall richness gradient, but they differed in how the underlying component contributions changed along the gradient (Fig 3). The trees had a relatively large SAD-effect; that is, even when the number of individuals was standardized, the diversity gradient remained quite strong. This suggests that the diversity gradient is mostly underlain by changes in the species pool and associated patterns of commonness and rarity (i.e., the SAD). Nonetheless, the N-effect also contributed to the total diversity gradient, as total tree abundance tended to increase as absolute latitude decreased. In contrast to the trees, the reef fish diversity gradient was strongly dominated by the N-effect. For a standardized number of individuals, the fish diversity gradient was relatively weak (see SAD-component). This reflects that species rich reef fish communities are often dominated by a few species, the number of which does not vary strongly along the gradient. For a constant sample size, the many rare species in diverse fish communities have little weight in the diversity estimate. That is, they mostly affect the diversity for communities with more individuals, and are captured more-individual effect.
The contrasting results between fishes and trees could reflect biological differences of the two groups. Fish move in a three-dimensional space, which allows for much stronger gradients in total abundance. In forests, on the other hand, stem density is likely more strongly limited by available space. This suggests that for forests, community assembly processes change more strongly along the gradient, leading to communities with high relative evenness in the tropics (Ulrich et al., 2016). This is reflected in the strong SAD-effect. Conversely, the schooling nature of some tropical fishes allows for the dominance of a few species. Additionally, the number of dominant fish species does not vary strongly along the gradient, whereas the number of rare species (which are affected by sampling effects) does. Hence, we find the large N-effect in fishes.