Size Spectra slopes
The maximum likelihood estimates of the power law exponent (e.g., size
spectra slopes) for individual collections ranged from -0.73 to -1.80.
The median of the posterior distribution of size spectra slopes from the
hierarchical Bayesian model was -1.25 (95% CrI (-1.39, -1.11); Table
2). Size-spectra slopes had a 98% probability of being negatively
related to mean annual temperature (coefficient estimate = -0.004, 95%
CrI (-0.007, 0.0004); Table 2), meaning that increasing temperature was
associated with relatively fewer larger body sizes compared to small
body sizes within a community (Figure 2A). However, the magnitude of
this decline was small, with median slopes changing from -1.2 to -1.3
across the 29°C temperature gradient. This range was smaller than the
among-sample range in measured slopes within sites (Figure 2A). Using
the posterior predictive distribution from our model to estimate
variation among sites, median slopes ranged from a low of -1.33 at site
GUIL (mean annual temperature = 25°C) to a high of -1.15 at site CARI
(mean annual temperature = -1°C; Figure 3A).