Discussion
We analyzed the largest collection of size spectra relationships in stream communities that we are aware of, and we found that slope estimates varied in response to a broad temperature gradient. Size-spectra slopes represent the efficiency of energy transfer from small, abundant, individuals to fewer large predators (Trebilco et al. 2013) with clear implications for ecosystem functioning (O’Gormanet al. 2012; Cross et al. 2015). Shallow slopes indicate efficient transfer of energy by supporting a relatively higher proportion of larger individuals, while steeper slopes indicate inefficient energy transfer with relatively fewer large individuals. The steeper slopes reported here with increasing mean annual temperatures implies that warmer sites have fewer large individuals relative to the number of small individuals within a site compared to colder sites.
These results help to resolve previous uncertainty in how size spectra slopes scale with temperature. Variation in size spectra slopes is driven by variation in body size distributions and body size is in turn altered by temperature, either through reductions in taxon-specific body size, species turnover, or through changes in community structure (Bergmann 1847; Atkinson 1994; Daufresne et al. 2009; Winderet al. 2009; O’Gorman et al. 2012). Thus, it is widely expected size-spectra slopes should vary across temperature gradients, though the direction of change is uncertain (Daufresne et al.2009; Dossena et al. 2012; O’Gorman et al. 2012). Theory predicts that warm environments should favor smaller individuals, and this is supported by empirical (James 1970; Atkinson 1994; Daufresneet al. 2009) and experimental observations (Yvon-Durocheret al. 2011; Dossena et al. 2012). However, O’Gormanet al. (2017) found that warmed Icelandic streams had shallower slopes, perhaps due to increased nutrient availability and changes in trophic transfer efficiency, leading to increased top-down effects of consumers on diatoms. In contrast, Dossena et al. (2012) found that slopes declined with temperature, but the effect varied over seasons. Mazurkiewicz et al. (2020) found no relationship between marine benthic size spectra and temperatures in arctic systems. These contrasting outcomes, derived from different experimental approaches, generate uncertainty in how size spectra slopes should scale with temperature across large spatial gradients. The results presented here support the hypothesis that slopes become steeper in response to increasing temperature.
However, while size spectra slopes scaled with temperature, the overall change was relatively small, with median slopes declining by only ~0.1 units across the temperature gradient. Direct comparisons of this effect size with previous studies of size spectra responses to temperature (e.g., Yvon-Durocher et al. 2012, O’Gorman et al. 2017) are hampered by the different approaches to estimating size spectra exponents (Edwards et al. 2017). However, in a 30-year dataset from the International Bottom Trawl Survey (ICES 2015), Edwards et al. (2020) found that size spectra slopes calculated using maximum likelihood varied ~0.4 units among years. In addition, among sample variation at NEON sites in this study was ~0.3 to 0.8 units, with measured slopes at one site varying from -1 to -1.8 among sample dates. Placing our results into this context, the influence of temperature appears small relative to variation due to other factors, including natural variation over time.
In contrast to the negative relationship of temperature with size-spectra slopes, community biomass was positively related to mean annual temperature. This is also in agreement with predicted effects of increasing environmental temperature supporting more small-bodied individuals. For example, an increase in community biomass could be driven by small-bodied individuals alone, if their increase was larger than the relative decrease in larger-bodied individuals. It is important to account for community biomass, as increased biomass at lower trophic levels may be able to support biomass at higher trophic levels (O’Gormanet al. 2012), even if trophic transfer efficiency is affected by temperatures (Trebilco et al. 2013). However, the magnitude of the random effects of site were large, and the effect of temperature was relatively small. Including additional predictor variables thought to affect community biomass, such as productivity or nutrient availability at the base of the food web (Morin et al.2001; Daan et al. 2005), may help explain additional variation.
Given the relatively small influence of temperature and the overlap in site-specific averages of size spectra (Figure 3A), our results provide an opportunity to use our range of size-spectra as a baseline indicator in studies of disturbance. Size-spectra relationships have been proposed as a universal indicator of ecological health, with deviations from “natural” size spectra representing disturbed systems (Jennings & Blanchard 2004; Petchey & Belgrano 2010; Trebilco et al. 2013). Defining “natural” is difficult without accounting for variation among broad spatial and temporal scales. By accounting for the effect of temperature on size spectra slopes in relatively undisturbed systems across 50 degrees of latitude over three years, our results reveal bounds that could help to gauge the severity of size spectra change in response to disturbance. For example, one approach would be to compare size spectra from disturbed sites to the posterior predictive distribution of size-spectra at a similar site in our study, with deviations outside of the expected range of natural variation indicating the level of disturbance. This may represent a powerful tool for assessing ecological condition. Indeed, as NEON data continues to be collected, it will be possible to compare our predictions to size-spectra collected after intense disturbances, such as extremely high or low flow events, temperature anomalies due to climate change, wildfires, flow debris, etc. This represents an exciting opportunity to test responses to disturbances at higher levels of organization, which has typically been difficult or impossible due to the large logistical efforts needed to collect community-wide data across broad spatial scales. Furthermore, data on post-disturbance size spectra within the NEON sites will provide valuable information on community recovery, and the magnitude, direction, and expected duration of altered size spectra.