A global synthesis of seasonal temperature–size responses in copepods

Introduction

Biologists have long been fascinated by body size variation (Bergmann, 1847; Schmidt-Nielsen, 1984), in part because this ‘master trait’ affects all vital rates, including feeding (Burns, 1968; Kiørboe, 2011), growth (Poston, 1990; Kiørboe & Hirst, 2014), metabolism (Peters, 1983; Glazier, 2005; Hirst et al., 2014) and reproduction (Honĕk, 1993; Arendt, 2011), as well as many other aspects of the biology of an organism (Andersen et al., 2016). Consequently, identifying and understanding what drives variation in body size is of fundamental biological importance. Body size is sensitive to environmental conditions, due to the temperature dependence of physiological processes, as well as other factors such as changes in food availability. Ectothermic species that have short life cycles and inhabit seasonal environments are typically subjected to varying environmental conditions across subsequent generations within a year. This is commonly linked to marked temporal shifts in adult body size over an annual cycle, as sequentially recruited adults are affected by different abiotic and biotic conditions over their ontogeny. Intraspecific variation in size related to seasonal variation in temperature has been found across a wide range of uni- and multicellular organisms, including bacteria (Chrzanowski et al., 1988), rotifers (Diéguez et al., 1998), copepods (Liang & Uye, 1997; Hirst et al., 1999; Riccardi & Mariotto, 2000; Dutz et al., 2012), cladocerans (Miyashita et al., 2011) and insects (Kari & Huey, 2000), yet broad-scale analyses of temporal changes in adult body size are lacking.

By contrast, intraspecific variation in size-at-stage has been well described in the laboratory under different conditions (Atkinson, 1994; Forster et al., 2012), and also spatially over latitude or across regions (Horne et al., 2015). The most frequently observed intraspecific response to warmer temperatures in ectotherms is a reduction in size-at-stage; this has been formalized as the temperature–size (T–S) rule (TSR) (Atkinson, 1994). This phenotypically plastic response can be achieved within a single generation (Forster & Hirst, 2012; Forster et al., 2013), and in many metazoans the proximate cause is attributed to differences in the temperature dependence of growth and development during ontogeny (van der Have & de Jong, 1996; Forster et al., 2011a, 2011b; Zuo et al., 2012). The ultimate cause of this outcome, however, may be a complex of several factors (e.g. see Forster et al. 2012; Horne et al. 2015). The degree to which these responses are found in natural field conditions, where multiple variables can act simultaneously to influence body size, is still uncertain. For instance, the relative contribution of food and temperature in determining seasonal shifts in adult size still needs to be resolved. Food availability affects size at maturity, but while slower growth at lower temperature is frequently coupled with an increase in adult size, slower growth with reduced food availability is typically associated with smaller size at maturity (Berrigan & Charnov, 1994). Further, food quality can dramatically alter the T–S response, even to the extent that the sign of the T–S response can be reversed under conditions of poor food quality (Diamond & Kingsolver, 2010). Identifying and understanding seasonal variation in body size will not only help to determine the ultimate causes of such variation, but will also aid in predicting future shifts associated with changes in climate (IPCC, 2014) and phenology (Visser & Both, 2005). Our study aims to synthesize and quantify seasonal patterns in adult size of multivoltine species, going beyond previous species- and location-specific studies, so that we might provide a broader understanding of such patterns.

A recent meta-analysis of terrestrial and aquatic arthropods identified an impressive match between T–S responses measured under controlled laboratory conditions and intraspecific body size clines observed in the field across latitudes (Horne et al., 2015). The magnitude and direction of these responses revealed consistent differences in the strength and sign of the response between aquatic and terrestrial species. These results suggest that laboratory T–S responses and latitudinal body size clines may be driven by similar selective pressures within arthropods – specifically, by voltinism and season length trade-offs in terrestrial species (Kozłowski et al., 2004; Walters & Hassall, 2006) and the need to balance oxygen demand and supply in larger aquatic species (Woods, 1999; Atkinson et al., 2006; Forster et al., 2012). However, in many small organisms, in which oxygen diffusion under normoxic conditions is likely to be adequate to meet metabolic demand, size reductions with warming are still very common; they are observed, for example, in bacteria, protists and small metazoans, such as copepods (Atkinson et al., 2003; Forster et al., 2012). Copepods are one of the most numerous metazoans on the planet; they are ecologically important and play a pivotal role in marine and freshwater biogeochemistry and trophodynamics (Banse, 1995). Reduction in size with increasing temperature, consistent with the TSR (Atkinson, 1994), has been shown in many copepod species, both in controlled laboratory experiments (Horne et al., 2015) and across seasons in the field (seasonal T–S responses) (Uye et al., 1982; Hirst et al., 1999; Riccardi & Mariotto, 2000; Drif et al., 2010). Furthermore, the strength of the laboratory T–S response varies widely between species, to the extent that Horne et al. (2015) observed an approximate 30-fold difference between the strongest and weakest copepod T–S responses in their dataset on arthropods. It would appear, therefore, that another factor (or other factors) may be responsible for the size reductions with warming observed in these smaller taxa, and identifying the likely causes is an important next step. Planktonic copepods are excellent model organisms in which to investigate seasonal size responses. Temporal changes in adult body size have commonly been examined in this taxon (Fig. 1), especially in mid-latitude environments which demonstrate strong shifts in temperature and food, while most species have multiple generations within a year and short generation times of >10 to <100 days (Hirst & Kiørboe, 2002). Thus, in this paper we present and test a number of alternative hypotheses that may help to explain the considerable variation observed in sensitivity of body size to warming in planktonic copepods.

Mature adult size is dependent in part upon obtaining sufficient food to meet maintenance and growth requirements, and size at maturity is controlled by different body size scaling of catabolism and anabolism (von Bertalanffy, 1957; Perrin, 1995):

where s > 0 and l > 0 are exponents for energy supply and loss respectively, and and represent the temperature dependence of the intercept terms on a log–log scale. The point at which metabolic supply and demand intersect defines an organism's asymptotic mass:

In mathematical terms, the asymptotic mass, M_A, is given by

Temperature changes will affect both energy supply and expenditure, forcing the organism into a new asymptotic mass. Hence, we can predict the induced relative change in asymptotic mass per degree Celsius, noting that f′(x)/f(x) = f′ [log(x)]:

Thus, the temperature dependence scales inversely with the difference in the mass scaling of supply and demand (l – s), and is also influenced by the temperature dependence of the intercepts. Moreover, within this framework, the strength of the T–S cline should be independent of body mass.

Despite overwhelming evidence in favour of the TSR in a diverse range of ectotherms, there remains considerable unexplained variation in the strength of the response between species and taxonomic groups, which can most likely be attributed to key differences in life-history traits and their associated metabolic constraints. In copepods and many other small zooplankton, food acquisition is governed by prey availability and uptake. With a few exceptions, species within the order Calanoida largely utilize feeding currents to entrain and capture prey (Kiørboe, 2011); by contrast, species within the non-calanoid orders, i.e. the Harpacticoida, Cyclopoida and Poecilostomatoida, lack the ability to produce a feeding current and are either ambush feeders (Cyclopoida; Paffenhöfer, 1993), feed on surfaces, which in the planktonic environment are provided by marine snow aggregates (Harpacticoida; Koski et al., 2005), or are parasitic (e.g. many Poecilostomatoida; Huys & Boxshall, 1991). There is evidence that feeding mode is an important correlate of metabolic rate (respiration), and clearance, growth and ingestion rates (Kiørboe & Hirst, 2014). Ignoring parasitic copepods, the body mass dependence of clearance rate differs between feeding current feeders and more passive ambush and surface feeders (Kiørboe, 2011), suggesting a possible difference in the temperature–body size sensitivity between different feeding behaviours. We cannot yet predict the magnitude and direction of the T–S response since we do not know how metabolic rates change with mass during ontogeny, and we also do not know how the intercept terms vary with temperature. However, these considerations lead us to suggest that some of the variability in T–S responses may be due to differences in feeding behaviour.

Another potential influence on the T–S response is reproductive strategy. In copepods, reproductive strategy can be divided into broadcast spawning and sac spawning. Sac spawners carry eggs in external sacs and have much lower fecundity rates than broadcast spawners (Hirst & Kiørboe, 2002). Sac spawners commonly do not lay the next batch of eggs until the previous batch has hatched from the attached sac(s) (Ward & Hirst, 2007), hence egg production is limited by the egg hatching time (Hirst & Bunker, 2003). By contrast, broadcasters have much higher fecundity rates, and are less likely to be limited by clutch size or egg hatching rates in the same way. The potentially different thermal sensitivities of egg development versus egg production rates may produce different solutions for size at maturity (and in turn its temperature dependence) between these two reproductive strategies. However, even in the absence of clear evidence of such a difference in thermal sensitivity of egg production and hatching, optimum size may change to different degrees if the cost of carrying versus not carrying egg sacs is temperature dependent. For example, feeding rates of ectotherms, including predators such as fish, typically increase with warming (Barneche et al., 2008), and such an increased risk of mortality to prey organisms may amplify any small differences in size- and fecundity-related trade-offs observed between broadcast and sac spawners at cooler temperatures. In principle, therefore, differences in the optimum body size response to temperature between the two spawning strategies can be hypothesized.

Our study therefore aims to: (1) quantify and synthesize for the first time the seasonal temperature–size responses of a wide range of planktonic copepod species, and to compare these with responses under controlled laboratory conditions; (2) examine the temperature dependence of size at maturity in copepods, based around major differences between taxonomic orders, species body sizes, modes of feeding (feeding current versus active ambush feeding), and reproductive strategy; (3) assess the relative importance of food concentration and temperature in driving seasonal size change.

Methods

We searched the literature for studies in which the adult body size of planktonic copepods was assessed on multiple occasions during an annual cycle. In addition to temperature we also recorded the concentration of the phytoplankton pigment chlorophyll-a (Chl-a) when this was reported. Chl-a concentration is commonly used as a proxy for phytoplankton biomass and food availability; indeed, adult fecundity and juvenile growth in many copepods correlates to this term (Hirst & Bunker, 2003; Bunker & Hirst, 2004). To reduce potential sampling bias in the sizes of animals collected, only those studies in which the adults were sampled across the entire depth of the water column, or across most of the depth range of the species, were included. Adult size data were collected as either lengths or dry, wet or carbon mass. These measurements were subsequently converted to dry mass (mg) using published intraspecific regressions. If these were not available, regressions for closely related species, or more general interspecific regressions, were used. A list of the data sources is given in Appendix 1. All raw data and conversions are detailed in Data S1 in the Supporting Information. Taxonomic order and family were confirmed for each species using the World Registry of Marine Species (WoRMS Editorial Board, 2015).

In order to test which form of equation best describes the relationship between mass and temperature within a species, we used the Akaike information criterion (AIC) to compare linear, quadratic, exponential and allometric fits to the data. We found that the exponential equation form was overwhelmingly favoured for modelling seasonal T–S responses, as judged by Akaike weights (see Table S1 for full details). We therefore used an exponential equation form to model the seasonal T–S response for each species from each study in our dataset, separating responses by sex. Species-specific slopes of the natural log (ln) of dry mass versus temperature were determined and transformed into percentage change in dry mass per degree Celsius, using the formula [exp(slope) − 1] × 100 = % change in mass per  °C (Forster et al., 2012). This value represents the seasonal T–S response, with a negative value showing a reduction in body mass with increasing temperature, hence following the same trend as the TSR. Size responses from multiple studies of the same species were then combined into a simple mean to generate a single species-specific seasonal T–S value, separated by sex.

To quantify relationships between body mass and Chl-a, the species-specific slopes of ln(dry mass) versus Chl-a concentration (µg l⁻¹) were determined for all individuals and transformed into percentage change in dry mass per µg l⁻¹, again using the formula [exp(slope) − 1) × 100 = % change in mass per µg l⁻¹, to generate a chlorophyll–size (C–S) response. The mean and 95% confidence intervals of T–S and C–S responses, calculated from the 95% confidence intervals of the individual estimated slopes, are presented for each order in Table S2. For all datasets containing both a measure of temperature and Chl-a concentration (n = 80), we compared the coefficient of determination (R²) of both parameters (i.e. by comparing the R² of each seasonal T–S response with its corresponding C–S response), to determine whether one consistently explained significantly more of the variation in seasonal body size clines than the other. Given that temperature is a mechanistic driver of variation in primary productivity, we also utilized an alternative approach to examine these relationships; first we regressed body mass against temperature and then subsequently regressed the residuals from this on Chl-a concentration, to determine how much of the seasonal variation in body size could be attributed to Chl-a after accounting for temperature.

All statistical analyses were conducted using the free statistical software package R (R Core Team, 2014). We derived several candidate models to determine the best predictors of seasonal T–S responses based on the AIC. In order to determine whether species body size affects the T–S response, we included log₁₀ species mass at a reference temperature (15 °C) as a predictor, as such an allometric relationship has previously been shown to be significant (Forster et al., 2012; Horne et al., 2015). Taxonomic order, log₁₀ body mass (at 15 °C, calculated using species- and sex-specific slopes) and sex were incorporated as fixed variables in a global linear mixed effects model (using the package lme4), with species nested within family, and latitude included as a random effect on the intercept. When selecting our random effects, we considered the estimates of variance explained by each of our proposed random variables [environment type (marine versus freshwater), latitude and species nested within family] and used stepwise elimination of non-significant terms to determine which parameters to include in the final model. All possible combinations of the global model terms were compared using the dredge function in the MuMIn package in R. The best model was identified as that with lowest small-sample corrected AIC (AICc), and Akaike weights (w_i) were used to determine the probability (0–1) of each candidate model being the best fit model (i.e. if w_i = 0.9, there is a 90% probability that a given model is the best fit model among those considered and given the data available). Where the difference between a model's AICc and the lowest AICc (i.e. ΔAICc) is less than two, a set of best fit models, rather than a single best model, can be assumed, and model averaging may be used to identify the best predictor variables across the top candidate models and determine their relative importance (computed for each variable as the sum of the Akaike weights from all models in which they appear). In addition to AIC, a series of F tests (using the ‘anova’ function in R) were used to verify the significance (P < 0.05) of each parameter's effect on the strength of the seasonal T–S response. Post hoc comparisons were made using the Tukey honest significant difference (HSD) test.

To compare seasonal T–S responses with laboratory T–S responses, we used the extensive T–S response dataset of Horne et al. (2015), supplementing this where possible with newly identified data following identical methods for acquiring data.

Results

We derived a total of 140 seasonal T–S responses from 33 different global locations (Fig. 1) within the latitudinal range of 25° to 61°, hence largely falling around mid-latitudes (with a dominance of Northern Hemisphere locations). This range in part reflects well-studied temperate environments with strong seasonality while also being inhabited by copepod species with multiple generations in a year. The data set included 48 planktonic copepod species from four taxonomic orders (Calanoida, Cyclopoida, Harpacticoida, Poecilostomatoida). These species-specific seasonal T–S responses had negative slopes in 87% of cases, with a mean (±95% CI) reduction in size of −2.87 ± 0.65% body mass  °C⁻¹ (Fig. 2), reinforcing the generality of the negative T–S response in copepods. The overall strength or direction of the seasonal T–S response did not vary significantly across latitudes (F_1,138 = 1.20, P = 0.27). Of the 80 seasonal body size clines for which we had a measure of Chl-a concentration (corresponding to 33 species), across all orders we observe a mean body mass response of 0.98 ± 2.01% per µg l⁻¹, which does not differ significantly from zero (t₇₉ = 0.97, P = 0.34) (Fig. 2). On average, across all taxonomic orders temperature explained more of the variation in seasonal body mass than Chl-a concentration: this is inferred from the mean R² values of each parameter when both were modelled separately (0.44 ± 0.07 vs. 0.22 ± 0.05, respectively), and also when comparing body mass–temperature regressions and the residuals from these against Chl-a concentration (0.44 ± 0.07 vs. 0.07 ± 0.03, respectively) (see Fig. S1). Considering each of the four orders separately, temperature always explained more of the variation in adult body mass than did Chl-a concentration.

In explaining variation in the strength of the seasonal T–S response among planktonic copepods, the model with the lowest AICc includes only taxonomic order as a fixed variable, whilst all other candidate models have ΔAICc > 2 (Table 1). Thus, given the data available, we may reject the other candidate models in favour of a single best fit model in which taxonomic order has a significant independent effect on the strength of the seasonal T–S response (F_3,82 = 9.43, P < 0.001). Post hoc comparisons (Tukey HSD) show that Calanoida (n = 66, mean= −3.66 ± 0.70% body mass  °C⁻¹) have a significantly stronger negative seasonal T–S response than both Cyclopoida (n = 12, mean = −0.91 ± 0.59% body mass  °C⁻¹) and Poecilostomatoida (n = 6, mean = 1.36 ± 3.06% body mass  °C⁻¹), but not Harpacticoida (n = 2, mean = −1.19 ± 3.60% body mass  °C⁻¹), though our seasonal data for this order are sparse, including only male and female Euterpina acutifrons. We note specifically the different temperature response between the calanoids, which use feeding currents, and ambush-feeding cyclopoid copepods, with a four-fold difference in the strength of the seasonal T–S response observed between these two groups (Fig. 3). We find no significant change in the strength of the response with mean species body mass in either the Calanoida (F_1,101 = 0.11, P = 0.75) or non-calanoid orders (F_1,35 = 2.75, P = 0.11), supporting our prediction that any change in mature body size is independent of mean species body mass in these smaller taxa (Fig. 4).

Reproductive strategy also varies within and between orders; calanoid species can be either broadcast or sac spawners, but are more commonly the former (n = 44 vs. n = 22 for broadcast and sac spawners, respectively, in our dataset), whilst all species in the three remaining orders considered here are sac spawners. Given that taxonomic order and reproductive strategy correlate exactly in three of the four orders in our dataset, while in calanoids both reproductive strategies occur, we tested for differences in the seasonal T–S response between broadcasters and sac spawners exclusively in calanoids, finding no significant effect (F_1,64 = 0.71, P = 0.40). We also tested for order-level differences in the seasonal T–S response exclusively in sac spawners (i.e. by excluding any broadcast spawning calanoid species), and found significant differences in the strength of the seasonal T–S response between taxonomic orders, still observing a four-fold significant difference between calanoids and cyclopoids (t-test; t = −4.51, d.f. = 31, P < 0.0001). This leads us to suggest that reproductive strategy is not responsible for driving the observed differences in seasonal T–S responses between taxonomic orders and, hence, explains why we chose to exclude the latter from our global linear mixed effects model.

Despite the numerous other variables that may act to obscure the correlation between body mass and seasonal temperature, we find a strong match between the mean seasonal T–S response for Calanoida (−3.66 ± 0.70%) and the mean T–S response (−3.20 ± 0.49%) measured in the laboratory under conditions of excess food (t-test; t = −1.09, d.f. = 79, P = 0.28). However, we note that the two datasets comprise different species. Indeed, when we regress species-specific seasonal T–S responses against laboratory T–S responses for the small number of species for which we have both sets of data (n = 12), separating responses by sex, we observe much greater variation in seasonal T–S responses than those measured under controlled laboratory conditions (Fig. 5). This suggests that food quantity/quality, and potentially other environmental variables, are having an impact on the T–S response in the field. There appears to be no systematic difference in the strength of laboratory and seasonal T–S responses between the sexes, such that sex has no significant effect on the strength of the seasonal T–S response, either across species (F_1,84 = 0.03, P = 0.86) or intraspecifically (paired t-test; t = 1.35, d.f. = 37, P = 0.19). Unfortunately, we are unable to make further meaningful comparisons between field and laboratory responses. For example, we could not compare the broad differences between taxonomic orders we observe in the seasonal T–S data with laboratory data, as very few laboratory studies on species other than calanoids have been conducted; our dataset contains male and female laboratory T–S responses for just two planktonic cyclopoid species and a single harpacticoid species.

Discussion

Our work combines field data from numerous studies world-wide (Fig. 1), and goes beyond controlled laboratory-based T–S studies to demonstrate broad patterns in the thermal size responses of marine and freshwater planktonic copepods. Despite numerous other variables that may act to complicate the T–S signal in the field, we show that almost 90% of copepod species in our dataset follow the TSR in seasonal environments, maturing at a smaller adult body mass in warmer conditions. Yet, as we might expect, seasonal T–S responses appear to be much more variable than those measured under controlled conditions in the laboratory (Fig. 5), suggesting that environmental factors in addition to temperature may play a role in driving seasonal body size variation in the field. We should also consider that the temperature at which adults are collected is unlikely to correspond exactly to temperatures experienced during ontogeny, and this may be further complicated by the existence of a winter diapause, during which many copepods will cease recruitment over late winter to early spring. Throughout this period their prosome length will change little, and yet temperatures may vary considerably.

Food availability has also been shown to have a direct influence on body size (Berrigan & Charnov, 1994; Diamond & Kingsolver, 2010), though we find that Chl-a concentration explains very little of the seasonal variation in body mass, both when modelled independently and after accounting for the effects of temperature. This suggests that temperature is much more significant in driving body size responses in these natural populations. Greater quantities of food typically lead to larger size at maturity in ectotherms, and we observe a positive but non-significant percentage change in adult body mass with increasing Chl-a concentration on average (Fig. 2). Chl-a concentration commonly correlates significantly with juvenile growth and adult fecundity rates in many natural populations of planktonic copepods (Hirst & Bunker, 2003; Bunker & Hirst, 2004), and hence is generally considered a reasonable proxy of food availability. However, many copepods have an omnivorous diet that does not exclusively include prey containing this pigment (e.g. including heterotrophic ciliates and flagellates; Calbet & Saiz, 2005), and the proxy also fails to account for variation in prey quality (Pond et al., 1996), which has been shown to alter the T–S response, even reversing its sign at times (Diamond & Kingsolver, 2010). Here we find little evidence for sign reversal when comparing laboratory and field animals. Time lags might also obscure the correlation between Chl-a concentration and body size. As food availability commonly varies over a much shorter time-scale than generation time, whilst temperature varies over a relatively longer time-scale, correlations with the latter are likely to be much more reliable. Although greater chlorophyll concentration is often associated with increased growth (Hirst & Bunker, 2003; Bunker & Hirst, 2004), consumer abundance is also predicted to increase with primary productivity (O'Connor et al., 2009). Our analysis does not account for the abundance of the copepods, and hence we are unable to assess the role of food availability on a per capita basis. Assuming the metabolic rate has a Q₁₀ of 2.5 and scales with body mass^0.75 (Zuo et al., 2012), a simple calculation suggests that an organism would have to decrease its body mass by approximately 11.5%  °C⁻¹ of warming to offset the increase in metabolic rate associated with this temperature increase. Given that calanoid copepods on average reduce their body mass by only 3.7%  °C⁻¹, this compensates for approximately a third of the increase in metabolic rate  °C⁻¹ of warming. If resources were limiting and kept constant then population abundance would have to fall substantially with warming to accommodate the extra metabolic demand, even with a reduced individual body size.

In the field, beyond variation in temperature and food availability, we might expect predation by ectotherms to increase with warming (Kordas et al., 2011). This in turn may lead to increased copepod mortality, selecting for earlier maturation and resulting in a reduced adult body size. Copepods can also detect and perceive chemical signals released by predators, such as fish kairomones, the presence of which has been shown to trigger faster development and earlier maturation at a smaller body size in calanoids (Gutiérrez et al., 2010). Thus, increased predation risk in the warm and associated increases in mortality and the presence of chemical cues may amplify the T–S response in the field.

The relative strength of the seasonal T–S response does not vary significantly between the sexes in this study, evidence for which can also be observed in Fig. 5. These findings agree with the broader analysis across Arthropoda, for which T–S responses were not found to differ significantly between the sexes (Hirst et al., 2015). Rensch's rule suggests that male body size varies more than female body size, irrespective of which is the larger sex (Rensch, 1960). Applied within species, the rule would predict an increase in sexual size dimorphism (SSD) with increasing body size in species where males are the larger sex, and a decrease in SSD with body size in species where females are larger. Thus males should consistently have the greater size variation, yet we find no evidence to support this pattern at the intraspecific level. Our finding at the intraspecific level here concurs with there being isometry between male and female size seen across copepod species (Hirst & Kiørboe, 2014), suggesting that the selection pressures on the seasonal T–S response have been equally as strong for both males and females.

Though recent studies have begun to identify broad trends in both the magnitude and direction of the T–S response, for example between terrestrial and aquatic species (Forster et al., 2012; Horne et al., 2015), there remains a large amount of variation in the strength of the response that is yet to be explained. This is especially true for planktonic species, which are only a few millimetres or less in size, where in most conditions oxygen availability appears unlikely to be a driver. Indeed, our most compelling finding is the significant difference in the strength of seasonal T–S responses between species of calanoid and cyclopoid (Fig. 3), both of which typically employ different feeding modes. We find that calanoids exhibit much greater size plasticity upon temperature changes than non-calanoids. This is consistent with the hypothesis that feeding mode may influence the T–S response, since all calanoids can produce a feeding current to harvest prey, while none of the other orders do so. The extent to which the T–S response differs between these two feeding modes depends on the differences in both size scaling and the thermal response of feeding in relation to metabolism. Thus, in order to thoroughly test this hypothesis, one would need estimates of within-species mass scaling and temperature dependence of feeding and metabolism. While there are some estimates of between-species body mass scaling of respiration and feeding of the two groups (e.g., Kiørboe & Hirst, 2014), the body mass-dependent changes in vital rates during ontogeny are typically different (Hirst et al., 2014; Glazier et al., 2015), and thus needed for these groups. A further complication arises from the fact that feeding mode may change during ontogeny: while all cyclopoids are ambush feeders throughout their development, many calanoids are ambush feeders during the nauplius stage, and feeding current feeders during the copepodite stages; or they may switch between feeding modes in the copepodite stages (Kiørboe, 2011).

We note the association between taxonomic order and feeding mode in our dataset, and appreciate the potential difficulty in disentangling effects of feeding strategy from other order-specific differences in physiology and behaviour. For example, all cyclopoids in our dataset are sac spawners, whilst calanoid species can be either broadcast or sac spawners, but are more commonly the former. However, we find no substantial effect of reproductive strategy on the sensitivity of mature body mass to temperature. Whilst broadcast and sac spawning planktonic copepods have markedly different rates of adult fecundity (Bunker & Hirst, 2004), egg mass production rates (Hirst & Bunker, 2003) and egg mortality (Hirst & Kiørboe, 2002), they appear to have somewhat similar rates of juvenile growth, development and mortality (Hirst & Kiørboe, 2002; Hirst & Bunker, 2003). The T–S responses of species with determinate growth are largely generated during the juvenile phase of ontogeny (Forster & Hirst, 2012). Similarity of important life-history rates during the juvenile phase may therefore explain the lack of difference in the T–S responses within the calanoids based upon reproductive strategy. Expanding our analysis in future to consider ambush-feeding calanoid copepods, such as in the genera Tortanus and Pareuchaeta, will help to more definitively separate effects of feeding strategy from order-level differences. Unfortunately, at present, suitable data are not available on these taxa. We recommend that future experimental studies comparing species-specific size variation in response to temperature, both within and between taxonomic orders, should focus on those taxonomic groups that are currently data deficient.

Given that body size is an important predictor of fitness, and warming is a prominent feature of climate change, there is an urgent need to accurately predict changes in body size with temperature. This is particularly the case in zooplankton, which globally represent a primary resource for invertebrates and vertebrates, including fish (Ware & Thomson, 2005). Changes in body size will not only affect individual and population fitness, but may have an impact on feeding rates and alter food web connectivity given the size dependency of trophic processes (Hansen et al., 1994; Rice et al., 2015), as planktonic food webs are especially highly size-structured (Webb, 2012). Measuring and accounting for abundance in the field would also help to define the relationship between food availability per capita and adult body size under natural conditions. This may be particularly informative in light of the fact that the T–S response in the majority of ectotherms appears to compensate for only a small proportion of the predicted increase in metabolic rate with temperature, whilst metabolic rate in autotrophs (and thus primary productivity) increases substantially less with warming than metabolic rate in heterotrophs (Allen et al., 2005).