Transitional waters are ecotone ecosystems occurring at the interface between terrestrial, marine and freshwater ecosystems (Basset et al., 2013); they are highly productive and valuable aquatic ecosystems (Costanza et al., 2014; Madricardo et al., 2017), shaped by strong contrasting forces (Basset et al., 2013) and subject to multiple disturbance pressures (Elliott and Quintino, 2007).

The spatial and temporal variability of water salinity (Akin et al., 2005; Basset et al., 2013), the high nutrient supply and the strong gradients of both nutrient concentrations (Pérez-Ruzafa et al., 2005; Basset et al., 2013) and dissolved oxygen (Hull et al., 2008; Cozzoli et al., 2013; Shokri et al., 2014) are the main “natural” sources of disturbance. Potential coloniser species, being of freshwater and marine origin (Barnes, 1989; Ciotti et al., 2015), need to cope with these sources of disturbance and particularly with water salinity variability and the related osmoregulation costs, which drive species' standard metabolic rates.

Individual standard metabolic rate in relation to body size has been studied in many aquatic species, especially pelagic and non-pelagic invertebrates (Glazier, 2005) including crustaceans (Ivleva, 1980; Childress et al., 1990), insects (Gutiérrez and Menéndez, 1997) and molluscs (Xiao et al., 2014), as well as fish (Killen et al., 2010; Brucet et al., 2012). Among invertebrate species, the metabolic rates of crustacea such as Cladocera (Hart and Bycheck, 2011) and Amphipoda (mainly gammarids; Dorgelo, 1973) have received particular attention due to the relevance of these groups of species to the ecological status of aquatic ecosystems (Altermatt et al., 2014). Regarding gammarids, on which this study is focused, metabolic rates have primarily been addressed in a small number of species that have been proposed for use in bio-monitoring (e.g. the freshwater Gammarus pulex (Toman and Dall, 1998; Foucreau et al., 2014) and Gammarus minus (Lowenstein, 1935; Dorgelo, 1973); the transitional water Gammarus insensibilis (Gates, 2006); and the marine Gammarus oceanicus (Halcrow and Boyd, 1967)) and have tended to explore average metabolic rates rather than their variations, despite the relevance of gammarids to all types of aquatic ecosystems and the importance of individual metabolic rates as a key individual trait (Speakman et al., 2004). Gammarids include widely distributed amphipod species inhabiting freshwater, brackish and marine ecosystems (Tedengren et al., 1988) that are generally characterised by their higher energy requirements than other closely related Malacostraca (Hamburger and Dall, 1990).

Allometric variations of metabolic rate with individual body mass have been described (West et al., 1997, 2002; Brown and West 2000; Brown et al., 2004), and the consistency and deviation of the slope of the allometric equations with respect to the expected value of 3/4 has been discussed with reference to metabolic level (Glazier, 2005, 2009; Duncan et al., 2007; Sieg et al., 2009; White et al., 2009) and, particularly in ectotherms, metabolic state (White et al., 2004). In this regard, researchers have also considered ontogenic shifts in individual body shape (Hirst et al., 2014), the metabolic intensity of growth, reproduction and locomotion (Glazier, 2005), and body-size-dependent biological regulation (Glazier et al., 2015). However, the degree of variability of intra-population standard metabolic rates and the meaning and sources of such variability have received less attention, even when the implications of the allometric scaling of metabolic rates in the context of global warming are addressed (Forster et al., 2012). There is now growing interest in the ecological basis of metabolic rate variability at the individual level, which would allow a more accurate evaluation of the integrated responses of plant and animal communities to global warming (Bruno et al., 2015), ocean acidification (Lannig et al., 2010) and chemical contamination (Johnston et al., 2015). Moreover, there is increasing evidence that very small-scale, intra-individual variation may have manifold ecological consequences for population and community ecology (Herrera, 2017). In most animals, ecologically significant sub-individual variation is mostly sequential in nature, as it generally arises from ontogenetic and/or seasonal changes in individual traits (e.g. behaviour, colouration, digestive organ size) (Piersma and Lindström, 1997; Delhey and Kempenaers, 2006; Stamps et al., 2012; Herrera, 2017). Moreover, intra-individual variation has been observed to increase species' niche width (Sides et al., 2014), enhancing community-wide functional diversity (Siefert et al., 2015).

Here, we address individual-level variability in the metabolic rates of three species of gammarid amphipods: Echinogammarus olivii (H. Milne Edwards, 1830), Gammarus insensibilis (Stock, 1966) and Gammarus aequicauda (Martynov, 1931). The aim of this paper is to address the variability of individual metabolic rates at three different levels: (1) at the individual level, during periods of daylight; (2) within populations, across and within individual body size classes; and (3) among species occurring in the field in the three different categories of aquatic ecosystems, from freshwater to marine.

Within-population variation in standard metabolic rates is expected to be higher than both intra-individual and inter-species variation due to the well-known allometric scaling of metabolism with size (Kleiber, 1932; West et al., 1997; Glazier, 2005). However, earlier studies at the population level in aquatic crustaceans (Vignes et al., 2012) have also shown high intra-population variation that is independent of body size, and the relevance of this to both intra-individual and the inter-species variation is also addressed in this study.

The mass-specific standard metabolic rate (mass-specific SMR) at the individual level, defined as the resting energy expenditure per unit body mass per day, was computed here with individual body mass expressed as individual dry weight (Wilhelm et al., 2006; Glazier et al., 2011; Lagos et al., 2017), in accordance with the following formula: mass-specific standard metabolic rate=MO2DW, where MO2 is the individual standard metabolic rate and DW is the individual dry weight. The coefficient of variation was calculated from daily replicate SMR measurements for each species, and compared among species with one-way ANOVA.

We used parametric statistics (one-way ANOVA) to compare the mean values for weight, length and standard metabolic rates among the three species, considering that these tests were statistically independent and, therefore, individual p values were interpreted with caution.

Standard metabolic rates were compared among populations using ANCOVA with log-transformed data in order to linearise the relationship with body dry weight (covariable) and to meet model assumptions of the normality of residuals and homogeneity of variance. The assumption of slope homogeneity (interaction between populations and slope) was tested before conducting comparisons among populations.

Standard OLS regression analyses were used to measure the linear dependency of mass-specific standard individual metabolic rates on individual body condition (BCI) and body dry weight (DW), respectively.

Since the first few decades of the 20th century (Kleiber, 1932), it has been observed that common body-size-dependent scaling patterns of basal and standard metabolism, from unicellular to multicellular organisms and homeotherms, are established when global-scale patterns or large body size ranges are considered (Tilman et al., 2004). However, at the population and within-population scales, where body size ranges are small, deviations from the observed body-size scaling patterns are commonly observed. Specifically, both scaling coefficients (Vignes et al., 2012) and individual metabolic rates deviate from the best approximation of the body-size scaling patterns (Tilman et al., 2004).

Concerning the components of body-size scaling patterns at the intra-species and inter-species levels for the three gammarid species considered in the study, the results presented in this paper show the following: 1.

The short-term stochasticity of individual metabolic rates is low.

Our first conclusion above is supported by the lack of any significant temporal variation in individual metabolic rates during the experiment and by the very low average coefficient of variation of individual metabolic rates among replicated assessments of the same individuals. Flow-through techniques have been used in order to allow experimental measurement of the individual “resting” standard metabolic rate at fixed temperatures (Wrona and Davies, 1984). In our study, the experimental chambers allowed active swimming on the part of the individuals, which is a potential source of variability in individual standard metabolic rates, but generally the chambers were not large enough to exclude stress, which could also affect the standard metabolism of individuals.

Analyses of variation in the standard metabolic rate of single individuals during daylight periods are not common in the literature, since most papers, regardless of the methodology used, perform a single measurement per individual per day. However, our observations are consistent with those of Basset and Montalenti (1990) concerning Gammarus pulex, which also found a low coefficient of variation of metabolic rate among replicates and suggest that the experimental conditions and individual swimming behaviour allowed by the chamber size do not constitute substantial sources of variation in individual standard metabolic rates in the three species of gammarids considered in this study.

Our second conclusion is supported by the fact that body size explained only 12.2 % to 41.7 % of overall inter-individual variance in standard metabolic rates in the three studied species. Despite significant differences in both the intercepts of the allometric equations at the population level and the residuals of the pooled allometric regressions, the body weight of each of the three species exerted a strong influence on standard metabolism. Individuals of the two larger species, i.e. G. aequicauda and G. insensibilis, have significantly higher standard metabolic rates than those of the smaller species, i.e. E. olivii, which is supported by the global rule that the metabolic rate is higher in larger animals than smaller ones (Lampert, 1984). Based on the different intercepts of the weight–length relationships and the better body condition in E. olivii than the others, species shape could potentially interfere with metabolic rate.

Typically, the relationship between metabolic (respiration) rate (R) and body mass (M) is expressed as a power function, R=αMb. It is widely accepted that the scaling exponent, b, is 3/4, the so-called 3/4 power law (Brody, 1945; Hemmingsen, 1960; Kleiber, 1962; McMahon and Bonner, 1983; Peters, 1983; Calder, 1984; Schmidt-Nielsen, 1984; Blaxter, 1989; Brown and West, 2000; Savage et al., 2004). The results of this study confirm the size dependency of the metabolic rates of all the studied species in accordance with the allometric scaling model, with scaling coefficients ranging from 0.32 to 0.36. These scaling coefficients do not differ from each other, but they are significantly lower than the 0.75 value commonly observed (Kleiber, 1932; Peters, 1983) when interspecific comparisons are performed. Similar results to ours have also been observed in Lekanesphaera monodi along habitat productivity gradients (Vignes et al., 2012). Some of the most convincing evidence against the universality of 3/4 power scaling comes from studies of intraspecific (ontogenetic) metabolic scaling (Glazier, 2005). Indeed, at the intraspecific level the size range is much smaller than in interspecific comparisons, and it has already been shown that under such conditions, factors other than body mass can affect individual metabolic rates (West et al., 1997). Based on a review of the literature, Glazier (2005) found that scaling coefficients show departures from the Kleiber–Peters scaling exponent (b=0.75) even for species with a wide range of body sizes (up to 5 orders of magnitude). He also found a wide range of variation of the scaling coefficient, from b=-1.20 to b=2.05, for invertebrate species (Glazier, 2005). The metabolic rate allometric scaling coefficients observed for the three species in this study are well within this range, although they are lower than the average values that have been observed for amphipods generally.

Our third conclusion is supported by the comparison of residuals, which were calculated for the three species using a single pooled equation, demonstrating that the metabolic rate of E. olivii was significantly lower than those of G. aequicauda and G. insensibilis. The difference among species might be attributable to different osmoregulation and energetic costs along the salinity gradient between freshwater and marine ecosystems. This explanation was also proposed by Sutcliffe (1984). Glazier and Sparks (1997) showed that freshwater and estuarine gammarid species have significantly higher metabolic rates than those of marine species. In general, the mean metabolic rate of freshwater species is 60 %–65 % higher than that of brackish–marine species (Sutcliffe, 1984). Thienemann (1928) suggested that oxygen uptake is easier in water with higher salinity. Schlieper (1929) measured the metabolic rate of several brackish water animals at various salinities and found that the oxygen consumption of Carcinus generally increased with salinity. Within the gammarids, the brackish water species Gammarus chevreuxi has a higher metabolic rate than the marine Marinogammarus marinus (currently Echinogammarus marinus (Leach, 1815) according to the World Register of Marine Species , WoRMS), while the freshwater Gammarus pulex has a higher metabolic rate than either of them (Potts and Parry, 1964). Therefore, the results found in this paper are in agreement with most of the available evidence on amphipod gammarid species.

The metabolic rate per unit of mass in Echinogammarus olivii, which lives in marine ecosystems, was lower than other species in fresh and brackish waters. This could be related to their osmoregulation energy cost because the haemolymph of marine species is closer to seawater than fresh water, so species living in freshwater ecosystems usually consume more energy for osmoregulation. Similar results were also found by the Graetz (1931) study of the euryhaline stickleback, Gasterosteus, which, despite living for a long period in fresh water, has a higher metabolic rate in fresh water than in isosmotic seawater, where blood chloride equals environmental chloride. Wolvekamp and Waterman (1960) showed that oxygen uptake in freshwater, brackish and marine Gammarus species falls progressively while the salinity of their respective habitats rises.

Water salinity is only one of several abiotic factors that have been observed to affect individual metabolic rates in various species of Crustacea. These include the following: (i) water temperature, including short-term and prolonged fluctuations (Newell, 1969), e.g. various Idotea species (Adcock, 1982; Meyer and Phillipson, 1983; Vetter et al., 1999; Salomon and Buchholz, 2000); (ii) water salinity, with a direct response in terms of standard metabolic rates to increasing water salinity, e.g. Palaemon peringueyi (Allan et al., 2006) and Uca pugnax (Shock et al., 2009), but see also Gammarus oceanicus (Normant et al., 2004); (iii) dissolved oxygen concentration, e.g. Gammarus pseudolimnaeus (Hoback and Barnhart, 1996; a curvilinear “bell-shaped” relationship was observed); (iv) habitat productivity, with a direct response of standard metabolic rates to increasing habitat productivity, e.g. Lekanesphaera monodi (Vignes et al., 2012); and (v) resource quality, with a direct response of standard metabolic rates to increasing resource quality, e.g. Gammarus aequicauda fed on live prey or Flabellia petiolata or P. oceanica leaves (Remy et al., 2017).

The results of this paper show that most variation in the standard metabolic rates of gammarids occurs at the within-population level and is large enough to determine significant deviations from the expected 3/4 metabolic theory law, with much lower scaling coefficients for all species. Standard metabolic rate variation within individuals and between species was lower than within populations, probably since the three selected species were similar in size. The difference standard metabolic rates might also be due to non-adaptive reasons in the three studied species (Garland and Adolph, 1994). However, the lower metabolic cost of E. olivii with respect to both G. insensibilis and G. aequicauda might also suggest the need for a greater focus on osmoregulation costs for gammarid species along salinity gradients in order to explain the relevance of marine species as colonisers of transitional water ecosystems (Ciotti et al., 2015).

The results of this study seem to support the arguments of Glazier (2005), i.e. that it is essential to look beyond the 3/4 power law and consider the great diversity of metabolic scaling relationships that exist in the living world and that changes in metabolic rate with individual age can occur due to ontogenetic shifts in individual growth and locomotion costs. Additional studies, disentangling the influence of individual body size and age, are required in order to deepen our understanding of the causes of the observed deviation in the metabolic scaling coefficients of the three gammarid species studied here and more generally the observed deviations in studies at the population level.