Length-mass relationships of pond macroinvertebrates do not hold between Southern and Northern Europe

Overview of published length-mass relationships

We assessed the number of published length-mass relationships available in the literature for the freshwater communities of Portugal and Sweden, by performing a systematic search on the Google Scholar database in October 2023. For this, we used the following search terms: “length mass” OR “length weight” OR “size mass” OR “size weight” “relationship” AND “aquatic invertebrate” AND “Portugal” OR “Sweden”.

Study areas and focal taxa

We determined length-mass relationships for macroinvertebrate taxa present in ponds from Southwestern and Northern Europe (Fig. 1, Table 1). Situated in southwestern Europe, southwest Portugal is characterised by a termo-Mediterranean climate, with warm arid summers and cold windy winters; with air temperature typically ranging from 7 °C to 30 °C during the year. Situated in northern Europe, southern Sweden is characterised by a temperate oceanic (Baltic) climate, with partly cloudy summers and long, very cold, snowy, windy winters; with air temperature typically ranging from −2 °C to 21 °C during the year. The geographical separation between the study areas is large (over 3,000 km and 17° of latitude), and, consequently, the pond communities surveyed experience different climate conditions and environmental constraints. In Southern Europe, the life cycle of pond macroinvertebrates is mainly constrained by desiccation and high water temperatures, while in Northern Europe it is mainly constrained by icing and long periods of low water temperatures (Cedar Lake Ventures, Inc., 2019).

We obtained 18 new length-mass relationships for two families and seven genera of freshwater macroinvertebrates from Portugal (n = 1,020 specimens) and Sweden (n = 978 specimens) that were sufficiently abundant to develop the equations (see Table 2). While Portugal and Sweden shared most of the surveyed taxa, especially common and widespread taxa, there were a few exceptions. We developed a length-mass equation for the genus Asellus, a common shredder of the Isopoda order with dorsoventrally flattened body distributed throughout the temperate zone in Europe (Verovnik, Sket & Trontelj, 2005), that was only found in Sweden. We developed length-mass equations for two genera of gatherer collectors of the Ephemeroptera order, the cosmopolitan genus Cloeon, which is usually the most common mayfly in ponds; and the small sized genus Caenis, which occurs in high densities on the bottoms of shallow ponds (Bauernfeind & Soldan, 2013). Furthermore, we developed length-mass equations for predatory taxa from the orders Hemiptera (four genera), Odonata and Coleoptera. The taxa from the order Hemiptera comprised the genus Notonecta, with strongly dorsally convex and ventrally flattened body, which occurs in western and central Europe; the genera Corix and Sigara from the family Corixidae, which are predators or macrophyte piercers with flattened body that use the hind legs for swimming and are found in a wide range of habitats, often as the most common insect (Schuh & Slater, 1995); and the genus Naucoris, characteristic of well vegetated stagnant waters and mainly distributed in tropical and subtropical regions (Zettel, Nieser & Polhemus, 1999), that was only found in Portugal. The taxa from the order Odonata comprised the family Coenagrionidae, characterized by its elongated abdomen, which is the most widely distributed family of the suborder Zygoptera in Europe (De Knijf et al., 2024); two genera of the family Aeshnidae, Aeshna and Anax, which are insectivore taxa with similar morphology and occur in well vegetated still waters of southern and central Europe; and the genus Sympetrum, which is a small to medium-sized dragonfly commonly found across Europe amongst gravel, vegetation, detritus and mud (Dolný, Harabiš & Bárta, 2016). The overview of family, genus and species distribution of the surveyed taxa was based on the online databases Naturdata (2012) (Portugal) and Biodiversity4all (Tiago, 2020) (Sweden).

Macroinvertebrate collection

In southwestern Europe, macroinvertebrates were collected during spring, between the 1^st and the 4^th of May 2023, in the rural surroundings of Grândola (Portugal, 38°N), in eight ponds within a 6 km radius (Fig. 1, Table 1). In northern Europe, macroinvertebrates were collected during late spring, between the 1^st and the 7^th of June 2022, in the rural surroundings of Trelleborg (Sweden, 55°N), in five ponds within a 2 km radius (Fig. 1, Table 1). To cover the diversity of macroinvertebrate taxa and the natural variation in their body mass and length, sampling was conducted in all the available microhabitats and at different depths within each pond. We collected individuals from the orders Isopoda (genus Asellus), Ephemeroptera (genera Caenis, Cloeon), Hemiptera (genera Corixa, Sigara, Notonecta, Naucoris), Odonata (genera Sympetrum, Anax, Aeshna and family Coenagrionidae) and Coleoptera (family Dytiscidae). Individuals were collected with a D-shaped net (200-µm mesh size) using kick sampling (Wetzel & Likens, 2000). This method was supplemented with hand-collecting and sweeping with a kitchen strainer (19 cm in diameter; 1 mm mesh). Immediately after collection in the field, the individuals were preserved in a solution of 96% ethanol and transferred to plastic bottles (500 mL) for further processing in the laboratory. In each pond, we recorded water parameters—temperature (°C), pH, conductivity (µS.cm⁻¹) and dissolved oxygen (mg.L⁻¹)—in three replicates (Table 1), using a portable YSI multimeter (model 6909 ProDSS; YSI Environmental, Yellow Springs, OH, USA).

Sample processing

Macroinvertebrates were identified using the keys developed by Jansson (1969), Rozkošný (1980), Tachet et al. (2010), Dolný, Harabiš & Bárta (2016) and Kriska (2022). When genus-level identification was not possible, or the number of individuals of the genus was insufficient to estimate a length-mass relationship (n < 10, Johnston & Cunjak, 1999), individuals were grouped at the family level. The individuals of each taxon were sorted and separated into different size classes that consisted of 1 mm increments from the smallest to the largest body length. Then, one to three individuals were measured per size class, i.e., per 1 mm increment in body size, recording the body length of larval stages for all macroinvertebrate taxa, including full-grown larvae with black wing pads for Ephemeroptera. We also recorded the body length of adults of the order Hemiptera, as both larval stages and adults of this taxonomic group share similar morphological features and preferred habitats. We measured body length, a widely used morphometric trait in length-mass relationships in aquatic invertebrates, since its broader range provides higher determination coefficients than head width (Rosati, Barbone & Basset, 2012). The body length was defined as the distance from the frontal part of the head capsule to the posterior of the last abdominal segment, excluding the cerci and anal prolegs. Measurements were taken with a stereomicroscope (model SMZ800; Nikon, Tokyo, Japan) with an ocular micrometre of 0.05 mm accuracy (see Table S1). Overall, the number of individuals per size class was evenly distributed for most taxa, with two exceptions. The size distribution of the order Hemiptera was skewed towards small and medium-sized individuals, with only a small number of large adults included in the models (n = 76). Additionally, the size distribution of the family Dytiscidae displayed a few gaps in the size classes, that may have resulted from differences in the size ranges of the four subfamilies (five genera) composing this family.

After measuring total length, each individual was transferred into a pre-weighed aluminium cup, that was subsequently oven-dried at 60 °C for 24 h following standard protocol (Benke & Huryn, 2007). Then, their dry mass (DM) was determined to the nearest 0.01 mg on a high precision scale (model PX224M; Ohaus Pioneer, Zürich, Switzerland) (see Table S1). The number of individuals used to estimate the length-mass relationship for each taxon ranged between 33 for Aeshna and 300 for Coenagrionidae, both in Portugal (Table 2).

Statistical analysis

Dry mass (DM) was modelled as a function of body length (BL) using a natural log-linear function with a power model (ln DM = ln a + b × ln BL), where a and b represent the coefficients of the intercept and the slope of the relationship, respectively. We used this power model as it is the most commonly employed in studies developing length-mass relationships (Burgherr & Meyer, 1997; Johnston & Cunjak, 1999). We developed equations at the genus level for all surveyed taxa, except for two equations at the family level (Coenagrionidae and Dytiscidae). This approach allowed us to maximize the number of comparisons between countries, while undertaking an acceptable compromise in the accuracy of our models. Equations at the genus or family level are less accurate than equations at the species level, but it is common to find length-mass relationships at the genus level in the published literature (e.g., Smock, 1980; Burgherr & Meyer, 1997; Oliphant & Hyslop, 2020; Dekanová, Venarsky & Bunn, 2022). Moreover, a study by Giustini et al. (2008) showed small differences between estimated and measured biomass at the genus level in the genera Leuctra (Plecoptera) and Baetis (Ephemeroptera), suggesting that species share similar biomass-size growth patterns within a given genus.

We employed linear mixed-effects models (LMMs; Pinheiro & Bates, 2006) to compare the length-mass relationships. Each model included the fixed effects of body length and country of origin, as well as their interaction, to test for potential differences between Portugal and Sweden. Simpler length-mass models without specifying country of origin were used for two of the surveyed taxa, that were only found in either Sweden (Asellus) or Portugal (Naucoris). To account for the non-independence of data from individuals collected in the same pond, pond identity was treated as a random effect in the LMMs. The Akaike Information Criterion (AIC) was used to specify the random effect structure. Both random intercept and random slope models were fitted for each taxon and the one with the lower AIC value was further evaluated. We screened all models for homoscedasticity, linearity and normality using diagnostic plots of the residuals and no considerable violations of the model assumptions were detected (See Figs. S1 and S2). The statistical significance of the fixed effects was assessed using F tests with Satterthwaite’s approximation of degrees of freedom to account for correlations among repeated measurements taken from the same ponds (Kuznetsova, Brockhoff & Christensen, 2017). We calculated marginal pseudo-determination coefficients (R²_m; Nakagawa, Johnson & Schielzeth, 2017) as goodness-of-fit measures. All analyses and visualisations were performed in R version 4.2.2 (R Core Team, 2021) using the libraries DHARMa (Hartig, 2022), ggplot2 (Wickham, 2016), lme4 (Bates et al., 2014), lmerTest (Kuznetsova, Brockhoff & Christensen, 2017) and performance (Lüdecke et al., 2021).

Comparison of the length-mass relationships developed for Portugal and Sweden

To compare the allometric slopes (b) of the length-mass relationships developed in this study with those published in the literature for Portugal and Sweden, we extracted the following information from each article: the taxonomic level, fitting method (linear or non-linear regression), size-mass equation (a and b constants, determination coefficient, sample size, body size range), morphometric trait measured (e.g., body length, head width), type of mass measured (wet-mass, dry-mass, ash-free dry mass), type of habitat, and sample preservation method. To allow for comparison, the published equations were transformed into a log-linear equation with a power model (ln DM = ln a + b × ln BL) and compared visually by plotting body length (mm) versus dry mass (mg) data.

Literature review

For Portugal, we found 18 studies with published size-mass equations, developed for the classes Gastropoda and Bivalvia (e.g., Lillebø, Pardal & Marques, 1999; Gaspar, Santos & Vasconcelos, 2001; Vasconcelos et al., 2018), and the orders Amphipoda (e.g., Cunha, Moreira & Sorbe, 2000; Pardal et al., 2000; Marques et al., 2003) and Trichoptera (e.g., Canhoto, 1994; González & Graça, 2003; Azevedo-Pereira, Graça & González, 2006). Two studies developed but did not publish the length-mass equations (Sprung, 1994; França et al., 2009), while six studies used published equations, mostly to assess spatial and temporal variation in invertebrate biomass (e.g., Maranhão et al., 2001; Ferreira et al., 2016). For Sweden, only three studies published the size-mass equations, developed for the classes Gastropoda, Bivalvia, and Malacostraca—Rumohr, Brey & Ankar (1987) with 44 species, Perkins et al. (2010) with three species and Eklöf et al. (2017) with 14 species. A total of 17 studies developed, but did not publish, the length-mass equations (e.g., Hjelm et al., 2001; Persson & Brönmark, 2002; Svanbäck et al., 2008), while 20 studies used published equations, mostly to determine the contribution of macroinvertebrates to fish diets (e.g., Persson, 1997; Byström et al., 2004). For more information, see Table S2.

The database review of the composition of the families and genera surveyed showed some, but limited, taxonomical overlapping between the species lists of Portugal and Sweden, and a higher biodiversity at the species level in Portugal than in Sweden for most genera (see Table 3). The genus Asellus has only one species in both countries. The comparison of the genus Caenis was not possible since the data at the species level were not available for Portugal. Species diversity was only higher in Sweden than in Portugal for two genera, the genus Cloeon from the Ephemeroptera order and genus Aeshna from the Odonata. Among the taxa identified at the family level, the family Coenagrionidae has five genera occurring in Portugal and six genera in Sweden, of which four genera are shared between the two countries. For the family Dytiscidae, there are 26 genera confirmed in Portugal and 16 genera in Sweden, of which 11 genera are shared between the two countries.

Length-mass relationships

Overall, the models explained a high amount of variance in the mass data in all taxa (range of R²_m = 0.80‒0.95), except in the genus Caenis (R²_m = 0.63; see Table 4). The statistical comparison of the equations obtained for each taxon in Portugal and Sweden revealed similar length-mass relationships for only one out of the eight taxa compared. The potential transferability of the length-mass relationships of the genus Caenis was high between the two study areas, as the variable country had no significant effect on the curves obtained (P = 0.172; Table 4, Fig. 2). However, we found significantly different country-specific equations for the genera Cloeon, Sympetrum and Notonecta, and for the families Aeshnidae, Coenagrionidae, Corixidae and Dytiscidae (P < 0.05 in all cases; Table 4, Fig. 2), that indicate a low potential transferability of the equations obtained for these taxa between the two regions. We did not find general consistent trends for the country-specific effect. The slope of the length-mass equations obtained for the genera Sympetrum and Notonecta and for the families Aeshnidae and Corixidae was steeper for Portugal than for Sweden (Fig. 2). However, the slope of the length-mass equation obtained for the family Dytiscidae was steeper for Sweden than for Portugal (Fig. 2). Similarly, the slope of the length-mass equations obtained for the genus Cloeon and the family Coenagrionidae was steeper for Sweden, but in these cases the equations for the two countries crossed, respectively, at the body length of 8.6 and 16.2 mm (Fig. 2).

We were able to find published length-mass relationships in the literature, from either Portugal or Sweden, for a single taxon. In their study, Perkins et al. (2010) report a length-mass relationship for Asellus aquaticus using individuals collected in Umeä, Northern Sweden. The equation developed in our study closely matches the equation obtained by Perkins et al. (2010), especially in the smallest body size classes (Fig. 3). For larger body size classes, the equation published by Perkins et al. (2010) showed a steeper slope, being essentially parallel to the equation developed in this study and potentially transferrable upon application of a constant correction factor.

Literature review

The literature was scarce and patchy, indicating that the study of size-mass relationships in aquatic macroinvertebrates is still poorly developed, with equations available for only a restricted group of taxa and based on a multitude of size measurements across studies. Overall, we found a total of 40 studies developing length-mass relationships for the two countries. Interestingly, the equations were published in 18 out of the 20 studies found for Portugal, but only in three out of the 20 studies found for Sweden. Notably, borrowing equations from the literature was more common for studies conducted in Sweden (n = 21) than in Portugal (n = 6). This disparity may be related to the general focus of the studies, as Portuguese studies focused on spatial and temporal variation in invertebrate biomass, while Swedish studies focused on the contribution of invertebrates to fish diets. Except for the study by Marques et al. (2003), we were unable to find any studies comparing the length-mass relationships developed with those published for other countries. These authors compared the size-frequency distribution, growth and biomass of populations of one Amphipod species between Portugal, Tunisia and Italy, but did not compare the size-mass relationships.

The low number of studies with published length-mass relationships limits our ability to easily and accurately obtain biomass estimates for aquatic macroinvertebrates, forcing authors to develop their own equations. As an alternative to this time-consuming task, authors often borrow published equations from the literature, often developed in different geographic and climate regions. However, this practice can lead to gross errors in the determination of biomass estimates, with important consequences for the study of secondary production in aquatic habitats. Our review underscores the need to expand the array of readily available length-mass relationships in the literature and advocates for the publishing of the equations developed as a standard practice.

Our research on the similarity in the composition of the families and genera surveyed showed a small overlap between the species of the families Dytiscidae and Coenagrionidae in Portugal and Sweden, as could be expected at the family level. This was especially true for Dytiscidae, which is an extraordinarily big and morphologically and ecologically diverse family. On the contrary, the genera Cloeon and Notonecta showed a substantial overlap between the species occurring in the two countries, with all species of the genus Cloeon in Portugal and all species of the genus Notonecta in Sweden being shared with the other country. Similarly, most of the genus Sympetrum species in Sweden are shared with Portugal. However, despite the large overlap in the species pools, the length-mass equations of Cleon and Sympetrum differed between the two countries. Analysing the species overlapping for the genus Caenis, the only taxon with transferable equations between countries, could inform on the contribution of interspecific variation to our results, but unfortunately there are no available data for the genus Caenis in Portugal. Nevertheless, despite potential issues with intra- and inter-specific variation, our results indicate that the transferability of length-mass relationships is difficult to predict, even when there is a large overlap in the species composing a genus in the two countries (e.g., Cleon and Sympetrum). This, suggests that other factors may determine the transferability of length-mass equations, such as local environmental conditions and climate. However, future studies at lower taxonomic levels of determination could provide clarity on the contribution of intra- and interspecific variation to the transferability of length-mass equations.

Transferable length-mass relationships

The transferability between the equations developed for each taxon in Southern and Northern Europe was low. The genus Caenis was the only taxon with transferable length-mass equations between the two countries. High transferability of length-mass relationships in aquatic macroinvertebrates has been previously documented in Europe by Rosati, Barbone & Basset (2012), who showed invariance in the length-mass regressions obtained for macroinvertebrate taxa in the Danube delta and in Southern Italy. These authors argued that length-mass relationships are transferrable between ecoregions and can be used to estimate population biomass, provided that the environmental conditions are similar. Here, we showed that, although infrequently, length-mass relationships may be transferrable between ecoregions with different environmental conditions, suggesting that the potential transferability of length-mass relationships is taxon-dependent and difficult to predict. For instance, length-mass equations in the order Ephemeroptera were transferrable in the genus Caenis, but not in the genus Cloeon.

To our knowledge, only a single study reported a length-mass relationship for any of the macroinvertebrate taxa assessed here. Perkins et al. (2010) developed an equation for Asellus aquaticus (order Amphipoda) from Sweden using individuals collected in Umeä (n = 50), while we used individuals collected in Trelleborg (n = 211). Despite the lack of overlap between the published equation and the 95% confidence interval of the equation developed in this study, the graphical comparison shows very similar length-mass relationships for the two populations of A. aquaticus, especially at small body sizes. The similar intercept and the parallel slope of the two curves suggest a high potential transferability between the two equations. The average mass at length is slightly higher in the equation developed by Perkins et al. (2010), which may reflect the differences in the climatic conditions at the collection sites in the two studies, separated by a large geographical distance (ca. 1,000 km). Umeä is located in Northern Sweden (63°N) and experiences colder temperatures for longer periods with shorter growing seasons than Trelleborg, which is located in Southern Sweden (55°N). The seemingly high transferability of the length-mass relationships in A. aquaticus may be rooted in the ecology of this species, which is known to inhabit a wide range of freshwater ecosystems, including streams, rivers, ponds and lakes, and typically has two complete generations per year (Needham, 1938; Lafuente et al., 2021). This broad ecological tolerance suggests that environmental conditions may not impose strong constraints to its life history, leading to a comparable allometry across sites. Furthermore, Rosati, Barbone & Basset (2012) showed the length-mass relationships of A. asellus to be similar across a large biogeographical scale within the same category of habitats. Hence, the stable conditions of lentic ecosystems may lead to body mass stability, reducing variation in the length-mass relationships of A. asellus and explaining the seemingly high transferability of the length-mass equations in this taxon.

Non-transferable length-mass relationships

The length-mass relationships obtained were not transferrable between the two countries for seven out of the eight macroinvertebrate taxa compared in this study. Namely, the families Aeshnidae, Coenagrionidae, Corixidae and Dytiscidae, and the genera Cloeon, Sympetrum and Notonecta. As shown in Fig. 1, the non-transferability of the length-mass equations obtained was caused either by a different intercept (family Coenagrionidae), a different slope (family Aeshnidae and genus Notonecta) or a combination of both (genera Cloeon and Sympetrum; families Corixidae and Dytiscidae). This was expected and may be driven by many of factors, operating at different levels and with varying importance, that can alter length-mass relationships. While this study did not assess the effects of any specific driver, it illustrates well the potential variation in length-mass relationships within the same taxonomic group that may have been driven by different environmental conditions in Southern and Northern Europe, as a consequence of the large geographic distance and latitudinal cline between the two study areas.

Studies comparing length-mass relationships of freshwater macroinvertebrate taxa at the familial level have shown high variability in model predictions for different geographic locations (Johnston & Cunjak, 1999; Hajiesmaeili, Ayyoubzadeh & Abdoli, 2019). Furthermore, environmental conditions, such as water temperature (Schröder, 1987) and chemistry (Meyer, 1989; Basset, 1993; Burgherr & Meyer, 1997), as well as food availability (Gee, 1988; Basset & Glazier, 1995) and trophic interactions (Basset & Rossi, 1990; Hajiesmaeili, Ayyoubzadeh & Abdoli, 2019) can cause intra-specific variability within instar dry mass, sex, life-stage and growth rates, constituting important drivers of changes in length-mass relationships of freshwater macroinvertebrates (Benke et al., 1999; Giustini et al., 2008; Méthot et al., 2012; Hajiesmaeili, Ayyoubzadeh & Abdoli, 2019). Curiously, mass at length was more frequently higher in Portugal (families Aeshnidae and Corixidae, and genus Sympetrum and Notonecta), but this trend in the country-specific effect was not observed systematically across all taxa. Following the temperature-size rule (Atkinson, 1995), the higher temperatures in Portugal should lead to smaller adult or metamorphic body sizes, but higher temperatures also accelerate growth and developmental rates (Dallas & Ross-Gillespie, 2015). Thus, we hypothesize that at the same body size, individuals from Portugal may be at more advanced developmental stages, which would explain the higher mass at length observed for most of the taxa surveyed. This is supported by the change in the steepness of the slopes of the length-mass relationships, that typically increases with size, suggesting that the last stages of development are characterized by a sharper increase in mass than in size.

Despite the effects of the geographic cline in the growth and life cycles on the length-mass relationships of aquatic macroinvertebrates (Marques et al., 2003), it is often possible to apply specific correction factors to increase their transferability. For example, the smearing correction factor used by Mährlein et al. (2016) successfully removed the bias introduced by the log transformation of mass in the length-mass relationships developed for macroinvertebrates found in the littoral zone of different lakes, leading to a good transferability of the equations between geographical regions. Such correction factors might apply to the family Aeshnidae (DM_Portugal = DM_Sweden × exp(0.881)) and the genus Notonecta (DM_Portugal = DM_Sweden × exp(0.302)), where the estimated length-mass relationships were parallel. However, the length-by-country interactions in our models (Table 4) prevent the application of a constant correction factor to improve the transferability of the equations developed for the families Coenagrionidae, Libellulidae, Corixidae and Dytiscidae and the genus Cloeon.

Our findings underscore the recommendations of previous studies to avoid the use of size-mass relationships obtained in study areas that are not in close geographic proximity, particularly at the edges of species’ distribution ranges (Gowing & Recher, 1985; González, Basaguren & Pozo, 2002; Mährlein et al., 2016). However, our results suggest that at least for some taxa the size range of the individuals used to obtain the equations may determine the potential transferability of length-mass relationships. Our data shows variation with size in the error of the biomass estimates obtained through the application of the equations developed for Portugal and Sweden, which increased with size in the taxa for which length-mass equations had a similar intercept but a different slope (e.g., family Aeshnidae, and genera Sympetrum and Notonecta). This suggests that the potential error in biomass estimates obtained through the application of equations borrowed from the literature may be lower at smaller sizes, even if the equations are not transferrable across the whole size range.

Length-mass relationships are used to calculate secondary biomass and production and can be used for estimation of life history traits (Resh, 1979; Cressa, 1999; Walther et al., 2006; Martin, Proulx & Magnan, 2014). The use of inappropriate length-mass relationships can thus lead to biased estimations of energy flow in aquatic or aquatic-terrestrial food webs, population dynamics and fitness, flux of emerging species and deposition to land, as a source of transport energy and nutrients from water to land. For example, Dekanová (2014) detected significant over- and underestimation of the biomass of three macroinvertebrate species when comparing self-developed and borrowed length-mass relationships from different geographical regions.

Methodological factors influencing the estimation of length-mass relationships

The development of length-mass equations is prone to methodological biases (Meyer, Peterson & Whiles, 2011). The field sampling, drying, measuring, and preservation methods, together with the gut contents of the individuals, the range of body lengths measured, and the sample size greatly determine the accuracy of length-mass relationships (Morin & Nadon, 1991; Johnston & Cunjak, 1999; Dekanová et al., 2023; Mocq, Dekanová & Boukal, 2024). In our study, these methodological biases are unlikely to be responsible for differences in the length-mass relationships obtained, since they were developed using the same methods of collection, sample treatment and mathematical modelling. Thus, the low transferability of the length-mass relationships obtained at the family level, and to a lesser extent at the genus level, may stem from undesirable heterogeneity in the data (Cressa, 1999). It is generally recommended to determine length-mass relationships at lower taxonomic levels (see Poepperl, 1998; Miyasaka et al., 2008), and numerous studies show that the accuracy of biomass estimates obtained for aquatic insects increases with taxonomic specificity (e.g., Smock, 1980; Benke et al., 1999; Sabo, Bastow & Power, 2002). The grouping of individuals from different taxa may, at least partially, explain the large difference and high variation in the model predictions leading to the non-transferability of most of the length-mass relationships determined here. This issue is best exemplified at the family level in Dytiscidae, for which the large confidence intervals of the equations result from the grouping of two (Portugal) and three genera (Sweden) with substantially different morphology and size ranges. In contrast, the greater similarity in morphology of the genera composing the families Coenagrionidae and Corixidae lead to smaller differences in the equations obtained, adding further support to this argument.