The effect of preservation on biomass and length estimates and its variation within and between two mayfly species

Biomass estimation offers insight into aquatic ecosystem processes and represents a powerful tool for measuring biomass in communities. However, the accuracy of this estimate depends considerably on the sample preservation method and biological traits of species. We quantified the effect of commonly used preservatives (70% ethanol, 4% formaldehyde, freezing), preservation time and initial body length on the estimation bias of body length, wet mass and dry mass in two mayfly species with dorsoventrally flattened (Rhithrogena carpatoalpina) and cylindrically elongated (Habroleptoides confusa) bodies. Freezing caused the largest relative bias in the length estimates (up to 18%). Both ethanol and freezing led to significant underestimation of wet mass of preserved animals (up to 26%). The most severe biases were observed in dry weights, where all preservation methods reduced body mass by up to 89%. Importantly, initial body mass influenced the bias caused by preservation, and the larger animals were less affected by mass losses. In contrast, larger individuals were subject to greater length reduction. Additionally, the length–mass relationship for R. carpatoalpina, based on fresh mass of living animals, was significantly different from relationships based on preserved animals. Overall, formaldehyde caused the least biased estimates of length and biomass in both examined species. This case study indicates that direct measurements of unpreserved individuals or length–mass equations based on fresh material are likely the best ways to avoid severe underestimation of biomass in aquatic insects. However, more studies covering a broader range of taxa are needed to reach general conclusions.


Introduction
The estimation of biomass is a critical measurement in ecological studies. It is an essential step towards understanding the life history patterns, distribution of resources and energy fluxes in food webs at the population, community and ecosystem levels (Benke 2010;Benke and Huryn 2007). In freshwater research, sampling protocols usually involve field sampling, preservation and transport to the laboratory for subsequent processing. Standard methods for macrozoobenthos preservation require the use of fixation chemicals, usually 5-10% formaldehyde or 70-80% ethanol, or freezing (Gaston et al. 1996); however, several studies have proven the effect of chemical preservation on the body parameters of aquatic invertebrates (e.g., Howmiller 1972;Landahl and Nagell 1978;Leuven et al. 1985). The effects of chemical preservation are unclear for many aquatic species; nevertheless, there is a general opinion that preservation can cause a reduction in both body length and mass, which can lead to significant biases in biomass estimates (Edwards et al. 2009;Leuven et al. 1985). Both formaldehyde and ethanol might cause specimen body shrinkage by replacing water in the animal tissues (Tucker and Chester 1984). Chemical preservation causes the leaching of water and organic components such as enzymes or lipids from the body and long-term changes in tissue composition (Mills et al. 1982). Additionally, sample storage time might have significant effects on the length and mass of the preserved specimens (Howmiller 1 3 1972). Furthermore, the response of the organism's body to preservatives is complex, and the rate at which these changes occur varies across taxonomic groups (Edwards et al. 2009). The differences among taxonomic groups can be caused by the specific biochemical composition of bodies, such as encrustation (Moody et al. 2016), sclerotization (Howmiller 1972) and chitinization (Von Schiller and Solimini 2005). These differences could be partly attributed to the body size when larger animals act differently from smaller animals (Stanford 1973). Several authors also noticed the consequence of preservation on the body shape of preserved organisms (Järvinen et al. 2014) or some body parts (Beladjal and Mertens 1999); however, the relation between the initial shape of the body and the preservation effect is still unclear.
Changes in body mass and shape during preservation may lead to biased estimates of population characteristics such as biomass or secondary production. For instance, the length-mass relationships that are used for rapid estimation of individual mass from body dimensions (Benke et al. 1999;Burgherr and Meyer 1997;Johnston and Cuniak 1999) are prone to bias caused by preservationrelated changes in morphological traits. Ideally, the fresh mass and length of individuals should be determined under conditions of minimum body disturbance to obtain the most accurate measurements (but see Dermott and Paterson 1974). Nevertheless, mass and length determinations of fresh aquatic organisms are not always possible, and the body length and mass of organisms are thus estimated from measurements of preserved material (Benke and Huryn 2007;Johnston and Cuniak 1999;Wetzel et al. 2005). Consequently, length-mass equations derived from preserved material may be biased in a complex way depending on the preserving agents used, preservation time and biological traits of the preserved animals. Currently, little information exists on how preservation methods bias length-mass relationships.
This study aims to determine the magnitude of bias caused by interplay of the most commonly used preservation methods (ethanol, formaldehyde, and freezing), preservation time and body dimensions on the estimation of body length, wet mass and dry mass of mayflies with dorsoventrally flattened (Rhithrogena carpatoalpina) and cylindrically elongated (Habroleptoides confusa) bodies. We selected mayflies as a model group since they represent an important part of aquatic invertebrate biomass in almost every type of freshwater environment (Brittain 1982;Brittain and Sartori 2009), are considered sensitive bioindicators (Bauernfeind and Moog 2000), and provide a range of ecosystem services (Jacobus et al. 2019). Therefore, mayflies are often chosen for analysis in freshwater studies, while accurate estimation of their biomass is a crucial assumption for the correct interpretation of the results and drawing conclusions.

Study site
All individuals used for this study were taken from the 3 rd -order Breznický potok stream (48°30′28" N, 19°1′40" E, 645 m asl) located in central Slovakia. The average depth of the stream at the sampling site was 0.3 m, and the stream width varied from 0.8 to 1.3 m. The stream water was well oxygenated (9.0-9.6 mg O 2 L −1 ), with low conductivity (93-175 µS cm −1 ) and slightly alkaline pH (7.4-8.0) (Stoklasa and Očadlík 2009). The stream substrate comprised cobbles, pebbles and gravel in roughly equal proportions. The riparian zone was dominated by Fagus sylvatica and Alnus glutinosa.

Fieldwork
In this study, we selected two mayfly species that are abundant in Central Europe (Sartori and Jacob 1986;Klonowska et al. 1987): Rhithrogena carpatoalpina Klonowska, Olechowska, Sartori and Weichselbaumer 1987 and Habroleptoides confusa Sartori and Jacob 1986. The larvae were collected on May 2, 2013. The kicking technique was employed to collect invertebrates from shallow riffle and pool habitats using a D-shaped net with a 200-µm mesh size. Sampling was performed in different habitats and at different depths to cover natural variability in the body length and mass of larvae. The larvae were hand-collected in the field and immediately transported under cold conditions to the laboratory.

Laboratory work
Larvae were kept overnight in a refrigerator (8 °C) to empty their guts and diminish potential bias caused by varying amounts of food remains in larval intestines. Body length (BL, mm) of living R. carpatoalpina larvae (n = 154) and H. confusa larvae (n = 43) was measured as the distance from the frontal part of the head capsule to the posterior of the last abdominal segment, excluding the cerci, using an ocular micrometre with the accuracy of 0.05 mm. The samples included immature and full-grown larvae with black wing pads. The body length ranged from 6.2 to 10.3 mm in H. confusa and 8.3 to 13 mm in R. carpatoalpina.
To obtain fresh mass (FM, mg), every individual was placed on filtration paper, and after one minute, it was weighed on an analytical balance to the nearest 0.1 mg. Subsequently, randomly selected specimens of R. carpatoalpina (n = 19) and H. confusa (n = 10) were oven-dried at 60 °C for 24 h and weighed on an analytical balance to estimate dry mass (DM, mg). For both species, we fitted relationships between fresh and dry mass to estimate reference values of DM for each individual. These calibration models were significant for both R. carpatoalpina (F (1,17) = 29.1, p < 0.0001, R 2 = 0.63) and H. confusa (F (1,8) = 8.7, p = 0.018, R 2 = 0.52) and provided relatively accurate predictions (cross-validated mean absolute percentage error: 15% for R. carpatoalpina and 11% for H. confusa).
The remaining individuals were divided into groups following three preservation methods; freezing at −18 °C: R. carpatoalpina (n = 42) and H. confusa (n = 10), 4% formaldehyde: R. carpatoalpina (n = 50) and H. confusa (n = 12), 70% ethanol: R. carpatoalpina (n = 43) and H. confusa (n = 11). Subsamples were randomly selected from specimens after 10, 20, 46 and 158 days of the above preservation, and we measured their body length, wet mass and dry mass. Individuals of H. confusa were not measured at 46 days due to the small total number of specimens (Table S1).
The relative bias of body length, wet mass and dry mass was calculated by subtracting the reference values in a fresh (unpreserved) state from values observed after preservation and dividing by the reference values and expressed as a relative fraction of the reference values.

Data analysis
We used a general linear model (Kutner et al. 2004) to link the relative bias in body length, wet mass and dry mass estimates to the preservation methods, preservation time and initial body length of the two mayfly species. The models involved a three-way interaction of the preservation method (formaldehyde, ethanol, and freezing) with the preservation time (4 and 3 sampling dates for R. carpatoalpina and H. confusa, respectively) and the initial body length of specimens. Residuals of the models were carefully checked using diagnostic plots; no considerable departures from normality and homogeneity of variances were observed. Tukey tests were performed for pairwise comparisons.
Since changes in body mass and body shape during preservation might affect the accuracy of length-mass relationships, we also compared the length-mass relationships estimated for fresh and preserved individuals of R. carpatoalpina. Dry mass was modelled as a function of body length using the power equation where a and b are coefficients representing the intercept and slope of the relationship, respectively (e.g., Burgherr and Meyer 1997). Analysis of covariance (ANCOVA) was DM = a BL b employed to test for differences among the length-mass equations.

Changes in body length
The body length of R. carpatoalpina was significantly influenced by the preservation method, preservation time and initial body length of the unpreserved specimens (Table 1). Freezing caused the greatest length reduction (−16%), while the effect of formaldehyde and ethanol caused a relatively small reduction by 4% and 2%, rspectively ( Fig. 1a). Overall, the relative length reduction of R. carpatoalpina varied from 5 to 10% and was significantly more pronounced in larger individuals (Fig. 1b). The greatest body shrinkage was observed after the first 10 days of preservation (9% reduction) (Fig. 1c).
The body length of H. confusa was significantly affected by the preservation method only (Table 1). Freezing caused the greatest decrease by 18%, while the effect of ethanol and formaldehyde caused a smaller reduction by 8% and 4%, respectively (Fig. 1d). Neither initial body length (Fig. 1e) nor preservation time (Fig. 1f) affected significantly the body length estimation of H. confusa.

Changes in wet mass
The bias of the wet mass estimates in R. carpatoalpina was significantly influenced by the preservation method and initial body length of the fresh specimens (Table 1). Freezing caused the greatest reduction of mass by 26%, followed by preservation in ethanol by 19% and formaldehyde by 11% (Fig. 2a). The wet mass reduction was more obvious in smaller than in larger specimens with average relative bias ranging from 2 to 34% (Fig. 2b). The effect of preservation time was not significant (Fig. 2c).
Changes in the wet mass of H. confusa were significantly influenced by the preservation method, body length in the fresh state and their interaction; thus, the effect of preservation depended on body length (Table 1). Wet mass estimates of smaller specimens were biased equally by all preservation methods, while the mass estimate of larger specimens was more biased by freezing and ethanol than by formaldehyde (Fig. 2d). Bias in the wet mass estimates of H. confusa was also significantly affected by preservation time; a longer preservation time led to greater reduction (Fig. 2e).

Changes in dry mass
The bias in the dry mass estimate of R. carpatoalpina was significantly affected by the method of preservation and initial body length of the fresh material (Table 1). Ethanol caused the greatest reduction in dry mass by 40%, followed by freezing by 28% and formaldehyde by 26% (Fig. 3a). The dry mass of smaller specimens was reduced more significantly (by 53%) than that of the largest individuals (by 7%) (Fig. 3b). The effect of preservation time was not significant (Fig. 3c).
We did not find significant differences in relative dry mass loss between the preservation methods in H. confusa (Table 1), but all of them greatly underestimated the dry mass of unpreserved specimens (Fig. 3d). The bias of dry mass estimates was significantly affected by initial body length and was more pronounced in smaller individuals than in larger ones (Fig. 3e) and increased with preservation time (average relative reduction ranging from 26 to 41%) (Fig. 3f).

Effect of preservation on length-mass relationships
A comparison of the length-mass relationships estimated for fresh and preserved specimens was made only for R. carpatoalpina, since preservation time did not influence the dry mass of this species (Table 1). Thus, data from all dates were merged for length-mass relationship estimations and compared for fresh and preserved specimens without considering the effect of preservation time. The length-mass relationship developed for fresh specimens was significantly different from the relationships estimated using preserved specimens (ANCOVA: F 3,146 = 2.97, p = 0.03). The length-mass relationships based on the preserved specimens were much steeper than those calculated using fresh specimens, and consequently, the bias was negatively related to body length (Fig. 4a). The cross-validated median relative error of dry mass prediction based on the length-mass equation derived from fresh specimens was 16.6%, while the equations based on the preserved specimens yielded median prediction errors of 26% for formaldehyde, 28% for freezing and even 42% for ethanol (Fig. 4b).

Effect of preservation method
In both examined mayfly species, we found body shrinking and a decrease in wet and dry masses caused by preservation. Body shrinkage due to preservation was identified as a significant problem in several studies of aquatic macroinvertebrates, causing bias in body weight estimation (e.g., Edwards et al. 2009;González et al. 2002;Lasenby et al. 1994); however, some studies reported no influence (e.g., Giberson and Galloway 1985;Nolte 1990) or even an increase in body length (Lasenby et al. 1994). Several studies have also proven that preservatives cause changes in aquatic macroinvertebrate wet and dry body masses (e.g., Edwards et al. 2009;Howmiller 1972;Von Schiller and Solimini 2005). Other factors that can influence the body features of studied samples are the volume and combination of preservatives (Leuven et al. 1985). In our study, freezing caused the greatest reduction in body length and wet mass for both mayfly species (up to 18%). Similar results were found by Howmiller (1972) for worms, where the freezing method caused significantly greater length shrinkage and dry mass loss than preservation with other agents. These losses are probably associated with cellular and lipid changes during freezing and thawing processes (Farrant 1980;Morris 1972). Although the lack of studies on the freezing effect limits general conclusions, it seems that freezing is an unsuitable preservation method for studies involving the biomass or body length of mayflies (but see Cressa 1999).  Table 1. Note that a relative bias of 0% indicates reference values for fresh (unpreserved) specimens The effect of ethanol on body length was relatively small for both species and was comparable to the results of Lasenby et al. (1994). On the other hand, ethanol caused the largest decrease in dry mass of both species in all preservation methods (up to 40%). Similarly, Von Schiller and Solimini (2005) reported an underestimation of dry mass for mayflies in ethanol of approximately the same magnitude. In contrast, Dermott and Paterson (1974) noticed an increase in the percentage of dry mass, with the highest values observed for larvae when the preservative fluid was dried. In general, the main reasons for the decreased mass of ethanolpreserved organisms are losses of several components, such as large quantities of degraded lipids (Mills et al. 1982).
Formaldehyde caused a smaller bias of wet mass than ethanol, which is in agreement with the findings of other studies conducted on mayflies (e.g., Giberson and Galloway 1985;Heise et al. 1988;Leuven et al. 1985). Similar to our results, Heise et al. (1988) and Von Schiller and Solimini (2005) showed a relatively small influence of formaldehyde also on the body length of mayflies. Formaldehyde seems to cause less biased estimates of body biomass than other preservation agents in mayflies.
The main reason for why ethanol and formaldehyde have different effects on biomass is probably associated with the differences in the specific gravity of the agents, their evaporation rates before stabilization, osmotic strength and the chemical reactions they cause. Since ethanol may dissolve some lipids and formaldehyde may alter protein structure, invertebrates can respond differently (Landahl and Nagell 1978). Therefore, Donald and Paterson (1977) recommend using formaldehyde prepared with filtered ambient water as the most suitable solution. Nevertheless, this recommendation should be carefully considered because formaldehyde might still cause changes in specimen bodies, altering the water content and tissue composition (Von Schiller and Solimini 2005). Furthermore, in some studies, significant differences between ethanol and formaldehyde were not proven (e.g., Gaston et al. 1996;Wetzel et al. 2005), or formaldehyde even led to an increase in biomass (e.g., Landahl and Nagell 1978).  Table 1. Note that a relative bias of 0% indicates reference values for fresh (unpreserved) specimens  Table 1. Note that a relative bias of 0% indicates reference values for fresh (unpreserved) specimens

Effect of preservation time
We showed that the effect of preservation time on body length and biomass is species-specific. We found significant nonlinear temporal changes in the body length of R. carpatoalpina, where the largest biases were observed after 10-158 days, and a similar trend was described by Britt (1953) and Lasenby et al. (1994). For wet and dry mass, significant temporal changes were observed in R. carpatoalpina after a longer preservation time, similar to the reports of Giberson and Galloway (1985) and Heise et al. (1988). On the other hand, the body length of H. confusa was not affected by preservation time, which is in agreement with studies where the effect of preservation on body length after the first week was not observed (e.g., Giberson and Galloway 1985;Heise et al. 1988;Nolte 1990;Von Schiller and Solimini 2005). Also relatively small sample size and resulting low power of statistical tests might contribute to the non-significant effect on body length in H. confusa. Nevertheless, the response of body traits to preservation is complex, involving short-term changes in water content and long-term changes in tissue composition (Mills et al. 1982). In general, the rate of length and mass losses for freshwater invertebrates usually stabilises after approximately one month of preservation (Dermott and Paterson 1974;Donald and Paterson 1977;Edwards et al. 2009;Leuven et al. 1985;Mills et al. 1982).

Intraspecific variability of estimation bias
Regarding the relative bias in body wet and dry mass estimates, smaller individuals of both studied species showed greater mass loss during preservation than larger individuals (up to 53%), which is consistent with the findings of Edwards et al. (2009) andWetzel et al. (2005). Similarly, Stanford (1973) showed that wet mass loss was considerably greater in smaller larvae. Regarding the relative bias of body length estimation, larger specimens of R. carpatoalpina were subject to greater length reduction than the smaller specimens, as in Von Schiller and Solimini (2005). Nevertheless, this pattern does not seem to be general because in some studies, the relation between the relative bias of body mass and length during preservation and differences in body size for the same species was not proven (e.g., Donald and Paterson 1977;Heise et al. 1988).
We assume that larger larvae were less affected in terms of mass losses during preservation because of their specific biochemical composition, such as heavier encrustation, sclerotization and chitinization, which may buffer them against the preservation effects better than smaller larvae which have a larger proportion of soft body tissues (Howmiller 1972;Moody et al. 2016;Von Schiller and Solimini 2005).
Another reason might be the high degree of tissue degradation in smaller larvae (Nagy 2010). Thus, the ratio of the volume of preservative to the mass of animal tissue might explain why smaller larvae are more affected by the same volume of preservation agent than larger larvae (Howmiller 1972).

Interspecific variability of estimation bias
In our study, ethanol and freezing caused a more pronounced shrinking of body length in relatively subtle larvae of H. confusa than in the more robust and sclerotized individuals of R. carpatoalpina. The magnitude of the changes in body size measurements was species-specific (Lasenby et al. 1994) and related to the biochemical composition of the larvae (Kulka and Corey 1982;Mills et al. 1982), especially their body sclerotization (Howmiller 1972).
The species-specific response may be related to interspecific variation in protein and lipid contents, which are known to be diluted and altered by ethanol and formaldehyde, respectively (Gaston et al. 1996). For example, mayfly species vary considerably in their lipid content (e.g., Cavaletto et al. 2003;Meier et al. 2000;Sartori et al. 1992;Shipley et al. 2012), and this interspecific variation may lead to differences in relative bias of biological traits among species.
Dorsoventrally flattened R. carpatoalpina and cylindrically elongated H. confusa did not follow the same trend in terms of shrinking, and we assume that the species-specific response might be linked with differences in body shape. This means that the ratio of body length to mean larval width, the level of dorsoventral flattening and lateral broadening of the body segments might play important roles in the shrinkage of the studied larvae. However, further research with wider taxonomic coverage and a broader range of species' body shapes is needed to reach general conclusions. For example, Cressa (1999) assumed that the lack of a clear relationship in dry mass losses among preservatives between two mayfly species could indicate that leaching is dependent on the dimensions of the organisms. Additionally, Giguère et al. (1989) examined potential causes of variation in dry mass and showed that considering another metric for body size, such as body width, provided a better fit for the equation used to correct the mass of preserved organisms. However, studies that consider the relationship between invertebrate initial body shape and the shrinkage of preserved bodies are rare (but see Nolte 1990).
Although this study based on two species with different morphologies does not allow for the separation of phylogenetic and functional drivers of the interspecific differences, body shape should be considered when length-mass regressions are referred from the literature. Studies of a broader range of taxa and morphologies would be needed to disentangle the effects of phylogenetic relatedness and functional traits. Nevertheless, the considerable intraspecific variation precludes the use of some simple constants to correct the difference between unpreserved and preserved specimens, which would result in significant errors (Dermott and Paterson 1974;Wetzel et al. 2005).

Bias in length-mass relationships caused by preservation
We showed that the length-mass relationship of R. carpatoalpina estimated using fresh specimens was significantly different from the relationships based on preserved specimens (in order of effect size: formaldehyde < freezing < ethanol). However, researchers often assume that the masses of preserved and unpreserved specimens do not differ considerably and base the calculation of length-mass relationships on preserved specimens. Considering our model species, using preserved specimens to estimate the length-mass relationships of mayflies may lead to an error of more than 40% in dry mass prediction. Additionally, Cressa (1999), González et al. (2002) and Von Schiller and Solimini (2005) showed that the length-mass equations obtained from preserved mayfly specimens predicted lower dry mass than the equations derived from non-preserved specimens. The preservation-related length and mass losses cause underestimation of biomass and may represent a large source of bias in biomass and production estimates (e.g., Giberson and Galloway 1985;González et al. 2002;Heise and Galloway 1988;Leuven et al. 1985;Mills et al. 1982), and the bias can be even exacerbated when using length-mass relationships based on preserved specimens.

Conclusions
Our study showed that the preservation of samples causes permanent species-specific body changes in two mayfly species that are difficult to adjust by simple correction constants. Studies quantifying mayfly biomass should be based on unpreserved material, or the biomass should be estimated from length-mass relationships derived from unpreserved fresh specimens. If the use of fresh specimens is impossible or impractical, we recommend the following guidelines to minimize bias in biomass estimation: (1) Use formaldehyde, the preservative with the smallest effect on body length and mass, if preservation is needed; (2) If all samples cannot be processed immediately, the preservation time should be at least one month before processing to allow the rates of loss to stabilize; (3) Use the length-mass relationships estimated from non-preserved organisms; (4) Do not use simple correction constants because of the high inter-and intraspecific variability of the preservation effect on body traits; and (5) Report all details of preservation and sample processing (including preservation time) when publishing the results.
The preservation process affects the body length and mass of aquatic invertebrates and may considerably bias the accuracy of length-mass relationships, biomass and production estimates. Importantly, this study is based on two mayfly species, and given the large degrees of inter-and intraspecific heterogeneity, more studies covering a broader range of macroinvertebrate taxa are needed to reach general conclusions.