Introduction
Abundance is the most common measurement used to understand the functioning (e.g. fluxes of energy and nutrients in the food chain) of aquatic ecosystems (^{Callisto et al., 2001}; ^{Townsend et al., 2003}). However, individuals and taxa have different sizes and, consequently, different ecological importance, e.g. one large shredder (Triplectides) may be more important on leaf breakdown than two or more small shredders (^{Woodward et al., 2005}). Thus, biomass measurements are fundamental to understand processes (e.g. energy transformation) in aquatic ecosystems. Moreover, biomass is essential to evaluating secondary production, life history, trophic relationships between the functional groups, and the role of invertebrates in leaf decomposition (e.g., ^{Benke and Huryn, 2010}; ^{Brand and Miserendino, 2012}; ^{Luek et al., 2015}).
Biomass can be estimated directly (weighing individuals) or indirectly using body and/or case dimensions (^{Smock, 1980}; ^{Burgherr and Meyer, 1997}). Direct methods for determining biomass generally do not allow use of individuals for further experiments (^{Cressa, 1999a}). In addition, in determining biomass of aquatic invertebrates that have small size and are numerous require a considerable amount of time and resources. Thus, indirect methods may be the best alternative for estimating the biomass of these organisms (^{Smock, 1980}).
Among the indirect methods for determining biomass, regression models examining dry mass and body dimensions are the most often used (^{Smock, 1980}; ^{Meyer, 1989}). Regression models are accurate, fast and inexpensive; however, the accuracy of regression methods is typically taxon-specific and varies among populations (^{Benke et al., 1999}; ^{Johnston and Cunjak, 1999}). In general, low efficiency has been recorded in models created for different regions and even in models for closely related taxa (e.g., family). Moreover, general models are not effective for estimating biomass at the order level, such as in Trichoptera, mainly due to the high variability in the bodies of individuals (^{Sample et al., 1993}; ^{Cressa, 1999a}).
Models for estimating the biomass of aquatic insects have been developed primarily for populations in temperate regions (e.g., ^{Smock, 1980}; ^{Meyer, 1989}; ^{Benke et al., 1999}; ^{González et al., 2002}). In the Neotropics, length-mass relationship models have been created for insect populations in streams in Argentina (^{Miserendino, 2001}), Brazil (^{Becker et al., 2009}; ^{Martins et al., 2014}), and Venezuela (^{Cressa, 1999a}). In Brazil, biomass estimation models have been obtained mainly for shredder organisms, in particular Phylloicus (Trichoptera: Calamoceratidae; ^{Becker et al., 2009}; ^{Martins et al., 2014}).
Triplectides (Trichoptera: Leptoceridae) are also important shredders in the Amazon region (^{Landeiro et al., 2010}; ^{Martins et al., 2015}). Due to this feeding habit, this genus has the potential for use in laboratory studies to evaluate the role of invertebrates in leaf breakdown in Amazonian streams. Our objective is to provide predictive models based on body and case dimensions for estimating the biomass of Triplectides egleri^{Sattler, 1963}, which is the only species of this genus reported in the study area (Barro Branco stream).
Material and methods
Study area
This study was conducted with a population of T. egleri in the Barro Branco stream, located in the Ducke Reserve (2°55' and 03°01' S, 59 53' and 59°59' W), Manaus municipality (Amazonas state). The Ducke Reserve comprises 10,000 hectares (100 km^{2}), drained by a dense network of low-order streams and covered by “terra firme” (unflooded upland) tropical rain forest (^{Ribeiro et al., 1999}). The streams in this area are shaded by dense riparian vegetation and have acidic (pH = 4.63 ± 0.08) well-oxygenated (6.62 ± 0.06 mg L^{−1}) water, with a low capacity to conduct electrical current (10.71 ± 0.41 μS cm^{−1}), and an average temperature of 24.52 ± 0.52 °C (^{Martins et al., 2014}).
Sample collection
Triplectides egleri larvae were collected in July 2014. Sampling of individuals was performed manually in pool areas of the stream with accumulation of litter. We sampled 57 larvae of different sizes and transported them to the laboratory in a Styrofoam box with stream water and litter; in the laboratory they were removed from their cases and analyzed individually (^{Becker et al., 2009}). Because two larvae got out of their cases, after collection, the sample size for case dimensions was 55.
Body dimensions used to predict the biomass of individuals were: body length, head-capsule width and interocular distance (^{Becker, 2005}; ^{Becker et al., 2009}). Body length was measured as the distance between the anterior region of the head and the distal region of the abdomen. Head-capsule width was measured as the widest portion of the head. Interocular distance was measured at the smallest distance between the eyes.
Case length and width can be used to estimate biomass because these insects build a portable tubular case from small twigs (^{Sattler, 1963}). Triplectides egleri excavate the ventral region of the case at its opening, probably to facilitate locomotion. Case width was therefore measured as the extension of the excavated portion (Fig. 1). Case length was measured as the distance between the anterior (case opening) and the posterior portion of the case.
The body and case dimensions were determined from digital images (Leica stereomicroscope M165, accuracy = 0.001 mm) obtained with Leica Application Suite (LAS) software version 3.8. After measurement, larvae were placed in an oven (OdontoBras Mod-EL-1.6) for 72 h at 50 °C and cooled in a desiccator for 24 h. Biomass (dry weight) of each individual was measured with a high-precision digital scale (Metter Toledo, model AB265-S/FACT; accuracy = 0.01 mg).
Data analysis
To evaluate the relationship between biomass and the body or case dimensions of T. egleri, we used regression analysis based on the least squares method (^{Zar, 2010}). Three regression models were fitted to the data: (a) linear (y = a * x + b), (b) exponential (y = a * e^{bx}; in linear form: ln y = ln a + b * x), and (c) power (y = a * x^{b}; in linear form: ln y = ln a + b* ln x). In all models, y is the biomass (mg), x is the size of the body or case (mm), and a and b are regression constants describing the allometric relationship between variables y and x. The fit of the regression models was assessed by the coefficient of determination (R^{2}) expressed in proportion or percentage (%) (^{Zar, 2010}). The coefficient of variation was obtained according to equation: CV = (SD/mean)*100, where SD is the standard deviation. Results of CV were presented in percentage (%).
Cross-validation analyses were performed to assess the predictive power of the models obtained from our data. We follow the method used by ^{Martins et al. (2014)}. Thus, our data set was divided in half, one of the subsets was used to create models (training set) and the other was used to evaluate the performance of predictive models (test set). Moreover, to assess whether biomass estimation models proposed for other leptocerids (Nectopsyche genus) are satisfactory to estimate T. egleri biomass, we used the equations proposed by ^{Cressa (1999a}, ^{b)}, which were obtained to Trichoptera larvae from the Venezuelan Amazon. The reliability of cross-validation was assessed by comparison of predicted results with the biomass obtained through weighing of the larvae. The significance of our cross-validation was tested using a paired t-test. The cross-validation was repeated 100 times and the result was expressed as an average value. All analyses were performed with the R program (^{R Core Team, 2014}).
Results
Dry weight of T. egleri had a high coefficient of variation (CV = 108%), with values from 0.18 to 11.30 mg (Table 1). Among body dimensions, body length had the highest range (3.97–18.25 mm) and coefficient of variation (36%). Case dimensions also had high coefficients of variation as compared to body dimensions (Table 1).
Range | Mean | SD | CV | N | |
---|---|---|---|---|---|
Body dimensions | |||||
Body length (mm) | 3.97–18.25 | 10.34 | 3.68 | 35.60 | 57 |
Head-capsule width (mm) | 0.39–0.80 | 0.80 | 0.23 | 28.60 | 57 |
Interocular distance (mm) | 0.22–0.72 | 0.48 | 0.15 | 30.06 | 57 |
Case dimensions | |||||
Case length (mm) | 13.00–146.40 | 32.65 | 18.53 | 56.75 | 55 |
Case width (mm) | 0.56–3.88 | 2.03 | 0.69 | 34.10 | 55 |
Body mass | |||||
Dry mass (mg) | 0.18–11.30 | 2.69 | 2.91 | 108.09 | 57 |
All regression models used to estimate T. egleri biomass were significant (Table 2). The best fit between biomass and body or case dimensions was obtained using the power model, followed by the exponential and linear models. For the linear, exponential and power models, body length was the dimension that best predicted biomass. Among the case dimensions, case width provided the best estimate in all models (Table 2).
Function | Models | a | b | R^{2} | N | p |
---|---|---|---|---|---|---|
Linear | DM∼BL | −3.990 | 0.646 | 0.663 | 57 | <0.001 |
DM∼HW | −4.157 | 8.564 | 0.444 | 57 | <0.001 | |
DM∼ID | −4.093 | 13.996 | 0.482 | 57 | <0.001 | |
DM∼CL | 1.191 | 0.044 | 0.065 | 55 | 0.034 | |
DM∼CW | −2.857 | 2.697 | 0.424 | 55 | <0.001 | |
Exponential | ln(DM)∼BL | −2.352 | 0.270 | 0.816 | 57 | <0.001 |
ln(DM)∼HW | −2.808 | 4.060 | 0.711 | 57 | <0.001 | |
ln(DM)∼ID | −2.739 | 6.556 | 0.754 | 57 | <0.001 | |
ln(DM)∼CL | −0.409 | 0.026 | 0.176 | 55 | 0.001 | |
ln(DM)∼CW | −1.907 | 1.148 | 0.531 | 55 | <0.001 | |
Power | ln(DM)∼ln(BL) | −5.821 | 2.755 | 0.835 | 57 | <0.001 |
ln(DM)∼ln(HW) | 1.257 | 3.058 | 0.728 | 57 | <0.001 | |
ln(DM)∼ln(ID) | 2.648 | 2.852 | 0.746 | 57 | <0.001 | |
ln(DM)∼ln(CL) | −4.976 | 1.596 | 0.395 | 55 | <0.001 | |
ln(DM)∼ln(CW) | −1.026 | 2.238 | 0.546 | 55 | <0.001 |
DM, dry mass; BL, body length; HW, head-capsule width; ID, interocular distance; CL, case length; CW, case width.
To linear models, the predicted biomass using models for body length was underestimated (∼ −11%; Table 3). On the other hand, the predicted biomass using others body and case dimensions were overestimated (∼18–64%). To exponential and power, the predicted biomass was overestimated when we used body (Exponential: ∼12–27%; Power: ∼8–24%) and case (Exponential: ∼35–52%; Power: ∼32–49%) dimensions. To all models, no significant differences between predicted biomass using cross-validation models and observed biomass was recorded (Table 3). On the other hand, all models using the models proposed to other Neotropical leptocerids (Nectopsyche) and Trichoptera (^{Cressa, 1999a}, ^{b}) were significantly different in relation to T. egleri biomass and, values were underestimated up to 98% (Table 4).
Model | Body and case dimensions | Difference (%) | R^{2} | T-Test | ||
---|---|---|---|---|---|---|
t | Gl | p | ||||
Linear | BL | −10.68 | 0.59 | 1.84 | 28 | 0.428 |
ID | 18.12 | 0.43 | 1.62 | 28 | 0.435 | |
HW | 28.22 | 0.32 | 1.90 | 28 | 0.392 | |
CW | 33.83 | 0.38 | 4.16 | 27 | 0.089 | |
CL | 64.30 | 0.08 | 5.23 | 27 | 0.107 | |
Exponential | BL | 11.52 | 0.56 | 1.68 | 28 | 0.424 |
ID | 18.42 | 0.38 | 2.66 | 28 | 0.277 | |
HW | 27.07 | 0.24 | 2.34 | 28 | 0.339 | |
CW | 35.04 | 0.25 | 2.21 | 27 | 0.394 | |
CL | 52.30 | 0.12 | 2.92 | 27 | 0.218 | |
Power | BL | 8.43 | 0.67 | 2.31 | 28 | 0.345 |
ID | 17.42 | 0.41 | 2.60 | 28 | 0.285 | |
HW | 23.56 | 0.25 | 2.28 | 28 | 0.345 | |
CW | 32.33 | 0.33 | 3.44 | 27 | 0.228 | |
CL | 48.89 | 0.28 | 2.92 | 27 | 0.351 |
BL, body length; HW, head-capsule width; ID, interocular distance; CL, case length; CW, case width.
Reference | Taxon | Preservation method | Difference (%) | R^{2} | t-test | ||
---|---|---|---|---|---|---|---|
t | Gl | p | |||||
Cressa (1999a) | Nectopsyche | Not preserved | −85.03 | 0.26 | 19.17 | 28 | <0.001 |
Cressa (1999b) | Nectopsyche | Freezing | −95.00 | 0.28 | 20.66 | 28 | <0.001 |
Cressa (1999b) | Nectopsyche | Formaldehyde 4% | −98.49 | 0.32 | 22.54 | 28 | <0.001 |
Cressa (1999b) | Nectopsyche | Kahle | −98.23 | 0.30 | 21.61 | 28 | <0.001 |
Cressa (1999b) | Trichoptera | Freezing | −97.77 | 0.27 | 21.39 | 28 | <0.001 |
Cressa (1999b) | Trichoptera | Formaldehyde 4% | −96.87 | 0.28 | 21.69 | 28 | <0.001 |
Cressa (1999b) | Trichoptera | Kahle | −98.05 | 0.26 | 21.31 | 28 | <0.001 |
BL, body length; HW, head-capsule width; ID, interocular distance; CL, case length; CW, case width.
Discussion
In general, body length was the best dimension for estimating the biomass of Triplectides egleri, explaining 66–75% of biomass variation. Other studies have also reported that body length is an efficient dimension for estimating the biomass of other genera of Trichoptera (e.g., ^{Towers et al., 1994}; ^{Benke et al., 1999}; ^{Miserendino, 2001}; ^{Becker et al., 2009}; ^{Martins et al., 2014}). This dimension typically provides good estimates because most of the body of the insect larvae (holometabolous) is not sclerotized and tends to increase gradually in each instar, accompanying the changes in the mass of individuals (^{Johnston and Cunjak, 1999}).
Interocular distance and head-capsule width also provided good estimates of biomass (>44%). For example, the best predictor of biomass in the exponential model was the interocular distance. In general, sclerotized structures (e.g., head) show low variation within each instar and tend to have a poor fit with biomass as compared to body length (^{Towers et al., 1994}; ^{Burgherr and Meyer, 1997}). However, sclerotized structures are resistant to damage in terms of breaks or deformities induced by preservation methods, and may be very useful for biomass estimation of preserved or damaged individuals (^{Johnston and Cunjak, 1999}).
In our study, case dimensions provided poor fit with biomass as compared to body dimensions, not corroborating the studies of ^{Cressa (1999a)} and ^{Martins et al. (2014)}, who observed similar accuracy for case and body dimensions for predicting biomass of Nectopsyche and Phylloicus (Trichoptera). The strong relationship between case dimensions and biomass of Nectopsyche and Phylloicus may be related to use of sand grains and leaves to build their cases (^{Glover and Floyd, 2004}; ^{Prather, 2003}). Usage of these materials, which are small or can be cut to size determined by larvae, allows construction of a case according to the size of the larvae. On the other hand, Triplectides individuals build their cases by drilling small twigs (^{Sattler, 1963}), behavior which limits the fit between case and biomass because it depends on the larvae finding small twigs suitable to their size, which does not always occur (^{Camargos and Pes, 2011}).
The power model provided the best fit between the biomass and body and case dimensions of T. egleri. However, the difference in relation to the exponential model was less than 2% (except for case length). Other studies have reported that the difference between power and exponential models in Trichoptera biomass explanation is low – usually less than 6% (^{Becker et al., 2009}; ^{Martins et al., 2014}) – and that differences observed using different regression models tend to decrease with increasing number of observations (Wenzel et al., 1990).
In aquatic invertebrates, the “b” value of the mass-length curve adjusted by the power regression model is close to 3, ranging between 2 and 4 (^{Towers et al., 1994}; ^{Benke et al., 1999}; ^{González et al., 2002}). Specifically, for insects, the average value of “b” tends to be lower than 3 (^{Benke et al., 1999}; ^{Smock, 1980}). In this study, “b” values for the relationship between body dimensions and biomass followed this pattern, indicating that the body narrows with an increase in body length (^{Benke et al., 1999}). For other Leptoceridae genera (Oecetis and Nectopsyche) of temperate regions, the “b” value has been found to be greater than 3, indicating that bodies of these organisms become proportionally larger as the length increases (^{Benke et al., 1999}). In the tropics, ^{Cressa (1999a)} recorded a “b” value of 1.35 for Nectopsyche individuals. The “b” values of equations for biomass estimation of T. egleri and Nectopsyche indicates that larvae of both genres tend to be longer than wide as the individuals grow. However, the low “b” value obtained for Nectopsyche indicates that the body shape of this genus changes more intensely during the larval development. Moreover, may have a large ontogenetic variation between different species comprising the genus. These differences may account for the high underestimation of biomass T. egleri when we used the equations proposed by ^{Cressa (1999a}, ^{b)}. Thus, our results reinforce the importance of obtaining models for organisms (genus or species) studied from a particular geographic region (^{Martins et al., 2014}).
Compared to the results obtained in other studies, we conclude that our regression models (power and exponential) describe with high precision the relation between body dimensions and biomass of T. egleri. Thus, they are useful tools for determination of biomass and can be used in studies that aim understanding the role of shredder invertebrates on ecological processes in Amazonian streams.