Portal de Periódicos da CAPES

Estimation of Insect Biomass by Length and Width

1993; University of Notre Dame; Volume: 129; Issue: 2 Linguagem: Inglês

10.2307/2426503

1938-4238

Bradley E. Sample, Robert J. Cooper, Richard D. Greer, Robert C. Whitmore,

Morphological variations and asymmetry

-Length-weight regression models were generated for 10 orders and 35 families of adult and larval insects using a power model. Additional models were generated that incorporated width as an independent variable, to account for varying body morphology within insect taxa. Inclusion of width improved the generalized insect model and models at the order level, but was of less value in improving family level models. The predictive value of all models was high; estimates were generally within ?2 mg of the actual values. The parameter values for our models were similar to those produced by other researchers. INTRODUCTION Estimation of insect biomass is extremely important to studies of ecosystem processes because of the great abundance of insects, their ecological and taxonomic diversity (Stork, 1988) and their importance as a vertebrate food resource (Rotenberry, 1980). Traditionally, biomass values were derived by weighing several individuals within a taxa, calculating a mean biomass, and multiplying by the number of individuals collected. This procedure was tedious, time-consuming, and expensive (Rogers et al., 1977). Additionally, accurate measurement of biomass requires access to precision electronic balances, particularly for small specimens. Such accuracy may not be possible in field studies. The use of length-weight regression equations has been suggested as an alternative method for determining insect biomass. In addition to a general formula for all insects (Rogers et al., 1976), equations have been generated for specific orders and families of insects (Beaver and Baldwin, 1975; Rogers et al., 1977; Schoener, 1980; Jarosik, 1989). Other methods for estimating insect biomass include wing length indices (Miller, 1977), liquid displacement (Ciborowski, 1983) and optical measurement (Smiley and Wisdom, 1982). While use of length-weight formulae may simplify biomass estimation, Schoener (1980) cautions against lumping of insects of widely different body proportions into the same formula. Because longer insects tend to be more narrow, they tend to reduce the slope of the length-weight regression line, particularly at the upper end (Schoener, 1980). To reduce the effect of body shape on the formulae, the use of taxa-specific equations was suggested (Rogers et al., 1976; Sage, 1982; Schoener, 1980). In this paper we expand on the length-weight formulae of Rogers et al. (1976 and 1977) and Schoener (1980). In addition, we develop new models that include width, for use with taxa with differing body proportions. Actual vs. estimated biomass values are compared to evaluate the predictive value of these models. METHODS Adult and larval insects were collected with light traps, malaise traps, and foliage clipped from forested areas in eastern West Virginia between 1988 and 1991. All specimens were stored in a freezer until measured. The methods we employed were chosen to be representative of other studies of lengthPresent address: Department of Biology, Memphis State University, Memphis, Tennessee 38152 2 Present address: Exxon Biomedical Sciences Inc., Mettlers Rd., CN-2350, East Millstone, New Jersey 08875 234 This content downloaded from 157.55.39.203 on Sun, 11 Jun 2017 18:34:48 UTC All use subject to http://about.jstor.org/terms 1993 SAMPLE ET AL.: ESTIMATING INSECT BIOMASS 235 weight relations. Lengths and widths were measured to the nearest 0.01 mm with a dendrochronometer (Model 3, Fred C. Hanson Co., Mission Viejo, Calif. 92690) equipped with a 1-2.5 x zoom stereoscope. This device is designed for digital measurement of tree ring widths in age determination and was chosen over calipers and ocular micrometers for its accuracy and precision. Specimens were pinned to cardboard sheets, placed on the dendrochronometer platform, advanced beneath the stereoscope, and length or width values were recorded from the dendrochronometer display. Body length was measured from the frons to the tip of the abdomen. Antennae, ovipositors and wings extending beyond these points were not included in the total length measurement. Width was measured at the mesothorax for all insects, including larvae. All insects were dried in a 70 C drying oven for 48 h, allowed to cool, then weighed to the nearest 0.1 mg on an electronic balance (Model SWA-200-DR, Sargent-Welch Inc.). Specimens damaged at any state of the study (storage, measurement or drying) were excluded from analyses. We used the power model (Rogers et al., 1977) to describe the length-weight relationship of adult and larval insects. This model is: y = a(X)b Where: y = weight x = length or length*width Data were transformed to natural logarithms to linearize for regression analyses. The transformed model is: ln(y) = ln(a) + b ln(x). Regression coefficients and Pearson's correlation coefficients were calculated using PROC REG (SAS Institute Inc., 1988). Regression models were calculated at the class, order and family level for those taxa in which sample size was n 10. Predictive value of selected models was evaluated by comparing the esimtated to actual biomass for a sample of specimens that had been excluded from the data set from which the models had been generated. Matched-pair t-tests (Dowdy and Wearden, 1983) were used to determine if the difference between the calculated and actual biomass values was significantly different from 0. Because the chance of committing a Type I error increases as the number of tests increases, the alpha level (P < 0.05) was adjusted for the number of taxa-specific t-tests performed using the Bonferonni adjustment (Rice, 1989). To verify that the validation data originated from the same population as the model data, regression analyses identical to those performed on the model data were performed on the validation data. The slopes obtained from the validation data were compared to those from the model data using Z-tests (Dowdy and Wearden, 1983). Nonsignificant differences between slopes were assumed to indicate that model and validation data originated from the same population. Additional validation was performed by comparing the slope (power) parameters from our data to those of Rogers et al. (1976 and 1977) and Schoener (1980). Only models with length as the independent variable were evaluated. Z-tests (Dowdy and Wearden, 1983) were performed to determine if the slopes differed significantly. RESULTS AND DISCUSSION Data were collected for 1673 adult insects, representing 13 orders and 99 families. In total, 84 models were generated for adult insects (Table 1). Larval models were calculated This content downloaded from 157.55.39.203 on Sun, 11 Jun 2017 18:34:48 UTC All use subject to http://about.jstor.org/terms 236 THE AMERICAN MIDLAND NATURALIST 129(2) 00 'C ~~~~ C'C 'CO TTT'C LO C' C' tLo~ , C' 0 LO n LO rL(n 'C n 00 ( C' C'C\''CC'0C' 'C' C' C' nC'Cr-C C' C' C' C' C'CON0C' C'O mC' C' C'00 rn T~~r L O Lrn T-'t\C00 cn TLO LOn c ( C'CLOCr00 C' c100 c1 c00 VO