Estimating biomass of Neotropical spiders and other
arachnids (Araneae, Opiliones, Pseudoscorpiones, Ricinulei) by
masslength regressions.




Abstract: 
We sampled 505 specimens of 7 arachnid orders (313 Araneae, 65
Opiliones, 111 Pseudoscorpiones, 10 Ricinulei, 3 Schizomida, 1
Thelyphonida, 2 Scorpiones) in natural forest and agroforestry sites in
central Amazonia to analyze fresh and dry mass to body length relations.
The low number of schizomids, scorpions, and thelyphonids did not allow
statistical analyses, but the raw data are given, because these
represent the first data published for these groups from Amazonia. For
all other orders general masslength relationships for ecological
studies were determined. Nonlinear regressions with a power model
proved to describe the relations very well and are highly significant
for all taxa and groups analyzed. The resulting equations can thus be
used to estimate biomass of large samples of arachnids from Amazonia
based on individual body length measurements. Linear regressions of mass
to length with logtransformed data also described the relation
adequately, but using the resulting equations to estimate biomass of the
whole spider sample caused a higher bias. This is because small biases
of masslength relation of the largest spider individuals are
exponentiated. However, linear regressions behaved better for spiders
smaller than 8 mm. The ratio of dry to fresh mass was around 0.3 for
spiders; 0.4 for pseudoscorpions, schizomids, and thelyphonids; 0.44 for
opilionids; and 0.53 for Ricinulei. A second sample of 99 spiders from a
South Brazilian Atlantic Forest revealed similar masslength relations,
but a different dry to fresh mass ratio. For spiders, the usefulness of
general equations to determine the biomass of bulk samples from
ecological studies with certain precision requirements was further
explored by using the equations from the two datasets crosswise,
regarding the resulting bias and by applying equations to a further
dataset from an ecological investigation. In conclusion and accordance
to former studies, general equations derived from masslength
regressions of bulk samples including many specimens of different
families and guilds are appropriate for an estimation of the biomass of
bulk samples from ecological studies. Equations from masslength
regressions from the literature, resulting from spider samples in
temperate regions, should not be used to estimate biomass of samples
from neotropical spider assemblages, especially when absolute biomass is
of interest and when precision is required. They underestimate biomass
of tropical assemblages due to a strong bias in masslength relation of
tropical spiders larger than 10 mm. Depending on the distribution of
large spiders in samples, considerable biases in single samples could
affect ecological analyses. Analisamos as relacoes entre comprimento corporal e massa fresca e seca de 505 especimes de sete ordens de aracnideos (313 Araneae, 65 Opiliones, 111 Pseudoscorpiones, 10 Ricinulei, 3 Schizomida, 1 Thelyphonida, 2 Scorpiones) coletados em florestas e agroflorestas na Amazonia Central. Devido ao numero baixo de Schizomida, Scorpiones e Thelyphonida nenhuma analise estatistica foi possivel e os dados brutos sao apresentados a serem os primeiros dados publicados destes grupos para a Amazonia. Para as outras ordens analises de regressao foram feitas. Regressoes naolineares de modelo potencial demonstraram excelente descricao para as relacoes, sendo altamente significativas para os taxons e grupos analisados. Os coeficientes obtidos nestas regressoes poderao servir de base para o calculo de biomassa em amostras da Regiao Amazonica que contenham grande numero de aracnideos, utilizandose como medida somente o comprimento total de cada individuo. Utilizandose dados logaritmicamente transformados, regressoes lineares de massacomprimento tambem descreveram adequadamente a relacao. Todavia a utilizacao destes coeficientes, para estimar exclusivamente a biomassa da amostra total de aranhas, apresentou resultados tendenciosos em funcao do efeito forte da relacao exponencial a desvios pequenos em aranhas de grande porte. Regressoes lineares apresentaram um comportamento estatistico mais favoravel apenas para aranhas com menos de 8 mm de comprimento corporal. A relacao obtida para massa seca em relacao a massa fresca foi de cerca de 0.3 para aranhas, cerca de 0.4 para Pseudoscorpiones, Schizomida e Thelyphonida, 0.44 para Opiliones e 0.53 para Ricinulei. Uma segunda amostragem de 99 aranhas na regiao meridional da Mata Atlantica brasileira revelou relacoes de massacomprimento similares, porem, com uma relacao diferenciada de massa seca a massa fresca. Para a ordem de aranhas a utilidade de equacoes gerais para a determinacao da biomassa de amostras ecologicas com devida precisao foi analisada aplicando coeficientes resultando de amostragens de outras regioes. Concluimos que coeficientes de regressoes de massacomprimento sao apropriados para uso em relacao a assembleia inteira de aracnideos, desde que as amostras contenham especimes de varias familias e guildas diferentes. Os coeficientes obtidos na regressao da grande amostragem da Regiao Amazonica podem ser usadas para a assembleias de aranhas da Mata Atlantica, porem nao e aconselhavel uso reciproco, mais especificamente para estimativas de massa seca. A utilizacao de coeficientes de regressoes de massacomprimento disponiveis atualmente na literatura, resultante de amostragens em regioes temperadas, deveria ser evitada para a estimativa de biomassa em amostras de assembleias de aranhas neotropicais. Estes coeficientes subestimam a biomassa de assembleias tropicais devido a uma grande distorcao na relacao entre massa e comprimento corporal em aranhas maiores do que 10 mm. Desta maneira analises ecologicas podem ser altamente influenciadas pela distribuicao de grandes aranhas entre as amostras individuais com distorcao dos resultados. Keywords: Arachnida, masslength relationship, Brazil 


Article Type:  Report 
Subject: 
Primary productivity (Biology)
(Research) Spiders (Environmental aspects) Spiders (Physiological aspects) Biomass (Measurement) 
Authors: 
Hofer, Hubert Ott, Ricardo 
Pub Date:  05/01/2009 
Publication:  Name: Journal of Arachnology Publisher: American Arachnological Society Audience: Academic Format: Magazine/Journal Subject: Biological sciences; Zoology and wildlife conservation Copyright: COPYRIGHT 2009 American Arachnological Society ISSN: 01618202 
Issue:  Date: May, 2009 Source Volume: 37 Source Issue: 2 
Topic:  Event Code: 310 Science & research 
Geographic:  Geographic Scope: Brazil Geographic Code: 3BRAZ Brazil 
Accession Number:  238834222 
Full Text: 
Biomass data (in the sense of the weight of living animals per unit
area, Bornebusch 1930; Edwards 1966) for arthropods are needed in many
ecological studies, especially when these aim to analyze the role and
functions of these abundant animals in ecosystems and food webs. Biomass
of soil fauna is of special interest in studies of nutrient cycling
involving the role of the fauna in decomposition and organic matter
transformation. The importance of soil fauna has long been recognized
and their function is also being studied more frequently in Neotropical
ecosystems (Lavelle et al. 1997, 2001; Barros et al. 2003, 2006; Mathieu
et al. 2004). The context in which we needed to estimate biomass of
arachnids and other arthropods was given by two projects in the
BrazilianGerman research programme SHIFT (Studies on Human Impact on
Forests and Floodplains in the Tropics) studying the quantitative
contribution of soil fauna to decomposition in central Amazonian natural
forests and different agroforestry systems (Hofer et al. 2001; Hanagarth
et al. 2004; Martius et al. 2004; Brown et al. 2006). Biomass can be obtained by direct weighing of individual living arthropods with analytical balances, but this is a very time consuming task and for very active animals it is difficult or impossible to obtain precise data. Certainly direct weighing is not a practical method in the field and for larger samples in laboratories. Most specimens in ecological studies are trapped and killed in fluids such as ethanol and it is difficult to measure preserved animals on a balance. Also, weighing fresh weight of preserved animals may provide incorrect estimations as body weight may be altered during preservation. For most studies dry mass is easier to obtain, but drying specimens or bulk samples to a constant weight, usually at 65[degrees] C or more, makes it impossible to later identify them due to their fragility. An alternative method is to use statistically verified relationships of mass with easily measurable body dimensions, such as body length or width, to estimate the biomass of each specimen. Body length might even be measured in the field or estimated with live animals so animals may not even need to be collected. Regressions using a power model (mass = a [(size).sup.b]) usually adequately describe masslengthrelations for most arthropods (Rogers et al. 1976, 1977; Schoener 1980; Sample et al. 1993; Edwards 1996). They have also been shown to provide useful data for spiders from temperate regions (Breymeyer 1967; Norberg 1978; Clausen 1983; Edwards 1996; Henschel et al. 1996a; Lang et al. 1997; Edwards & Gabriel 1998). Spiders and to a lesser extent other arachnids (opilionids, pseudoscorpions) are abundant in all terrestrial environments and are often included in functional ecological studies due to their well defined position in the food web as (arthropod) predators and their usefulness to indicate habitat quality (Jocque 1981; Chen & Wise 1999; Wise et al. 1999; Lawrence & Wise 2000, 2004; Wise 2004). As Henschel et al. (1996a) state, it is useful and possible to use general equations for arachnid orders (e.g., spiders and opilionids) to estimate the biomass of single specimens for the whole assemblage, notwithstanding the different speciesspecific masslength relationships. They suggest their equations are valid for other regions and habitats in Europe, at least for community studies involving numerous families, genera and species. Our main interest was to derive an equation for a general relationship to estimate biomass of bulk samples to compare soil fauna biomass at different sites in tropical South America. Thus we sampled 505 specimens of spiders and other arachnids from one location in central Amazonia and analyzed masslength relations of this large collection (first data set) in order to obtain valid equations for the biomass estimates we needed for our studies of Amazonian forest and agroforestry systems. We tested whether these equations reliably estimated biomass of bulk samples of spiders or if different equations were necessary for different functional groups (e.g., wandering versus web building spiders), size classes (tiny spiderlings versus large mygalomorphs), or spiders with an extraordinary body shape (like Micrathena or Deinopis). A second sample of spiders (second data set) was obtained from another region and large scale forest ecosystem of Brazil, e.g., in the southern part of the Brazilian Atlantic Forest (Mata Atlantica) and analyzed in the same way. Having two large data sets on spiders at hand and given the numerous data for this arachnid order in the literature, we explored the usefulness and limitations of general equations to determine the biomass of bulk samples from ecological studies with the required precision. This was done in three steps: 1. Determining which biases would be introduced when using equations from outside the Neotropical region for the Amazonian sample; 2. Determining the bias introduced by using the equations from the first data set (Amazonia) for the second data set (Atlantic Forest) and vice versa; 3. Determining the bias introduced by applying different equations for data from one ecological study in Amazonia and one ecological study in the Atlantic forest (application data sets) and looking for an effect of the bias on the conclusions of these studies. METHODS A second data set including 99 spiders from a South Brazilian Atlantic Forest (Mata Atlantica) (Reserva do Cachoeira, Antonina, Parana: 25[degrees]25'S, 48[degrees]40'W) was obtained in 2007. Spiders (Table 1) were sampled manually at night and during the day along trails in secondary forests. Weighing and measuring procedures were the same as described above. Tests for the effects of the bias from different equations were done with two application data sets: one from Amazonia, where spiders were sampled from 16 replicate sites of each of 7 different plantation systems (EMBRAPA central Amazonia) by means of large soil cores; and one from the Atlantic Forest, where 10 litter samples (1 [m.sup.2]) were taken in each of three different regeneration stages of a submountain forest (Schmidt et al. 2008). From both collections all spider specimens (n = 441 and 276) were individually measured (body length), so that coefficients from different regression equations could be applied to estimate the total biomass per site. Data were analyzed with Statistica 7.1 (StatSoft 2005) and graphs prepared with SigmaPlot[C] 8.0.2 (SPSS 2002). RESULTS Analyses of masslength relations.Masslength relationships (for both fresh and dry mass) for the arachnid orders with enough specimens sampled in the Amazonian habitats (first data set) are very well correlated with a regression model of the nonlinear (power) form: mass = a [(length).sup.b]. Determination coefficients are usually > 0.9 (Tables 3, 4) and type I error probabilities are very low (< 0.001) for both parameters, with the exception of the rare Ricinulei (n = 10, P = 0.15 for coefficient a). The masslength relationship is almost equally well described with a linear model using logarithmic data for length and weight (In (mass) = a + b In (length)). Note that power regression results are often presented in doublelogarithmic plots, but the model parameters are not the same for a power model calculated on raw data and a linear model calculated on logtransformed data. In our dataset the linear model represents the most abundant small spiders better because the few large spiders have a very high influence in the power model (Fig. 1). However the fresh biomass of the whole sample (313 spiders) with a mean length of 4.83 mm when estimated with the power model was closer to the observed biomass (99.8%) as when estimated with the linear model (95.7%). The same is true for dry mass estimation (power: 97.6%, linear: 86.9% ofobserved mass). Because different bulk samples might predominantly consist of either small or large spiders, often influenced by the sampling method, it might be useful to use either the linear model or the power model. In some cases it might even be useful to split a sample by size and use the linear model for spiders < 8 mm and the power model for spiders > 8 mm. Therefore, we present the coefficients of both models (Tables 3, 4). [FIGURE 1 OMITTED] The 313 Amazonian spiders that were measured and weighed represent a large spectrum in terms of size, shape, and taxonomic and functional groups. This dataset includes tiny orbweavers like Theridiosomatidae and Anapidae; tiny, but longlegged Ochyroceratidae; tiny, but shortlegged wandering spiders like Oonopidae; mediansized jumping spiders; very small to large mygalomorphs; large ctenid hunters; as well as large, longlegged pholcids (Table 1). Very few spider specimens (the smallest spider an ochyroceratid, one ctenid, and most of the longlegged ochyroceratids) lay outside the 95% confidence limits of our regressions and their exclusion did not lead to considerable changes in the model parameters. Nevertheless we calculated separate regressions for small spiders, the families Ctenidae and Oonopidae, the main hunting (or wandering) guilds; and webbuilding spiders because these groups might be of special interest in ecological studies (see also below); and because they always received high determination coefficients and significances (Tables 3, 4). The strong correlations in some cases caused very high PRESS values (> 30,000 for fresh mass and > 500,000 for dry mass vs. length of spiders). The PRESS value (Predicted Residual Error Sum of Squares) is a gauge of how well a regression model predicts new data and often a hint to overfitting of a dataset, resulting in decreased usefulness for other datasets. To test this, we split the whole Amazonian data set by a random procedure in one learn and one test dataset (crossvalidation). For both fresh mass and dry mass the regression line of the test dataset was well inside the 95% confidence limits of the learn dataset. This shows that the strong correlation is not a result of overfitting and consequently the resulting formulae should be useful for an estimation of fresh or dry mass of bulk spider samples from the same region (central Amazonia). The other three orders (Opiliones, Pseudoscorpiones, Ricinulei) for which regression analyses were possible were much more uniform in size and shape (Table 1). Power and linear models performed equally well and the coefficients are presented in Tables 3, 4. Masslength relationships of these orders and also the single specimens of Schizomida, Scorpiones, and Thelyphonida are presented in Figure 3. The masslength regressions for spiders collected in the Mata Atlantica (second data set) were also strongly correlated and highly significant, but coefficients were slightly different (Tables 3, 4). Only one subadult deinopid and a twiglike Argyrodes specimen lay outside the 95% confidence limits, but they did not influence the coefficients of the power model, which produced very good estimates of fresh and dry mass (99.5% of observed value) for the whole sample. The linear model in contrast produced a considerable underestimate of fresh and dry mass (70.2% resp. 73.4%). Ratio dry/fresh mass.Fresh mass and dry mass of spiders were strongly correlated ([R.sup.2] = 0.99, P < 0.001) in both data sets; the ratio dry/fresh mass was on average 0.293 [+ or ] 0.055 for Amazonian spiders and 0.208 [+ or ] 0.06 for spiders from the Atlantic forest. There was no significant difference in ratios for the two main hunting and webbuilding spider guilds (ttest P = 0.4). Anapids (tiny orb weavers) show the smallest ratio (0.24), oonopids and zodariids (small hunters, mostly strongly chitinized) the highest ratio (0.34) (Table 2). The highest variation of dry/fresh mass ratio occurred in the lowest range of body size, which is considered an effect of the decreasing precision of both length and weight measurements with decreasing size of the spiders. There was no correlation between length and the ratio dry/fresh mass. The ratio dry/fresh for opilionids was 0.44 [+ or ] 0.06 and for pseudoscorpions 0.38 [+ or ] 0.06. Both correlations are strong (R2 > 0.95) and highly significant (P < 0.01). Mean ratio dry/fresh for the three ricinuleid specimens was 0.53, and for the other arachnids between 0.30 and 0.39 (Table 2). General usefulness of equations.Regarding the statistics of masslength relationships, one certainly gets good estimates of biomass by length measurements for the Amazonian fauna using the coefficients from our equations. But how large would be the bias when using coefficients from other samples for our data or our coefficients for other data? When using coefficients derived from spider samples from temperate regions (taken from the literature) the estimate of the total biomass of our sample of 313 spiders produced serious biases from the observed mass: 56% (fresh) and 58% (dry mass) with coefficients from the linear model of Edwards & Gabriel (1998; spiders from Massachusetts, USA); 43% (dry mass) with coefficients from the power model of Breymeyer (1967; spiders from Europe); 25% (dry mass) using the coefficients from the power model of Henschel et al. (1996a; spiders from Germany); 23% (fresh mass) from the power model of Norberg (1978; spiders from spruce in Sweden). These strong biases are caused by the relatively high number of spiders with a length over 12 mm (e.g., Ctenidae) and some very large individuals (2436 mm) in our samples and the underestimation of these large spiders by formulae from temperate spider faunas (Fig. 2), which only represent spiders up to a length of 10 mm (Henschel et al. 1996a) or 8 mm (Norberg 1978). The equation of Rogers et al. (1977) from spiders (0.712 mm) collected from a shrubsteppe in southcentral Washington suited our data set better (105% of observed dry mass). To answer the question whether our equations are generally applicable to samples from spider assemblages in the Neotropics we tested our Amazonian equation on a spider sample (second data set) from another forest Brazilian ecosystem (Mata Atlantica) situated further south, geographically in the subtropics, and vice versa. When applying the Amazonian coefficients, the fresh biomass of the Atlantic Forest spiders was relatively well estimated (113% with power model, 110% with linear model), but the dry mass estimate was considerably overestimated (143% and 121 %). This is most probably caused by the lower ratio dry/fresh mass (0.21) for the spiders sampled in the Atlantic Forest in comparison with the spiders from Amazonia (0.29) (Table 2). When using the coefficients from the Mata Atlantica data set for the Amazonian data set the following biases (underestimation) resulted for fresh respectively dry mass: 84.5% / 66.4% by power, 62.6% / 52.4% by linear model. To obtain an idea of the effect of such biases we used one application data set from Amazonia. Fig. 4 shows box plots with means, medians and variances (percentiles) of spider biomass samples from different plantation systems, calculated with different coefficients. For most (5) systems the biomass of spiders per plot estimated with the equation from Henschel et al. (1996a) was higher than the biomass calculated with our own coefficients and showed comparable relations between medians and means and similar variance. This is due to overestimation of the dominant small spiders (< 4 mm) by the Henschel equation (s.a.). In each of the systems 4 and 6, however, one larger spider (8 mm) was sampled, and these are underestimated by the Henschel equation. In consequence, for these two systems the relative position of the means change depending on the equation used. However, due to the generally high variance of spider abundance between the replicates there are no significant differences between the systems, no matter if tested on means (ANOVA) or ranks (KruskallWallis) and by both equations. [FIGURE 2 OMITTED] [FIGURE 3 OMITTED] [FIGURE 4 OMITTED] We also applied the coefficients derived from the Amazonian data set in comparison with the coefficients derived from the Atlantic Forest data set to a second application data set: 30 litter samples taken in three different regeneration stages of an Atlantic submountain forest (Schmidt et al. in press). Mean dry mass values of spiders calculated by the Amazonian formula were 2.8, 5.3, 14.4 mg [m.sup.2] and calculated by the Atlantic forest formula 1.5, 2.9, 8.6 mg [m.sup.2]. Biomass values were significantly different (overestimated by the Amazonian formula, paired ttest P < 0.01), but ANOVA for the effects of the regeneration stage on biomass gave no significant effect. DISCUSSION Masslength regressions are a formidable solution for estimating biomass without having to destroy the specimens or handle them tediously on a microbalance, which is timeconsuming and expensive. Literature and our investigation show clearly that this can be made with one measurement of body length, which can be precisely taken with a micrometer eyepiece or a vernier caliper, even for live arthropods. In view of the very high determination coefficients and very low error probabilities, power regressions of length to estimate fresh or dry mass absolutely satisfy the needs, and no further effort is necessary to estimate volume by measurements of several body dimensions. A model should also not be overfitted (see below) since it would lose its applicability to new datasets. As mass is expected to be proportional to length cubed, in regression formulae the power (b) in a uniformly proportioned series of animals is expected to be close to 3. The fresh mass of spiders generally followed this relation, whereas dry mass of spiders and fresh and dry mass of opilionids increased with a power greater than 3. For pseudoscorpions, ricinuleids, and the oonopid spiders the power was less than 3. Schoener (1980) explained a power smaller than 3 for insects by a trend of longer species tending to be thinner. For our data set we suppose this to be due to different body densities (mass per volume), because all three groups represent more strongly chitinized rather than thinner animals in comparison to the other groups. When the aim is to estimate the biomass of bulk samples including many different spider species of different sizes and shapes, one formula can be used for all spiders, although a few very extraordinary shapes (e.g., very long and thin like some Argyrodes or Deinopis) may lie outside acceptable confidence limits. Especially for tropical soil fauna communities where most specimens are not readily identifiable, often not even to genus or family level, it is desirable, if not necessary, to have one regression equation covering the taxonomic level to which the organisms can be identified (sorted) easily, which most often is the order level for arthropods (Schoener 1980; Sample et al. 1993; Henschel et al. 1996b). Although not appearing very different, the coefficients given by other authors for estimation of spider biomass from length measurements when applied to our data produced slightly different values for single specimens, which result in considerable biases for bulk samples. The adequate precision of a single masslength regression depends upon the scientific question, and especially the variance included in the data set (e.g., how many different taxa with different body shapes were included and how strong the abundances vary in reality and in samples). As more masslength relations of different specimens/species are included, the coefficient of determination [R.sup.2] gets smaller, but unless it remains large enough to explain a considerable portion of the variation (> 0.8) and as long as the probability of being wrong in concluding that the coefficient is not zero remains small (P < 0.05), the regression model gains in predictability. In community ecology data sets, the variances in invertebrate abundance between different samples and study sites are usually high (standard deviation >100% of the mean) and thus precision of regression factors to calculate the biomass of groups of the community must not be very high, thus allowing relatively fast and rough measures. However, a systematic bias towards certain samples should be avoided. The comparison of the coefficients extracted from the two different models fitting our own data has already shown a possible cause for such a bias: a different proportion of very small or very large spiders in different samples treated with the same equation. In our tests, bias due to the "wrong" equation used for an estimation of biomass did not produce different ecological results. If no equation for the spider assemblage of interest is available, coefficients from an equation based on samples from other regions can be used if the size distributions do not differ strongly, which is obviously the case comparing spider assemblages from temperate and tropical regions. Attention must be given to individual, very large spiders in a sample, which in addition to its already problematic outlier position can produce a kingsize bias due to the power effect of the regression. But this should be resolved by statistical procedures in the ecological study. We have shown that it is difficult if not impossible to estimate biomass from different studies (regions) using the same equation and compare the absolute values. Even within the Neotropical rainforest realm, considerable bias can result from the estimation with nonautochthonous coefficients. We conclude from our results that our equations from the Amazonian sample are useful for biomass estimation of bulk arachnid samples from ecological studies in Amazonian rainforests and, with some restrictions, also for other neotropical forest spider assemblages. As these are often rich in species, which are represented by several developmental stages, it is valuable to have an idea of the distribution of size classes in the samples. If a wide range of sizes is represented, including spiders larger than 15 mm, the coefficients of the power functions should be used. If only smaller spiders were collected, which is often the case in soil or litter samples, the coefficients of the linear models would be more adequate or the equation resulting from the subsample of spiders < 2.5 mm should be used. We also present the coefficients for specific (abundant) taxa (ctenids, oonopids) and the guilds of hunting and webbuilding spiders, which can be used in studies of these specific groups. ACKNOWLEDGMENTS The sampling in Amazonia was conducted within the framework of the program SHIFT, the sampling in Parana within the framework of the program MATA ATLANTICA. Both BrazilianGerman research programs were funded by the German Federal Ministry for Education and Research (BMBF) and the Brazilian Council on Research and Technology (CNPq). We thank the Brazilian institutions EMBRAPA Amazonia Occidental (Manaus) and Federal University of Parana (UFPR) and the NGO Society for Wildlife Research and Environmental Education (SPVS) for the permission to use their sites and laboratories. We are very grateful to our friend Werner Hanagarth for assistance in sampling the arachnids in Amazonia and we thank Florian Raub and Ludger Scheuermann for their help in sampling the spiders in the Mata Atlantica. Manuscript received 25 February 2008, revised 5 November 2008. LITERATURE CITED Barros, E., J. Mathieu, S.C. TapiaCoral, A.R.L. Nascimento & P. Lavelle. 2006. Soil macrofauna communities in Brazilian Amazonia. Pp. 4355. In Soil Biodiversity in Amazonian and Other Brazilian Ecosystems. (F.M.S. Moreira, J.O. Siqueira & L. Brussaard, eds.). CAB International Publishing, Wallingford, UK. Barros, E., A. Neves, E. Blanchart, E.C.M. Fernandes, E. Wandelli & P. Lavelle. 2003. Development of the soil macrofauna community under silvopastoral and agrosilvicultural systems in Amazonia. Pedobiologia 47:273280. Bornebusch, C.H. 1930. The Fauna of Forest Soil. Nielsen and Lydiche, Copenhagen. 224 pp. Breymeyer, A. 1967. Correlations between dry weight of spiders and their length and fresh weight. Bulletin de l'Academie Polonaise des Sciences 15:263265. Brown, G.G., J. Rombke, H. Hofer, M. Verhaagh, K.D. Sautter & D.L.Q. Santana. 2006. Biodiversity and function of soil animals in Brazilian agroforestry systems. Pp. 217242. In Sistemas Agro florestais: Bases cienticias para o desenvolvimento sustentado. (A.C. GamaRodrigues, E.F. GamaRodrigues, M.S. Viana, M.S. Freitas, J.M. Marciano, J.M. Jasmin, N.F. Barros & J.G.A. Carneiro, eds.). UENF, Campos dos Goytacazes, Rio de Janeiro. Chen, B. & D.H. Wise. 1999. Bottomup limitation of predaceous arthropods in a detritusbased terrestrial food web. Ecology 80:761772. Clausen, I.H.S. 1983. Weightlength relations of eight species of spiders (Araneae) from Denmark. Entomologiske Meddelelser 50:139144. Edwards, C.A. 1966. Relationships between weights, volumes and numbers of soil animals. Colloquium on Dynamics of Soil Communities, 510 September 1966, BraunschweigVolkenrode. Friedr.Vieweg & Sohn GmbH, Braunschweig. Edwards, R.L. 1996. Estimating live spider weight using preserved specimens. J urnal f Arachn l gy 24:161166. Edwards, R.L. & W.L. Gabriel. 1998. Dry weight f fresh and preserved spiders (Araneida: Labid gnatha). Ent m l gical News 109:6674. Hanagarth, W., H. Hofer, C. Martius, M.V.B. Garcia & J. Rombke. 2004. Soil fauna densities and fluctuations in central Amazonian forests and polycultures as affected by the El Nino and La Nina events in the years 19971999. Environtropica 1:118. Henschel, J.R., D. Mahsberg & H. Stumpf. 1996a. Masslength relationships of spiders and harvestmen (Araneae and Opiliones). In Proceedings of the XIIIth International Congress of Arachnology, Geneve, September 1995. (V. Mahnert, ed.). Revue Suisse de Zoologie, vol. hors serie 1:265268. Henschel, J.R., H. Stumpf & D. Mahsberg. 1996b. Increase of arachnid abundance and biomass of water shores. In Proceedings of the XIIIth International Congress of Arachnology, Geneve, September 1995. (V. Mahnert, ed.). Revue Suisse de Zoologie, vol. hors serie 1:269278. Heifer, H. & A.D. Brescovit. 2001. Species and guild structure of a Neotropical spider assemblage (Araneae; Reserva Ducke, Amazo nas, Brazil). Andrias 15:99120. Hofer, H., W. Hanagarth, M.V.B. Garcia, C. Martius, E.N. Franklin, J. Rombke & L. Beck. 2001. Structure and function of soil fauna communities in Amazonian anthropogenic and natural ecosystems. European Journal of Soil Biology 37:229235. Jocque, R. 1981. Size and weight variations in spiders and their ecological significance. Biologish Jaarboek 49:155165. Lang, A., S. Krooss & H. Stumpf. 1997. Masslength relationships of epigeal arthropod predators in arable land (Araneae, Chilopoda, Coleoptera). Pedobiologia 41:327333. Lavelle, P., E. Barros, E. Blanchart, G.G. Brown, T. Desjardins, L. Mariani & J.P. Rossi. 2001. SOM management in the tropics: Why feeding the soil macrofauna? Pp. 5362. In Managing Organic Matter in Tropical Soils: Scope and Limitations. (C. Martius, H. Tiessen & P.L.G. Vlek, eds.). Kluwer Academic Publishers, Dordrecht, Netherlands. Lavelle, P., D. Bignell, M. Lepage, V. Wolters, P. Roger, P. Ineson, O.W. Heal & S. Dhillion. 1997. Soil function in a changing world: the role of invertebrate ecosystem engineers. European Journal of Soil Biology 33:159193. Lawrence, K.L. & D.H. Wise. 2000. Spider predation on forestfloor Collembola and evidence for indirect effects on decomposition. Pedobiologia 44:3339. Lawrence, K.L. & D.H. Wise. 2004. Unexpected indirect effect of spiders on the rate of litter disappearance in a deciduous forest. Pedobiologia 48:149157. Martius, C., H. Hofer, M.V.B. Garcia, J. Rombke & W. Hanagarth. 2004. Litter fall, litter stocks and decomposition rates in rain forest and agroforestry sites in central Amazonia. Nutrient Cycling in Agroecosystems 68:137154. Mathieu, J., J.P. Rossi, M. Grimaldi, P. Mora, P. Lavelle & C. Rouland. 2004. A multiscale study of soil macrofauna biodiversity in Amazonian pastures. Biology and Fertility of Soils 40:300305. Norberg, R.A. 1978. Energy content of some spiders and insects on branches of spruce (Picea abies) in winter; prey of certain passerine birds. Oikos 31:222229. Rogers, L.E., R.L. Buschbom & C.R. Watson. 1977. Lengthweight relationships of shrubsteppe invertebrates. Annals of the Entomological Society of America 70:5153. Rogers, L.E., W.T. Hinds & R.L. Buschbom. 1976. A general weight vs. length relationship for insects. Annals of the Entomological Society of America 69:387389. Sample, B.E., R.J. Cooper, R.D. Greer & R.C. Whitmore. 1993. Estimation of insect biomass by length and width. American Midland Naturalist 129:234241. Schmidt, P., K. Dickow, A.A. Rocha, R. Marques, L. Scheuermann, J. Rombke, B. Forster & H. Hofer. 2008. Soil macrofauna and decomposition rates in Southern Brazilian Atlantic rainforests. Ecotropica 14:89100. Schoener, T.W. 1980. Lengthweight regressions in tropical and temperate forestunderstory insects. Annals of the Entomological Society of America 73:106109. SPSS. 2002. SigmaPlot[R] 8.0.2. SPSS Inc, Chicago. StatSoft. 2005. Statistica 7.1. StatSoft, Inc, Tulsa, Oklahoma. Wise, D.H. 2004. Wandering spiders limit densities of a major microbidetritivore in the forestfloor food web. Pedobiologia 48:181188. Wise, D.H., W.E. Snyder, P. Tuntibunpakul & J. Halaj. 1999. Spiders in decomposition food webs of agroecosystems: theory and evidence. Journal of Arachnology 27:363370. Hubert Heifer: Department of Zoology, Staatliches Museum fur Naturkunde Karlsruhe, Erbprinzenstrasse 13, D76133 Karlsruhe, Germany. Email: hubert.hoefer@smnk.de Ricardo Ott: Museu de Ciencias Naturais, Fundacao Zoobotanica do Rio Grande do Sul, Porto Alegre, Brazil Table 1.Number of specimens measured and weighed for lengthmass regression, mean and range of body length (minimum and maximum in brackets) from seven arachnid orders. First data set (Amazonia) Order/Infraorder/Family Specimens Length (mm) Araneae 313 4.83 (0.5636.0) Infraorder Mygalomorphae 43 3.17 (0.7819.1) Infraorder Araneomorphae: Amaurobiidae 1 Anapidae 1 1.07 Anyphaenidae Araneidae 8 1.89 (0.813.40) Corinnidae 11 5.85 (1.8513.9) Ctenidae 74 12.43 (1.3036.0) Deinopidae Linyphiidae 9 1.60 (1.201.90) Lycosidae Mysmenidae Ochyroceratidae 24 1.40 (0.562.40) Oecobiidae 2 1.75 (1.701.80) Oonopidae 68 1.46 (0.672.50) Palpimanidae 4 3.04 (1.524.00) Pholcidae 8 2.00 (1.074.30) Pisauridae 2 4.67 (3.965.40) Salticidae 39 3.40 (1.126.60) Scytodidae 5 2.57 (1.603.10) Selenopidae Sparassidae 3 6.00 (5.906.10) Tetragnathidae Theridiidae 5 1.33 (1.002.00) Theridiosomatidae 3 0.75 (0.620.83) Thomisidae Trechaleidae Uloboridae Zodariidae 4 3.60 (2.004.50) Zoridae Opiliones 65 2.12 (0.576.90) Pseudoscorpiones 111 1.38 (0.862.10) Ricinulei 10 4.46 (2.105.60) Schizomida 3 1.62 (1.451.88) Scorpiones 2 16.30 (3.6029.0) Thelyphonida 1 7.00 Second data set (Mata Atlantica) Order/Infraorder/Family Specimens Length (mm) Araneae 99 7.08 (1.3528.0) Infraorder Mygalomorphae Infraorder Araneomorphae: Amaurobiidae 1 8.27 Anapidae Anyphaenidae 2 6.87 (6.836.92) Araneidae 18 5.77 (2.6910.67) Corinnidae 2 4.57 (4.524.62) Ctenidae 18 16.52 (4.2328.0) Deinopidae 1 16.50 Linyphiidae 1 2.30 Lycosidae 2 16.85 (7.6926.0) Mysmenidae 3 1.73 (1.352.40) Ochyroceratidae 1 1.83 Oecobiidae Oonopidae 1 2.31 Palpimanidae Pholcidae 6 2.79 (1.923.94) Pisauridae 1 4.61 Salticidae 4 4.86 (3.655.77) Scytodidae Selenopidae 1 5.00 Sparassidae 2 5.58 (3.567.60) Tetragnathidae 2 5.86 (4.337.40) Theridiidae 20 3.02 (1.6310.0) Theridiosomatidae 3 1.91 (1.492.69) Thomisidae 1 7.60 Trechaleidae 3 12.53 (5.2925.0) Uloboridae 1 5.38 Zodariidae Zoridae 4 4.23 (3.854.81) Opiliones Pseudoscorpiones Ricinulei Schizomida Scorpiones Thelyphonida Table 2.Ratios dry/fresh mass for arachnid orders. Order Ratio dry/fresh mass First data set Second data set Family Guild (Amazonia) (Mata Atlantica) Araneae 0.29 0.21 Mygalomorphae hunting 0.29 Araneomorphae Anyphaenidae hunting 0.25 Amaurobiidae hunting 0.12 Corinnidae hunting 0.29 0.27 Ctenidae hunting 0.26 0.19 Lycosidae hunting 0.19 Oonopidae hunting 0.34 0.19 Oxyopidae hunting 0.24 Palpimanidae hunting 0.32 Pisauridae hunting 0.28 0.22 Salticidae hunting 0.28 0.21 Scytodidae hunters 0.29 Selenopidae hunting 0.16 Sparassidae hunting 0.26 0.19 Thomisidae hunting 0.18 Trechaleidae hunting 0.20 Zodariidae hunting 0.34 Zoridae hunting 0.20 Anapidae webbuilding 0.24 Araneidae webbuilding 0.25 0.22 Deinopidae webbuilding 0.16 Linyphiidae webbuilding 0.33 0.19 Mysmenidae webbuilding 0.20 Ochyroceratidae webbuilding 0.31 0.19 Oecobiidae webbuilding 0.29 Pholcidae webbuilding 0.27 0.20 Tetragnathidae webbuilding 0.29 Theridiidae webbuilding 0.28 0.21 Theridiosomatidae webbuilding 0.29 0.18 Uloboridae webbuilding 0.18 Opiliones 0.41 Pseudoscorpiones 0.38 Ricinulei 0.53 Schizomida 0.37 Scorpiones 0.30 Thelyphonida 0.39 Table 3.Regression coefficients (a, b) and coefficient of determination in regressions of fresh mass to body length (left: power model: mass [mg] = a body length [mm] (b), right: linear model: In mass [mg] = a + ln body length [mm] b) for arachnids from Amazonia (first data set) and Mata Atlantica (second data set) (n = sample size, se = standard error, [R.sup.2] = coefficient of determination). All regressions are highly significant (P < 0.001). Power model n a [+ or ] se b [+ or ] se Mata Atlantica: 99 0.066 [+ or ] 0.025 3.160 [+ or ] 0.118 all Araneae Amazonia: all 313 0.169 [+ or ] 0.009 2.899 [+ or ] 0.016 Araneae Araneae < 2.5 mm 225 0.085 [+ or ] 0.010 3.288 [+ or ] 0.081 Ctenidae 74 0.177 [+ or ] 0.020 2.886 [+ or ] 0.034 Oonopidae 68 0.131 [+ or ] 0.007 2.682 [+ or ] 0.076 Hunting spiders 253 0.169 [+ or ] 0.010 2.899 [+ or ] 0.018 Webbuilding 60 0.072 [+ or ] 0.011 3.710 [+ or ] 0.114 Opiliones 65 0.147 [+ or ] 0.028 3.622 [+ or ] 0.105 Pseudoscorpiones 111 0.156 [+ or ] 0.006 2.453 [+ or ] 0.071 Ricinulei 10 0.225 [+ or ] 0.146 2.760 [+ or ] 0.387 Power model Linear model [R.sup.2] a [+ or ] se Mata Atlantica: 0.98  2.166 [+ or ] 0.175 all Araneae Amazonia: all 0.99  2.058 [+ or ] 0.029 Araneae Araneae < 2.5 mm 0.94  1.958 [+ or ] 0.037 Ctenidae 0.99  1.758 [+ or ] 0.096 Oonopidae 0.94  2.039 [+ or ] 0.042 Hunting spiders 0.99  2.108 [+ or ] 0.023 Webbuilding 0.97  1.784 [+ or ] 0.092 Opiliones 0.98  0.899 [+ or ] 0.048 Pseudoscorpiones 0.92  1.892 [+ or ] 0.027 Ricinulei 0.93  1.907 [+ or ] 0.192 Linear model b [+ or ] se [R.sup.2] Mata Atlantica: 2.872 [+ or ] 0.097 0.90 all Araneae Amazonia: all 2.980 [+ or ] 0.020 0.99 Araneae Araneae < 2.5 mm 2.746 [+ or ] 0.053 0.92 Ctenidae 2.894 [+ or ] 0.039 0.99 Oonopidae 2.666 [+ or ] 0.099 0.96 Hunting spiders 3.017 [+ or ] 0.015 0.99 Webbuilding 2.255 [+ or ] 0.169 0.75 Opiliones 2.984 [+ or ] 0.060 0.97 Pseudoscorpiones 2.515 [+ or ] 0.073 0.91 Ricinulei 3.014 [+ or ] 0.130 0.98 Table 4.Regression coefficients (a, b) and coefficient of determination in regressions of dry mass to body length (left: power model: mass [mg] = a body length [mm] (b), right: linear model: In mass [mg] = a + In body length [mm] b) for arachnids from Amazonia (first data set) and Mata Atlantica (second data set) (n = sample size, se = standard error, [R.sup.2] = coefficient of determination). All regressions are highly significant (P < 0.001). Power model n Mata Atlantica: all 99 0.0067 [+ or ] 0.005 Araneae Amazonia: all 313 0.0165 [+ or ] 0.001 Araneae Araneae < 2.5 mm 225 0.028 [+ or ] 0.003 Ctenidae 74 0.017 [+ or ] 0.002 Oonopidae 68 0.050 [+ or ] 0.003 Hunting spiders 253 0.0165 [+ or ] 0.001 Webbuilding 60 0.017 [+ or ] 0003 Opiliones 65 0.042 [+ or ] 0.009 Pseudoscorpiones 111 0.057 [+ or ] 0.003 Power model b [+ or ] se [R.sup.2] Mata Atlantica: all 3.413 [+ or ] 0.245 0.96 Araneae Amazonia: all 3.242 [+ or ] 0.014 0.99 Araneae Araneae < 2.5 mm 3.180 [+ or ] 0.079 0.94 Ctenidae 3.232 [+ or ] 0.029 0.99 Oonopidae 2.459 [+ or ] 0.094 0.90 Hunting spiders 3.242 [+ or ] 0.016 0.99 Webbuilding 3.881 [+ or ] 0.123 0.97 Opiliones 3.879 [+ or ] 0.119 0.98 Pseudoscorpiones 2.589 [+ or ] 0.103 0.86 Linear model a [+ or ] se b [+ or ] se Mata Atlantica: all  3.860 [+ or ] 0.224 2.950 [+ or ] 0.092 Araneae Amazonia: all  3.213 [+ or ] 0.029 2.902 [+ or ] 0.021 Araneae Araneae < 2.5 mm  3.121 [+ or ] 0.038 2.680 [+ or ] 0.054 Ctenidae  3.197 [+ or ] 0.096 2.921 [+ or ] 0.039 Oonopidae  3.162 [+ or ] 0.046 2.767 [+ or ] 0.108 Hunting spiders  3.237 [+ or ] 0.025 2.926 [+ or ] 0.016 Webbuilding  2.997 [+ or ] 0.093 2.199 [+ or ] 0.172 Opiliones  1.862 [+ or ] 0.049 3.069 [+ or ] 0.062 Pseudoscorpiones  2.967 [+ or ] 0.037 2.771 [+ or ] 0.100 Linear model [R.sup.2] Mata Atlantica: all 0.93 Araneae Amazonia: all 0.98 Araneae Araneae < 2.5 mm 0.92 Ctenidae 0.99 Oonopidae 0.95 Hunting spiders 0.99 Webbuilding 0.74 Opiliones 0.97 Pseudoscorpiones 0.87 
Gale Copyright:  Copyright 2009 Gale, Cengage Learning. All rights reserved. 