- Top of page
Over the past 40 years, suction samplers have become an established method for the collection of above-ground grassland invertebrates in low-lying vegetation, such as grasslands and agroecosystems (Morris & Rispin 1987; Gibson, Hambler & Brown 1992; Haughton et al. 2003; Otway, Hector & Lawton 2005; Woodcock et al. 2007, 2008). Suction samplers use a petrol engine to drive the rapid intake of air through a wide bore sampling tube. When placed over vegetation, insects are sucked into a collecting chamber from which they can be subsequently removed and identified. As the sampling area is equivalent to the aperture of the sampling tube, suction sampling has been treated as a quantitative method. Indeed, this has been commonly stated as its main advantage over pitfall trapping where the samples are biased both by individual species’ activity and by the pitfall traps which do not collect from a defined area (Southwood & Henderson 2000). Pitfall traps are also poorly suited to the collection of invertebrates from plant structures, and are better for collection from the soil surface (Mommertz et al. 1996; Borges & Brown 2003).
Given the current popularity of the suction sampling method, there is a pressing need for experimental verification of its efficiency. Much of the research to date has focused on correlative analyses of existing data sets, normally of a restricted number of invertebrate taxa. To address this paucity of information, we consider the effectiveness of suction sampling for the collection of beetles (Coleoptera), true bugs (Hemiptera: Heteroptera), planthoppers (Hemiptera: Auchenorrhyncha) and spiders (Araneae). The choice of taxa represents the dominant taxonomic groups cited in the literature as being collected using this sampling method. These groups also represent a major proportion of the total abundance of invertebrates active within the vegetation sward. Although, other taxa are present within low lying vegetation, they are not considered here as they are either infrequently collected by suction sampling or are too fragile or small to be effectively sampled, (e.g. Collembola; Southwood & Henderson 2000). In the present study, we aim to address the following questions: (i) What is the effect of sampling effort on suction sampler efficiency; specifically, how does increasing the duration or the numbers of sub-samples increase the accumulation of individuals and species? (ii) Does increasing sward height decrease capture efficiency? (iii) Does suction sampling collect invertebrate assemblages that are comparable to a more absolute sampling method?
- Top of page
This study was undertaken on two grasslands: Bradenham Estate (051°40′09″ N, 000°48′07″ W), a managed species-rich chalk grassland of c. 0·7 ha, and a mesotrophic hay meadow of c. 1·0 ha on the University of Reading campus (051°26′19″ N, 000°56′15″ W).
Suction sampling was undertaken with a commercially available Vortis suction sampler (Burkard Manufacturing Co. Ltd, UK) whose use has been cited extensively within the literature (e.g. Haughton et al. 2003; Otway, Hector & Lawton 2005; Woodcock et al. 2007, 2008). As suction samplers operate on similar principles, this study will be relevant to other models. With the throttle set at full, each individual sample is collected through a tube with a diameter 15·7 cm. All sampling was undertaken on dry days between 10·00 h and 16·00 h. The sampled groups represent four of the most numerically dominant grassland invertebrates (Duffey et al. 1974; Tscharntke & Greiler 1995).
experiment 1: sampling effort
As the ability of different species to resist the vacuum varies, increasing the duration of suction samples is predicted to increase the chance of collecting all individuals. We assess the effect of the duration of individual suction samples on the efficiency in collecting both individuals and species of the beetles, true bugs, planthoppers and spiders within the mesotrophic hay meadow. Seven time periods (in seconds) were tested: 1, 5, 10, 15, 20, 30 and 45. Ten replicates of each time period were conducted at randomly chosen locations.
Conventionally, an individual experimental unit will be sampled multiple times on one sampling date (e.g. Haughton et al. 2003; Otway, Hector & Lawton 2005; Woodcock et al. 2007, 2008). This sub-sampling approach is needed because many grassland invertebrates have aggregated distributions, and sampling should be designed to reflect this (Dennis, Aspinall & Gordon 2002). Species-accumulation curves have been used to account for insufficient sampling effort (Colwell & Coddington 1994) and they remain useful to assess sub-sampling levels that best assess invertebrate populations. In the mesotrophic hay meadow, six sub-sampling levels were tested: 1, 5, 10, 15, 20 and 30 sub-samples. Each of these was replicated 12 times at randomly chosen points and held over a fixed location for a duration of 10 s, reflecting normal procedure (e.g. Duffey 1980; Haughton et al. 2003; Woodcock et al. 2007, 2008).
experiment 2: sward height
The impedance of suction sampler airflow in dense or tall vegetation could reduce capture efficiency (Sunderland & Topping 1995; Hossain, Gurr & Wratten 1999). To account for this problem, we used small plastic beads, similar in size to many small grassland invertebrates (0·024 g; diameter, 2·5 mm; Rogers, Hinds & Buschbom 1976). The beads characterize these effects without being confounded by species variation in morphology and life history. In both grasslands, 1 g of beads were scattered within a tube equivalent in diameter to the suction sampler nozzle. The suction sampler was lowered into the tube and the proportional recapture of the beads was recorded for a 10-s suction period. The sward height adjoining the sample position was measured using a drop disc (Stewart, Bourn & Thomas 2001). At both sites, 30 sample positions were selected from individual areas showing high levels of variation in sward height.
experiment 3: comparison of suction and absolute sampling methods
Turf removal is a destructive method of sampling grassland invertebrates (Morris & Rispin 1987). Individual turfs were sampled using a rigid plastic tube, the same diameter as the suction sampler nozzle (15·7 cm), attached to a large plastic bag. The tube was placed over vegetation, then the turf was cut around the circumference to a depth of 5 cm and the contents emptied into the bag. This approach provides the most accurate measure of the density of above-ground invertebrates within a defined area. By comparing suction sampling with turf removal, it was possible to assess accuracy of the suction sampler. Individual turf samples were paired with single suction samples taken for a duration of 10 s at randomly chosen locations within the chalk grassland site.
For experiment 1, abundance and species accumulation curves in response to sampling effort (time and number of sub-samples) were fitted using a linear dependence model (Soberón & Llorente 1993; Moreno & Halffter 2000). This model assumes that the accumulation of either individuals or species decreases linearly with increased sampling effort. For both abundance and species accumulation curves, the linear dependency model was defined as:
- X(n) = a/b[1 − exp(−bn)],
where n is a measure of sampling effort, either in terms of duration of each suction sample, or the number of sub-samples; X(n) is the predicted number of either individuals or species at sampling effort n; a is the rate of increase in either individuals or species at the start of sampling; b is species accumulation; and a/b represents the asymptotic number of species or individuals. Parameter estimation of these models was undertaken in sas 9·01 using non-linear regression (PROC NLIN). Pseudo-R2 was used to assess model fit (Schabenberger & Pierce 2002). Both abundance and species accumulation curves were fitted for the duration of suction sampling. Only species accumulation curves were fitted where sampling effort was defined by the number of sub-samples. The sampling effort required to collect a defined proportion q of the invertebrate fauna given by a/b was also assessed by calculating (nq):
where q is the desired proportion of the total invertebrate fauna to be sampled. As the linear dependence model would require infinite sampling effort to collect all individuals, the sampling effort required to collect 90% of all individuals or species was assessed (Moreno & Halffter 2000).
For experiment 2, the proportional recapture of beads in response to sward height was evaluated using generalized linear models (GLM), with binomial error structures and logit link function (Schabenberger & Pierce 2002). The proportion of recaptured beads was evaluated with the explanatory variables of sward height, site and the interaction term of sward height × site. Model simplification was by deletion of least significant terms or covariates from this fully saturated model. All GLM analyses were carried in sas 9·01.
For experiment 3, redundancy analysis (RDA) in canoco 4·5, was used to determine whether the structure of invertebrate assemblages differed significantly for the two sampling methods. The choice of RDA reflected the relatively short gradient lengths of the four invertebrate taxa (< 3 SD; ter Braak & Šmilauer 2002). Abundance values were Log10 N + 1-transformed and covariables coding for the paired structure of the experimental design were included as a blocking factor. Sampling method was coded for by two nominal environmental variables and tested using Monte Carlo permutations of both canonical axes (1000 permutations).
- Top of page
Summed over all three experiments, spiders were the most numerous taxa [abundance (N) = 4806, species richness (SR) = 26], followed by planthoppers (N = 3730; SR = 18), beetles (N = 2857, SR = 102) and true bugs (N = 235, SR = 18). Mean sward height of the mesotrophic and chalk grasslands was respectively 78·4 cm ± (SE ± 4·86) and 59·6 cm (SE ± 3·07), and at both sites floral species richness was relatively high with c. 12 species m−2 or more.
experiment 1: sampling effort
The duration of sampling was significantly related to the accumulation of species and individuals for all four taxa (Table 1, Fig. 1). The variance explained by these models (pseudo-R2), however, showed a high degree of variability (Table 1). For abundance and species accumulation curves, the linear dependence models explained > 70% of the data variance for the beetles, planthoppers and spiders but was considerably lower (> 50%) for the true bugs. The sampling effort required to capture 90% of individuals was less than 2 s for the true bugs, planthoppers and spiders, although for the beetles, this time was 15·6 s. For all four taxa, however, the model predicted a sampling duration of 3 s to be sufficient to capture 90% of the species. Differences between sampling duration and the number of sub-samples to collect 90% of the species are related to scale.
Table 1. Parameter estimates and model fit parameters for the individual or species accumulation curves (based on linear dependence models) used to assess the effect of duration of suction sampling (n). Individual and species accumulation curves are assessed for four dominant grassland invertebrate taxa: the beetles, true bugs, planthoppers and spiders. Pseudo-R2 is provided as a measure of model fit. Sampling effort (duration of suction samples) that would be required to collect 90% of either the individuals or species is given, where a is the slope at the initiation of sampling; b, the parameter defining the rate of species accumulation; n, the sampling effort; X(n), the predicted number of individuals or species collected for sampling effort n; and ***, significant P < 0·001
|Taxa||Linear dependence model X(n) = a/b[1 − exp(−bn)]||90 % sampling effort (seconds)|
| Beetles|| 4·14||0·14||0·84||181·0***||15·6|
| True bugs|| 4·21||1·93||0·36||39·4***||1·1|
| True bugs||2·01||2·10||0·45||57·1***||1·1|
Figure 1. The effect of sampling effort (duration of suction samples) on the accumulation of either individuals or species when suction- sampling for the beetles (a and b), true bugs (c and d), planthoppers (e and f) and spiders (g and h). Fitted accumulation curves are based on linear dependence models of the form: X(n) = a/b[1 − exp(−bn)], where X(n) is the predicted number of species collected for sampling effort n; and a and b are fitted parameters for each model.
Download figure to PowerPoint
The number of sub-samples was significantly related to the accumulation of species and individuals for all four taxa (Table 2, Fig. 2). In each case, species accumulation curves explained at least 70% of the variance in the species data (Table 2). There was, however, considerable between-taxa variation in the number of suction samples required to collect 90% of the species (Table 2, Fig. 2). Both the planthoppers and spiders required the collection of c. 16 samples (Table 2), whereas over three times greater sampling effort was needed to collect 90% of the beetle species at 54·8 sub-samples. The true bugs required the greatest degree of sampling effort to collect 90% of the species at 109·6 sub-samples.
Table 2. Parameter estimates and model fit parameters for the species accumulation curves (based on linear dependence models) used to assess the effect of number of suction sub-samples (n). Species accumulation curves are assessed for four dominant grassland invertebrate taxa: the beetles, true bugs, planthoppers and spiders. Pseudo-R2 is provided as a measure of model fit. Sampling effort (defined by the number of suction samples) that would be required to collect 90% of the species is given where a is the slope at the initiation of sampling; b, the parameter defining rate of species accumulation; n, the sampling effort; X(n), the predicted number of species collected for sampling effort n; and ***, significant P < 0·001
|Taxa||Linear dependency model X(n) = a/b[1 − exp(−bn)]||90% sampling effort (no. of sub-samples)|
|True bugs||0·085||0·021||0·71|| 87·6***||109·6|
Figure 2. The effect of sampling effort (number of suction sub-samples) on the accumulation of species when suction sampling for the beetles (a), true bugs (b), planthoppers (c) and spiders (d). Fitted accumulation curves are based on linear dependence models of the form: X(n) = a/b[1 − exp(−bn)], where X(n) is the predicted number of species collected for sampling effort n; and a and b are fitted parameters for each model.
Download figure to PowerPoint
experiment 2: sward height
Sward height varied between the two sites, with the mesotrophic grassland showing the greatest range from 35 to 128 cm, compared to the generally shorter chalk grassland (range = 24–105 cm). The proportional recapture of beads (used to model effectiveness of suction sampling) was negatively correlated with sward height (F1,58 = 29·9, P < 0·001; Fig. 3). There was no significant influence of either site (F1,57 = 1·20, P > 0·05) or the interaction between sward height and site (F1,56 = 2·60, P > 0·05) on bead recapture.
Figure 3. The relationship between sward height and the proportional rates of recapture of a known number of plastic beads (surrogate invertebrates) from two lowland hay meadows. Trend line is back transformed from the original binomial model where y = 1/1 + (1/exp(a) * exp(bx)), and a = 1·12, b = −0·02.
Download figure to PowerPoint
experiment 3: comparison of suction and absolute sampling methods
Compared with the previous two experiments, a much lower number of individuals were collected during the comparison of the turf removal and suction sampling methods (spiders: N = 523, SR = 14; beetle: N = 360, SR = 36; planthoppers N = 287, SR = 12; true bugs: N = 77, SR = 11). RDA revealed no significant difference in assemblage structure between the turf removal and suction sampling methods for beetle, spider or planthopper communities (Table 3). In the case of the true bugs, however, assemblage structure differed between the two sampling methods (Table 3), accounting for 11·8% of the variance in the species data.
Table 3. Redundancy analysis comparing assemblage similarity of the beetles, spiders, true bugs and planthoppers collected by two sampling methods, the suction sampler and a destructive absolute sampling method based on the complete removal of turfs and accompanying vegetation. F ratio and P value obtained by Monte Carlo permutation tests (1000 permutations), where % expl. 1st axis is the percentage species variability explained by axis 1; and r 1st axis is the species environment correlation on axis 1
|Taxa||% expl. 1st axis||r 1st axis||F||P|
- Top of page
This study demonstrates that the duration of suction sampling required to collect both individuals and species was related to life-history characteristics of the four taxa considered. This was illustrated by the beetles, which in contrast to the true bugs, planthoppers and spiders, were characterized by many species that were either epigeal or dwelling within tussock-forming vegetation (e.g. c. 33% of species were predominantly epigeal ground and rove beetles; Luff 1966; Dennis, Young & Gordon 1998). Compression of vegetation by the suction sampler is likely to impair the efficiency of capturing near-ground-dwelling species (Hossain, Gurr & Wratten 1999; Southwood & Henderson 2000). As a result, the duration of suction sampling required to collect 90% of beetle individuals was over 10 times higher (16 s) than that of the true bugs, planthoppers or spiders. In the case of the true bugs, planthoppers and spiders, as well as the majority of phytophagous beetles (e.g. weevils or leaf beetles), a much greater proportion of species (> 70%) persist within the vegetation itself (Naumann 1994). In contrast, species accumulation of beetles, true bugs, planthoppers and spiders was rapid in all cases, with 90% of the species from all four taxa collected at the sampling point within 3 s of suction sampling. Accordingly, we propose that the sampling effort of each suction sample is fixed at c. 15 s duration when sampling beetles. Shorter sampling periods can be used when targeting other taxa; however, it is worth noting that many spiders and true bugs are found on the soil surface, and for local assemblages that are dominated by such species, sampling times should be increased appropriately (e.g. to 10 s). While an increase in sward height would be likely to raise the time threshold (Hossain, Gurr & Wratten 1999), this experiment was performed in a sward both tall (> 75 cm) and structurally complex, and therefore close to the upper limit of vegetation heights suitable for suction sampling. As such, our threshold is likely to be a conservative figure.
There was considerable variation between taxa in the number of sub-samples required to collect 90% of the species within the grassland, ranging from 16 samples for the spiders and planthoppers, up to 110 for the true bugs. As the Vortis suction sampler has a relatively small sampling aperture (< 16 cm), the sampling efforts reported are applicable to other models with larger apertures.
The extent to which populations of insects are aggregated is likely to show both inter- and intra-specific variation between habitats (Dennis, Aspinall & Gordon 2002; Crist, Pradhan-Devare & Summerville 2006). For this reason, the sub-sampling levels identified here should be used as a baseline. Where insufficient sampling effort has been used, rarefaction techniques could well prove useful in providing adjusted measures of species richness (Colwell & Coddington 1994; Standen 2000). These curves are limited in their utility, for example, where comparisons between experimental units consider assemblage structure, rather than just species richness. In such cases, the level of sampling effort should be carefully considered.
The effect of sward height on the efficiency of suction samplers is perhaps the most well-recognized of the problems associated with their use (e.g. Hossain, Gurr & Wratten 1999; Standen 2000; Southwood & Henderson 2000), and this is corroborated here. The recapture of beads (surrogate invertebrates) showed a negative correlation with sward height. This represents a problem where suction samplers are used to compare experimental treatments that differ greatly in sward height. This is particularly relevant to many experimental treatments that impact on sward height such as cutting, grazing or fertilization (Woodcock et al. 2007). The inclusion of sward height as a covariate is an obvious and simple way of controlling for the effects of this confounding variable in analyses (Schabenberger & Pierce 2002).
The use of beads as a model to assess recapture rates of invertebrates, whilst being independent of species specific behavioural effects was, however, likely to be a conservative estimate of the effect of sward height on capture efficiency. The beads falling in most cases to the soil surface would have experienced the maximum reduction of recapture rates due to interference by above-ground vegetation. As many grassland invertebrates inhabit the vegetation rather than the soil surface, the reduction in suction sampler efficiency for the beads may be something of an overestimate. This is supported by the species accumulation curves described in experiment 1 where 90% of individuals were collected in under 15 s for all taxa from tall, structurally complex grasslands.
For the beetles, planthoppers and spiders, there were no differences in the assemblage structure revealed by the turf removal and suction sampling techniques. Only in the case of the true bugs did suction sampling prove to be an inadequate sampling method, although this may be biased by the low abundances of the true bugs collected in this study. It should be noted that this comparison was within a single grassland site, and other types of grassland may not have resulted in such a favourable comparison.
Suction sampling is an effective quantitative tool for the measurement of invertebrate diversity and assemblage structure providing sward height is included as a covariate. The effective sampling of the beetles, true bugs, planthoppers and spiders altogether would require a minimum sampling effort of 110 sub-samples of duration of 16 s. Such sampling intensities could be adjusted depending on the taxa sampled (e.g. excluding true bugs reduces the number of sub-samples to c. 55). The inadequate sampling of true bug species composition raises questions as to the suitability of suction sampling for this taxon in this instance. Taking such precautions into account, suction sampling should remain an important component in the toolbox of experimental techniques used during the sampling of invertebrate communities.