Northern Prairie Wildlife Research Center
SATs and ATs were of similar size (Figure 1a,b) within constraints imposed by shape differences between the two trap types. Funnel aperture sizes for ATs were 176.72 and 6.61 cm² (outside and inside, respectively) compared to 387.10 (154.84 cm² above, 232.26 cm² below the water surface) and 32.26 cm² (12.90 cm² above, 19.36 cm² below the water surface) for SATs. Smaller inner funnel openings of ATs may have restricted entrance of some invertebrates, thus influencing trap performance. However, constructing both trap types with funnel openings of identical area was impractical because resulting inside openings of SATs would have been so narrow as to exclude many macroinvertebrates. We constructed SATs from 0.48-cm-thick transparent plexiglass. All pieces were professionally laser-cut to specified dimensions. Traps were assembled at NPWRC using methylene chloride. Assembly time was approximately 15 minutes per trap. We adjoined SAT strata using small office binder clips to allow disassembly during sample collection; alternatively, strata could be permanently joined using methylene chloride. We fitted the rear portion of each SAT stratum with a removable insert so trap contents could be removed easily (Figure 1b); inserts were held in place with rubber bands but could be installed permanently. Cost of material for each SAT was approximately $15 (U.S.); this contrasts with about $5 (U.S.) for each AT constructed after the general design of Ross and Murkin (1989).
|Figure 1. Design of (a) conventional activity trap (AT) and (b) surface activity trap (SAT) for collecting aquatic invertebrates (drawings are not to scale): 1. 1-L glass jar; 2., 9. hardware cloth fish screen; 3. hose clamp; 4. clip; 5. rubber band; 6. hanger wire; 7. transparent powder funnel, polymethylpentene; 8. eye screw; 10. wing nut; 11. binder clip; and 12. removable PVC back.|
To satisfy objectives of a concurrent study, fathead minnows (Pimephales promelas [Rafinseque]) were added to 10 randomly chosen study wetlands each year (10 wetlands remained fishless). To exclude fish and limit depredation of invertebrates within ATs and SATs (Murkin et al. 1983, Elmberg et al. 1992), we placed 1.4-cm-diameter wire mesh over inside openings of all funnels (ATs and SATs), leaving a 1-cm gap between inside funnel openings and mesh (Figure 1). Most invertebrate taxa were still capable of entering traps (including adult Dytiscidae, Zygoptera, and Lethocerus [Belostomatidae]); fish were captured only rarely.
Invertebrates were sampled 3 times (periods 1-3) each year between 10 June and 15 July using horizontally-positioned ATs and SATs (Figure 1a,b). Both trap types were deployed concurrently for 24-hr exposure periods. In contrast to conventional ATs, SATs were comprised of three compartments and thus were designed to sample discrete vertical zones from 10.16 cm above to 15.24 cm below the water surface (Figure 1b). All trap contents were condensed by passing samples through a 0.4mm funnel and preserved with 70% ethanol. In the lab, invertebrates were identified to the lowest feasible taxonomic group using Merritt and Cummins (1984) and Pennak (1989). Taxonomic resolution varied, but identification was usually made to family or genus.
Matched pairs of traps were deployed horizontally from PVC frames fastened in sediments along a single randomly chosen linear transect in each wetland. Each PVC frame held one SAT and one AT approximately 0.3 m beneath the water surface (Figure 2). Four PVC frames, hence 4 paired AT and SAT combinations (trap pairs) were deployed simultaneously in each wetland for 24 hrs. One frame (thus one trap pair) was deployed at each of 4 depth locations (0.3, 0.6, 0.9, and 1.2 m) in each wetland. Based on funnel dimensions, ATs sampled the upper 15 cm of the wetland water column; likewise, SATs sampled approximately the upper 15 cm of the water column as well as the water surface and region 10.16 cm above. This design yielded 720 samples from independent matched pairs of traps over the 3 years of our study.
|Figure 2. General orientation used in deploying surface activity traps (SAT) and conventional activity traps (AT) in study wetlands: 1. 0.3-cm-ID PVC pipe; 2. PVC elbow; 3. PVC Tee; 4. bolt with wing nut; 5. SAT; and 6. AT.|
Non-detection Rates. We assessed abilities of SATs and ATs to detect presence of each of 13 invertebrate taxa. Each trap was scored 1/0 for each taxon based on presence/absence of at least 1 individual. Our question of interest was whether non-detection probabilities (for each taxon) were greater for ATs than for SATs in each matched pair. We tested this hypothesis using McNemar's test (Agresti 1996:227-228) separately by sampling period and fish presence/absence. First, we calculated the proportion (with 95% confidence limits) of SATs that did not detect any organisms, then the additional proportion (with 95% confidence limits) of ATs that failed to detect any organisms. The latter was always positive and represents the extent to which SATs improve on ATs (i.e., the estimated sampling device effect). Essentially, McNemar's test evaluates whether these device effects differed significantly from zero (Agresti 1996: 227-228). To assess seasonal and/or fish-related differences in strength of device effects, a dichotomous score was created for each trap pair such that a pair received a score of 1 if the SAT detected a taxon, but the AT did not. Pairs were assigned a score of 0 for all other outcomes. For each taxon, we modeled the 0,1 score as a linear function of the fixed effects of the fish treatment and period; wetlands-within-years was treated as a random effect. We used the SAS GLIMMIX macro (after Littell et al. 1996) to fit this generalized linear mixed model (Breslow and Clayton 1993) with a binomial error structure.
Relative Abundance of Invertebrates. We assessed paired differences in numbers of organisms (each of 13 selected taxa and taxon richness) captured by ATs and SATs, conditional on at least 1 pair member trapping ≥1 organism from each taxon. We excluded from our analysis trap pairs wherein individual taxa were not detected by either device because these pairs contributed no information on differential trapping ability. We based our analysis for taxon richness on all taxa captured, not just the 13 selected groups used in assessing non-detection and relative abundance.
We analyzed paired differences in log-relative abundance of each taxon, plus taxon richness of the trap contents. Log transformations were required because our relative abundance data were severely skewed. We fit a linear model to the paired log-differences, modeling year and pond as random block effects and sampling period and fish presence/absence as fixed effects. As in our non-detection rate analysis, we obtained estimates of paired differences for each of the 6 sample period/fish fixed effects combinations and tested the null-hypothesis that each difference was zero. Differences were back-transformed and reported along with 95% confidence intervals. The back-transformed value of a log-difference has special utility for assessing effects associated with a new treatment (e.g., SAT) relative to a more traditional one (e.g., AT) (Keene 1995). Algebraically our improvement ratios were
where: SAT/AT = ratio of the estimated geometric treatment means.
This improvement ratio measures multiplicative improvement in the capture efficiency associated with using an SAT instead of an AT. Ratios whose 95% confidence limits include 1.0 imply no improvement, whereas ratios with lower bounds on their 95% confidence limits >1.0 imply significant improvement in capture efficiency due to use of SATs. We used SAS PROC MIXED (Littell et al. 1996) to fit the model and to obtain (1) adjusted catch-ratio estimates, (2) paired t-tests, and (3) F-tests for improvement-ratio differences resulting from sampling period and/or presence of fathead minnows.
Vertical Position Within SATs. SATs were initially developed to facilitate comparisons among catches of organisms associated with discrete water-column strata. We assessed potential differences (top, middle, bottom strata of each SAT) in non-detection probabilities and paired differences in log-relative abundance of each of our 13 invertebrate taxa using McNemar's test (Agresti 1996: 227-228) and SAS PROC MIXED (Littell et al. 1996) as described above. Again, we modeled year and pond as random block effects and sampling period, fish presence/absence, and vertical position (top, middle, bottom) as fixed effects.