A Comparison of Herpetofaunal Sampling Effectiveness of Pitfall, Single-ended, and Double-ended Funnel Traps Used with Drift Fences

We assessed the relative effectiveness of pitfal ls , s ingle-ended, and double-ended funnel traps at 12 replicate sites in sand pine scrub using drift fence arrays. Pitfalls captured fewer species but yielded more individuals of many species and higher average species r ichness than funnel traps. Pit fal ls and funnel traps exhibited differential capture bias probably due to differences in behavior or morphology. More surface-active lizards, frogs, and small semifossoriai herpetofaunal species were captured in pitfalls whereas captures of large snake species were restricted to funnel traps. Double-ended funnel traps captured twice as many large snakes as single-ended funnel traps. All three trap types yielded similar estimates of relat ive abundance of l izards and frogs but not snakes. Est imates of relat ive abundance of large snakes were higher for double-ended funnel traps than pitfal ls or s ingle-ended funnel traps. Pitfal l and funnel traps yield complementary results, and choice of type(s) depends on target species and sampling goals. Drif t fences with pit fal l and single-ended (1E) or double-ended (2E) funnel traps are an effective sampling method of herpetofaunal communities. Applications include inventory, estimation of species relative abundance, longterm monitoring, determination of activity cycles, intercommunity comparisons, and other experimental purposes . However, as for al l sampl ing techniques , there are potent ial biases from select ive sampling. Smal l , surface-act ive species are more easily captured by pitfall or funnel traps than are large snakes and turtles (Campbell and Christman, 1982; Enge and Marion, 1986). Tree frogs (Jones, 1986; Dodd, 1991) or other arboreal species are less likely to be captured on the ground by either trap type (Gibbons and Semlitsch, 1982). Species possessing cl imbing or jumping ab i l i t i es are more l ike ly to escape or trespass drift fences than are terrestrial herpetofauna (Franz et al., unpubl. data; Dodd, 1991; Corn, 1994). Home range size, daily and seasonal movement patterns, and microhabitat f idel i ty also inf luence capture effect iveness (Gibbons and Semlitsch, 1982; Bury and Corn, 1987; Corn and Bury, 1990). Differences in relative effectiveness of trap types or the arrangement of arrays can also bias herpetofaunal sampling. Funnel traps are more effective than pitfalls in capturing snakes (Campbell and Christman, 1982; Gibbons and Semlitsch, 1982; Vogt and Hine, 1982; Bury and Corn, 1987). Pitfall trapping may be enhanced by the use of dr i f t fences , especia l ly for rept i les (Corn and Bury, 1990), and especially snake captures (Bury and Raphael 1983; Raphael, 1988). Fence length, numbers, height, and arrangement can affect results (Campbell and Christman, 1982; Vogt and Hine, 1982; Jones, 1986; Bury and Corn, 1987; Corn and Bury, 1990). No single trapping system captures all species in proport ions representat ive of their actual abundance, rendering estimates of population or relat ive abundance and diversity among habitats difficult (Corn, 1994). Although several studies discuss advantages and disadvantages of different capture techniques, the effectiveness of different trap types in yielding similar est imates of species r ichness and relat ive abundance of herpetofauna has not been examined thoroughly. Here, we compare the relative effectiveness of pitfall, lE, and 2E funnel traps used with drift fences for sampling herpetofauna of the sand pine scrub of central Florida. We hypothesize that there are differences in relative effectiveness of sampling taxonomic categories, species richness, numbers of individuals , or relat ive abundance among the three trap types. These results could have useful implications for the selection of trap type(s) in relat ion to target species or sampling ob ject ives , as well as in the interpretation of capture data using these techniques. 3 2 0 C. H. GREENBERG ET AL. MATERIALS AND METHODS Study Area.-The study was conducted in sand pine scrub in the Ocala National Forest , Marion County, Florida. Sand pine scrub is a sclerophyllous, shrub-dominated ecosystem nearly restricted to Florida. In its natural state, peninsular scrub is subject to high intensity, low frequency wildlife. Post-fire recovery is rapid, with pre-disturbance dominant plants regaining site dominance within a few years (Abrahamson, 1984a, b; Givens et al., 1984). In the mature forest, the canopy is limited to a single tree species, sand pine (Pinus clausa). The shrub stratum is dominated by myrtle oak (QUETCUS myrtifolia), sand live oak (Q. geminata), Chapman’s oak (Q. chapmunii), fetterbush (Lyonia ferruginea), and two palmetto species (Serenou repens and Subul etoniu). Soi ls support ing sand pine scrub are excess ively drained aeol ian or marine sands (Kalisz and Stone, 1984). This area receives approximately 130 cm of rainfall annually, with over half fal l ing between June and September. Average temperatures range from 20-32 C between April and October and 11-23 C between November and March (Aydelott et al., 1975). Methods.-We established a single trapping array in each of 12 randomly selected sites in the Ocala National Forest. Nine sites were 5to 7-year-old open-scrub sites recently disturbed by catastrophic burning or c learcutt ing and fol lowed by highor low-intensity site preparation methods; and three were mature (2 55 yearold) sand pine forest s i tes . General s i te se lect ion cr i ter ia were (1) s imi lar e levat ion, topographic , and soil characteristics; (2) stand area 28.5 ha; and (3) 10.9 km from any known water source (temporary or permanent) or other habitat type. Arrays were located at least 25 m from roads or stand edges (except for two drift fences of one array) . Trapping arrays (Fig. 1) used material equivalent to two standard Campbell and Christman (1982) arrays, but consisted of eight 7.6-m lengths of erect, 0.5-m-high galvanized metal flashing arranged in an “L” pattern with a 7.6-m space between each length. All arrays were oriented with one arm (four drift fences) running north-south and the other east-west. Drift fences were buried 4-6 cm into the ground for support . Two black 18.9-L plastic paint buckets with 1.25-cm holes dr i l led into the bot tom for dra inage were sunk flush to the ground at both ends of each fence (N = 16 per site). Sticks were jammed into the drill holes to prevent animal escape. A sponge was placed into each bucket and dampened at each visit to reduce probability of desiccation. Funnel traps consisted of aluminum window screen (76 cm wide) rolled into a cylinder and stapled, with a screen wire funnel inserted into one (1E) or both (2E) ends pointing inward (Campbell and Christman, 1982). Funnel openings were approximately 35 cm in diameter. One 1E and one 2E funnel trap were randomly placed along either s ide of and adjacent to each fence (N = 8 per site each). Buckets were shaded by squares of Masonite pegboard slanted over the opening. Trapping arrays were opened and closed simultaneously for al ternating two-week periods from August 1991 through September 1992. We checked open traps every 2 to 3 d. Animals were marked for identification by toe (lizards and frogs) or scale (snakes) c l ipping and released at the point of capture. Pitfall traps were closed by fitting pegboard squares over the buckets and covering them with sand for a tighter seal. Funnel traps were closed by stuffing sponges into funnel openings. We excluded recaptures from the data set for this analysis. Odd-numbered buckets were eliminated from the data set in order to create a balanced design (N = 8 pitfalls, lE, and 2E funnel traps, respectively, per site). Trap success was calculated based on array-nights, defined as captures per trapping array per 24 h. We calculated numbers of individuals trapped, numbers of commonly trapped taxonomic groups and species, relative abundance, species richness, and Shannon’s divesity indices (Brower and Zar, 1977) for each trap type based on 12 si tes .