Résumé
Terrestrial gastropods are sometimes sampled for ecological and parasitological studies using simple traps of cardboard placed on the ground. This method enables the collection of large numbers of individuals with relatively little effort. However, the violation of assumptions of the trap method may mean that samples are biased. I examined the reliability of this method by comparing species composition, richness and diversity of gastropod collections from traps to paired hand-searched plots in Algonquin Path, Ontario, Canada. Ten samples for each method were made from five replicate sites in three habitat types, white birch, hardwood and logged forest stands. Of the 18 species found overall, 17 and 16 species were found using hand search and trap methods respectively. Some species, especially Arion circumscriptus, were consistently over-represented using traps, while most other species had greater representation in hand-search collections. In general, hand-search collections had significantly greater species richness and diversity than trap collections. Further, the similarity in species composition of the two collections seemed to depend, at least in white birch and hardwood habitats, on the level of site diversity; as site diversity increased the percent similarity decreased. However, estimated richness, calculated using a jackknife estimator that accounts for heterogeneity in species detection, was not different for the two collection methods. There was no difference in average species' detection per sample between collections; the probability of sampling a species using a single sample within a site was low for both methods (16-34% of species). However, overall detection of species on a site (average across all samples within a site) tended to be higher for hand-searched plots, especially in more complex habitats. Thus, as trap sampling does not provide representative gastropod collections, this method may not be appropriate to use when physical collections or detailed information on species abundances are required. Conversely, as there is no difference between methods in estimated richness, traps may be suitable to estimate gastropod community metrics, provided appropriate estimation models are used.