Validation of a randomization procedure to assess animal habitat preferences: microhabitat use of tiger sharks in a seagrass ecosystem.

1. Tiger sharks Galeocerdo cuvier are important predators in a variety of nearshore communities, including the seagrass ecosystem of Shark Bay, Western Australia. Because tiger sharks are known to influence spatial distributions of multiple prey species, it is important to understand how they use habitats at a variety of spatial scales. We used a combination of catch rates and acoustic tracking to determine tiger shark microhabitat use in Shark Bay. 2. Comparing habitat-use data from tracking against the null hypothesis of no habitat preference is hindered in Shark Bay, as elsewhere, by the difficulty of defining expected habitat use given random movement. We used randomization procedures to generate expected habitat use in the absence of habitat preference and expected habitat use differences among groups (e.g. males and females). We tested the performance of these protocols using simulated data sets with known habitat preferences. 3. The technique correctly classified sets of simulated tracks as displaying a preference or not and was a conservative test for differences in habitat preferences between subgroups of tracks (e.g. males vs. females). 4. Sharks preferred shallow habitats over deep ones, and preferred shallow edge microhabitats over shallow interior ones. The use of shallow edges likely increases encounter rates with potential prey and may have profound consequences for the dynamics of Shark Bay's seagrass ecosystem through indirect effects transmitted by grazers that are common prey of tiger sharks. 5. Females showed a greater tendency to use shallow edge microhabitats than did males; this pattern was not detected by traditional analysis techniques. 6. The randomization procedures presented here are applicable to many field studies that use tracking by allowing researchers both to determine overall habitat preferences and to identify differences in habitat use between groups within their sample.