EVALUATION OF A SONIC TELEMETRY SYSTEM IN 3 HABITATS OF AN ESTUARINE ENVIRONMENT

A directional sonic telemetry system in small embayment, seagrass, and channel habitats in Core Sound, North Carolina was evaluated. We comp ared point location estimates calculated after correcting for system b ias in three different ways: using test station (i.e. known location t ransmitting-receiving stations) angle errors, and using a within test site and an among test site mean angle error estimate. Estimates are n ecessary because, when tracking an animal, system bias cannot be corre cted for using test station angle errors. In addition, telemetered ani mals may move beyond test areas or into different habitats. We found n o significant difference (P > 0.05) among point location estimates, su ggesting that a within or an among mean angle error was an acceptable estimate. Choosing the appropriate angle error estimate must be done c arefully because both presented limitations. A within test site mean a ngle error was the more conservative approach, avoiding biases caused by significant (P<0.05) interhabitat angle error variability. An estim ate containing interhabitat variability (i.e. among test site angle er ror) might be more robust for correcting system bias when the instrume nted animal moves outside test areas or into a heterogeneous area. Sea grass habitat polygons in southern Core Sound range in size from 0.1 t o 3189 ha. Attained levels of accuracy and precision from this study s uggest that work could be conducted in areas where polygons are greate r than or equal to 6.9 ha, which represents > 97% of the seagrass habi tat in Core Sound. Although the majority (80%) of the polygons are sma ll (< 10.0 ha), they represent < 5% of the total seagrass area. In add ition, classifying use of habitat in areas where polygons are less tha n or equal to 6.9 ha is possible because small polygons have a contagi ous distribution; hence, their areas may be additive. Risks of misclas sifying use of habitat can be reduced also by controlling the size of confidence areas (A,) by adjusting the distance between observers and the tracked animal. The confidence area as a function of distance can be predicted because location error varied linearly and significantly with geometric mean distance (D-g). On the basis of this relationship, D-g must be < 326 m for A(e)95 to be < 10.0 ha. (C) 1997 Elsevier Sci ence B.V.