COMPARISON OF 3 NONLINEAR MODELS TO DESCRIBE LONG-TERM TAG SHEDDING BY LAKE TROUT

We estimated long-term tag-shedding rates for lake trout Salvelinus na maycush using two existing models and a model we developed to account for the observed permanence of some tags. Because tag design changed o ver the course of the study, we examined tag-shedding rates for three types of numbered anchor tags (Floy tags FD-67, FD-67C, and FD-68BC) a nd an unprinted anchor tag (FD-67F). Lake trout from the Gull Island S hoal region, Lake Superior, were double-tagged, and subsequent recaptu res were monitored in annual surveys conducted from 1974 to 1992. We m odeled tag-shedding rates, using time at liberty and probabilities of tag shedding estimated from fish released in 1974 and 1978-1983 and la ter recaptured. Long-term shedding of numbered anchor tags in lake tro ut was best described by a nonlinear model with two parameters: an ins tantaneous tag-shedding rate and a constant representing the proportio n of tags that were never shed. Although our estimates of annual shedd ing rates varied with tag type (0.300 for FD-67, 0.441 for FD-67C, and 0.656 for FD-68BC), differences were not significant. About 36% of ta gs remained permanently affixed to the fish. Of the numbered tags that were shed (about 64%), two mechanisms contributed to tag loss: disint egration and dislodgment. Tags from about 11% of recaptured fish had d isintegrated, but most tags were dislodged. Unprinted tags were shed a t a significant but low rate immediately after release, but the long-t erm, annual shedding rate of these tags was only 0.013. Compared with unprinted tags, numbered tags dislodged at higher annual rates; we hyp othesized that this was due to the greater frictional drag associated with the larger cross-sectional area of numbered tags.