Estimating density of Florida Key deer

Abstract

Florida Key deer (Odocoileus virginianus clavium) were listed as endangered by the U.S. Fish and Wildlife Service (USFWS) in 1967. A variety of survey methods have been used in estimating deer density and/or changes in population trends for this species since 1968; however, a need to evaluate the precision of existing and alternative survey methods (i.e., road counts, mark-recapture, infrared-triggered cameras [ITC]) was desired by USFWS. I evaluated density estimates from unbaited ITCs and road surveys. Road surveys (n = 253) were conducted along a standardized 4-km route each week between January 1999–December 2000 (total deer observed, n = 4,078). During this same period, 11 ITC stations (1 camera/42 ha) collected 5,511 deer exposures. Study results found a difference (P < 0.001) between methods with road survey estimates lower (76 deer) than ITC estimates (166 deer). Comparing the proportion of marked deer, I observed a higher (P < 0.001) proportion from road surveys (0.266) than from ITC estimates (0.146). Lower road survey estimates are attributed to (1) urban deer behavior resulting in a high proportion of marked deer observations, and (2) inadequate sample area coverage. I suggest that ITC estimates are a reliable and precise alternative to road surveys for estimating Key deer densities on outer islands. I also evaluated density estimates from 3 road survey methods. Road survey methods (n = 100) were conducted along a standardized 31-km route where markresight, strip-transect, and distance sampling data were collected between June 2003– May 2004. I found mark-resight estimates to be lower ( x = 384, 95% CI = 346–421) than strip-transect estimates ( x = 854, 95% CI = 806–902) and distance estimates ( x = 523, 95% CI = 488–557). I attribute low mark-resight estimates to urban deer behavior resulting in a higher proportion of marked deer observations along roadways. High strip-transect estimates also are attributed to urban deer behavior and a reduced effective strip width due to dense vegetation. I propose that estimates using distance sampling eliminate some of these biases, and recommend their use in the future.