Efficiency and Bias in Strip and Line Transect Sampling

摘要: Efficiency and relative bias of strip transect line (Fourier series estimator) survey methods were compared. Comparisons involved the negative exponential half-normal detection functions for expected sample sizes 40, 60, 100 three levels spatial aggregation. computed as width, w. Final comparisons based on mean square errors (MSE), also a function The estimator always had smaller MSE than when was exponential. Therefore, sampling is preferred detectability drops off rapidly with increasing distance from line. most combinations aggregation w, halfnormal function. In general, results indicate preference method over basis efficiency. J. WILDL. MANAGE. 49(4):1012-1018 Applied ecologists faced problem estimating size (N) or density (D) population frequently consider use (see Burnham Anderson 1984 comparison these two methods). Strip transects are merely long narrow plots quadrats elementary theory applies; this has existed several decades Gates 1979 Seber 1982 reviews). contrast, perhaps one exception (Hayne 1949), serious attempts toward rigorous analysis began much later papers by Eberhardt (1968) et al. (1968). Reviews examples present state application found in (1979), (1980), Ralph Scott (1981), Hayes Buckland (1983). (1980:34) raised issue, there more fundamental question whether any [line transect] technique actually effective using relatively transect" (also see 1979: 24). One faces trade-offs compromises between statistical efficiency attempting to answer posed Eberhardt. For example, waterfowl nests reported Pospahala (1970). Narrow (2.46 m each side observer) searched, This content downloaded 157.55.39.25 Tue, 02 Aug 2016 04:10:05 UTC All subject http://about.jstor.org/terms Wildl. Manage. 49(4):1985 EFFICIENCY AND BIAS IN TRANSECT SAMPLING * 1013 total number observed (n = size) expanded an estimate N D area interest. Fortunately, perpendicular data taken allow theory. allowed test critical assumption required that no gone undetected. Approximately 13% undetected, representing significant 1978b:16-17 Pollock 1978 further analyses data). (1984) estimated width would have be reduced 1.54 observer assure essentially all been detected. extremely inefficient impractical (Eberhardt 1979:25). A may objects detected, hence, counted. will produce unbiased density, D. However, it not reliable because could small, which means large variance. As increases, opposing phenomena occur: variance decreases, but increases can become severe. By contrast avoid, reduce, estimators employ "adjust" decreasing probability distance. (effectively unbounded) counted (sample size, n), typically greater transect. An additional parameter must transects. wonders about (defined here ratio standard estimators) vs. schemes. sampling, only var(n) D, usually counts replicate strips. var( f(0)), distances, evaluated at zero. purpose paper examine width. From theoretical results, we suggest some guidelines useful design conduct field