Wildl. Soc. Bull. 20:455-456, 1992 WILDLIFE SOFTWARE TRANSAN: LINE TRANSECT ESTIMATES BASED ON SHAPE RESTRICTIONS R. D. ROUTLEDGE, Department of Mathematics and Statistics and Department of Biological Sciences, Simon Fraser University, Burnaby, BC V5A 1S6, Canada D. A. FYFE,1 Department of Mathematics and Statistics and Department of Biological Sciences, Simon Fraser University, Burnaby, BC V5A IS6, Canada 1 Present address: 27098 Wagner Circle, Kingston WA 98346. Line transects often are used to estimate the abundance of wildlife populations. Johnson and Routledge (1985) and Routledge and Fyfe (1992) have discussed a procedure for constructing such an estimator (called the shape-restricted estimator) and associated confidence limits. We herein describe the computer program TRANSAN for calculating the shape-restricted estimator. An existing program, TRANSECT, calculates several of the competing line-transect estimators described by Burnham et al. (1980) and references therein. Line-transect estimators are typically based on distance measurements, the perpendicular distances of each detected individual from a transect line. TRANSAN will accept distance data either directly or as frequencies for equally sized distance classes. To calculate confidence interval estimates, the program requires data from at least 4 independent transects. Line-transect estimators involve modelling the probability of detecting an individual as a function of its perpendicular distance from the transect line. The graph of this function is called the detection curve. The shape-restricted estimator calculated by TRANSAN imposes the restrictions that this detection curve decline from 1-0 without ever increasing, that it initially curve downwards from a horizontal shoulder to some point of inflection, and that it then level out as the detection probability approaches 0. The location of the inflection point can be either specified by the user or estimated by TRANSAN subject to certain limitations. Other restrictions on the detection curve - for example, the width of its horizontal shoulder or the height of its tail--also can be imposed. Hence, the estimator is adaptable to a variety of situations. The shape-restricted estimator also behaves more stably than its chief competitors. Johnson and Routledge (1985) demonstrated that other estimators typically perform well for specific shapes of detection curves, but poorly for others. The half-normal parametric estimator can produce a bias of > 25% for detection functions deviating from the presumed half-normal shape. Although nonparametric estimators such as the Fourier series estimator are relatively successful at controlling the bias, they often do so at the cost of low precision of the estimator. The shape-restricted estimator has shown the most consistency in maintaining a standard error that is competitive with the parametric estimators (Johnson and Routledge 1985). With its default constraints on the shape of the detection curve, the shape-restricted estimator performs consistently well. Furthermore, the program offers an opportunity to improve this performance. For example, if the observer is certain no animals are missed to some width around the transect line, the shoulder width could be set to a value larger than the default. This would typically reduce the standard error of the density estimator. However, if the observer were in reality likely to miss some animals within this width, then the shoulder width would be set at too large a value and the estimator would be biased. TRANSAN contains a simulator option that can be used to assess the effect of altering the default constraints. For example, to assess the bias, the user can choose a detection curve, a set of transect lengths, etc., and then specify alterations to the default constraints. The program then simulates sets of observations, calculates a density estimate for each, and uses these density estimates to estimate the bias. The program also can be instructed to simulate confidence limits that can be used to further assess the reliability of the confidence interval procedure. Users contemplating altering the default constraints are encouraged to use the simulation option and are particularly cautioned to explore the consequences of overextending the shoulder width. PROGRAM AVAILABILITY AND SYSTEM REQUIREMENTS Copies of the program and related documentation are available on the Bird Monitor bulletin board (~01-498-0402). A technical appendix discussing the algorithm and related mathematical theory is available at a nominal charge from the Depository of Unpublished Data, CISTI, Building M55, National Research Council of Canada, Ottawa, ON KlA 0S2, Canada. TRANSAN is available on the bulletin board as a FORTRAN 77 source code and in executable form for direct use on IBM-compatible microcomputers. The program is accompanied by an electronically stored manual and by auxiliary files, including data files containing the sea star (Pisaster sp.) observations described in Routledge and Fyfe (1992) and examples of the simulation capabilities of the program. The system requirements are minimal: an IBM-PC-compatible with a double-sided disk drive. However, the confidence interval procedure is time consuming. On a system with an 80286 chip running at 12 MHz, the confidence limits for the sea star example of Routledge and Fyfe (1992) with default constraints can be obtained in approximately 2 hours. A 486 machine running at 33 MHz can reduce this time to 20 minutes. Users with a mathematical coprocessor can increase the speed by recompiling the source code and linking with a mathematical library that accesses the coprocessor. Those contemplating extensive simulation runs would benefit from access to a more powerful computer. On an SGI 4D/320S computer running IRIX System V Release 3.3n, the above task takes <2 minutes. Acknowledgments. - This project was partially supported by a grant from the Natural Sciences and Engineering Research Council of Canada and was improved by constructive comments from J. Laake, the editorial staff, and anonymous reviewers. LITERATURE CITED BURNHAM, K. P., D A. ANDERSON, AND J. L. LAAKE. 1980. Estimation of density from line transect sampling of biological populations. Wildl Monogr.72. 202pp. J0HNSON, E. G., AND R. D. ROUTLEDGE. 1985. The line transect method: a nonparametric estimator based on shape restrictions. Biometrics 41:669-679. ROUTLEDGE, R. D., AND D. A. FYFE. 1992. Confidence limits for line transect estimates based on shape restrictions. J. Wildl. Manage. 56:402-407. Received 30 October 1991. Accepted 22 April 1992. Software Editor: Rexstad.