Hamel_1_pif09.indd

Partners in Flight Conference: Tundra to Tropics TEST OF PARTNERS IN FLIGHT EFFECTIVE DETECTION DISTANCE PAUL B. HAMEL,1,4 MELINDA J. WELTON,2 CARL G. SMITH, III,1 AND ROBERT P. FORD3 1Center for Bottomland Hardwoods Research, P. O. Box 227, 432 Stoneville Road, Stoneville, Mississippi 38776, USA;2Gulf Coast Bird Observatory, 5241 Old Harding Road, Franklin, Tennessee 37064, USA; and3U.S. Fish and Wildlife Service, The University of Memphis, South Campus, Memphis, Abstract. Estimation of population sizes of North American avian species has been attempted in the North American Landbird Conservation Plan. Such estimated numbers have considerable conserva- tion value as starting points to estimate extinction probability, as was done for Cerulean Warbler (Dendroica cerulea) during the U.S. Fish and Wildlife Service evaluation of the petition to list the spe- cies as Threatened. Population estimates presented in the Flight Plan refl ect assumptions applied to counts reported by observers on Breeding Bird Survey routes. One of these assumptions is the assign- ment of species to effective detection distance radii. We chose to test the assumption that effective detection distance of 125 m for Cerulean Warbler was an adequate value in bottomland hardwood and other forests in the species’ breeding range. We randomly selected roadside and off-road loca- tions, visited each multiple times with multiple observers, and used hand-held Global Positioning System units to measure the distance between count station and birds detected aurally. We used multiple covariate distance sampling to analyze these data in Program Distance. Our best estimate of effective detection distance is 94 m (95% CI 88–101 m), signifi cantly lower than 125 m. Consequently, the total population estimate of Cerulean Warbler in the North American Landbird Conservation Plan, 560 000, should be revised to approximately 875 000; assuming all other factors involved in the calculation of total population remain equal. Key Words: Cerulean Warbler, Dendroica cerulea distance sampling, effective detection distance, popu- EVALUACIÓN DE DISTANCIA EFECTIVA DE DETECCIÓN PARA LA REINITA CERÚLEA (DENDROICA CERULEA) SELECCIONADA POR Resumen. Estimar el tamaño de las poblaciones de aves en Norte América forma parte del Plan de Vuelo de la organización Compañeros en Vuelo. El valor de conservación de estos estimados reside en su utilidad como punto de partida para poder estimar probabilidad de extinción de especies como la Reinita Cerúlea (Dendroica cerulea). Esos estimados fueron utilizados durante el proceso de revisión que llevo a cabo el Servicio Federal de Pesca y Vida Silvestre para atender la solicitud de listado para la especie bajo el Acta de Especies en Peligro de Extinción. Los estimados poblacionales presentados en el Plan de Vuelo refl ejan premisas aplicables a los conteos reportados en el censo anual de aves durante la época reproductiva (BBS, por sus siglas en ingles). Una de estas premisas consta de asignar cada especie a un radio efectivo de detección. En este trabajo evaluamos la validez del radio de detec- ción de 125 m establecido para la Reinita Cerúlea en distintos tipos de bosque dentro de su rango reproductivo. Seleccionamos localidades al azahar dentro y fuera de caminos las cuales visitamos en varias ocasiones con múltiple observadores y usamos unidades portátiles de GPS para determinar la distancia entre las estaciones de conteo y los individuos detectados auditivamente. Utilizamos muestreo a distancia en un diseño de covariables múltiples analizado en el programa Distance. El mejor estimado para distancia efectiva de detección fue 94 m (95% CI 88–101 m) lo que es signifi cati- vamente menor a 125 m. Por tanto, los estimados poblacionales para la Reinita Cerúlea presentados en el Plan de Vuelo de 560 000 individuos deben ser modifi cados a unos 875 000, asumiendo que todos los otros factores envueltos en este cálculo permanezcan igual. Effective Detection Distance for Cerulean Warbler—Hamel et al. distance of 125 m for Cerulean Warbler was an adequate value in bottomland hardwood and Estimation of global population size for bio- other forests in the species’ breeding range.
logical species is a task generally restricted to species with small populations of limited geo- graphic scope. Accurate estimation of popula- tion size is useful for assessing risks of various We identifi ed random locations in Hatchie threats to populations in terms of mortality fac- tors, and in estimating time to extinction based Tennessee; Chickasaw National Wildlife Refuge, upon modeled population growth rates. The Lauderdale Co., Tennessee; Meeman Shelby utility of making such population estimates for Forest State Park and Wildlife Management the variety of species in an entire fauna has been Area, Shelby Co., Tennessee; and Center Hill further demonstrated in the Partners in Flight Lake Recreation Area, DeKalb, Co., Tennessee. North American Landbird Conservation Plan Each of the study areas is characterized by large Taking advantage of the continent-wide cov- Roadside locations in each of the areas were erage of the Breeding Bird Survey (BBS; Robbins established systematically from a random start- et al. 1986), Rich et al. (2004) produced a set of ing point, and placed 800 m apart when surveys population estimates for most North American were conducted from vehicles, and 500 m apart landbirds. The procedure for calculating the when surveys were conducted on foot. estimates is described in Rich et al. (2004), elab- Off-road locations were established on exist- orated by Rosenberg and Blancher (2005), and ing study areas in Cerulean Warbler habitat revised in Blancher et al. (2007). Thogmartin et in Chickasaw National Wildlife Refuge and al. (2006) provide an independent evaluation of Meeman Shelby Forest State Park and Wildlife the procedure. The method involves evaluating Management Area. In each of these study areas, the average number of registrations recorded we selected grid points on existing 50 x 50-m on BBS routes within a particular area, correct- grids such that points were 250 m apart. Off- ing or adjusting that number of registrations to road locations on Hatchie National Wildlife account for detectability, and then applying the Refuge were established at randomly selected resulting estimate of point density to the entire grid intersections of a 300 x 300-m grid laid across areas in the Refuge in which Cerulean Embedded in this procedure is an assumed Warblers were known to occur.
effective detection distance, believed to be At each selected off-road or roadside loca- typically larger than the effective detection dis- tion, one or two of the coauthors visited the tance arrived at by analytical methods such as location during the morning hours in May or those produced by distance sampling analy- June of 2007 or 2008. On arrival at the count- ses (Buckland et al. 2001), as exemplifi ed in ing station, we established a waypoint using a Program Distance (Thomas et al. 2005). Thomas Garmin Global Positioning System (GPS) unit et al. (2002) defi ne effective detection distance, and listened for three minutes, conducting a or effective strip (half) width, as that distance count in standard Breeding Bird Survey pro- from the counting station “for which as many tocol. (Note: The use of trade or fi rm names in objects are detected beyond as are missed this publication is for reader information and within”; we follow that usage here. The area does not imply endorsement by the United around each counting station, and hence the States Department of Agriculture of any prod- total survey area calculated using this assumed uct or service.) When a Cerulean Warbler distance has a profound effect on the ultimate was detected, the observer went to a location size of estimated populations. Thogmartin et al. directly below the singing male bird, and either (2006) point out that few empirical data exist for 1) established a second GPS waypoint, or 2) appropriate estimation of detection distances noted the distance and measurement error from the counting station as displayed in the GPS Such estimated population numbers have unit. We thus avoid the potential additional considerable conservation value as starting source of error caused by observer variability points to estimate extinction probability, as in distance estimation (Alldredge et al. 2007). was done for Cerulean Warbler (Dendroica ceru- Observers independently registered the birds; lea) during the U.S. Fish and Wildlife Service when only one of a pair of observers heard the evaluation of the petition to list the species bird, only that individual was credited with an as Threatened (U.S. Fish and Wildlife Service observation. In cases in which a bird had moved 2006a, 2006b). We chose to test the assump- during the interval between initial detection tion in Rich et al. (2004) that effective detection and the observer reaching its location, we did Proceedings of the Fourth International Partners in Flight Conference not record a distance. Such occasions were rare, in only 4 of 204 observations were we unable to Distance and evaluated using information the- record a distance. While we did not make spe- oretic methods listed above. Signifi cance level cifi c note of bird movements between detection was set at α = 0.10 to evaluate goodness of fi t and distance measurement, such movements tests, given the modest sample sizes achieved. beyond minor movements within the same tree, Our test of the null hypothesis that effective detection distance of Cerulean Warbler did not Radial distance measures were calculated differ from the 125 m posited by Rich et al. (2004) by distances between waypoints, or directly as was conducted by assessing whether 125 m fell the recorded distance made in the fi eld. Each within the 95% confi dence interval around the distance was identifi ed by its type, off-road or mean detection distance determined by the var- roadside, the date, time, and the observer(s) ious models. When it did not, we rejected the The set of radial distances recorded in this way was subjected to Multiple Covariate Distance radial detection distances for Cerulean Warblers Sampling analysis in Program Distance (Thomas in Tennessee in this project (Table 1), approxi- et al. 2005). A priori models were established mately equally divided between roadside and to evaluate appropriate modeling functions, off-road detections. Among several models to in which data were fi tted to half-normal distri- evaluate the detection distance, each fi t the data bution function with cosine adjustment or as well (Table 1), with non-signifi cant probability of hazard-rate function with a polynomial adjust- the model values differing from those observed. ment. We wished to test the null hypotheses that The best model for evaluating the effective neither observer nor type of registration had an detection distance was the half-normal approxi- effect on effective detection distance of Cerulean mation, without cosine adjustments (Table 2). Warbler. We further wished to test the null An alternative formulation with hazard-rate hypothesis that effective detection distance of function produced a model with a ΔAIC value Cerulean Warbler did not differ from the 125 m more than two units higher than the half-normal posited by Rich et al. (2004). We used 2nd order function. Neither observer nor type of registra- AIC corrected for small sample sizes (AIC ) as tion contributed signifi cantly to improving the our criterion to evaluate models in the candidate fi t of the model function to the distance data; set, using AIC differences (ΔAIC ) and Akaike including observer did improve the AIC above weights (w ) to make the comparisons. Upon the constant model, but the improvement was selection of appropriate model from this set within two units of AIC and the models can thus using information theoretic approaches, further be considered to be equivalent.
evaluations were conducted to identify poten- Effective detection distances for Cerulean tial improvements in models when observer and Warbler in Tennessee resulted in estimates that type of registration were included as covariates varied from 88–104 m depending upon observer in the models. For covariates that improved the and registration type (Table 2). The best-sup- null, constant detection distance model, individ- ported model, in which observer effects were ual analyses of these factor combinations were included in addition to the half-normal func- further conducted to produce covariate-specifi c tional form, produced an estimate of effective estimates of effective detection distance.
detection distance of 94 m. The constant model, TABLE 1. COMPARISON OF MODELS OF DISTRIBUTION OF 204 DETECTION DISTANCES FOR CERULEAN WARBLERS MEASURED IN Half normal with cosine, Registration type, Observer Half normal with cosine, Registration type c AIC = –2 log L + 2K + 2K(K + 1) / (n – K – 1).
d w = exp[–{ΔAIC / 2}] / Σ exp[–{ΔAIC / 2}].
e Goodness-of-Fit Probability of modeled function to original data, evaluated with Cramer von Mises tests.
Effective Detection Distance for Cerulean Warbler—Hamel et al. TABLE 2. RADIAL DISTANCES OF CERULEAN WARBLERS DETECTED IN ROADSIDE AND OFF-ROAD COUNTS BY THREE OBSERVERS IN Observer size (n) distance (m) a Values presented result from models calculated in Program Distance with the appropriate covariates specifi ed and using the half-normal function. For example, results for the Count Type=Off-Road, Observer=MJW row indicate a model in which analysis was confi ned to all Off-Road observations in which no effects beyond the parameters of the We found no effect of type of registration, half-normal form were included, was virtually roadside or off-road, on effective detection dis- equivalent to that best-supported model. The tance for Cerulean Warbler in this study, sug- constant model produced an effective detection gesting that Breeding Bird Survey methods may distance of 95 m. All estimates of effective detec- provide an adequate representation of detec- tion distance included 100 m in the 95% confi - tion distance within habitats. Effect of differ- dence interval; only estimates based upon 44 or ence among observers, while representing an fewer observations included 125 m in the 95% improvement in AIC over the constant detection confi dence interval. We present the descriptive distance model, did not meet the criterion of an results of all the a priori models we developed so improvement of two AIC units over the constant that the modest variations in effective detection detection distance model. Thus, in this study, observer effects were minor or not signifi cant.
Our result, that the effective detection distance of Cerulean Warbler is 100 m rather than 125 m, indicates that the population estimate of this spe- Our data do not support the assumption cies provided in Rich et al. (2004) of 560 000 may in Rich et al. (2004) that the effective detec- be an underestimate. The difference between the tion distance for Cerulean Warbler is 125 m. assumed 125 m and our 100 m effective detection A more appropriate detection distance is 95 m distance is a reduction in sampled area of 36%. (Table 2), or 100 m, a value included in the Applying this reduction in area to the estimate 95% confidence interval of all of our estimates. in Rich et al. (2004) results in a population esti- Blancher et al. (2007) indicated that estimates mate of 875 000 (range: 858 000–1 130 000, given of effective detection distance presented in 94 m estimate with 88–101 m CI). Our estimate, Rich et al. (2004) are intended to be conserva- 100 m, is one that is suffi cient for some variation tive or robust. They further indicate that an in observers and for different areas in bottom- order of magnitude resolution is appropriate land and upland forest in Tennessee. The study for the use of the population size estimates areas we chose in middle and west Tennessee produced by their methods. Inasmuch as this represent a range of topography and vegetation was a study of a single species, in which we composition refl ective of Cerulean Warbler habi- attempted specifically to register Cerulean tats in that state, and perhaps elsewhere as well. Warblers, our estimate of effective detec- Nevertheless, our estimated effective detection tion distance is likely also conservative rela- distance should not be considered to apply to the tive to one developed during counts in which entire breeding range of the species without fur- the intent was to register all. Thus, the actual ther specifi c evaluation in additional areas. effective detection distance applicable to the BBS data used by Rich et al. (2004) may be www.fws.gov/midwest/Eco%5FServ/soc/ birds/cerw/cerw06rapp5c.pdf; 27 March 2009) Proceedings of the Fourth International Partners in Flight Conference estimated the probability of a 90% decline in RICH, C. M. RUSTAY, J. M. RUTH, AND T. C. Cerulean Warbler populations at perhaps 90% WILL. 2007. Guide to the Partners in Flight within a century of the 1995 estimate of popula- tion at 560 000. His work further suggested that if the population was assumed to be 50% higher in 1995 than the 560 000 value given by Rich et Series No 5. [Online.] <http://www.part- al. (2004), the probability of population reduc- nersinfl ight.org/> (29 July 2008).
tion would be reduced, and the estimated time BUCKLAND, S. T., D. R. ANDERSON, K. P. BURNHAM, to reach the 90% probability of a 90% decline J. L. LAAKE, D. L. BORCHERS, AND L. THOMAS. would be more than a century. Based upon 2001. Introduction to Distance Sampling. the fi ndings reported here, which imply that Oxford University Press. Oxford, UK.
the 1995 population was perhaps 50% higher PETERJOHN, B., AND K. PARDIECK. 2002. A bibliog- than the Rich et al. (2004) estimate, we suggest raphy for the North American Breeding Bird that the period of time in which conservation Survey. U.S. Geological Survey, Patuxent actions can be applied to reverse the declines in Wildlife Research Center. [Online.] <http:// the species numbers is longer than previously RICH, T. D., C. J. BEARDMORE, H. BERLANGA, P. J. BUTCHER, D. W. DEMAREST, E. H. DUNN, W. C. We tested the assumed effective detection HUNTER, E. E. IÑIGO-ELIAS, J. A. KENNEDY, A. M. distance for Cerulean Warbler presented in MARTELL, A. O. PANJABI, D. N. PASHLEY, K. V. the North American Landbird Conservation ROSENBERG, C. M. RUSTAY, J. S. WENDT, T. C. Plan (Rich et al. 2004) and found it to be closer to 100 m than the assumed 125 m. We found no difference between the effective detection distance recorded on off-road versus roadside ROBBINS, C. S., D. A. BYSTRAK, AND P. H. GEISSLER. point counts in Tennessee, and only a small 1986. The Breeding Bird Survey: its fi rst effect of observer variability on the estimated fi fteen years, 1965–1979. U.S. Department effective detection distance. We suggest that of Interior, U.S. Fish and Wildlife Service the time for applying conservation action to Resource Publication 157. Washington, D.C.
maintain and increase populations of this spe- ROSENBERG, K. V., AND P. J. BLANCHER. 2005. cies of conservation concern, though finite, Setting Numerical Population Objectives for is considerably longer than previously sup- Priority Landbird Species, pp. 57–67. In C. J. Ralph and T. D. Rich [eds.], Bird conserva- tion and implementation in the Americas: Partners in Flight Conference. Vol. 1. United States Department of Agriculture, Forest staff of the U.S. Fish and Wildlife Service West Service, Pacifi c Southwest Research Station, Tennessee Refuges Complex for contributing funds to this work. Our ideas were improved in discussion with John Sauer during the McAllen THOGMARTIN, W. E., F. P. HOWE, F. C. JAMES, D. H. JOHNSON, E. T. REED, J. R. SAUER, AND through reviews by Scott Barras, T.J. Benson, F. R. THOMPSON, III. 2006. A review of the THOMAS, L., S. T. BUCKLAND, K. P. BURNHAM, ALLDREDGE, M. W., T. R. SIMONS, AND K. H. D. R. ANDERSON, J. L. LAAKE, D. L. BORCHERS, POLLOCK. 2007. A fi eld evaluation of dis- tance measurement error in auditory avian pling, pp. 544–552. In A. H. El-Shaarawi point count surveys. Journal of Wildlife and W. W. Piegorsch [eds.], Encyclopedia of Environmetrics., Vol. 1. J. Wiley & Sons, BLANCHER, P. J., K. V. ROSENBERG, A. O. PANJABI, B. ALTMAN, J. BART, C. J. BEARDMORE, G. S. THOMAS, L., J. L. LAAKE, S. STRINDBERG, F. F. C. BUTCHER, D. DEMAREST, R. DETTMERS, E. H. MARQUES, S. T. BUCKLAND, D. L. BORCHERS, D. R. IÑIGO-ELIAS, D. N. PASHLEY, C. J. RALPH, T. D. J. H. POLLARD, J. R. B. BISHOP, AND T. A. Effective Detection Distance for Cerulean Warbler—Hamel et al. MARQUES. 2005. Distance 5.0. Release “Beta 5”. Research Unit for Wildlife Population Assessment, University of St. Andrews, U.S. FISH AND WILDLIFE SERVICE. 2006b. UK. [Online.] <http://www.ruwpa.st-and.
ac.uk/distance/> (27 March 2009).
Plants; 12-Month Finding on a Petition To List the Cerulean Warbler (Dendroica cerulea) as Threatened With Critical Habitat. Federal Risk Assessment & Conservation Planning Workshop. [Online.] <http://www.fws.

Source: http://www.partnersinflight.org/pubs/McAllenProc/articles/PIF09_Decision%20Support%20Tools/Hamel_1_PIF09.pdf