"RESPONSE TO FARMER (2008): LIMITATIONS OF STATISTICALLY DERIVED POPULATION ESTIMATES, AND SUGGESTIONS FOR DERIVING NATIONAL POPULATION ESTIMATES FOR SHOREBIRDS

We commend Adrian (Farmer 2008) for initiating a debate about how best to update bird-conservation lists and, in particular, how the shorebird community evaluates population sizes of shorebirds (e.g., Morrison et al. 2006). We concur with him that many of the initial efforts to estimate shorebird population sizes in North America (e.g., Donaldson et al. 2000, Brown et al. 2001, Morrison et al. 2001) were made by species experts who, admittedly, were providing their best estimates of the size of breeding populations. This best-information approach was deemed acceptable, given the greater good expected from development of the national shorebird plans. Recognizing these shortcomings, substantial effort has been and is being made to refine population estimates of shorebirds. We ad-vocate that new estimates be subject to rigorous review such that future plans rely on increasingly accurate estimates. Rejection of new estimates should be viewed not as acceptance or preference of prior best estimates, but as an honest assessment of the merits of each esti-mate as it is presented. In our view, inserting new, potentially equally or more inaccurate estimates simply continues the propagation of wrong numbers, and may lead to erroneous changes, confusing a management community charged with recovering threatened bird popula-tions. We strongly disagree with several points presented by Farmer (2008:xx). Perhaps most serious is his assertion that species experts are prone to maintain low population estimates "because, in one way or another, we benefit from studying high-profile, 'declining' bird spe-cies." This implies that biologists reject new population estimates to benefit either their careers or their ability to acquire funds. We believe this is untrue. We think (Farmer 2008) fails to recognize—or at least minimizes—that much of the discussion on population size centers around the fact that species experts have low confidence in the "statistically derived" estimates because of limitations in the study design, field methods, or analyses used to generate them. In our view, "thorough" and "statistically derived" estimates are not automatically "credi-ble," as (Farmer 2008) implies, but must be critically evaluated. Moreover, there are philosophical issues related to establishing a population estimate. Many bi-ologists believe that a conservative estimate should be used, especially for species of concern, so as to avoid a Type II error in managing the species. This is the approach used by (Jorgensen et al. 2008), and we concur with it. (Farmer 2008) presents an alternative philosophical view—one that assumes that statistically derived estimates are always accurate estimates and should be used without reservation. We do not believe that such an approach is the correct way to proceed, given the many uncertainties often inherent in range-wide surveys that rely on large-scale extrapolations. Farmer (2008:yy) uses recent data collected on Buff-breasted Sandpiper (Tryngites subruficollis) and Long-billed Curlew (Numenius americanus) to demonstrate how science can "establish institutional research priorities that form the basis for lists of high-priority birds." He goes on to say that in both cases there was a fair amount of "anchoring" (Tversky and Kahneman 1974) and that new results were overly scrutinized and rejected. Although we agree that these studies were critically reviewed, this analy-sis was appropriate and, contrary to what (Farmer 2008) suggests, the results of these studies were used to derive final population esti-mates by (Morrison et al. 2006). (Farmer 2008) discusses three studies in which results of statistically designed surveys were extrapolated to derive estimates of population size. The first study was carried out by (Lanctot et al. 2002) on the wintering grounds in South America. (Lanctot et al. 2004) dismissed the possibility of developing an accurate population estimate because of the presence of weak habitat–bird associations and the inability to identify key habitats in satellite images, which made defining suitable habitat impossible. The second and third studies mentioned by (Farmer 2008) were surveys conducted during the species' migration in Nebraska (Jorgen-sen et al. 2008 and unpubl. data) and in Texas and Louisiana (W. Norling et al. unpubl. data). The Jorgensen study had not been published at the time (Morrison et al. 2006) published their revised population estimate for the species, which forced them to rely on information presented by Jorgensen at a shorebird meeting in Boulder, Colorado (Jorgensen et al. unpubl. data). The Norling et al. study has not been properly peer-reviewed to date; we believe that it must be before the data can be fully considered. (Farmer 2008) neglects this fact and presents the results as on par with the other two studies. Care-ful consideration of these studies indicated several potential limitations, which species experts felt needed to be taken into account when national population estimates were developed (Morrison et al. 2006). Limitations included (1) potential bias from conducting surveys from roads, (2) the fact that sightings during migration indicate that the species is sparsely distributed and can aggregate in large flocks whose detection (or nondetection) could increase or decrease population density estimates, and (3) that no information on turnover rate was available. Although Jorgensen et al. (2006, 2008) evaluated some of these limitations, we do not think that essential natural-history infor-mation was available for them to fully evaluate these factors. In the end, (Jorgensen et al. 2008) generated mean estimates between ~24,000 and 78,960 in 2004 and between 20,000 and 35,000 in 2005, and Norling et al. (unpubl. data) generated population estimates of 229–687 in 1996 and 28,058–84,174 in 1998. (Farmer 2008) mentioned only a range of 34,000–78,000 for (Jorgensen et al. 2008), and "at least 84,000" for Norling et al. (unpubl. data). Given the uncertainties above, and the need for additional information on the species' natural history to make informed decisions on population size, the population size was increased from 15,000 to "30,000 birds or more" (Morrison et al. 2006). This value was at the high end of Jorgensen's original estimates provided at the Boulder meeting (16,000–32,000; Jorgensen et al. unpubl. data) and at the low end of Norling's suggested numbers (28,000–84,000). This middle-of-the-road value is contrary to the way (Farmer 2008) reports how this species' population size estimate was developed (i.e., "anchoring" to prior estimates). Concern over the recently published population estimate for the Long-billed Curlew centers around many of the same factors mentioned above. First, dependence on pre-existing roads made the sampling effort for the range-wide survey of Long-billed Curlews decidedly nonrandom. In our opinion, Stanley and Skagen's (2007) attempt to assess the roadside-bias issue fell far short. Roadside-sampling bias was compounded by extrapolation over vast areas of unsuitable habitat. Another area of concern was survey timing, which likely overestimated the breeding population in some cases and underestimated it in others. Further, (Stanley and Skagen 2007) extrapolated very small numbers of observed birds (172 and 153 Long-billed Curlews in 2004 and 2005, respectively) from a very small area (~0.2% of estimated breeding range surveyed) to a very large geographic area (slightly over 180 million ha each year), yielding widely varying ex-trapolated estimates of 164,515 ± 42,047 (SE) and 109,523 ± 31,060 Long-billed Curlews breeding in the United States in 2004 and 2005, respectively. Such large-scale extrapolations would be fine if the species were evenly or randomly distributed and if transect locations were chosen randomly—but this was not the case. The large confidence intervals (and coefficients of variation above 25%) around the population estimates speaks to the uncertainty in Stanley and Skagen's (2007) population estimates. (Farmer 2008) criticizes (Morrison et al. 2006) for estimating the Long-billed Curlew population at one standard deviation below the mean value estimated by (Stanley and Skagen 2007) but never indicates the value (Morrison et al. 2006) finally gave as their official estimate (i.e., 55,000–123,000, one of the few interval estimates presented). The 55,000 represented expert opinion of the number of birds on the breeding and wintering grounds, and the 123,000 reflected the assumption that there were 23,000 birds in Canada and 100,000 birds in the United States. This 100,000 value represented (1) the approximate midpoint of the two "conservative" (mean minus 1 SE) estimates, (2) is close to the mean estimate for 2005, and (3) is certainly much higher (3–6×) than the prior estimate of 20,000 for the species. Again, we believe that this is not anchoring (Farmer 2008) but, rather, careful consideration of data available for the species and an attempt to take a conservative approach. This approach by no means precludes adoption of some of the higher estimates at some future point as results are further evaluated and more information becomes available. We hope that we have adequately illustrated why (Morrison et al. 2006) used the estimates presented in the studies discussed above but also relied on other sources of information to obtain final population estimates. Thus, although we agree that the approach used by (Morrison et al. 2006) may seem "ad-hoc" (Farmer 2008), it was deemed to be the best approach for generating estimates, given all information available at the time, with the expectation that further evaluation of methods and new results will lead to revised estimates. For better or worse, bird-conservation initiatives have adopted a population-based management–conservation approach. This is unrealistic for shorebirds at present, given the lack of a broad-based program to survey shorebirds at breeding, migration, or wintering areas. Nevertheless, we recognize that population estimates are needed in some cases—for example, to identify wetlands of key importance under various reserve networks; to assess the value of critical areas; and to assess status, set recovery targets, and judge success of recovery programs for endangered or declining species (Morrison et al. 2006). In these cases, careful consideration should be given to where (breeding, migration, or wintering area) and how (sampling design, survey method) a survey is conducted. Recommendations on appropriate methods for various species were outlined in technical documents accompanying the U.S. Shorebird Conservation Plan (Howe et al. 2000). We suggest that exten-sive studies be conducted before or concurrent with formal surveys so that assumptions (e.g., length of stay, roadside survey bias) of the survey design can be evaluated and incorporated into the final estimate. In addition to counting birds, surveyors should collect information on bird–habitat relationships, breeding chronology, timing of migratory movements, and other relevant natural-history information across the species' geographic range. Such efforts will improve current and future survey efforts. We also hope that once survey data are collected, careful con-sideration is given to the quality of the data so that reports do not exceed what is justified. It may not be realistic to assume that species-specific population surveys can be conducted and provide credible population estimates, especially with one or two years of funding. Further, we en-courage more coordination between biologists designing population-monitoring studies and those conducting detailed site-specific studies (of-ten with marked populations). Finally, in view of limited funds, benefits of obtaining population estimates should be weighed against the need for collecting other information (i.e., demographics, causes of declines) or investing in direct conservation. We concur with (Farmer 2008) that we should consider establishing a more transparent, national review process that rigorously evalu-ates new knowledge as it pertains to established population estimates. This could involve establishing formal committees within the U.S. and Canadian Shorebird Plan councils to develop a process for updating estimates and evaluating trend reports and other data that affect ranking factors used to prioritize species. It should be recognized, however, that a system for generating population estimates and updates is already in place, though it does not operate formally. We also argue that the informal committee has done an effective job in collecting and assessing available information on shorebirds in North America (Morrison et al. 2001, 2006). The committee consists of a volunteer panel of experts who assemble information on and provide expertise related to a wide range of species. The updates are currently conducted through a process coordinated by the International Wader Study Group and Wetlands International as part of a worldwide network that assesses waterbird populations at three-year intervals for the Ramsar Convention (Delany and Scott 2006). We strongly disagree that spe-cies experts should be excluded from such committees because of potential conflicts of interest. To sideline them would be to lose invalu-able knowledge that could be used to interpret new and future population estimates. It may be useful to have an independent panel review future estimates once the formal shorebird committees have generated new population estimates. Adoption of some of the protocols suggested by (Farmer 2008) could improve the transparency of how estimates are assessed and facilitate allocation of accuracy ratings to estimates. One potential pitfall of making these changes is the danger that differences of opinion between committee and reviewers could lead to an endless series of discussions about how to treat data. In some ways, it is this sort of disagreement that is at the root of the present discussion. In the end, someone will have to provide an "official" estimate. We are not confident that pro-viding an interval estimate will resolve these issues, because organizations are likely to simply adopt the midpoint. Interval estimates, however, would reflect the uncertainty inherent in all the data available for a species. Given the desirability of adopting a more formal process to ensure that future population updates are made in a timely manner by suitably qualified people, careful considera-tion needs to be given to how this procedure is designed. Whatever the process, we advise documenting decision rules on how estimates were generated so that the process is as transparent as possible. Finally, we emphasize that population estimates are only one of six criteria used to prioritize species (Brown et al. 2001). Acknowledgments.—We thank J. Almeida, S. Brown, A. Farmer, J. Jorgensen, S. Matsuoka, J. McCarty, W. Norling, B. Peterjohn, S. Skagen, L. Wolfenbarger, and one anonymous reviewer for commenting on this manuscript. We also thank W. Norling and J. Jorgensen for allowing us to cite their unpublished data.—RICHARD B. LANCTOT, U.S. Fish and Wildlife Service, Migratory Bird Management Division, 1011 East Tudor Road, MS 201, An-chorage, Alaska 99503, USA (e-mail: richard_lanctot@fws.gov); ALEX HARTMAN, Program in Ecology, Evolution and Con-servation Biology, University of Nevada, Reno, 1000 Valley Road, Reno, Nevada 89512, USA; LEWIS W. ORING, Department of Natural Resources and Environmental Science, University of Nevada, Reno, 1000 Valley Road, Reno, Nevada 89512, USA; and R. I. GUY MORRISON, Canadian Wildlife Service, National Wildlife Re-search Centre, Carleton University, 1125 Colonel By Drive (Raven Road), Ottawa, Ontario K1A OH3, Canada.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.