Commentary A Note on Opportunism and Parsimony in Data Collection RICHARD BISCHOF, 1 Norwegian University of Life Sciences, Department of Ecology and Natural Resources, P.O. Box 5003, NO-1432 A ˚ s, Norway ANDREAS ZEDROSSER, Norwegian University of Life Sciences, Department of Ecology and Natural Resources, P.O. Box 5003, NO-1432 A ˚ s, Norway; Department of Integrative Biology and Biodiversity Research, Institute of Wildlife Biology and Game Management, University of Natural Resources and Applied Life Sciences, Vienna, Gregor-Mendel Str. 33, A - 1180 Vienna, Austria SVEN BRUNBERG, Scandinavian Brown Bear Research Project, Kareliusva ¨g 2, S-79498 Orsa, Sweden JON E. SWENSON, Norwegian University of Life Sciences, Department of Ecology and Natural Resources, P.O. Box 5003, NO-1432 A ˚ s, Norway; Norwegian Institute for Nature Research, NO-7485 Trondheim, Norway ABSTRACT Out of precaution, opportunism, and a general tendency towards thoroughness, researchers studying wildlife often collect multiple, sometimes highly correlated measurements or samples. Although such redundancy has its benefits in terms of quality control, increased resolution, and unforeseen future utility, it also comes at a cost if animal welfare (e.g., duration of handling) or time and resource limitation are a concern. Using principle components analysis and bootstrapping, we analyzed sets of morphometric measurements collected on 171 brown bears in Sweden during a long-term monitoring study (1984–2006). We show that of 11 measurements, 7 were so similar in terms of their predictive power for an overall size index that each individual measurement provided little additional information. We argue that when multiple research objectives or data collection goals compete for a limited amount of time or resources, it is advisable to critically evaluate the amount of additional information contributed by extra measurements. We recommend that wildlife researchers look critically at the data they collect not just in terms of quality but also in terms of need. (JOURNAL OF WILDLIFE MANAGEMENT 73(6):1021–1024; 2009) DOI: 10.2193/2008-509 KEY WORDS animal welfare, brown bear, data collection, morphometrics, principal components analysis, Sweden, Ursus arctos. When capturing wildlife during monitoring or other studies, we are often inclined to collect as many types of data (e.g., measurements, samples, observations) on each individual as possible, particularly when collection of these data is relatively noninvasive and capture itself is effort and cost- intensive. We may be motivated by a need for redundancy, quality control, uncertainty about which of a set of measurements or samples will be the most appropriate in subsequent analyses, or a more or less vague hunch that it may be useful in a future study. Occasionally, the motivation for collecting a certain type of data is simply ‘‘Why not?’’ Although the question’s intention is rhetorical, there are answers, including 1) animal welfare concerns (Arnemo et al. 2006, Cattet et al. 2008); even nonintrusive data collection methods prolong handling time, and cumulative fitness effects of manipulation may not be discountable, 2) time constraints; one measurement is often collected at the expense of other measurements, because in most situations handling time (e.g., anesthesia duration) is limited, and 3) a general effort to work efficiently. Fluctuations in financial support, shifts in priorities, changes in personnel, and technological advancement can make longitudinal studies dynamic affairs. Often the number of different types of data collected increases over time, as new procedures are added with ease. On the other hand, most researchers are reluctant to omit a data type that has been collected for a long time, even if its current or future utility is not apparent. When it becomes clear that some reduction or change in processes is in order, how should one decide which types of data to keep and which to drop? The Scandinavian Brown Bear Research Project’s (SBBRP) long-term monitoring program on brown bears (Ursus arctos) in Sweden (e.g., Swenson et al. 1994, Zedrosser et al. 2007a) is no exception from the aforementioned patterns of procedural congestion. During the 24 years of its existence, the capture–mark–recapture study has seen its share of protocol modifications and a proliferation of measurements taken and samples collected. At the time of writing, processing (e.g., marking, data, and sample collection) takes up a substantial portion (30–45 min, depending on surgical procedures for transmitter implantation) of the period during which an anesthetized bear can safely be manipulated (60 min) given the current anesthesia regime. Meanwhile, various new research questions and pilot studies compete for timeslots and manpower allocated for processing anaesthe- tized bears. Furthermore, an ever-growing list of protocol items and manipulations increases the risk of distraction and errors or omissions. Using data from the SBBRP as an example, specifically the collection of morphometric data intended for size and growth estimation, we illustrate how one may identify measurements that add little additional information and, thus, are potential candidates for elimina- tion or replacement. We hope that this example will motivate others working with wildlife to critically evaluate the level of redundancy in their data collection protocols. STUDY AREA Monitoring occurred in 2 study areas located in northern and south-central Sweden. The northern study area (North, 67uN, 18uE) encompassed 12,000 km 2 ; the other site (South, 61uN, 18uE) was 11,500 km 2 . These areas were based on genetically distinct subpopulations that correspond with geographical clusters of bears with no or very little interchange of females (Manel et al. 2004). Both study areas were within the southern, intermediate, and northern boreal vegetation zones (Nordiska ministerra ˚det 1984, Bernes 1 E-mail: richard.bischof@umb.no Bischof et al. N Data Collection Parsimony 1021