Research design considerations to ensure detection of all species in an avian community Maggi Sliwinski 1 *, Larkin Powell 1 , Nicola Koper 2 , Matthew Giovanni 3,4 and Walter Schacht 5 1 School of Natural Resources, University of Nebraska-Lincoln, 3310 Holdrege Street, Lincoln, NE 68583, USA; 2 Natural Resources Institute, University of Manitoba, 303-70 Dysart Road, Winnipeg, MB R3T 2M6, Canada; 3 School of Natural Resources, Department of Agronomy and Horticulture, University of Nebraska-Lincoln, 3310 Holdrege Street, Lincoln, NE 68583, USA; 4 Stantec Inc., 209 Commerce Pkwy, Cottage Grove, WI 53527, USA; 5 Department of Agronomy and Horticulture, University of Nebraska-Lincoln, 312 Keim Hall, Lincoln, NE 68583, USA Summary 1. Recent advances in the estimation of species richness from count data have allowed avian ecologists to incor- porate incomplete detectability of species when comparing richness across space or time. Raw counts from single or repeated visits to sample point(s) are nonetheless still used for assessing community composition, and the fail- ure to account for detectability when making these evaluations may lead to incorrect inferences about the com- munity. 2. We estimated detection probabilities (P) for a suite of bird species and used these detection probabilities to determine the minimum number of visits at a single point and the minimum number of points in a grid required to confidently (≥90%) detect the full community of birds for rare, moderately rare, and common species. We used occupancy modelling to estimate the detection probabilities for species from two study sites in Nebraska and Sas- katchewan. 3. Some common or highly detectable species were confidently detected in a single visit to a point, whereas others with low detection probabilities (P <020) required more than ten visits to be confidently detected at a point. The grid size required to detect a species in an area varied from a single point for a common highly detectable spe- cies to over 30 points for a rare species with low detectability. 4. Detection probabilities of the least detectable species in a study area can be used to determine the number of visits to a single point or the number of points in a grid to be confident that the full community is detected. Biolo- gists can conclude that a species is most likely absent from the community if it remains undetected using the appropriate sampling effort. Key-words: community, grassland birds, occupancy modelling, research design, sampling effort, study design Introduction Species richness and diversity are widely studied phenomena in ecology (Bestelmeyer, Miller & Wiens 2003; K ery et al. 2009) and are important because they provide a useful way of com- paring different sites and management strategies for groups of birds or other wildlife, rather than single species (e.g. Fritcher, Rumble & Flake 2004; Johnson et al. 2011). Biologists com- monly use statistical methods to estimate detection probabili- ties and adjust density or abundance estimates for individual species (e.g. Buckland et al. 1993; Farnsworth et al. 2002; Royle 2004; Efford & Dawson 2009). More recently, capture- recapture and occupancy methods have been modified to account for detection during assessments of richness and diver- sity of communities (Hines et al. 1999; Dorazio & Royle 2005). However, richness is not useful to evaluate the composition of species within communities, as it is impossible to describe the species that were not observed. Is a threatened species pre- sent at a site (Chad es et al. 2008)? Has the guild of grassland or forest obligates responded after management (e.g. MacNally 1997)? Studies designed to answer these questions should maxi- mize confidence in detecting or confirming the absence of rare or elusive species. In fact, knowing which species are missing from a community can be as important as knowing the species that are present (P€ artel, Szava-Kovats & Zobel 2011) because knowledge of the absent species provides insights for assessing management for functional ecosystems. Thus, it is important to design studies that account for detection probabilities of rare or less-detectable species. Point counts are a common method of sampling bird abun- dances and communities (Diefenbach, Brauning & Mattice 2003). Based on the law of diminishing returns, the number of visits suggested for studies using point counts is between two and five, although these suggestions are based mostly on forest bird studies (Ralph, Sauer & Droege 1995; Grant, Madden & Berkey 2004; Field, Tyre & Possingham 2005; Purcell, Mori & *Correspondence author. E-mail: maggi.sliwinski@gmail.com © 2015 The Authors. Methods in Ecology and Evolution © 2015 British Ecological Society Methods in Ecology and Evolution 2015 doi: 10.1111/2041-210X.12506