Evaluating the power of surface attendance counts to detect long-term trends in populations of crevice-nesting auklets

• Published in: Environmental Monitoring and Assessment
• Main Authors: Renner, Heather M., Reynolds, Joel H., Sims, Michelle, Renner, Martin
• Format: Article in Journal/Newspaper
• Language: English
• Published: 2011
• Subjects: Aethia cristatella; Aethia pusilla; Bering Sea; Hierarchical models; Mixed-effects; Population trends; Seabirds; Statistical power; Zero-inflated
• Online Access: https://hdl.handle.net/1983/0f36ab2c-7026-4d15-b43b-f89b28547ef5; https://research-information.bris.ac.uk/en/publications/0f36ab2c-7026-4d15-b43b-f89b28547ef5; https://doi.org/10.1007/s10661-010-1664-4; http://www.scopus.com/inward/record.url?scp=79959747168&partnerID=8YFLogxK

Power analyses are essential when developing a long-term monitoring program for a target species whose observation is logistically challenging and expensive. These analyses can be complicated when the observations have a complex variance structure reflecting many factors. Crevice-nesting seabirds such as least and crested auklets Aethia pusilla and Aethia cristatella illustrate both this need and these challenges. They are ecosystem indicators for the Bering Sea, a system expected to undergo large changes. Unfortunately, they are difficult to monitor as colonies occur on remote, hard to access islands in the Aleutians and Bering Sea, and nests occur in crevices underground, preventing direct observation. Current monitoring consists of breeding-season counts of auklets standing on the surface of sample plots in the colony; logically, a substantial decline in nesting population guarantees an eventual substantial decline in surface attendants. Yet, it remains debatable whether these highly variable counts can be used to statistically detect biologically relevant declines in the attending population let alone the nesting population. Subsequently, existing monitoring programs vary widely in survey design, effort levels, and daily summary statistics. The power of different survey designs was assessed by simulating observations from a state model developed from 11 years of observations using mixed-effects models and zero-inflated Poisson-lognormal regression. The analyses illustrate the process required for any monitoring program whose observations are described inadequately by standard statistical models. State model development revealed survey design refinements that reduce sampling variation. For least auklets, current sampling efforts provided 90% power to detect annual declines of 11% ("Critically Endangered" using IUCN Red List criteria), 4.5% ("Endangered"), or 2.4% ("Vulnerable") in two, four, or six generations, respectively; crested auklets took a few years longer. Power was more sensitive to number of days ...