They're staying how long? Methods of and complications in determining stopover estimates using banding data
They're staying how long? Methods of and complications in determining stopover estimates using banding data.
SARA R. MORRIS and H. DAVID SHEETS, Canisius Coll., Buffalo, NY.
The importance of stopover sites for migrant birds is recognized widely and thus conservation of these sites is receiving substantial attention. However, many questions still remain about stopover ecology, how migrants use various stopover sites, and how to compare the use and quality of different stopover habitats and/or sites. Stopover length, the amount of time migrants spend at stopover sites, is one potential statistic that can be used to compare sites. The simplest, and generally most conservative, estimate of stopover length is minimum stopover. Because only a few birds are used in calculating the minimum stopover estimate, more complex open population models have been recommended to estimate stopover length. Unfortunately, these models require large numbers of migrants captured per day to effectively estimate stopover duration. We have investigated three methods of reducing the amount of data needed to use these models: pooling (combining several days into a single new interval), multiple-day constancy (MDC, holding parameters constant for several days), and truncation (excluding periods of limited captures at the beginning and/or end of the banding season). Pooling of our banding data resulted in an upward bias in stopover estimates, and thus should be avoided. We instead recommend MDC to reduce the number of estimates parameters and therefore effectively model banding data using open population models. Truncation resulted in modest increases in the utility of these models, but resulted in a slight upward bias in minimum stopover length and an increase in the standard error in model-based stopover estimates. Our results suggest that there are still difficulties in deriving accurate stopover length estimates, but relative stopover length for comparison among species or locations does seem possible with existing methods.