Estimating Nest Density When Detectability Is Incomplete: Variation In Nest Attendance And Response To Disturbance By Western Meadowlarks

Authors

Matthew D. Giovanni, University of Nebraska - Lincoln
Max Post van der Burg, University of Nebraska-LincolnFollow
Lars C. Anderson, University of Nebraska-LincolnFollow
Larkin A. Powell, University of Nebraska-LincolnFollow
Walter H. Schacht, University of Nebraska - LincolnFollow
Andrew J. Tyre, University of Nebraska at LincolnFollow

Document Type

Article

Date of this Version

2011

Citation

The Condor 113(1):223–232; DOI: 10.1525/cond.2011.100076

Comments

Copyright The Cooper Ornithological Society 2011

Abstract

Researchers commonly model nest density as a function of ecological variables, but nests, like birds, can be undetected while present. In the Nebraska Sandhills in 2007, we used the rope-drag method on previously located Western Meadowlark (Sturnella neglecta) nests to model nest-detection error and nest density. Detecting nests by rope dragging (commonly used for nests of grassland birds) is conditional on two primary sources of nest availability, adults attending nests and adults flushing from nests in response to disturbance from the rope, the behavioral cue necessary for nest detection. On the basis of our trials with rope dragging, the probability of adults attending nests was 0.46. The probability of trials eliciting flight responses from adults was 0.19, indicating that observed estimates of nest density were subject to detection error and negatively biased. We compared our ability to detect differences in nest densities between two habitat types with a statistical power analysis, using (1) t-tests with estimates of apparent, uncorrected density and (2) a Bayesian hierarchical model incorporating nest detection error. Our power analysis indicates that accounting for detection error in a hierarchical model increased statistical power for revealing differences in nest densities by habitat type. We discuss mechanisms underlying detection error for nests of grassland birds and make recommendations for sampling and modeling to maximize the reality of models for nest detectability and density.