Rigby, Elizabeth2018-02-132018-02-132016-11https://hdl.handle.net/11299/193439University of Minnesota Ph.D. dissertation. November 2016. Major: Natural Resources Science and Management. Advisor: Douglas Johnson. 1 computer file (PDF); x, 324 pages.Counts obtained from point count bird surveys can be treated as an index to bird abundance, but imperfect detectability can complicate inferences about abundance. Adjustment analysis methods, including double-observer, replicated counts, removal, and distance sampling methods, have been developed to estimate detection in addition to abundance. These methods require additional information to estimate detection, which may entail added logistical costs or be additional sources of error. It is not clear when or if adjustment methods outperform index methods, or how the benefits of adjustment methods compare to their costs. I simulated point counts of birds, modeling birds spatially as moving within bivariate normal territories, modeling song production as an autocorrelated process, and modeling perceptibility as a logit function of distance to the observer. In Chapter 1, I simulated counts using a test scenario with parameters reflecting surveys of black-throated blue warblers (BTBW, Setophaga caerulescens), analyzed counts using index and adjustment analysis methods, then evaluated and compared the performance of analysis methods. Estimates from index methods underestimated true density of birds (Dp) for all survey types, but were highly correlated with true density. Adjusted estimates from distance sampling and removal analysis methods showed a reduction in bias as compared to index estimates, but had reduced correlation with true density. Adjusted estimates from double-observer analysis methods were nearly unchanged from index estimates. Adjusted estimates from replicated counts analysis methods were susceptible to highly inflated density estimates, resulting in extremely high bias and low correlation with true density. Index methods, while biased, were better correlated with true density and would provide better information about changes in abundance than an adjustment analysis method for the BTBW scenario. If detection is constant and relative abundance is sufficient to meet survey objectives, using an index method is often preferable. For systems with variable detection probability where inference about absolute abundance is necessary to meet objectives, practitioners should select adjustment methods suited to model the source of imperfect detection in their system. Ill-suited adjustment methods will not improve inference and are no more useful than an index. In Chapter 2, I used the model to simulate counts for scenarios with high or low availability and high or low perceptibility. I also included scenarios where abundance was confounded with perceptibility, and scenarios where they were independent. I then analyzed count data using index methods and adjustment methods. Although index methods were biased and only had a strong correlation with true density when detectability was high, adjustment methods generally did not offer an improvement. As compared to index methods, adjustment method performance ranged from far worse (replicated counts), to no added value (double-observer) to moderate improvement (in bias only, for removal and distance sampling in specific scenarios). Practitioners should carefully consider the sources of variation in detection probability in their system. If detection components are unknown or known to be variable, I advise practitioners to perform a pilot study to estimate detection components. Additionally, practitioners should standardize their methods to increase availability and perceptibility in their surveys and to lower the variation in these detection components. In Chapter 3, I conducted simulated bird surveys using recorded bird songs to assess factors affecting detection probability in grassland bird point counts. I used mixed effects logistic regression models to estimate factors affecting detection probability and to estimate and visualize the variation in the area around the observer where birds can be perceived (the perceptible area). I conducted simulated surveys with 8926 binary opportunities for detection in Minnesota grasslands in 2011 and 2012. Species, distance to the observer, wind speed and direction, observer, and density of vegetation all affected detection of recorded bird songs. Species had a strong effect; the size of the predicted perceptible area around the observer differed by more than 10-fold among species. Wind also had a strong effect on detection. As wind speed increased, probability of detection downwind of the observer was reduced and the perceptible area around the observer became smaller and more asymmetrical. The effective distance at which an observer is more likely to detect a bird than to not detect it may differ among species and angles to the wind, even within the same survey. I recommend using fixed-radius counts for bird surveys in grasslands and reducing the variation in detection probability by standardizing surveys across wind conditions.enabundancedetectionmodelingsurveyThesis or Dissertation