Electronic Thesis and Dissertation Repository

Degree

Doctor of Philosophy

Program

Biology

Supervisor

Dr. Philip Taylor

2nd Supervisor

Dr. Chris Guglielmo

Co-Supervisor

Abstract

Counts of migrating animals are used to monitor populations, particularly for species that are not well sampled by breeding and wintering surveys. The use of migration counts for population monitoring relies on the assumptions that new individuals are detected each day, and that probability of detecting those individuals remains constant over time. The impact of violating these assumptions on our ability to estimate reliable population trends is not well understood. Further, on a broad spatial scale, our ability to combine data across sites to estimate regional or national trends has been limited by the possibility that trends vary regionally in an unknown way. Using simulated migration count data with known trend, I tested whether sampling effort (daily vs. non-daily sampling) and a temporal change in stopover duration (and thus detection probability) influenced our ability to estimate the known trend. I also tested whether analyzing data as hourly, daily or annual counts, or accounting for random error using analytical techniques, could improve accuracy and precision of estimated trends by reducing or better modeling variation in counts, respectively. Further, using model selection analytical techniques, I tested whether we could detect when trends vary regionally using current or increased number of sampling sites in a region. My findings show that trends can be improved for species with highly variable daily counts by sampling less frequently than daily or by aggregating hourly counts to annual totals. Commonly and rarely detected species were better analyzed as daily counts, collected daily throughout the migration. A linear increase in stopover duration over time biased trends and lead to a high probability of detecting an incorrect trend, which is only improved by both reducing sampling effort and including a covariate for stopover duration in regression analyses. Regional variation in trends can be detected, and increasing the length of the time series was more efficient for improving accuracy and precision of regional trends than increasing the number of sites sampled. Continued advancement of our knowledge of breeding origins and stopover duration of migrants are priorities for the further refinement of trends estimated using migration counts.


Share

COinS