Citizen science, which uses volunteer observers in research, is fast becoming standard practice in ecology. In this thesis, I begin with an essay reviewing the benefits and limitations of citizen science, and then measure the influence of several forms of observer error that might bias ornithological citizen science. Using an internet-based survey, I first found that observer skill level can predict the nature of false-positive detections, where self-identified experts tend to falsely detect more rare species and moderately-skilled observers tend to falsely detect more common species. I also found that overconfidence is widespread among all skill levels, and hence that observer confidence is an unreliable indication of data quality. Using existing North American databases, I then found that older observers tend to detect fewer birds than younger observers -- especially if the birds' peak call frequencies exceed 6 kHz -- and that published long-term population trend estimates and high-pitched (>= 6 kHz) peak bird vocalization frequencies are negatively correlated. Taken together, these data suggest that both hearing loss and other sensory changes might be negatively biasing long-term trend estimates. In the next chapter, I measured how observer experience can bias detection data. In solitary observers, I found that detections tend to increase over the first 5 years of service (e.g. learning effects), after which they decline consistently (e.g. observer senescence). Conversely, among survey groups that may be motivated to exceed a previous year's species count, I found that species richness tends to increase consistently with consecutive survey years. In this case, individual sensory deficits may be offset by group participation. Lastly, I re-evaluated the established assumption that the quality of new volunteers on North American Breeding Bird Survey routes is increasing over time. I showed that the existing measure of “quality” ignores variable lengths of observer service, and that, after accounting for this variable, “quality” is unchanging. Throughout this thesis, I also show how generalized additive mixed models can address these biases statistically. My findings offer new opportunities to improve the accuracy and relevance of citizen science, and by extension, the effectiveness of wildlife conservation and management.

This study constructed habitat suitability criteria for fishes of the South Fork of the Shenandoah River, in Virginia. The criteria will be used in an IFIM study to produce estimates of the discharge required by fishes in the South Fork. Chi-square tests were used to evaluate whether criteria described habitat use to a statistically significant degree. Secondly, chi-square tests were used to test transferability. The criteria described the habitat use of seven taxa commonly found in the South Fork to a statistically significant degree. Habitat criteria for two taxa did not describe their habitat use to a statistically significant degree. One set of criteria from the North Fork of the Shenandoah transferred to the fish observed in the South Fork. Secondly, this paper examined observer effects of underwater observation. It was of interest to explore how observer effects influenced habitat suitability criteria.