Avian mortality at communication towers in the United States and Canada: which species, how many, and where?

Abstract

Birds migrating to and from breeding grounds in the United States and Canada are killed by the millions in collisions with lighted towers and their guy wires. Avian mortality at towers is highly variable across species, and the importance to each population depends on its size and trajectory. Building on our previous estimate of avian mortality at communication towers, we calculated mortality by species and by regions. To do this, we constructed a database of mortality by species at towers from available records and calculated the mean proportion of each species killed at towers within aggregated Bird Conservation Regions. These proportions were combined with mortality estimates that we previously calculated for those regions. We then compared our estimated bird mortality rates to the estimated populations of these species in the United States and Canada. Neotropical migrants suffer the greatest mortality; 97.4% of birds killed are passerines, mostly warblers (Parulidae, 58.4%), vireos (Vireonidae, 13.4%), thrushes (Turdidae, 7.7%), and sparrows (Emberizidae, 5.8%). Thirteen birds of conservation concern in the United States or Canada suffer annual mortality of 1–9% of their estimated total population. Of these, estimated annual mortality is >2% for Yellow Rail (Coturnicops noveboracensis), Swainson’s Warbler (Limnothlypis swainsonii), Pied-billed Grebe (Podilymbus podiceps), Bay-breasted Warbler (Setophaga castanea), Golden-winged Warbler (Vermivora chrysoptera), Worm-eating Warbler (Helmitheros vermivorum), Prairie Warbler (Setophaga discolor), and Ovenbird (Seiurus aurocapilla). Avian mortality from anthropogenic sources is almost always reported in the aggregate (“number of birds killed”), which cannot detect the species-level effects necessary to make conservation assessments. Our approach to per species estimates could be undertaken for other sources of chronic anthropogenic mortality.

Introduction

Avian mortality from collisions with human-made structures is an issue of ongoing conservation concern (Drewitt and Langston, 2008, Longcore et al., 2008, Longcore et al., 2012, Manville, 2005, Manville, 2009). Mortality at communication towers has generated long-term studies at single sites (e.g., Crawford and Engstrom, 2001, Kemper, 1996), many incidental observations (Avery et al., 1980, Kerlinger, 2000, Trapp, 1998, Weir, 1976), and comparative studies across towers in several regions (Gehring et al., 2009, Johnston and Haines, 1957, Morris et al., 2003, Seets and Bohlen, 1977). The U.S. Fish and Wildlife Service (USFWS) has estimated avian mortality from communication towers at 4–5 million birds per year and released guidelines designed to minimize such mortality (U.S. Fish and Wildlife Service, 2000). We derived an updated estimate of 6.8 million birds per year with a tower height–mortality regression and the characteristics of >70,000 towers demonstrating that mortality increases predictably with tower height (Longcore et al., 2012). The USFWS has made recommendations to the Federal Communications Commission (FCC) on how to further reduce incidental take (Manville, 2007) and Environment Canada is currently assessing incidental mortality of migratory bird species at towers as part of a comprehensive effort to address all sources of incidental mortality.

Avian mortality at communication towers occurs most frequently when nocturnal migrants are attracted to tower lights. Birds that enter the zone of influence of lights then circle the towers and are at risk of death from exhaustion, collision with the tower and its guy wires, and collisions with each other (Gauthreaux and Belser, 2006). This usually occurs in inclement weather when other navigational cues are obscured and around the time of passage of cold fronts that drive birds down to altitudes where they are more likely to encounter towers and their lights (Avery et al., 1976).

Estimates of mortality for individual species are needed to assess biological significance of avian mortality at communication towers (Longcore et al., 2005, Longcore et al., 2012). The term biological significance is not formally defined in the context of environmental impact assessment, but a logical definition might be that a biologically significant impact would adversely affect a species or its habitat and could be expected to affect the population growth or stability of the species and influence the population’s long-term viability. Others have concluded that what constitutes a biologically significant population change is not easy to define (Reed and Blaustein, 1997). It may be important to understand the degree to which population growth is suppressed by a mortality source (Loss et al., 2012). Any change in a population has some biological consequence to other species, and therefore any population decline could be important and determining whether it is “significant” may be arbitrary. Biological significance in this context should not be confused with a statistically significant trend in a biological variable. Although statistical significance may influence the judgment about whether an impact is biologically significant, it is not a prerequisite.

To evaluate the biological significance of mortality, species or populations should be the unit of analysis in most instances. For example, barbed wire fences kill a relatively small proportion of birds compared with such hazards as windows and free-roaming cats, but barbed wire fences are a biologically significant source of mortality for Whooping Cranes (Grus americana), an endangered species (Allen and Ramirez, 1990). Higher taxonomic groups, such as families or even guilds that cut across taxonomic groups, may be the appropriate unit of analysis if something is known about the conservation status of the units as a whole. For example, oil pits (pits where oil producers dispose of waste fluids) kill an estimated 500,000–1,000,000 birds per year (Trail, 2006). This raw number can be interpreted with the knowledge that 162 species have been killed in oil pits, of which 63% were ground-feeding birds, including several species of conservation concern (Trail, 2006). Mortality at communication towers, up to this point, has been a conservation issue because the species predominantly killed at towers are Neotropical migratory songbirds, which are of conservation concern as a group. Beyond this general observation, however, only crude estimates have been made of the species composition of the millions of birds killed annually at communication towers (Arnold and Zink, 2011, Shire et al., 2000).

Arnold and Zink (2011) performed an analysis of the proportion of birds killed at towers and regressed the relative risk of collision against 30-year population trends calculated from Breeding Bird Survey data. They concluded from this regression that tower mortality had no discernible effect on population trajectories and claimed that their methods had statistical power to detect as little as a 4.1% contribution to the observed trends. Arnold and Zink (2011) have been criticized for their methods (Schaub et al., 2011) and for the scope of their inferences (Klem et al., 2012), and we have several additional concerns about their analysis. First, they used a flawed secondary data source (Shire et al., 2000) as their raw data for tower mortality. Shire et al. (2000) included a single list of the number of each species killed at towers, which they obtained by summing the results from 47 towers for which they found data. This unpublished report, however, did not exhaustively cover the literature available at the time, contained tabulation errors, and is now dated. It also presents raw sums, which are heavily influenced by the length of the various studies and do not account for regional variation in mortality. Arnold and Zink (2011) identified species that were killed more or less frequently than expected based on population sizes, but because they failed to obtain the primary sources, their mortality proportions contain the errors inherent in the Shire et al. (2000) report and do not account for regional variation or provide a mechanism to combine studies of different lengths in a way that keeps large datasets from overwhelming smaller ones. Failing to account for geographic variability leads to the unrealistic assumption that each tower in North America kills exactly the same proportion of each species of bird. Furthermore, we are unconvinced that impacts of collision mortality would be seen across hundreds of species in the manner assumed by Arnold and Zink (2011). Rather, it is much more likely that tower mortality represents one of an array of stressors affecting the population trajectories of a more limited number of species. In short, we doubt the ability of their method to definitively identify the cumulative impacts of avian mortality at towers and buildings, and make no such sweeping claim for the approach we develop here.

To better understand the effects of avian mortality at communication towers, we combine our previous geographically stratified estimate of total avian mortality at communication towers (Longcore et al., 2012) with estimates of the proportion of each bird species killed within different regions to develop geographically explicit tallies of avian mortality at communication towers by species. We chose geographically specific estimates because avian mortality and tower height vary regionally, and this additional information should be incorporated into any estimates. We then compare these per species mortality estimates with population estimates for these species to gauge the magnitude of this mortality source on a species-by-species basis.