Continent-wide analysis of how urbanization affects bird-window collision mortality in North America

I’m late on this, but I figure late is better than never. Recently a paper which I collaborated on (Continent-wide analysis of how urbanization affects bird-window collision mortality in North America) was published in Biological Conservation. The first and second authors (and organizers of the entire project) put together a press release and asked the other collaborators to disseminate it.

The short version is that we were looking at the factors that lead to birds killing themselves hitting windows. It’s not always appreciated just how many birds kill themselves hitting windows because birds tend to be small and their carcasses are both easily overlooked and easily removed or eaten whole by scavengers.

What we found was that building size is linked to bird mortality. This isn’t surprising since a larger building has more killing surfaces on it. What is more interesting is that a building of a given size was more dangerous to birds if it was in a region of low urbanization. (It also helped if the building itself was in a more “natural” area with grass or trees – again, not surprising, since these areas attract birds.) This may suggest that birds in urbanized areas behave differently and are perhaps more cautious around buildings. It may also suggest that certain sorts of birds may be more at risk and avoid urban areas. For instance, if migratory birds are more likely to kill themselves on windows (which seems reasonable, since they haven’t learned the area) and are also more likely to avoid cities then rural areas should see higher numbers of fatal bird strikes on buildings. Another hypothesis (and none of these are mutually exclusive) is more macabre. It posits that urban areas are not areas in which birds are less likely to die but that they are areas where birds have so many places to kill themselves that no one building becomes Bird Murder Central. Meanwhile, in rural areas all the birds that are going to try to fly through windows survive until they hit the one set of available windows, concentrating bird-murder around single buildings.

Anyway, all of that and more is available through the press release package and the paper itself. (Although the paper is paywalled, and some of you may not be able to access it1.) What I think I can add in this post is what it was like for me to be part of this project.

I came into this project because a friend of mine mentioned to me that they knew of someone at Duke who was doing it. I’m going to guess that this person was Nicolette Cagle, who is listed right after me (because of alphabetization and not geography). This friend then passed on some information and I contacted Steve Hager and offered to participate. At the time I assumed that I would be able to find several undergraduate students who would be interested in participating in a study this large and that I would be able to mostly supervise the project.

I began the project by selecting sites to survey (which included our tallest dormitory building, the tallest building on campus, and also two houses used as office buildings). I had to alter the sites twice. First, I selected a site that I was told I would have access to but, when the time came, I was told I couldn’t access the back of the building anymore. This building was actually just off campus and so I relocated to a building controlled by the university. Second, the buildings all needed to be 100 meters apart. Initially I selected a building that (unusually for our campus) had a large set of plate glass windows. I hoped by selecting this building that we could get enough variation in window size to try to get some real sense for how important this variable is. However, when I checked the measurements one last time before we started I realized that I needed 100 meters between the outer walls of each building, not the center of each building, and so I had to switch to a different building with smaller windows. As a result, we didn’t survey any buildings with particularly large windows on our campus.

Once the semester started I started recruiting students. I had thought I would get several volunteers, but I didn’t. Instead, one student (who I will not name without specific permission) volunteered. This particular student was an exceptionally dependable person and I sometimes wonder if they initially volunteered simply to be helpful.

At this point the student and I taught ourselves the sampling protocols and did some dry-runs. There was a protocol for one worker and one for two workers. We practiced both, and ended up needing both, since on several days we weren’t both available. The first thing we needed to do was clean up any dead birds already present around the buildings so they wouldn’t appear as fresh kills in the surveys. Most of the carcasses we found in clean-up were thoroughly rotten anyway. We then spent 21 consecutive days checking each of our six buildings at around 2-3 pm. When you come on to campus on the weekends looking for dead birds this stirs up a lot of excitement. One day I was stopped by another faculty member who didn’t recognize me and was sure I was some crazy person up to no good. Another Saturday an enthusiastic student took me on the random dead animal tour of campus (“And there’s a dead cat that got hit by a car and crawled under the dumpster over here….”).

All in all we found no freshly-dead birds. At some point a half-rotten starling showed up in our sample zone but I am pretty sure it blew off a roof because nothing rots that fast. This is in marked contrast to some other sites where dead birds showed up regularly. It was also somewhat disappointing in that while “no fatalities” is good data and good for birds it feels a lot like we found nothing.

Having turned in the data we got a request to collect data at the same time the following year. I didn’t do this. I feel sort of bad about this, but the reality was that I had assumed I would be directing a relatively large group of students, none of whom would need to take more than half the shifts. Instead, I ended up taking a lot of shifts myself. I just didn’t (and still don’t) have that sort of time in the afternoons when a lot of the classes I teach run. I should point out that I’ve been nothing but happy about the way Steve Hager, who has been the primary contact person for the study, has handled everything. From that angle my work on this project was really good. However, if I were to re-do this experience I would change two things.

First, I would line up student workers in advance. This would have been an easier study, and would have been done better (two surveyors per site every day), with more students. We could have managed two years, and I could have done more of the data analysis on my end instead of passing it off to Steve, with more workers.

Second, I would have put more effort in the summer into making sure I had my sites right early on and done some of the satellite image work early on.

It’s worth thinking about the fact that a lot of these issues were rookie issues. I wasn’t running a lab with seven undergraduates back then and I didn’t have a good handle on how to recruit undergraduate researchers, how to vet undergraduate researchers (this is linked to recruiting, because some recruitment techniques get you lots of volunteers at the expense of quality), and what you could realistically expect undergraduates to handle without your direct supervision. I lucked out on this last one because I had a really solid undergraduate assistant but I have now supervised plenty of really enthusiastic, hard working students who would not have been able to take that task on and get it done correctly until they had more research experience. I also hadn’t done some of the analysis types before and assumed they would be straightforward. These days I generally make what I think is a reasonable estimate and then assume that some unforeseen issue will triple the time required.

However, I think that ultimately this sort of collaborative venture can be an excellent way to do some research at a primarily-teaching institute. By splitting up the data collection load no one person is stuck holding up the project forever (unlike some of my projects, which are held up until I can personally find time to go to a field site that is not now very close to where I live). This one worked best when at least one undergraduate was involved (since the best sampling protocol involved two people) which also gives the undergraduate student some real research experience.