Any raw count of individual birds, bats, etc. that die from colliding with some structure is a conservative estimate of the actual mortality at that structure. Here’s why:
1) I could just plain not see the carcass during my check. Maybe it’s hidden by some vegetation, someone has kicked it under a bush, or the animal landed outside my search zone, say 10m from the building as opposed to the 2-3m zone I’m checking. Thanks to some of Dan Klem’s excellent work, we also have an idea that about 50% of the birds that strike windows but manage to fly away only temporarily stunned will also die, and potentially at quite a distance from where the collision took place. In fact, many bird deaths attributed to cats may actually be made possible by stunned birds that have hit windows.
2) Some predator has scavenged a carcass before I get there to record it. Cats, raccoons, opossums, skunks – any of these common predators would be happy to snatch a dead bird lying conspicuously on a sidewalk near a building.
3) Some human – perhaps someone doing a study like mine – has “scavenged” a carcass before I get there to record it. Groundskeeping crews do occasionally remove dead birds from sidewalks as well.
All of this means that if I count 50 dead birds around a building over the course of an observation period, that 50 is some fraction of the real total number of birds that died from colliding with that building during that period. Okay, but is 50 a large fraction or a small one?
I could calculate observer detection rate by having a partner place dead birds around the building and then going through my normal routine to determine how many I miss. Let’s say that I have a high detection rate of 95%, which is likely for my building because I have easy visual access to the entire perimeter. I would then modify my estimate of 50 by adding 5% (2.5) to my raw count. OK, so now we think that 52.5 birds died.
We can also test removal rate by putting out carcasses, chicken wings, etc. and seeing how many are removed over a set period of time. Let’s say that 25% are removed. We now add another 25% (13.125) to the estimate and conclude that 66 individuals died at the location, even though only 50 were actually counted.
But there are some problems with making these estimates. The first is that they greatly increase the complexity of gathering the data.
I’ve actually been a long-time opponent of Herculean efforts to quantify “removal rate” of carcasses around things. We invest a huge amount of time, money, and effort to do things like improve our accountability in detection a few percentage points. The real story is not whether 50 or 66 birds died by colliding with some structure, it’s that a lot of birds died by colliding with the structure. That’s the message that needs to get out; estimating that there were 16 more casualties than actually counted isn’t going to change anyone’s mind about the severity of the problem. What’s more, in cases in which detectability is low and estimating things like removal rates bumps up the estimate a great deal (say, 10 carcasses found but 100 estimated deaths), the public can become dismissive of the problem because they’re skeptical of the means used to derive the estimate.
The bigger problem, however, lies in attaching a static rate of removal to a raw count. Determining “the” scavenging rate is not only extremely labor and time-intensive, it’s silly. Scavenging rates likely vary over time at a location. It only takes one cat or crow or raccoon to learn that dead birds appear at the base of some building every night and the scavenging rate can change dramatically over a short time frame. Especially given that we study scavenging rate by placing food around the base of buildings, we need to be sensitive to how our own actions as researchers could be teaching scavengers about a reliable food source.
The solution? I advocate a healthy dose of common sense. If you check your building daily, you’re a lot less likely to have carcasses removed before you get there. If you leave carcasses in place after you identify them, you can get a sense for the removal rate by determining how long the carcass lasts. Right now, I’ve had three carcasses present at the NRC – all still plainly visible to me – that have been in place for more than 1 week. This tells me that scavenging rate is probably quite low.
Lastly, recognize that the number is not the most important part of the story when it comes to collision mortality. I’m far more interested in the species and life history composition among the dead birds I find. For example, abundant residents like starlings and House Sparrows are almost never counted among the casualties at a site like the NRC. In contrast, the great majority of victims are long distance migrants. So far this fall I’ve counted 7 victims of which 6 were Neotropical migrants – the 7th was a migrant too, a Grasshopper Sparrow. This is where I think the most interesting questions lie regarding collision mortality. The risk of collision varies by species, and this may mean that some species are disproportionately affected by window collisions.