Casualty #357 on this project was not a magnum, it was more of a micro: this apparent HY male Ruby-throated Hummingbird at the southeast alcove.
I found a pre-scavenged Mourning Dove at the main north entrance today. This is another great example of the difference between scavenging and removal.
The carcass was scavenged even before I found it. The whole point of working to determine scavenging rate is a matter of detectability, i.e., that our raw counts will always underestimate mortality because some carcasses are scavenged before they can be found. But scavenging isn’t the issue per se, removal is. If the carcass is scavenged but not completely removed, then it is still detectable. Therefore, the act of scavenging was irrelevant to my ability to detect the carcass, and thus the event.
We can do some field trials with known specimens and determine that our observers detect, for example, 95% of the carcasses in their search area. We can also do removal trials by setting out specimens and determining what proportion of them are removed in a set period of time. For example, let’s say 25%.
If we do some window-collision monitoring and find 10 dead birds at a building, we can modify our estimate according to our imperfect detection rate: 10.00/0.95 = 10.53. That’s the detection-adjusted estimate of mortality. The removal rate of 0.25 suggests that another 2.5 carcasses were removed before they could be detected (or at least before 95% of them could be detected). Removal rate bias then bumps our estimate from the raw count of 10 to an adjusted count of 12.5. Factoring in the detection rate on that estimate increases our adjusted mortality to 13.16 from the raw count of 10 carcasses we actually found.
This matters naught if our objective is to highlight the total number of casualties. It’s is not the case that public outcry to help solve the problem of window collision mortality with be louder for 13.16 casualties than it is for 10. For comparisons among studies, however, it is important to have this information presented and standardized. If, for example, two sites are compared according to their respective landscaping or lighting influence on mortality, that analysis would be corrupted if there was an unaccounted stark difference in removal rate between the two sites. So it is important to quantify rates of detection and removal in monitoring so that our efforts can be of greatest use.
In this long-term monitoring project, I have approached removal rate differently. I leave some carcasses in place to determine for how long they are detectable. Some are removed before I ever find them; some are immediately scavenged but not removed so I can detect them for weeks after the event. Some are never removed and their feathers and bones can still be detected months afterwards. On average, carcasses in my my study last about 10 days on the ground, and I conduct my surveys every 1–2 days. This means that, on average, I have 5–10 opportunities to detect a carcass before it is removed.
That’s pretty good.