One day, a few years ago, I was drinking with some fellow biologists in “Seminar Room A” opposite the Natural History Museum in Helsinki. The Sibelius Academy is next door to the museum, so we shared the bar with musicians. Once of the musicians came up to us with a plastic bag, and gave it to us saying that they thought we’d know what to do with it. In the bag was a dead gull.
Well, the other biologists knew what to do with it: it had a band on it so they took that off its leg, and dumped the body. The band was sent to the local banding centre, so they would know that the bird had died. And another datum was added to the store of human knowledge.
But what’s the point of banding birds and collected the bands after the bird’s dead? Because it helps us learn about how their populations are faring. We can learn a lot about survival from this sort of data, particularly if it’s combined with observations of birds when they’re alive. But this data is quite messy, so it needs some seriously fun statistics to get anything out of it.
The basic problem is that we don’t see every bird ever year. So, if we didn’t see a bird in one year, how do we know it was still alive? We can’t be sure, but we might be able to put a probability on it. It could be that it was dead, or it was alive but not seen. We can estimate the probability of not seeing the bird from observations of other birds. For example, if we look at all of the birds seen in the previous and subsequent years (i.e. ones we know exist and survived), then the proportion of those that we see is an estimate of the probability of seeing the bird if it was alive.
Once we know the probability of not seeing a bird, we can estimate the probability of it surviving (because we can write down the probabilities of death and of not being seen each year and sum these up to get the probability of never seeing the bird again. Seriously, it does work). Hence, we can estimate the probability of survival. This is helped if we know when some of the birds have died, because that tells us a lot more about the probability of dying and being recovered.
Now, in practice it gets more complicated, because we want to estimate survival that varies, e.g. because of the poor weather we’ve been having or because survival changes with age. Statistics has some standard approaches to do this, but they make it more complicated to estimate anything. So, either one needs to write your own computer code to do the analysis (which takes time, most biologist would also have to start of by learning to programme). Or someone has to write a programme that can take some straightforward inputs and do the complicated stuff itself.
All of which brings us to this video, expounding on the virtues of a computer package to do just that:
The package is based on an amazing piece of statistical software called R, which has taking over as the tool to use. This means biologists only have to learn to use R and they have a lot of tools at their fingertips for them to abuse.
The applications of these methods go beyond ecology. In the video they mention estimating historical survival of humans, but the methods have also been used to estimate HIV infection, and the numbers of political killings in Kosovo. If we have several incomplete lists of individuals in a population, these methods can help us estimate how incomplete they are, and if they are ordered in time we can estimate when individuals drop out, either through death or just moving away.
So, if you see a dead banded bird, pop over to the nearest ornithologist-filled pub and give them the band, or the whole bird if you really want to make their day.