Way back…when I was undergraduate…in order to graduate, I had to complete an internship. You’re never going to believe this but that internship was to survey deer on the Rachel Carson National Wildlife Refuge. I was an environment science major, and this was my initiation into the deer mafia. We all know how that story ends.
But it began innocently enough running spot light counts where I met the strange little bird known as the American woodcock (Scolopax minor). They remain one of my favorite birds, vying for top spot with the golden-crowned kinglet (Regulus satrapa). Besides spot light counts, we conducted line transects. Line transects are routes located randomly within some area to be surveyed. They provide the data for distance sampling methods to estimate population density.
In distance sampling, observers perform standardized surveys along a series of line transects searching for the critter of interest. In our case, it was deer (of course). For each animal detected, the distance from the established line is recorded. The fundamental assumption of distance sampling is that ALL critters on and near the line are detected.
Animals further from the line are harder to detect by the observer and will be missed. Distance sampling models this probability of detection to estimate the population. The detection function (blue line in figure below) is key to distance sampling.
Figure 1. Distance sampling models how the probability of detection decreases the further an animal is from the observer (transect line where distance = 0).
Distance sampling requires that transect lines to be located randomly with respect to the distribution of animals, animals are detected at their initial location (i.e. animals do not move in response to observer), distances are measured exactly, and observations are independent events (i.e. flushing one animal does not cause another to flush). If these assumptions are met, distance data can be used to estimate the detection function.
While there is calculus involved, it seems simple enough. That is until you are actually in the woods walking one of these RANDOMLY placed transects. Besides woodcocks, which I remember fondly, I also remember walking those transects. Let’s just say I do NOT remember those as fondly. Lines are drawn on a map. It does not matter if that line cuts through a river, a salt marsh, a greenbrier thicket, a downed tree, or any other obstacle. Randomized line transects are the bane of fieldwork. They crush souls.
In order to avoid this soul crushing, some people suggested placing transects along roads. That’s a big no-no. Animals, like deer, are known to actively avoid roads and roads certainly are not randomly placed on the landscape. In fact, below is a plot of deer locations on our Susquehannock State Forest study areas relative to distance from the nearest road ½ hour before sunset to 3 hours after sunset. Keep in mind nearly all these roads are unpaved forest roads. Notice how there are much fewer locations less than 44 yards (40 m) from a road.
Theoretically, one could take this information about the distribution of radio-collared deer to correct distance sampling estimates obtained from road transects. That is something that we are working on in The Deer-Forest Study, but it’s a work in progress.
Remember, the whole point of population estimation is to count those animals that are NOT seen. No matter what, you are not going to count every deer from a randomly placed transect in the woods or a road. You will only see some of the deer, and even fewer from a road.
So does distance samples adequately represent reality? It can. But to do it for deer, we need to work out the kinks. Then answer the question of whether it is an improvement over the current system given time, personnel, and money.
-Duane Diefenbach and Jeannine Fleegle
If you would like to receive email alerts of new blog posts, subscribe here.
And Follow us on Twitter @WTDresearch
This post is part of a series on Population Estimation
Over the River (you are here)