Monday, November 19, 2007

Counting populations: Superferries and tropicbirds

The August Superferry protests on Kaua'i challenged my crowd-counting skills, and reinforced the value of accurate population assessments.
(Image: Red-tailed tropicbird and chick. U.S. Fish and Wildlife Service photo.)

Counting individuals, or development techniques to develop reasonably accurate counts, are valuable in assessing the size of crowds of humans, but also crowds of critters, as we will see later in this article.

In my case, the work was comparatively simple. One media account said there were 1,000 people at the protest of Superferry Alakai's first sailing to Kaua'i. Superferry representatives were claiming the crowd on the Nāwiliwili Jetty was no larger than 150.

I've been counting crowds for nearly 40 years in lots of circumstances. It's fairly easy when they're in a stadium or auditorium because you have a known number of seats, which brackets the possible crowd count. (If the place seats 5,000 and the fire department's on hand, you won't get more than 5,000. You can ask about ticket sales, count full seats, count empty seats, or do representative samplings to get a number. Or use a combination of these technques.)

Milling crowds in the open are more difficult.

I used a couple of different crowd-counting techniques for the Superferry crowd, employing one method as a check on the other. I came up with close to 300, and that's the number I used to describe the crowd on the first day of the protest.

But because the numbers from other media and those asserted by Superferry supporters were so different, on Protest Day Two, I did my calculations and then backed those calculations up with an actual census by simply walking from one end to another and counting every single individual. The Nāwiliwili Jetty crowd was strung out for about a quarter mile, so the counting was pretty easy and only took a few minutes.

My initial calculated estimates on the second day of the protest came in at about 250, and my actual count came to 310—the difference largely because when I walked the crowd I found there was a clutch of a few dozen protesters at the mauka end inside Niumalu Park, a group I hadn't seen earlier.

Crowd-counting ain't rocket science, but it does require a little attention to detail.

I had what you might call a “high degree of confidence” in my numbers, but we still had people calling the newsroom to tell us our numbers were wrong. You get a lot of that—criticism from people who either have an ax to grind, who are mistaken or who are pathetically lacking a clue.

For scientists counting wildlife, the numbers can be more important. You want them to be accurate enough that you can distinguish long-term trends.

In a November issue of the journal Biological Conservation, Nathaniel Seavy and Michelle Reynolds, both of the U.S. Geological Survey's Pacific Islands Ecosystem Research Center, Kïlauea Field Station, on the Big Island, reviewed population counts for red-tailed tropicbirds at Tern Island in the Northwestern Hawaiian Islands.

Their paper is entitled, “Is statistical power to detect trends a good assessment of population monitoring?”

One of the things they were trying to determine: If the population of a species drops 50 percent over time, are your counts accurate enough to detect it?

In this situation, you need to understand that at Tern Island, seabirds don't come in individuals. They come in clouds. Of some species, there are tens of thousands. How do one or two people wandering a little island count those kinds of numbers?

With birds, there are all kinds of difficulties. At any given time, some of the population might be sitting on the nest, but some of the population might be feeding. The proportions can change with the time of day, the progress of the breeding cycle and the time of year, along with weather and lots of other things (like the presence of a human with a clipboard causing birds to change their behavior).

“Trend and power analyses alone are sensitive to the sampling period, sampling methods, and the statistical model used (and they) should not be the only tool to evaluate population status,” Reynolds said in an email.

And how can you tell whether changes you detect in a bird population are real, or perhaps the result of what Seavy and Reynolds call “observation error?”

The authors recommend close attention to monitoring methods, but also to the biology of the birds. You can spend too much time focusing on a trend without knowing whether the trend is significant, they say.

In some cases, “statistical power to detect trends is less important than understanding the long-range variability of the population.”

The upshot is that you never trust your methods entirely, you back them up with alternative counting systems when possible, and you constantly look for errors in your methods or ways to improve them.

© 2007 Jan W. TenBruggencate

No comments: