The Breeding Bird Survey in Tennessee - 45 years of data

The North American Breeding Bird Survey (BBS) represents one of the most extensive long-term datasets in existence for tracking the populations of North American birds over the decades. If you are not familiar with it, the link will give you a detailed description of what it is and how it operates. In short, it consists of many hundreds of roadside routes all across the U.S. and southern Canada, which are sampled every year during the nesting season using a standardized protocol. The survey began in 1966, and many routes have been sampled continuously every year since then.

The strengths and weaknesses of the BBS data have been discussed and analyzed at length over the years. The BBS has developed sophisticated statistical methods to extract long-term trends from the data while compensating for complications such as differences between observers and routes that are not run every year. Sometimes, though, the numbers than come out of complex statistical models can feel a bit divorced from the real, in-the-field experiences of the actual observers who stood there for three minutes a stop, 50 stops per route, over five decades. I decided I wanted to have a more direct look at the basic raw numbers and how they have changed over the last 45 years, in my adopted home state of Tennessee.

Probably the major complicating factor when looking at BBS data is the observer effect. Even birders of seemingly similar skill levels will produce quite different data on the same route. We all have our unique biases -- some people tune out Cardinals but will key in to a Cerulean Warbler at 100 yards, others can spot a Cooper’s Hawk in a split second but might miss the soft murmuring of a Gnatcatcher. Many statisticians have labored for many years to find ways to account for this. My simpler take on the matter, though, is that so long as you are averaging over enough different routes, and so long as the average biases of the whole pool of obervers have not changed substantially over time, then this will come out in the wash. I was not birding in 1966, but I was in 1974 and I ran my first BBS route in 1976. So I have been around for a large portion of the BBS era. To be sure, field ID skills have changed over that time, and in substantial ways they have improved. But in a place like Tennessee, the bulk of BBS skill boils down to one thing: How well can hear and correctly identify the typical vocalizations of the expected common and uncommon species of birds in your area? The birders I learned from in 1974 might not have yet figured out all the structural differences that help tell silent Empidonaces apart, and hybrid gulls were all but Terra Incognita. But they knew perfectly well how to pick out singing Acadian, Least, and Willow Flycatchers from the morning din. I do not really think that the birders of the 1960s and 1970s were all that different, on average, from the birders I go out with now in the 3rd Millenium when it comes to aural accuity and the ability to hear and identify the songs and calls of their local avifauna. While engaging in state-of-the-art deliberations about the molt sequence of a mystery gull that might have been passed off with a glance 40 years ago, it still seems to be the same proportion who will catch the distant chattering of a Red-headed Woodpecker, and the same proportion who will overlook it.

The second major complicating factor is than not all routes are run each year. While some gold-star routes have a complete 45 year sequence, many others have missing years, in some cases long chunks of missing years. Again, for the most part this will come out in the wash if the coverage lapses are not heavily biased to particular regions. For routes in most of the state, where the avifauna is not dramatically different between areas, this seems to be the case. There are some exceptions, however.

Three routes in particular have a disproportionate effect on the averages -- two in the far northwest part of the state and one in the far northeast. In the west, Tiptonville (001) and PawPaw (002) are the only Tennessee BBS routes that sample the Mississippi Alluvial Valley (MAV), which is biogeographically quite distinct from the rest of the state. Tiptonville runs along the south shore of Reelfoot Lake for many miles, and the nearby PawPaw route covers large areas of open cropland with many intermittent ponds, passing very close to the Mississippi main channel near Island 13. Statewide totals for many species of wetland birds are almost entirely dominated by counts from these two routes, with the Moscow Route (003) in the southwest contributing some as well. Year-to-year variation in these counts tends to be quite high with the small number of routes, the extreme variability in wetland habitat conditions between years, and the random hit-and-miss of large flocks. As a for instance, I currently run the Tiptonville route myself, and in 2011 after extensive flooding in the area I had the first record for Snowy Egret on the route -- 20 of them, all in one flock in one flooded field! Clearly these sorts of incidents have very little bearing on real Statewide population trends, but they can make for impressive spikes in a graph.

In east Tennessee the higher mountains also are ecologically quite different from the rest of the state. For the first several decades of the BBS, these high elevation habitats were only minimally sampled. In the early 1990s several new routes were added and coverage stepped up on existing routes, creating an “Appalachian bump” on counts of many warblers and other woodland species when averaged statewide. To help reduce this effect, I excluded the five 900-series routes that were first added in 1989 specifically to increase coverage of the high elevation public lands. Even so, improvements and changes in coverage on the other existing routes in this region contribute some to a residual “Appalachian bump” that is still visible in many graphs, peaking around 1993-1995. After excluding the 900-series routes, there remains only one route that samples the highest elevation birds such as Winter Wrens, Veeries, and Canada Warblers -- Fish Springs (042). This route runs from Watauga Lake to south of Roan Mountain with extensive areas above 1000m elevation. As it has been sampled steadily since 1966 with only a few missing years, I included it in the tallies. Still, given its unique habitat it also shows disproprotionately in the graphs for these high mountain species, with year-to-year sampling variability and changes in observers causing quite bumpy lines! Again, the significance of these sampling bumps as indicators of real population trends is minimal. In each of these cases (the MAV and Fish Springs Effects and the “Appalachian Bump”) I will point out on a species-by-species basis when they are dominating the patterns.

My approach was very basic: Take the statewide total for each species for each year (excluding the 900-series routes), and divide it by the total number of routes surveyed. Presto, a statewide average for individuals counted per route. I then plotted these numbers out, both the individual years and a 5-year moving average to give a smoother trend line. To get an overall summary number for the long-term trend, I took the difference between the average for the first 10 years (1966-1975) and the last 10 years (2001-2010; 2011 data are not yet available). I made an error estimate for this 35-year change by calculating the standard error for the 10 paired 35-year differences (1966 versus 2001, 1967 versue 2002, etc.) and then giving the 95% (2-tailed) confidence interval.

I’ll present these data in taxonomic order in batches, starting with the next post.

Tennessee BBS index:
Introduction
Next: 1: Waterfowl to Herons
2: Vultures to Doves
3: Cuckoos to Woodpeckers
4: Flycatchers to Corvids
5: Larks to Wrens
6: Gnatcatchers to Waxwings
7: Wood Warblers
8: Towhees to Buntings
9: Icterids to House Sparrow
Ups and Downs
Habitats
Inflection Points
Summary