As part of the atlas, a small group of observers are conducting point counts. In this article we introduce this wing of the atlas effort, explain how it’s gathering critical information for conservation efforts in the state, and report on some preliminary results from 2018.
What is a point count?
Point Counts are bird surveys conducted at a single stationary location. Observers stand in one spot for a fixed period of time (in our case, 10 minutes), and record all the birds they detect. Counts are typically aimed at songbirds and conducted in early morning during the breeding season, so the majority of birds are detected by ear.
Why do point counts?
The atlas is already doing an excellent job filling out maps showing where a species occurs in the state, which is the primary goal of most atlas projects. But what the atlas doesn’t do as well is determine where within that distribution a species is most or least abundant. Adding information on relative abundance atop the information collected by general atlasing gives us the strongest picture we’ve ever had about our state’s breeding birds.
Here’s a map of Red-eyed Vireo from the first atlas conducted from 1995–2000. It looks like this species occurs across almost the entire state. But is it equally abundant everywhere within that distribution?
Here’s a map (often called a “thunderstorm map” for its resemblance to weather radar) showing the relative abundance of Red-eyed Vireo within that distribution. Clearly Red-eyed Vireo is NOT equally abundant across the state, instead appearing to be much more common in the northern forest, peaking in the northeast.
All thunderstorm maps in this article were produced by Dr. Jenny McCabe, a postdoc at UW-Madison working with Dr. Ben Zuckerberg. These maps are the result of complex models that use bird observations from point counts, and relate these data to other variables such as latitude and longitude, precipitation, temperature, time, date, and landscape factors (e.g., grassland or deciduous forest). We can then use models to predict across the state which areas are likely to have birds at certain densities. The density scale is different for each map, and can be thought of as the density per 10 acres, ranging from the red value (varies per model) to near zero (but not exactly zero) in the white areas. These are preliminary models, and as we refine our modeling techniques, we will be able to bring in additional information we collected during the point counts.
In addition, outputs from models like these can provide population size estimates for many of our breeding bird species, an important component to assessing species status and setting conservation goals.
One downside of point counts is that if a species is not very detectable during a 10-minute roadside count, it does not get well sampled. So for the most part, we will get very good information on many songbirds, but other birds will often get missed, such as species that don’t call much during daylight (Great Horned Owl, Eastern Whip-poor-will), don’t call much at all (Spruce Grouse, Virginia Rail), are uncommon breeders (LeConte’s Sparrow, Cape May Warbler), or often occur far from roads (Lincoln’s Sparrow, Acadian Flycatcher). These species will be better sampled through traditional atlasing methods that allow for more time in the field, and more off-road coverage, or after the atlas period in targeted off-road surveys.
How did we do it?
To accomplish the goals of our point count effort, we determined we needed to survey 5 roadside points per block in every other block across the state (See here for the technical details of how we determined this). That works out to over 16,000 points in more than 500 atlas blocks. And given our project timeline, we had to do it in just four years, meaning we’d need to survey 4000+ points per year!
Within each block to be surveyed, the five points were drawn randomly and moved to the nearest road, with backup points drawn in case the initial points were inaccessible. The overall plan was fairly similar to methodology used in past second bird atlases in Ontario, Pennsylvania, and Ohio.
So this year, we assembled a team of 25 trained and certified surveyors (8 Wisconsin DNR employees, 11 hired technicians, and 6 skilled volunteers) and got to work. We followed a regionally-standardized monitoring protocol (Knutson et al. 2008, Landbird Monitoring Protocol for the U.S. Fish and wildlife Service, Midwest and Northeast Regions) and recorded the distance band and minute of first detection for each individual bird.
What did we find?
We conducted 4,095 point counts statewide between May 24 and July 10, and detected 112,357 individual birds of 187 species (not all of them breeders).
The points conducted in 2018 are yellow, 2017 are red, and 2016 are blue. Green points will be sampled in 2019.
Here are the Top 10 Species Detected on Point Counts, the number that we detected, and how the list compared to past years:
Obviously these are 10 very common species, but they are also species that can be detected on roadside counts and are either very visible or sing often. Interestingly, 9 of 10 species in the top 10 have been the same in all 3 years. In 2018, Canada Goose jumped up the standings (quite variable, given you can get a flyover flock that boosts your numbers), and American Crow slid down the standings (no explanation for this, but ranks 6-8 are pretty close together so it may not mean anything). Each year we hit different points across the state, so it’s actually quite interesting how consistent the top species are.
On the other hand, here is a list of species detected only once on a point count:
This list mostly contains birds that are rare breeders or difficult to detect on point counts. No White-winged Doves were detected during the point counts this year (2016 had one on a point, and 2017 had one travelling between points). There are a few good warblers on this list!
During the planning process, we identified that certain songbirds were poorly sampled by existing surveys like the federal Breeding Bird Survey, specifically species inhabiting northern coniferous forest, and southern interior forest. For this reason, we randomly scattered an additional few hundred points into northern and southern forest habitats to try to get better information on this group of species. While results were mixed, we did boost the numbers for a few species, which should improve our understanding of their breeding abundance in the state. Here is the list of our targeted forest species, showing how many we found to date on point counts (including the regular and special forest points) — our current estimate is that 25 birds per year is the “magic number” as far as being able to potentially model their distribution, so 75 birds at this point may be enough.
This is a list of species for which we are lacking good information based on federal BBS (Breeding Bird Survey) data. Highlighted species are birds we project we will be able to model, based on numbers observed so far.
After getting 11 Red Crossbills in the first 2 years, a major invasion of multiple types this year yielded 455 on point counts! Unfortunately due to their generally early nesting, we won’t be able to model this species using the techniques we use to model singing songbirds on territory – many observations of Red Crossbills were flocks with young or flyover flocks. Always remember to try to capture audio or video whenever encountering Red Crossbills in Wisconsin, so we can assign them a type. Many of us were point counting with cameras or phones at the ready on the hood of our car to record video clips for typing if crossbills flew over!
We will input the point count records into Atlas eBird to supplement other atlas observations, however, there is an extensive and lengthy data-checking process before that can happen, and most of the records from point counting will be at the Possible (S or H) level. So far, the 2016 point count data are in eBird, and the 2017 and 2018 datasets are still being checked, with the help of undergraduates at Northland College and UW-Madison.
Here are a few more maps. Remember, these are very preliminary maps, and will continue to be refined as more data come in and we improve the models.
Eastern Meadowlarks are sparse in the North Central Forest, Central Sands, and urban areas, but are predicted to occur at low levels throughout the open 2/3 rds of the state, reaching highest abundances in the southwest, and Buena Vista grasslands really stands out too.
White-throated Sparrow attains highest abundances across the northern forest. The main portion of its range tracks the tension zone, becoming sparse by Burnett and Washburn counties in the northwest, but reaching high numbers in highly forested Menominee and northern Oconto in the east. Some of the flecks in central and southern Wisconsin are probably not correct (none at Horicon) but the splotch in eastern Jackson/SW Wood counties is correct.
Canada Warbler is probably among the least abundant species we’ll be able to model. This map is decent but perhaps not particularly refined, predicting high abundances in lowland conifer.
Northern Cardinal is strongly associated with suburban habitats, and sparsely occurs in the northern quarter of the state. Due to the scale of the habitat covariates used in this version of the model, it’s probably missing slightly elevated numbers in most small towns across the state.
With 1 more year — and 4,000 points — to go, we at Atlas Central are excited to see the results of this project, which will undoubtedly shed new light on the state of Wisconsin’s breeding birds.