The main goal of the breeding atlas was to map evidence for breeding occurrence across the state for each bird species. Estimating species abundance accurately is far more difficult, but also extremely useful for conservation planning. Consequently, we augmented the block-based volunteer data collection, with standardized point count surveys designed to provide the data that are necessary to estimate abundance patterns.

Point count surveys were conducted from May to July in 2018, 2019, and 2021. Survey locations were selected using two-stage cluster sampling with stratification by major habitat types. To do this we first used the Create Fishnet tool in ArcMap to create a grid across the state and identify points at the center of each cell.  We used a cell size of 500 m x 500 m to ensure that selected points were not located too close to each other. We then randomly selected 200 atlas blocks, and generated a stratified random sample of 10 points per block by selecting points from the the ArcMap grid. For blocks along the coast, or on the state boundary, we reduced the number of points to reflect the area of the non-marine proportion of the cell that lies with the state.

Sampling was stratified to produce a weighted sample that represented nine major landcover classes based on the North American Land Cover Change Monitoring System (NALCC 2015): Agricultural, Central Oak-Pine, Coastal Grassland & Shrubland, Emergent Marsh, Northern Hardwood & Conifer, Ruderal Shrubland & Grassland, Tidal Marsh, Urban/Suburban Built, Wet Meadow/Shrub Marsh. To ensure adequate sampling of all landcover types, we revisited the original pool of available points for those with fewer than 50 points after the initial random selection. For these landcovers, we either added points until we reach the target minimum of 50, or used all points if fewer than 50 were available.

Once we had selected points, we randomly assigned the atlas blocks to three groups, with a goal of sampling one-third of the points in each sampling year. Due to logistical constraints, not all selected points were sampled, but we finished the project with data from 1,151 points, each of which was visited twice, once before and once after June 17, during one of the three breeding seasons.

Surveys were conducted by trained field technicians and each survey comprised four consecutive 2-minute periods, during which every bird detected was recorded, along with information on its estimated distance from the observer, the method of detection (song, call, visual, or some combination), and a variety of ancillary information. Surveys were done between just before dawn and 11:00 and were not conducted during periods of high wind or precipitation. We did not limit the distance a bird could be from the observer for it to be counted, but prior to statistical analyses we removed all detections that were reported as greater than 500 m from the survey point. The complete survey protocol given to technicians is available here.