Bird communities were surveyed at 69 sites located on public and private property. The study area was the upper French Broad River basin in western North Carolina, USA. This basin is in the southern Appalachian Mountains. Sites were stratified by elevation and development intensity (e.g., building density). The same sites were used to collect wildflower community data for a concurrent study and detailed site selection methods are published (Graves et al. 2017).
To protect privacy of private landowners, a random value between -0.01 to +0.01 degree was added to the latitude and longitude coordinates of those sites. Thus, the actual sites were located within the 1.0 km2 marked by the coordinates provided in this dataset.
Bird surveys were conducted at least once every three weeks at each site, and a subset of sites was visited weekly, from April 1 to August 8, 2014. Surveys consisted of standard 10-min point counts conducted between 05:45 and 10:30. One of 3 trained observers performed each point count, and all sites were surveyed by at least 2 different observers throughout the study. Surveys were not conducted on rainy days or days with high wind. Each bird detected by sight or song within 100 m of the point location (i.e., site center) was identified to species and recorded. Birds observed outside the 100-m radius were recorded but not included in the primary analysis.
Bird species were classified based on (1) migratory status (mig =migrants vs. resident) and if migratory (migrate = short- or long-distance migrants), and (2) synanthrope status (synclass: C = casual,T = tangential, or F = full synanthrope;, Johnston 2001) and synanthrope class (synanth: 1 = synclass of C, T, F; 0 = otherwise) . We calculated an index of relative species rarity (rarity_ebird) using the observation frequency of each species within the study region using the eBird dataset, an online citizen science bird monitoring project (Sullivan et al. 2014). This index had values of 1 (lowest) to 4 (highest) based on quartiles of the frequencies of observations in the study region. Finally, species of greatest conservation concern were identified from the NC Wildlife Action Plan (NCWRC 2015).
Remotely sensed and GIS data were used to derive environmental variables: local and landscape building density, land-cover diversity, tree cover, vegetation structural diversity, estimated annual productivity, and elevation. Annual productivity and elevation were extracted at the center point of each study site. The remaining variables were extracted using buffers of 100, 200, and 1000 m, depending upon the scale at which each variable had the strongest relationship with bird detections.
Annual vegetation productivity was extracted from a smoothed and gap-filled MODIS Normalized Difference Vegetation Index (NDVI) dataset (Spruce et al. 2016). We calculated the 10-year (2004–2014) median of annual vegetation productivity for each study site. Elevation was extracted from the National Elevational Dataset-Digital Elevation Model (NED-DEM, USGS 2017).
Building density (building units per hectare) was quantified by counting the number of buildings located within 100 and 1000 m of the center of each study site to account for local and landscape scale effects of development intensity. Vegetation structure and tree cover were calculated from LIDAR (light detection and ranging) data within 100 m of the site center. Vegetation structural diversity was calculated using the Shannon Evenness index using the proportion (pi) of LIDAR returns in each of four vegetation strata (S = 4, i.e., herb, shrub, subcanopy, and canopy layers). Tree cover was recorded as the proportion of LIDAR returns within subcanopy or canopy layers ([2.0 m above ground) (Graves et al. 2017a). Land-cover diversity was calculated using Simpson’s diversity index with six land-cover categories (grassland/herb, shrubland, cropland, forest, developed, and other/water) within 200 m of each study site. SIDI ranges from 0 to 1.0 and describes the probability of two points chosen at random within a given area being in different land-cover types (McGarigal et al. 2012). We used the 2014 Cropland Data Layer (CDL, USDA-NASS 2014) and calculated SIDI using Fragstats (McGarigal et al. 2012).
See Graves et al. (2019) for more details of data collection and resulting analysis.
==================== Data Sources =========================
Title: DayMet
URL: https://doi.org/10.3334/ORNLDAAC/1345
Creator: Thorton et al. 2016
Contact: ORNL DAAC uso@daac.ornl.gov
Title: MODIS NDVI Data, Smoothed and Gap-filled, for the Conterminous US: 2000-2015
URL: https://doi.org/10.3334/ORNLDAAC/1299
Creator: Spruce, J.P., G.E. Gasser, and W.W. Hargrove (2016)
Contact: ORNL DAAC uso@daac.ornl.gov
Title: National Elevation Dataset, 1/3rd arc-second Digital Elevation Models
URL: https://viewer.nationalmap.gov/
Creator: U.S. Geological Survey 2016.
Title: NC Flood Mapping: LIDAR Phase 3 all-returns data
URL: http://fris.nc.gov/fris/Download.aspx?ST=NC
Creator: Floodplain Mapping Program; Raleigh, North Carolina. NCDEM 2006.
Title: Cropland Data Layer
URL: https://nassgeodata.gmu.edu/CropScape/
Creator: USDA National Agricultural Statistics Service. USDA NASS 2014.
Title: North Carolina Wildlife Action Plan
URL: http://www.ncwildlife.org/plan.aspx
Creator: North Carolina Wildlife Resources Commission. NCWRC 2015.
===========================================================