Home >  Methods >  Avian Data Analysis >  Density

DENSITY ESTIMATION

BAM developed estimates of avian densities (males per hectare) for 70 species of boreal forest songbirds. These estimates were based on off-road survey data only; BBS route data were not included.  The results are summarised by species under RESULTS.

Our earlier methodological research provides the foundation for these estimates that enables us to adjust the survey counts for the many, substantial, observational biases inherent in the BAM dataset (see Avian Data Analysis). Biases occur because of incomplete detection rates of birds during surveys, variability in detection rates relative to habitats and the times of day and year that surveys occur, variability in the survey protocols used by different studies, and uneven sampling of habitats and geographic strata.

We developed a hierarchical model that accounts for these sources of observation bias. Our models implemented many of the recommendations made to improve Partners in Flight's (PIF) estimates of landbird population size in North America by better accounting for sampling effects related to avian detection rates and uneven geographic and habitat coverage of surveys.

The model estimates breeding densities and population sizes of 70 species of songbirds as functions of geographic strata and habitat classes, and variation in habitat use among geographic units.

Characteristics of the Breeding Density Model

The hierarchical model that we used to estimate avian densities included the following specifications:

  • Generalised linear model with Poisson or Negative Binomial error.
  • An offset that adjusts the mean count for detection probability and how it relates to species' singing rates, species' specific effective detection radii (EDR), and variability in the length of the counting period and the count radius
  • Geographic strata defined by the combination of Bird Conservation Region and jurisdiction (province, territory, state).
  • Habitat classes based on the Land Cover Map of Canada 2005 collapsed into 14 categories by BAM.
  • Breeding densities were estimated for each combination of BCR, jurisdiction, and habitat class.

Estimating avian density and population is a complicated task. No single, definitive approach exists to calculate avian densities. Our density values, like all estimates, include an inherent level of uncertainty. They were calculated using one possible approach to correcting for the sampling issues that influence bird counts, based on these assumptions:

  • All birds present but not detected during surveys were accounted for by our adjustments for detection rates,
  • Observers accurately estimated the distance to birds and the time intervals that birds first sing,
  • Individual birds are not double-counted within or among survey points,
  • Observers correctly identified the species detected during the surveys,
  • Bird movements were negligible during surveys,
  • Bird survey locations were spatially and temporally representative of jurisdictions,
  • Habitat strata were classified correctly.