The geographic and habitat coverage of avian surveys collated by BAM are highly uneven across the boreal region. Therefore bird habitats are not sampled in proportion to their availability across the boreal forest. More accurate prediction of species populations will require that corrections be made to account for this uneven sampling.
This uneven distribution of surveys results from access to avian habitats by the road network that is limited and concentrated to the southern boreal forest. Additional surveys in northern boreal regions would help fill gaps in survey coverage. Coverage by the roadside BBS and off-road surveys compiled by BAM both mirror the uneven geographic and habitat coverage of roads across the boreal (Fig.). For example, survey coverage is strongly biased towards southern ecozones which make up 49% of the boreal region but received 89% of the surveys.
BAM currently uses a combination of geographic and habitat strata to minimize the problem of uneven sampling. This improves our predictions of avian abundance in under-sampled areas and habitats. BAM's next generation of abundance models will also include climate covariates to increase the resolution of our spatial predictions of avian abundance.
The figure shows the bias in geographic and habitat coverage of roadside surveys by the North American Breeding Bird Survey (BBS) and off-road avian surveys compiled by the Boreal Avian Modelling Project (BAM). We also present bias coverage by the road network (roads) to emphasize where patterns in bias in avian surveys closely follow that of roads. Bars represent sampling bias calculated as the difference in proportion sampling (use) and proportional availability.
Poorly-sampled habitats include conifer forests and woodlands, tundra- and lichen-dominated areas, wetlands, and areas with recent burns.