Modelling avian habitat suitability in boreal forest using structural and spectral remote sensing data

Avian species are often highlighted as key indicator species as they are easily and reliably studied, are sensitive to environmental change, and occupy a wide range of ecological niches. A link between structural vegetation indices and avian diversity has been widely established. Remote sensing technologies such as airborne laser scanning (ALS) can quantify forest structure and have been widely used in bird/habitat studies. However, its use to quantify avian habitat suitability in managed boreal forest has received less investigation. We developed models by best subset regression of avian habitat suitability in western Newfoundland. Data included traditional vegetation plot surveys, satellite spectral information, topographic, landscape, climatic and ALS datasets. Models were assessed using goodness of fit to provide insight into how forest structure data can improve the efficacy of habitat suitability models. Additionally, the probability of presence and abundance of several key species relative to key metrics was examined. In 2018, we observed 371 individuals of 38 bird species in 29 plots over a 10 day period. Field plot metrics produced significant (α = 0.05) models of avian presence-absence and abundance. ALS metrics alone produced significant models, but with a poorer fit than field plot metrics, while spectral metrics alone were unable to produce significant models. The addition of the other environmental data produced significant results but also had poorer fit than field plot models. While field plot data may outperform remote sensing metrics from a single sensor; spatial models, combining multi-sensor and environmental data were found to be more significant. Results can aid monitoring the environmental impacts of management decisions and climate change.