Usage
  • 105 views
  • 106 downloads

Evaluating remote sensing covariates for understanding habitat selection by boreal forest birds

  • Author / Creator
    Casey, Brendan
  • Boreal forests are changing in response to climate change and shifts in disturbance regimes. Statistical models that link distribution, abundance, and community structure to select environmental variables have been used to understand how birds respond to these changes. However, model performance is influenced by the choice of spatial covariates. Remote sensing, like Light Detection and Ranging (LiDAR), and satellite photogrammetry, can improve bird-habitat models by introducing novel biologically relevant spatial covariates at fine resolutions. This thesis presents methods to use and evaluate modern remote sensing tools to refine our understanding of species-habitat relationships. It also explores how bird communities respond to forest harvesting in the boreal. First, LiDAR can improve bird-habitat models by providing novel vegetation structure covariates. However, temporal misalignment between LiDAR acquisitions and point count surveys may influence the predictive power of models that use LiDAR predictor variables. As vegetation undergoes successional changes, LiDAR data that is temporally restricted may cease to reflect habitat conditions, thus compromising the usefulness of LiDAR predictor variables. To evaluate this, I examined how the time-lag between LiDAR acquisitions and bird surveys influenced model robustness for early-successional, mature-forest, and forest generalist birds. The results indicated that for species occupying older, more stable forests, a time difference of up to 15 years has a negligible impact on the predictive power of LiDAR based bird-habitat models. For early-successional birds, the findings suggest that a time difference of 5-13 years between LIDAR and bird data may decrease model performance. Next, I compared the suitability of covariates from LiDAR, a Landsat time series, and forest resource inventories for predicting bird response to forest harvesting in Alberta. These covariates were used to predict the abundance of twenty species associated with different foraging and nesting strata, within harvest areas across a chronosequence of recovery. The results suggest that integrating LiDAR and Landsat spectral change covariates improves model performance over models built using forest resource inventory data alone. Additionally, spectral estimates of harvest intensity and time since disturbance explained most of the variation in species abundance models. Finally, I used a spectral change detection algorithm, point count data, acoustic monitoring tools, and mixed-effects regression models to evaluate the impact of the interaction between forest harvest intensity and recovery time on the taxonomic and functional diversity of birds. The findings suggest that harvest residuals can mitigate the short-term effects of forest harvesting on bird communities. Furthermore, I demonstrate that metrics derived from a time-series of Normalized Burn Ratio (NBR) are a promising alternative to conventional categorical harvest intensity metrics included in many classified land cover maps. Collectively, this work shows that supplementing classified land cover data with LiDAR and Landsat time-series data can improve the performance of bird-habitat models while avoiding the costs of ground-based habitat surveys.

  • Subjects / Keywords
  • Graduation date
    Fall 2023
  • Type of Item
    Thesis
  • Degree
    Doctor of Philosophy
  • DOI
    https://doi.org/10.7939/r3-s2km-4621
  • License
    This thesis is made available by the University of Alberta Libraries with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.