Detecting Climate Impacts on Rainforest Birds Using Bayesian Spatiotemporal Modelling

Published:

Problem

Tropical montane bird populations are increasingly threatened by climate change and extreme events, but detecting climate-driven signals in noisy, long-term monitoring data is challenging.
Goal: Use Bayesian hierarchical spatiotemporal models to quantify how climate variables and cyclone impacts drive population change, integrating multi-scale environmental predictors from remote sensing.

Approach

  • Assembled multi-decadal bird abundance datasets across rainforest sites in the Australian Wet Tropics.
  • Derived spatiotemporal climate predictors, including temperature, precipitation, and cyclone exposure indices, at the site-year level.
  • Processed high-resolution satellite imagery to quantify cyclone-induced changes in rainforest vegetation structure.
  • Integrated climate and vegetation metrics into a hierarchical Bayesian framework to model abundance in a multidimensional space:
    • State process: latent population dynamics across space and time.
    • Observation process: detection probability from repeated surveys.
  • Ran models in JAGS with spatial and temporal random effects, quantifying effect sizes, uncertainty, and spatial heterogeneity in climate impacts.

Stack

  • Bayesian spatiotemporal modelling: spatial random effects, temporal trends, and covariate integration.
  • Remote sensing integration: processed satellite imagery to derive vegetation change metrics.
  • Advanced statistical modelling: detection–abundance separation, credible interval estimation.
  • Data workflows: large-scale data cleaning, spatial joins, reproducible analysis pipelines.
  • Implementation: conducted in R for data processing/visualisation and JAGS for model specification and inference, under version control.

Results

  • Identified significant negative population responses to cyclone-driven vegetation loss and to climate warming in several species.
  • Revealed spatial heterogeneity in climate impact strength, with higher elevations often showing stronger declines.
  • Quantified uncertainty, enabling robust interpretation for conservation planning.

Impact

  • Provided direct evidence linking extreme climatic events and long-term warming to bird population declines in the Wet Tropics.
  • Informed adaptive management strategies and reinforced the case for targeted conservation action in climate-vulnerable habitats.

Links & Resources

Role

  • Designed and implemented the spatiotemporal Bayesian modelling framework.
  • Processed and integrated satellite-derived vegetation change metrics.
  • Conducted model fitting, validation, and uncertainty quantification.
  • Interpreted results in the context of climate change impacts and co-authored the manuscript.