Predicting bird species distributions in reconstructed landscapes

Area of Science:

• Ecology
• Conservation Biology
• Spatial Modeling

Background:

• Biodiversity conservation in fragmented landscapes necessitates accurate species distribution modeling.
• Box-ironbark forests in Victoria, Australia, face fragmentation challenges impacting bird populations.
• Predictive models are crucial for guiding revegetation strategies in cleared or degraded areas.

Purpose of the Study:

• To predict the occurrence probabilities of 61 bird species in fragmented box-ironbark forests.
• To evaluate the effectiveness of Bayesian model averaging for species distribution prediction.
• To inform landscape optimization for biodiversity by prioritizing revegetation efforts.

Main Methods:

• Utilized Bayesian model averaging with logistic regression to predict species occurrence.
• Employed topographic, edaphic, and climatic variables as predictors for model application in revegetation scenarios.
• Validated models using independent data from remnant native vegetation blocks.

Main Results:

• Accurate predictions were achieved for 18 of 45 woodland bird species (40%).
• Bayesian model averaging outperformed single best models in predictive accuracy.
• Models were more successful for smaller-bodied species dependent on specific vegetation types.
• Predictions for larger, generalist, and migratory species were less successful due to distribution changes over time.

Conclusions:

• Validated models can project species occurrence probabilities to guide landscape-scale revegetation.
• Bayesian model averaging is a robust method for predicting species distributions in fragmented habitats.
• Longer-term data are required to improve model accuracy by accounting for temporal dynamics in species distributions.