Forecasting Population Viability with Bayesian Hierarchical Models

Problem

Understanding population sustainability is critical to conservation prioritisation—but count data are often imperfect and biased by detection issues.
Goal: Build robust forecasts of population viability using Bayesian hierarchical models that explicitly account for detection probability, to inform elevated conservation listing at national and international levels.

Approach

• Collated long-term count data with imperfect detection from population monitoring of targeted species.
• Built Bayesian hierarchical models: observation process (detection component) separated from true state process (abundance/trend).
• Integrated prior knowledge and error structures to model latent population trends and forecast future viability under current and projected conditions.
• Derived population viability metrics (e.g., extinction probability, trend trajectories), feeding decisions for national/international conservation listings.

Stack

• Bayesian hierarchical modelling: detection–abundance partitioning, forecasting of latent trends with credible intervals.
• Forecasting workflows: dynamic prediction of future population trajectories under forecasted climate change, uncertainty quantification.
• Data workflows: count data cleaning, variable development (climate change predictors), model fitting, posterior analysis, reproducible scripting and reporting.
• Implementation: carried out in R (data processing, analysis, visualisation) and JAGS (Bayesian model specification and MCMC sampling), with full version control for transparency.

Results

• Forecasted strong declines in target species with credible uncertainty bounds.
• Identified species with high extinction risk over relevant time horizons.
• Results directly contributed to elevating conservation priority status for those species under national and international protection lists.

Impact

• Strengthened the scientific basis for conservation policy decisions by delivering rigorous, uncertainty-aware forecasts.
• Demonstrated the value of integrating detection-corrected Bayesian models into species viability assessments.

Links & Resources

• 📄 Paper: Diversity & Distributions article
• 💾 Repository: Dryad dataset

Role

• Conceptualised and developed the hierarchical Bayesian framework.
• Cleaned and structured count/detection data and climate change covariates (heatwaves and warming) for robust inference.
• Ran forecasting models and interpreted posterior outputs.
• Produced insights used in elevated conservation recommendations.
• Wrote the manuscript and communicated findings to conservation authorities.