Combining multiple data sources with different biases in state-space models for population dynamics.

Area of Science:

• Ecology
• Population Dynamics
• Statistical Modeling

Background:

• Accurate animal population modeling requires high-resolution data, often from multiple life stages, enabling seasonal dynamics descriptions.
• Abundance estimates used in models can suffer from random and systematic errors, particularly unknown observation biases.
• State-space models (SSMs) offer a framework to differentiate process variation from observation error, allowing for the inclusion of varying biases across datasets.

Purpose of the Study:

• To investigate the consequences of including or excluding unknown bias parameters in sequential life stage population dynamics SSMs.
• To evaluate the impact of bias parameters on the inference of population processes like recruitment and survival.
• To explore strategies for addressing parameter redundancy and characterizing process uncertainty when bias is present.

Main Methods:

• Utilized a sequential life stage population dynamics state-space model (SSM).
• Employed a combination of theoretical analysis, simulation experiments, and an empirical case study.
• Compared model performance with and without bias parameters, including scenarios with fixed bias parameters.

Main Results:

• Excluding bias parameters in unbiased data leads to increased precision.
• When data are biased and biases are not estimated, recruitment, survival, and process variance estimates are inaccurate.
• Including bias parameters substantially reduces estimation problems, even when one parameter is fixed incorrectly.
• Models with bias parameters may exhibit parameter redundancy, posing inferential challenges.

Conclusions:

• Combining multiple datasets via bias parameters for rescaling can significantly enhance population model inference and diagnostics.
• Careful consideration and strategies are needed to manage process uncertainty confounded by bias parameters.
• Estimating bias parameters is dataset-specific and may require higher precision than typically available in ecological data.