From theory to application: Elasticity-consistent aggregation of Leslie matrix population models for comparative demography

Technical Summary: Elasticity-Consistent Aggregation of Leslie Matrix Population Models

Problem Statement
The field of ecology has entered a "big data" era, facilitated by comprehensive databases like COMADRE and COMPADRE, which enable comparative demography across the tree of life. However, a significant methodological barrier exists: matrix population models within these datasets vary widely in sampling frequency, life cycle stage width, complexity, and census type (e.g., pre- vs. post-reproduction). Furthermore, multiple matrix representations can describe the same population, and the choice of representation directly influences key demographic parameters. Consequently, a primary challenge in leveraging these large datasets is the ability to perform fair comparisons between models constructed with different projection intervals and structural complexities.

Methodology
To address the difficulty of comparing models of varying complexity, the authors focus on the aggregation of larger models down to the dimensionality of smaller models. While a "standard aggregator" exists that successfully preserves stable growth rates and stable stage distributions, it fails to maintain consistency in reproductive values and elasticities.

To overcome this, the authors extend and validate an existing elasticity-consistent aggregator. Their methodological improvements include:

• Derivation via Flows: The aggregator is derived using balancing and interstage flows. This approach removes the mathematical requirement of matrix primitivity, which was a limitation in previous formulations.
• Biological Constraints: The method restricts aggregation specifically to Leslie-to-Leslie models. This constraint is critical for preventing biologically infeasible outcomes, such as survival probabilities exceeding one.
• Validation Metrics: The authors introduce explicit metrics to quantify the effectiveness of the aggregation process.
• Empirical Application: The approach was applied to 12 Leslie matrix population models representing animal populations from diverse taxonomic classes.

Key Results
The study demonstrates that the elasticity-consistent aggregator successfully returns a Leslie matrix that preserves essential properties of the original matrix, including irreducibility and primitivity. Crucially, it yields consistent estimates for:

• Population growth rate
• Stable age distribution
• Reproductive values

In a comparative analysis across the 12 examined animal populations, the elasticity-consistent aggregator outperformed the standard aggregator in the majority of cases:

• 86% of generation time estimates were more accurate.
• 76% of net reproductive rate estimates were more accurate.
• 60% of Demetrius entropy estimates were more accurate.

Significance and Claims
The paper posits that by preserving the key properties of the original model, the elasticity-consistent aggregator provides a robust framework for comparing matrix population models of varying complexity within comparative demography. The authors claim that this methodological advancement is essential for realizing the potential of big data in ecology. It enables the discovery of universal principles across the tree of life, spanning microbes, fungi, plants, and animals, by ensuring that demographic comparisons are not confounded by structural differences in model construction.