Bias in population-level chemical risk assessment when neglecting seasonality from matrix population models

Abstract

Matrix population models are widely used to understand how chemical exposure affects population dynamics. In species used in ecotoxicological tests, matrix population models are often built with subannual time steps and population dynamics are projected by iterating a single matrix corresponding to a test result, which may ignore cyclic seasonal changes in vital rates. Here, we examined how neglecting seasonality from matrix population models affects population-level risk estimation. We developed a periodic matrix population model incorporating seasonal variation in vital rates and compared the expected percent decline in population growth rate under chemical exposure with that of the aseasonal projection. We found that while the aseasonal projections could provide more precautionary assessments under most scenarios, they indicated lower risk relative to the seasonal projections when chemicals affected the survival rate of overwintering stages. Elasticity analysis revealed that survival during the non-breeding season was the key determinant of annual population persistence, which was not explicitly represented in aseasonal projections and thus resulted in lower risk indices compared with the seasonal model. These results suggest that incorporating life-history seasonality into matrix population models may be warranted under specific conditions to better achieve precautionary risk assessment at the population level.