UCPH Statistics Seminar: Emily Simmonds
Speaker:
Emily Simmonds (University of Edinburgh)
Title:
Embracing uncertainty in predictive ecological models: gaps, importance, and opportunities
Abstract:
Predicting population responses to environmental changes is a pressing global challenge. Ecological forecasting is a rapidly growing sub-field of ecology, which rises to this challenge and implements predictive models for ecological processes. Uncertainty accumulates at multiple levels in these predictive models and failing to account for all sources leads to a large amount of ignored variability in the model’s predicted outcomes. This talk will cover an overview of quantitative uncertainty reporting across seven scientific fields before zooming in on uncertainty consideration in predictive population models in ecology (matrix population models). Despite the importance of uncertainty for matrix population models, we have found that complete uncertainty propagation is rarely achieved (31% of papers). But how important is this omission? Our simulation study demonstrated that even with moderate levels of uncertainty, incomplete propagation introduces bias for predicted population growth rates and can substantially alter conclusions. I will finish with a forward look at what we might be able to do better in the future.