Circumventing structural uncertainty: A Bayesian perspective on nonlinear forecasting for ecology

Lenfest Published Paper September 7, 2016

Munch, S. B., Poynor, V., & Arriaza, J. L. (2016). Circumventing structural uncertainty: A Bayesian perspective on nonlinear forecasting for ecology. Ecological Complexity. doi:10.1016/j.ecocom.2016.08.006

Summary

Bottom line: This paper develops a method to improve ecological forecasts by extending an approach known as nonlinear time series analysis. The method may be useful in ecosystem-based fisheries management and for predicting how fisheries will shift as a result of climate change.
Background:
- Ecological models are often used to predict the results of proposed management changes. However, seemingly small inaccuracies in a model can lead to large errors of prediction.
- One way to circumvent this issue is to use nonlinear time series approaches, which describe how a system tends to change over time without specifying a mechanism for why it happens that way. However, these methods typically require long time series of data, such as abundance estimates for the same population of organisms over many decades. They also typically require “stationary” systems, meaning that the causal relationships driving the system—perhaps the effect of temperature on growth rates or the feeding rate of predators on prey—do not change over time. These requirements are rarely met in ecological settings.
- Methods: To address these two problems, the paper introduces statistical methods that have not previously been applied to nonlinear time series analysis—specifically, hierarchical Bayesian modeling .
- Findings: The paper integrates these techniques into a nonlinear approach and demonstrates that it can improve forecasts across a wide range of ecological scenarios. This methodology may be useful in ecosystem-based fisheries management when some of the important driving factors are poorly understood. It may also be useful in predicting how ocean ecosystems will respond to the non-stationary conditions created by climate change.

The full publication is available here:

http://www.sciencedirect.com/science/article/pii/S1476945X16300708.