Predicting our predictions: Reporting uncertainty in forecasting tools

under development for ecosystem-based fisheries management

Kerim Aydin

Alaska Fisheries Science Center, NOAA Fisheries

Sustainable fisheries management in large marine ecosystems (LMEs) relies on both short-term and long-term predictions. For short-term forecasts, sustainable year-to-year exploitation relies on the assumption that fishing is reversible, and models currently in use tend to predict that, over a wide range of exploitation rates, a reduction of fishing would allow an ecosystem to return to its "natural state." In the long-term, climate scenario modeling released by the Integovernmental Panel on Climate Change (IPCC) is leading to the development of long-range forecasting tools for fisheries. For example, between 2000 and 2005, an unprecedented lack of ice cover in the southeast Bering Sea may have lead to substantial changes in plankton production and possible consequences for upper trophic levels; attempts to incorporate IPCC predictions of ice cover into forecasts of ecosystem production are currently underway. However, in light of past fisheries collapses and "environmental regime changes," it is important to evaluate the way we view and communicate predictions and uncertainty. Here, I review concepts and models of fisheries sustainability, specifically with reference to the current state-of-the-art practices in single-species and multi-species population modeling and statistical analysis as used within the Alaskan groundfish fisheries management system. Are we using the right statistics in predicting catch and abundance trends, and (even if we report

our confidence limits) are we successfully communicating our uncertainty to stakeholders and decision-makers? As a partial answer, I outline metrics of variability beyond those in current use which may increase our chances of success in practicing ecosystem-based management.