Better ignorant than misled: Including uncertainty in forecasts supporting management and policy

Background/Question/Methods

The fundamental challenge of management is to choose actions that will meet goals for the future. Ecological modeling can play a role in supporting these choices by forecasting the future state of ecological systems under different alternatives for management. These forecasts must properly communicate uncertainty to decision makers. 

Results/Conclusions

I describe a broadly useful approach for communicating results of ecological forecasts used to support decisions on environmental management and policy. The central problem in managing the environment is to choose among alternative actions based on their ability to meet multiple goals, goals that may compete with one another. For each goal, there is often a range of outcomes that are acceptable. Bayesian state space models provide a useful way to evaluate the ability of actions to meet goals. A key feature of these models is the ability to separate uncertainty that arises from variance in environmental processes from uncertainty arising from errors in observing those processes. This separation enables honest forecasts represented as the probability distribution of model estimates of future states of the system being managed. The probability of meeting goals can be deduced by calculating the area of these distributions corresponding to acceptable values of an ecological state. It is important to compare the probability of meeting a goal given some action relative to the probability of meeting the goal if we do nothing. Doing nothing serves as a null model for assessing the value of taking action.

The approach has several desirable features. Increasing uncertainty in forecasts favors prudence because the probability of meeting goals given no action converges on the probability given an action. Actions will differ in their ability to meet different goals. Adaptive management and continuous evaluation of management actions are encouraged because uncertainty propagates with time, reducing the ability to make distinctions among potential actions far into the future. Model forecasts enhance the conversation among decision makers and stakeholders by plainly revealing the risks of action and inaction. Understanding these risks allows human values to enter the discussion about which actions should be taken, an outcome that is inevitable and desirable.