Dynamic models and large scale field experiments in environmental impact assessment and management

Dynamic numerical models and field experiments play important roles in impact assessment and management. Unfortunately, extreme and simplistic views have developed about whether and how to use these tools, so their complementary values to the manager are often not recognized. We often hear the outrageous claim (or hope) that numerical models can synthesize ‘all’ relevant information for predicting the impact of policy choice, hence making experimental experience unnecessary. From experimentalists, we hear the equally naive criticism that ecological systems are so complex that nothing is predictable without experimental experience. What we usually get from the proponents of these extremes are either models that are dangerously unreliable, or experiments that provide nice scientific answers to the wrong questions. Wise use of modelling begins with the following points: (i) explicit modelling is an excellent way to clarify policy concerns and identify processes that are most likely to be important in making predictions about policy effects; (ii) we can do a very good job of modelling some processes and relationships, particularly those having to do with basic spatial and temporal scales of impact as related to physical transport, chemical transformations, and life history characteristics of indicator populations (longevity, delays and response times due to age structured rates of reproduction and mortality); and (iii) there are some important dynamic processes, such as long‐term accumulation of toxic materials in the environment, that unfold over such large space and time scales as to preclude direct experimental study (leaving only the issue of which models to use in making predictions, not whether to model ‐ unless the processes are simply ignored). But points (i) and (ii) represent steps that a good experimentalist will take anyway: be clear about what practical results an experiment is intended to produce, and do not waste effort on experiments to measure things that can be predicted reliably from existing knowledge. The key to successful use of modelling and experimentation in management is in making good judgements about the interface between points (ii) and (iii); that is, in making good judgements about both what variables cannot be reliably predicted, and of these, which to treat experimentally and which to gamble on predicting from models.