Strategies for ecological extrapolation

Errors associated with the spatial extrapolation of information are an important issue in ecology. Although improvements in instrumentation and software have increased our ability to represent complex behaviors, a measurable reduction in prediction errors is not always achieved with increased complexity. This difficulty in generating accurate predictions is due, in part, to trade-offs in error propagation that exist between simple and complex approaches: simple models often exclude key processes resulting in unknown prediction biases while complex approaches reduce prediction bias at the expense of increased estimation and measurement error. Our inability to anticipate and measure these trade-offs impedes our ability to define and control prediction errors.