Uncertainty analysis applied to supervised control of aphids and brown rust in winter wheat. Part 2. Relative importance of different components of uncertainty

Abstract

The components of an existing model for supervised control of aphids (especially Sitobion avenae) and brown rust ( Puccinia recondita) in winter wheat contain uncertainty. Their contribution to uncertainty about model output is assessed. The model simulates financial loss associated with a time sequence of decisions on chemical control as a function of crop development, population growth, and damage. Four sources of uncertainty were quantified: model parameters, incidence sample estimates, future average daily temperature, and white noise. Uncertainty about the first two sources is controllable because it decreases when more information is collected. Uncertainty about the last two sources is uncontrollable, given the structure of the model. Uncertainty about model output, characterized by its variance, is calculated by repeatedly drawing realizations of the various sources of uncertainty, and calculating financial loss after each draw. By processing new realizations of these sources one by one, the contribution of each component to total variance can be assessed using an adapted Monte Carlo procedure. For most relevant initial conditions and decision strategies the sources of uncontrollable uncertainty cause more than half of the uncertainty about model output. White noise in the relative growth rates of aphids and brown rust is the most important source of uncertainty. Resources for improvement of the model are most effectively allocated to studies of the population dynamics of aphids and brown rust.