Evaluation of an agricultural innovation in the presence of severe parametric uncertainty: An application of robust counterpart optimisation

Constrained optimisation models can efficiently predict the impacts of innovations on complex, agricultural systems. However, cost restraints, data limitations, and prediction errors typically constrain the accurate definition of coefficients or their distributions in such models. This paper employs the optimisation of a robust counterpart model to proactively deal with severe parametric uncertainty through defining uncertain parameters as members of closed intervals. This approach immunises the feasibility of mathematical—programming solutions against parametric uncertainty. A novel method of robust optimisation—developed by Bertsimas and Sim (Operations Research, Vol. 52 (2004), pp. 35–53)—allows solution using linear programming and manipulation of the conservatism inherent in optimal solutions. The identification of these stable solutions—instead of isolated, precise optima—represents a paradigmatic shift in the use of constrained optimisation in farm management modelling, yet is highly relevant since agricultural and economic systems are typically extremely dynamic, very heterogenous, and modelling them precisely is difficult due to data constraints. The utility of robust optimisation, relative to a deterministic approach, is demonstrated in the evaluation of a new perennial pasture species for Western Australian agriculture. It is shown using both approaches that this species is a profitable addition to farming systems given its tolerance of poor soils and provision of valuable feed at a time when other supplies are scarce.

► Optimisation models can provide erroneous output given parametric uncertainty.
► Robust optimisation identifies solutions robust to parametric uncertainty.
► Stable solutions more suited to broad dissemination given farm heterogeneity.
► Inherent conservatism beneficial for evaluation of agricultural innovations.