Exploring profit - Sustainability trade-offs in cropping systems using evolutionary algorithms

Models that implement the bio-physical components of agro-ecosystems are ideally suited for exploring sustainability issues in cropping systems. Sustainability may be represented as a number of objectives to be maximized or minimized. Models provide a means to evaluate system performance for alternative strategic, tactical and operational decisions. The full decision space is usually very large and simplifications based on insight into agronomic relations and farming practice are necessary to safeguard computational feasibility. Different optimisation approaches have been proposed in the literature, usually based on mathematical programming techniques. The disadvantage of these techniques is that they can only deal with greatly simplified system descriptions, thus excluding the direct use of bio-physical system models. Here, we present a global search approach based on an Evolutionary Algorithm (EA). We introduce a multiobjective evaluation technique within this EA framework, linking the optimisation procedure to the APSIM cropping systems model, and perform simulations in parallel over an existing computer network. A case study addressing crop choice and sowing rules in north-east Australian cropping systems is used to illustrate the EA methodology. Sustainability of these systems is evaluated in terms of economic performance (gross margin, financial risk) and resource use (erosion). Because of the limited size of the problem, the quality of the EA optimisation can be assessed by comparison to a large sample of the full problem domain. Results demonstrate that the EA procedure, parameterized with generic parameters from the literature converges to a useable solution set within a reasonable amount of time. Frontier “peels” or pareto-optimal solutions as described by the multiobjective evaluation procedure provide useful information for discussion on trade-offs between conflicting objectives. Application to a more realistic decision problem is recommended to evaluate the usefulness of the approach in informing discussions on agro-ecosystem design.