Simulation of whole farm management decisions

Abstract

The simulation model APSFarm is an extended configuration of the systems model APSIM, and has been in development since 2005 to support the analysis of farm case studies in environments ranging from Central Queensland to the Victorian Riverina. One of the major extensions to APSIM was the implementation of dynamic farm management as a set of state transition networks, with each network representing the operation of a particular paddock or management unit. Each paddock has a current state (eg. fallow, crop), “rules” that allow transition to adjacent states, and “actions” that are taken when such a transition is made. These rules represent feasibility (eg. whether it is the correct planting season for this particular crop, whether machinery is available), tactics (eg fertiliser management), and strategy (eg. crop sequencing and mix of enterprise). Each day, the model examines all paths leading away from the current state to adjacent states. Should the mathematical product of all rules associated with a path be non-zero, the path becomes a candidate for action. Should more than one candidate be present, the highest ranking path is chosen, and its associated management actions to change state are undertaken (eg. sowing or harvesting a crop). There may be cascading events that flow from a state change, so the process is repeated until nothing more can be done for that day. The benefits of the network approach are multifold; gross patterns can be identified from the network structure that allows comparison between farmers and farm types. From a software perspective, details of the transition rules and their associated actions are readily accessible, instead of being encoded in complex logic code constructs. Encapsulating management strategy as a data structure instead of a series of specific instructions allows the construction of dynamic analysis tools in which complex decisions can be more clearly described as sets of simple and measurable rules of thumb. Further, this data structure can make the task of providing a cohesive user interface much easier, facilitating graphical representations of farm management. This paper presents a software tool used in studying diverse management systems drawn from ongoing case studies, discusses how the management systems are elucidated from farm managers and demonstrates some of the diagnostics available from such tools. © MODSIM 2009.All rights reserved.