249The whole-farm model APSFarm (Rodriguez et al. 2011 ) developed over the last
10 years is an extended configuration of APSIM, which allows users to simulate the
impacts (i.e. economic, financial, environmental) of the alternative allocation of
limited production resources (e.g. land, labour, time, irrigation water, livestock,
machinery, and finance) across alternative farm enterprises at the whole-farm level.
Example APSFarm simulation projects were first released in the APSIM 7.2 release,
including its application for large-scale commercial rainfed cropping (Rodriguez
et al. 2011 ), irrigated cropping (Power et al. 2011 ) and mixed grain and grazing
farm businesses (Rodriguez et al. 2014 ) (Fig. 7 ).
2.7 The APSFarm Technology
Modelling farms (compared to fields) to study resource allocation questions brings
further complications, and simulating those complications require a change from
fixed ʻcalendar-drivenʼ management rules to dynamic sets of rules that operate
across the whole spectrum of farm activity: land, labour, water, livestock and
machinery, to name a few. The introduction of APSIM modules implementing lan-
guages that support higher level data manipulation allowed APSFarm developers to
simulate farming system management as a set of state and transition networks, or
finite state automata, allowing rules for farm management to be studied in detail
without affecting higher level function. In APSFarm, each field has a current state
(e.g. fallow, wheat, sorghum) and ‘rules’ that allow transitions between adjacent
states (e.g. wheat – fallow – sorghum). These rules represent the capacity (e.g. avail-
ability of machinery, land, labour), capabilities (e.g. agronomic and technical skills)
and preferences (e.g. farm business strategies, risk attitude of the farm manager).
Rules can be expressed as a boolean value (true for feasible, false for otherwise) and
can take real values with higher values representing the desirability of a particular
action. Each day, the model examines all paths leading from the current state to
adjacent states, and if the product of all rules associated with a path is non-zero, it
becomes a candidate for action. The highest-ranking path is taken and the process
repeats until nothing more can be done that day. Rules can represent farm-level
criteria such as sowing windows for each crop, temporal definitions of ʻbreak of the
seasonʼ such as mm of rainfall over a defined time period, physical limitations of
farm components such as machinery operation, and preferential behaviour such as
the maximum farm area that could be sown to each crop – farmersʼ ʻrisk aversion’.
Rules of field-level criteria include minimum extractable soil water (ESW, mm)
required to sow a crop, definitions of a ʻsowing opportunity’, the cropping history
of that paddock that may force the inclusion of a break crop, the level of ground
cover, etc.
The results from actions having economic implications (e.g. variable costs from
fertiliser use, the need for seeds to plant a crop, or profits from crop harvesting) are
calculated based on a list of expected costs and prices provided by the participating
farmers. Fixed farm operational costs are also obtained from the participating farmer
Modelling Dryland Agricultural Systems