Crop model – this model was based on the CropWat model (Smith 1992)
developed by the Food and Agricultural Organization of the United Nations
(FAO). Actual yields can be estimated in the model using the methodology proposed by Doorenbos and Kassam (1979), in which the actual yield is linearly related
to the evapotranspiration deficit, which is determined by the ratio between cumulative values of actual evapotranspiration (ETA) and maximum evapotranspiration
(ETM) during the growing period.
Soil erosion model – this model was based on the Universal Soil Loss Equation
model (USLE model) proposed by Wischmeier and Smith (1978). In this model, at
each year in the simulation, soil loss in the study area was quantified subject to five
significant factors, namely rainfall erosivity, soil erodibility, slope length and slope
gradient, crop management and conservation practices.
For the socio-economic component, the CatchScapeFS model included farm
household agents and other social elements based on a farming systems approach
(Potchanasin 2008). To generate farm agents, cluster analysis and qualitative
analysis were applied, together with Monte Carlo techniques (Schreinemachers
2006; Potchanasin 2008). These analyses allowed a population of 60 agents to be
created, which was statistically consistent with the 30 sample households from the
2004 survey. The average amount of resources allocated to the agents, such as the
number of persons per household, land holding, heads of livestock, cash, debt and
stored rice, was not significantly different from the average for the farm households
in the survey, which indicates a close statistical fit between the agent population and
the sample population.
Each simulation time step corresponded to a 10 day time interval, and the model
simulated six dynamic phases, which were: biophysical dynamics, farm household
activities, socio-economic dynamics, information exchange and result
arrangements. Each phase contained model processes that were performed in
sequence; for instance, the farm household activities phase contained farm agent
decision processes, such as the selection of which crops to grow. Farm household
activities consisted of eight sub-phases: knowledge base updating, household
resource updating, cropping activities, household activities, harvesting, the selling
of farm products, financial activities and livestock activities. At each time step,
farm agents executed all the processes, some of which required agents to interact
with other agents or objects; for instance, asking for price information and crop
alternatives, asking for loans and changing crop properties on their plots. In the subphase of the cropping activities, farm agents examined their own plots; whether
they were ready to be cropped – such as having already been left fallow, or used to
produce the main crop from the previous year (to keep cropping patterns in line with
reality). Then, if plot properties reached the test conditions, the farm agent
interacted with the abstract object, a form of behavior used to select crops using
their own strategies. For example, if an agent was risk-seeking, it would execute its
own risky strategy to select the crop, whereby a crop which could generate a higher
level of income would be preferable to a crop grown with government support or for
subsistence purposes. After that, the agent tested or considered the other resources
available and the subsistence conditions, such as cash, labor and consumption
400
C. Marohn et al.
developed by the Food and Agricultural Organization of the United Nations
(FAO). Actual yields can be estimated in the model using the methodology proposed by Doorenbos and Kassam (1979), in which the actual yield is linearly related
to the evapotranspiration deficit, which is determined by the ratio between cumulative values of actual evapotranspiration (ETA) and maximum evapotranspiration
(ETM) during the growing period.
Soil erosion model – this model was based on the Universal Soil Loss Equation
model (USLE model) proposed by Wischmeier and Smith (1978). In this model, at
each year in the simulation, soil loss in the study area was quantified subject to five
significant factors, namely rainfall erosivity, soil erodibility, slope length and slope
gradient, crop management and conservation practices.
For the socio-economic component, the CatchScapeFS model included farm
household agents and other social elements based on a farming systems approach
(Potchanasin 2008). To generate farm agents, cluster analysis and qualitative
analysis were applied, together with Monte Carlo techniques (Schreinemachers
2006; Potchanasin 2008). These analyses allowed a population of 60 agents to be
created, which was statistically consistent with the 30 sample households from the
2004 survey. The average amount of resources allocated to the agents, such as the
number of persons per household, land holding, heads of livestock, cash, debt and
stored rice, was not significantly different from the average for the farm households
in the survey, which indicates a close statistical fit between the agent population and
the sample population.
Each simulation time step corresponded to a 10 day time interval, and the model
simulated six dynamic phases, which were: biophysical dynamics, farm household
activities, socio-economic dynamics, information exchange and result
arrangements. Each phase contained model processes that were performed in
sequence; for instance, the farm household activities phase contained farm agent
decision processes, such as the selection of which crops to grow. Farm household
activities consisted of eight sub-phases: knowledge base updating, household
resource updating, cropping activities, household activities, harvesting, the selling
of farm products, financial activities and livestock activities. At each time step,
farm agents executed all the processes, some of which required agents to interact
with other agents or objects; for instance, asking for price information and crop
alternatives, asking for loans and changing crop properties on their plots. In the subphase of the cropping activities, farm agents examined their own plots; whether
they were ready to be cropped – such as having already been left fallow, or used to
produce the main crop from the previous year (to keep cropping patterns in line with
reality). Then, if plot properties reached the test conditions, the farm agent
interacted with the abstract object, a form of behavior used to select crops using
their own strategies. For example, if an agent was risk-seeking, it would execute its
own risky strategy to select the crop, whereby a crop which could generate a higher
level of income would be preferable to a crop grown with government support or for
subsistence purposes. After that, the agent tested or considered the other resources
available and the subsistence conditions, such as cash, labor and consumption
400
C. Marohn et al.