10.7.3 The Model
10.7.3.1 Model Components and Dynamics
The application of MP-MAS here had four components:
• Agent decision-making is at the nucleus of MP-MAS. In this study, each agent
represented an individual farm household and there were an estimated total of
1,309 farm households in the Mae Sa watershed – the study area (see Chap. 1).
The land use decisions of farm households were simulated by optimizing net
household incomes under resource and knowledge constraints, including
monthly land, labor and water constraints, an annual cash constraint and the
level of knowledge on innovations. Agents decided what crops to plant on what
type of land, as well as what inputs (fertilizers, pesticides, labor and irrigation)
and how much of each to apply. Decision-making was separated in terms of
investment, production and expenditure decisions, as discussed in
Schreinemachers and Berger (2006, 2011). Prices were assumed constant
throughout the simulation, which was run for 15 years. Agents annually updated
their expectations about crop yields, rainfall and irrigated water supply, based on
the theory of adaptive expectations first implemented in MP-MAS by Berger
(2001).
• The landscape represented was the 140 km
2 Mae Sa watershed area. Within the
landscape, agricultural fields were represented as pixels, 40x40 m in size, and
these were divided into twelve types of agricultural land, determined by
combinations of average slope gradient (less than 8 %, 8–19 %, 20–35 % and
above 35 %) and average altitude (below 650 m, 650–1,000 m and above
1,000 m). The decision-model constrained what types of land were suitable for
what crops; for instance, litchi could only be grown on pixels above 650 m above
sea level, and steep areas were assumed to require more labor than flat areas.
Other than these physical characteristics, soil fertility was not considered in this
application.
• Crop production was modeled as based on the FAO CropWat model (Smith
1992), which assumes that if average monthly crop water demand is not met by
sufficient crop water supply then the crop yield reduces proportionally. The crop
water supply was modeled as the sum of effective rainfall and irrigation water
supply (from groundwater pumping or reservoirs), and for each farm household
was approximated using a backward calculation based on the observed land use,
irrigation methods and effective rainfall, and assuming no water shortage in the
year of data collection.
• Litchi fruit yields are a function of the age of a tree, the intensity of the
management applied and the water supply. Three levels of management were
defined in the model, namely unmanaged, poorly managed and well managed.
Artificial flower induction (AFI), drip irrigation, and AFI plus drip irrigation
were introduced as three alternative management options for well-managed
orchards. The model allowed agents to switch between management levels and
10 Integrated Modeling of Agricultural Systems in Mountainous Areas
415
10.7.3.1 Model Components and Dynamics
The application of MP-MAS here had four components:
• Agent decision-making is at the nucleus of MP-MAS. In this study, each agent
represented an individual farm household and there were an estimated total of
1,309 farm households in the Mae Sa watershed – the study area (see Chap. 1).
The land use decisions of farm households were simulated by optimizing net
household incomes under resource and knowledge constraints, including
monthly land, labor and water constraints, an annual cash constraint and the
level of knowledge on innovations. Agents decided what crops to plant on what
type of land, as well as what inputs (fertilizers, pesticides, labor and irrigation)
and how much of each to apply. Decision-making was separated in terms of
investment, production and expenditure decisions, as discussed in
Schreinemachers and Berger (2006, 2011). Prices were assumed constant
throughout the simulation, which was run for 15 years. Agents annually updated
their expectations about crop yields, rainfall and irrigated water supply, based on
the theory of adaptive expectations first implemented in MP-MAS by Berger
(2001).
• The landscape represented was the 140 km
2 Mae Sa watershed area. Within the
landscape, agricultural fields were represented as pixels, 40x40 m in size, and
these were divided into twelve types of agricultural land, determined by
combinations of average slope gradient (less than 8 %, 8–19 %, 20–35 % and
above 35 %) and average altitude (below 650 m, 650–1,000 m and above
1,000 m). The decision-model constrained what types of land were suitable for
what crops; for instance, litchi could only be grown on pixels above 650 m above
sea level, and steep areas were assumed to require more labor than flat areas.
Other than these physical characteristics, soil fertility was not considered in this
application.
• Crop production was modeled as based on the FAO CropWat model (Smith
1992), which assumes that if average monthly crop water demand is not met by
sufficient crop water supply then the crop yield reduces proportionally. The crop
water supply was modeled as the sum of effective rainfall and irrigation water
supply (from groundwater pumping or reservoirs), and for each farm household
was approximated using a backward calculation based on the observed land use,
irrigation methods and effective rainfall, and assuming no water shortage in the
year of data collection.
• Litchi fruit yields are a function of the age of a tree, the intensity of the
management applied and the water supply. Three levels of management were
defined in the model, namely unmanaged, poorly managed and well managed.
Artificial flower induction (AFI), drip irrigation, and AFI plus drip irrigation
were introduced as three alternative management options for well-managed
orchards. The model allowed agents to switch between management levels and
10 Integrated Modeling of Agricultural Systems in Mountainous Areas
415