We will build an agent-based model (ABM) of farm households responding to policy, market, and environmental signals to examine how the evolution of agricultural production systems on the tropical forest frontier makes farmers and the regional economy more or less vulnerable to shifts in green and blue water availability. We hypothesize that including interactions among farmers improves the predictions of landscape scale models.
Our proposed ABM framework will model farm households as boundedly rational, future discounting, expected utility maximizers who choose portfolios of land use and investments. Farmers’ utility coefficients across elements of this portfolio (including labor and other farm inputs) will be estimated from discrete choice experiments. Framed field experiments will also provide other key preference parameters as identified in the focus groups.
Methods and Activities
Farm households are assumed to function independently, but added together they influence soil and surface water flows, commodity markets, and their peers’ decision-making. Farmers indirectly interact with each other by making decisions in responses to water availability, market prices and structure, and considering what their peers have done. Subproject 4 is focused in determining if including farmer group interactions improves the predictions of landscape-scale models.Our prior research in the region suggests that individual farmer decisions about upstream deforestation impact on downstream productivity in milk production. Our analyses in Subproject 3 will identify the most significant influences on adoption of new production systems, including measures of water availability as affected by land-use and land-cover changes, market access and competitiveness, and the decisions of other farmers in formal and informal social networks.
We operationalize vulnerability in the ABM as the ratio of i) the coefficient of variation in realized utility to ii) the coefficients of variation in different dimensions of the hydroclimate including precipitation, soil moisture, and stream baseflow. Similarly, we operationalize the concept of adaptive capacity as the ratio of i) the difference between realized utility under the current portfolio and hypothetical utility in the same time period under previous portfolios (had they been kept), and ii) the costs associated with moving from these prior portfolios to the current portfolio. These key outcomes of vulnerability and adaptive capacity are relative measures that will allow us to compare conditions across farmers in the same simulation, as well as within individual farmers along time and across different scenarios and parameterizations.