An Agent-Based Model (ABM) to simulate the influence of farmers’ social network

Alper Bayram, Antonino Marvuglia, Tomas Navarrete

Contact: alperbayram2@gmail.com

Agricultural activities cause relevant environmental impacts and the dynamics of agricultural systems can be very complex. The decision-making regulating those activities could benefit from using tools that explicitly include farmers’ behavioural aspects and sustainability information from LCA point of view. The mechanisms to improve the sustainability of farming systems and the ways to support farmers in their choices based on a trade-off between economic revenue and environmental impacts via Agent Based Models(ABMs) have been discussed in previous work. ABM is a promising tool to capture interactions between farmers and their individual behaviour in a couple framework with LCA. Understanding real-life interactions between the agents is crucial to create a viable model. While farmers interact with each other during fairs, they share info with their spatial neighbours more frequently. We apply constrained allocation of fields to approximate the geographical distribution of farms in Luxembourg and Wallonia, which is unavailable due to confidentiality. The knowledge of geolocation of farms is then used to model the transactions and information transfer between the farmers. Alongside farmer's age, farm type or size the behaviour of neighbor farmers also affect the decision-making process in our simulations. The environment for the analyses is the agriculture sector in Luxembourg and Wallonia. Agents try to maximize the overall profit of the country while taking individual green consciousness(GC) into consideration. GC is an attribute that works as a proxy for awareness towards sustainability metrics and used as part of decision making of the farmer as explained in. Simulation steps are: 1) Initialize the attributes using data from and crop prices for the duration of simulation. 2) Build network to realize the interactions. The network of farmers is initialized via Watts–Strogatz model, and the strength of connections evolve according to the farmers’ dynamic or static attributes (age, geospatial info, risk aversion etc.). They influence decisions, which reflect on the LCIA results of their activities. 3) Let agents react to stimuli and get the areas under each crop. Using the simulated land use changes, the environmental impact for different crop types are calculated using the ILCD LCIA method. 4) Update GC according to farmers' degree centrality and other attributes. 5) Current model only takes crop production into account, but the modelling of dairy and meat production is an ongoing work. In the case of Luxembourg in most of the farms dairy farming and crop production co-exist. The current agents' network only includes farmers. In the future we aim to add farmers' associations, consultants and governmental organizations into the picture. Since these organizations directly influence agents' behaviour, the members of associations can be clustered as one community. The real behaviour of the farmers depends on a multitude of factors that introduce uncertainties and model aims at reproducing these, based on real data (e.g. risk aversion is inferred based on a survey). The goal is giving the farmers and policy makers an insight on the possible impacts of the choices they make.

← Schedule