< Back to previous page
High spatial resolution agent-based modelling at the country scale: an application of farming dynamics in Belgium.
Book - Dissertation
During the last decades, ongoing economic pressure as a result of increasing specialisation, mechanisation and globalisation have led to a continuous decrease in farmers numbers in many parts of the world. This process is leading to profound changes in the socio-ecological system of rural areas worldwide and confronts policy makers and rural planners with new challenges.To better understand such complex systems that are, to a certain extent, the outcome of the individual decisions of interacting agents, agent-based modelling (ABM) is a promising simulation tool. Reliable model simulations could provide more insight in current processes and in possible future evolutions in rural areas and could support decision making processes in rural planning. Since agent-based models (ABMs) require the simulation of the behaviour of every individual agent in the system, they need a large amount of data. Therefore, until now, the application of ABMs has been limited to small regions, or when applied to larger areas, with a great loss of detail due to strong generalization. There is therefore a gap between the level at which ABMs are designed to be used (the detailed, individual level) and the level that is relevant for policy making and planning (the regional or national level). This research aims to bridge that gap by developing and applying ADAM (Agricultural Dynamics through Agent-based Modelling): a simple agent-based farming model that operates at national scale but with the spatial resolution of individual fields. Belgium, that holds many different agricultural landscapes and farming types on a relatively small area, was used as a case study throughout this dissertation. In the first part of this work the current situation and trends of agriculture in Belgium are analysed and positioned in a global context. This enabled us to define relevant characteristics and key processes of the farming activities in Belgium. In the next part, these characteristics and key processes were generalized and put in to a conceptual framework that led to the development of ADAM. ADAM firstly estimates the drop-out and succession of farmers depending on both the characteristics of the farmer and his land. Farmlands without a successor are redistributed among neighbouring farmers or abandoned. The evolution of the agricultural population in ADAM was calibrated and validated with data from agricultural censuses for the period 2000-2010, resulting in an relative RMSE of 5.11% for the number of farmers and 46.4% for the evolution in the number of farmers when validated at the level of individual municipalities. The validation process showed a clear impact of urban expansion processes on the obtained results for Belgium. This impact can be direct through urban expansion on farm land or indirect when the farmland is used for suburban activities such as recreation or hobby farming. To incorporate the impacts of urban expansion, in the third part of this research, the original model was coupled with a constrained cellular automata land use change model.With this coupled model structure various scenarios on possible futures for Belgium's rural areas in 2035 were run. All scenario's show a continuous decrease of the number of farms and an increase in average farm size. The simulations show a very distinct spatial pattern with the highest decrease in farm numbers in the central part of the country and in the east of the country. In the last part, the results of the scenarios were used as an input for a species distribution model on bumblebees. The use of the high thematic resolution land use data as input allowed for a higher accuracy when modelling the distribution patterns of bumblebees. The added value of using these high thematic resolution land use data as input was seen when modelling more localized species as opposed to widespread bumblebee species, making the added value of the high thematic land use data dependent on the specific use case.