Sustainable farming strategies for mixed crop-livestock farms in Luxembourg simulated with a hybrid agent-based and life-cycle assessment model

Agent-based models (ABMs) are particularly suitable to simulate human-natural systems since they allow modelers to consider the behavioral aspects of individuals. Life Cycle Assessment (LCA) has also been widely used in research, industry, and policy to assess systems' environmental sustainability. In this paper, we introduce a coupled ABM-LCA model to simulate mixed crop-livestock farming activities in the Grand Duchy of Luxembourg. The simulator considers a wide range of aspects related to typical farming business activities and allows the calculation of the environmental and economic impacts of the decisions taken by agents as well. The paper simulates different scenarios. Scenario A is the baseline scenario. Scenario B considers reducing the stocking rate from 1.6 to 1.3 livestock units per ha (LSU/ha). Scenario C aims to reduce the soybean ratio in animals’ feed rations to the minimum possible level that is feasible for each farm. Scenario D simulates an increase in soybean autonomy: the farmer chooses to adopt local soybean (instead of imported one) into the crop rotation if its green consciousness exceeds a pre-fixed threshold. As expected, in scenario B, all impact categories show improvement with respect to the baseline scenario, the highest ones being an almost 25% reduction in freshwater eutrophication, 21% in climate change - human health, and 19% in freshwater ecotoxicity. For natural land transformation, the most significant improvements are obtained in scenarios C (11% reduction) and D (13% reduction). On the other hand, in scenario C, the change in feed composition, combined with an expected decrease in stocking rates, also has a positive effect (about 16% reduction compared to the baseline) on agricultural land occupation, due to the utilization of pasture and locally produced crops. An analysis of the systemic uncertainty calculated at the endpoint indicators level showed low coefficients of variation (CV) for all the scenarios, with scenario A always having the lowest values of CV in every impact category, scenario C having the highest CV in the ecosystem quality and the human health categories and scenario B having the highest CV in the resources category. The consideration of subsidy schemes in the Grand Duchy of Luxembourg allows the modelers to better interpret the resulting revenue and cost structures of different scenarios. The results show that the farm profitability stays on the same level when the stocking rate is reduced, and the subsidy granted for reducing nitrogen emissions is in place at the same time. Moreover, the soybean autarky in Luxembourg can be increased to 17% if the farmers are willing to incorporate soybean into their rotation scheme, thus reducing the impacts due to the transportation and production of soybean in South American countries.