Agent-based modelling as a time machine to assess nutrient cycling reorganization during past agrarian transitions in West Africa

Although in agronomy most model-based simulations explore future agro-ecosystems, we used a simulation model to explore past agro-ecosystems, which was particularly useful in a context where historical quantitative data on nutrient flows are lacking. The aim of this study was to assess the impact of an agrarian transition on the reorganization of nutrient cycles in agro-sylvo-pastoral systems (ASPS) in West Africa. The model was an agent-based model called TERROIR (TERRoir level Organic matter Interactions and Recycling model) that analyzes nutrient cycles at three levels of organization: plot, household, and landscape. Simulated scenarios were defined based on the agrarian transition that occurred in the 1920-2010 period in the groundnut basin of Senegal, a rapid transition caused by strong population growth, high climate variability and fluctuating cash crop markets. The main trends of the simulated agrarian transition are expansion of croplands onto rangelands and a shift from subsistence to market oriented farming systems, resulting in a shift from a livestock feeding system based on free grazing to a system based on the increasing use of feed concentrates and crop residues after harvest. The transition led to intensification of nitrogen flows and to reduced nitrogen use efficiency, due to the accumulation of nutrients in the area near the homestead and incomplete return of nutrients to cultivated plots. Two major properties appear to be outlasting the transition: (i) independence towards external inputs, based on crop-livestock integration, i.e. high biomass and nutrient recycling within the system; (ii) spatial heterogeneity due to nutrient transfers from peripheral land units to core land units, mainly through livestock. We argue that the persistence of these two emerging properties is a key pattern of past trajectories that can be used to make assumptions on and explore future ASPS trajectories.