SWORM: an agent-based model to simulate the effect of earthworms on soil structure

International audience Soil structure can be defined as the spatial organization of solid mineral and organic particles, and pore space. It is of great importance for soil functioning as it drives ecosystem functions (carbon sequestration, emission of greenhouse gases, nutrient cycling, primary productivity, etc.). Soil structure results from biotic and abiotic factors. Among biotic factors, numerous studies have shown the importance of organic matter, microorganisms, roots and invertebrates. Earthworms are known to play a key role in soil structure formation and maintenance through a continuous production of biogenic structures (casts and burrows). As far as we know, no models describe or quantify the effect of soil invertebrates on soil aggregation and porosity. It is a challenge to describe the physical soil environment for purposes of modelling because a soil is a multi-scale heterogeneous, three-dimensional and dynamic environment. An approach based on fractal theory (often used in soil sciences) was chosen to model such a real complex environment; it was integrated into a multi-agent system (MAS), which allows us to simulate agents (e.g. earthworms) situated in a virtual world (e.g. soil). It is a bottom-up approach that allows us to describe a system at a micro level (e.g. earthworms and their local soil environment) in order to observe, during simulations, macroscopic changes (e.g. soil structure evolution, organic matter dynamics, and microbial functions). In this paper we describe the SWORM (for 'Simulated Worms') model and the simulator, and present the results of the simulation applied to a case study. The effect of compacting and decompacting earthworm species on the structure of humid savanna soil at Lamto in Cote d'Ivoire has been widely studied. Quantitative and graphical outputs (e.g. thin sections of the virtual soil) indicate that the simulator was able to reproduce the effects of both compacting and decompacting species. Different ways to improve the model are discussed.