The MEPSOM project was supported financially by ANR Syscomm from 2010 to 2013 and selected as a flagship project of this call for proposal.

1. Experimental results obtained in the MEPSOM project, based on a series of microcosms experiments, showed the importance of the habitat of soil microorganisms, and especially how physical characteristics (pore sizes, connectivity) control the decomposition of organic substrates (Juarez et al. 2013a ; Ruamps et al., 2013). Our results even suggest that accessibility to substrates is more important than microbial diversity and microbial metabolic potential (Juarez et al., 2013 b, Pinheiro et al., 2015).

2. A second achievement of the MEPSOM project was to develop a suite of methods and models to explicitly describe 2D or  3D soil heterogeneity at scales relevant for microorganisms, i.e., micrometers:  the spatial distribution of solid and voids using µCT (Houston et al. 2013), the spatial distribution of water using synchrotron-based µCT and a lattice Boltzmann model (Pot et al., 2014), the spatial distribution of organic matter using a combination of soil organic matter staining with heavy metals and synchrotron-based µCT (Peth et al., 2014) and the spatial distribution of microorganisms in soils using a modeling approach that was tested with soil thin sections observed via fluorescence microscopy, (Raynaud and Nunan, 2014) as shown in Figure 1.

3. MEPSOM has contributed successfully to the development of three very complementary 3D models (Table 2) able to simulate for the first time the microbial degradation of organic matter at the scale of microhabitats in soil using real 3D images of soils.

(1) the Mosaic model is based on a geometric approaches to simulate soil structure, (Ngom et al., 2012, Monga et al. 2014)

(2) the LBioS model (Vogel et al., 2015) is based on lattice Boltzmann approach

(3) the µFun model  (Falconer et al. 2012) simulates fungal spreading in soil pores

Regarding pore architecture representation, LBios and µFun use directly the voxel based description extracted from 3D Computed Tomography images. By contrast, Mosaic takes advantage of advanced 3D computer vision and shape modelling algorithms to approximate pore space in a compact and intrinsic way using a limited number of geometrical primitives. These 3D models also offer various approaches used to simulate water physics and biological processes (Vogel et al., 2005): accurate description of diffusion is performed by the classical finite element method (FEM) for solving the partial differential equations (PDE) of diffusion-reactivity for µFun and the lattice Boltzmann method for LBioS while a simplified graph-based method is used by MOSAIC. Both Mosaic and LBioS simulate the decomposition of organic matter by bacteria while µFun simulates the growth of fungi. These new models, presented in Table 1 are all able now to describe the pore space distribution using 3-D computed tomography (CT) images of soil and to simulate numerous spatial interactions within and between pores such as water distribution, diffusion of carbon, microbial growth, and CO₂ efflux due to respiration. The Mosaic model was tested using experimental data of simple sand systems (Monga et al. 2014). Management of large numerical data is one of the key factors determining the ability to run the models in 3D using real soil images.

4. A sensitivity analysis of the 3D models, µFun (Falconer et al., 2012) and LBioS (Vogel et al., 2015) was carried out using 3D CT images of real soils. The sensitivity analyses have highlighted the role of bacteria location as single factor and the role of interactions among bacteria location, water content and architectures of pores as combined factors to explain respectively, 28% and up to 51% of the variance of the simulated fluxes of CO2, µB and DOC (Vogel et al., 2015). In line, the water content decreased the colonization efficiency of the fungi and the architecture of the soils combined to water distribution had an effect on the general morphology of the hyphal network (Falconer et al., 2012). These results demonstrate thus the role of the soil micro-heterogeneities in explaining soil microbial activity.