Improving the sustainability of agriculture has become crucial to deal with tomorrow's challenges such as supplying food to a continuously growing world population while mitigating its environmental impacts (e.g. climate changes). Recycling organic wastes to substitute chemical fertilizers for various organic ones (e.g. sewage sludge, household refuses, plant residues, livestock manures, agro-food industrial wastes) is one of the ways towards this end. Addressing this calls for the coordinated use of heterogeneous knowledge on both the biophysical (i.e. organic products, soils, crops) and managerial (i.e. farmers' practices) components of the whole production systems. Computer models, encompassing various pieces of that knowledge, are built to represent these systems as linked production and consumption units spread over a territory. These models are used for simulating management scenarios and assessing their performances against agronomical and environmental criteria. This paper describes our main achievements: (i) a methodology for modeling and analyzing material flows on a territory scale; (ii) a conceptual modeling framework of farming systems; (iii) a way of representing human activity in farming systems based on the 'situated action' theory. It points also out two remaining issues: (iv) assessing simulated management scenarios; (v) using models with stakeholders to support their management practices.