The strong link between poverty, natural resources and the environment is apparent in smallholder agriculture: farmers are making repeated land use and management decisions while facing diverse resource endowments and significant environmental constraints on production. To investigate the likely effects of changes in agricultural practices on the natural resource base and on farmer welfare, we develop a bio-economic dynamic model of agricultural households in the western Kenya highlands. Our modeling framework extends economic farm household models to incorporate the dynamic nature of natural resource management and its implications for household welfare, and to permit a meaningful interface with biophysical processes through soil carbon management. Using an eight-year panel data set, the model combines econometrically estimated production and soil carbon flow equations in a dynamic programming framework. We use the model to determine the optimal management of the farming system over time in terms of the quantity of mineral fertilizer and crop residues to apply, taking into consideration initial resource endowments and prices. Understanding how soil resources respond to the combined applications of mineral and organic resources is important for improved resource allocation at the farm level and for national agricultural policy decisions.