The paper proposes the use of meta models to determine the optimal intertemporal management of soil and phosphorus losses from agricultural land. This approach allows finding a equilibrium of the economic and biophysical system simultaneously. In contrast to the existing literature the model takes account of nonlinear relationships and of a large number of agricultural activities. The mathematical problems arising from this complex setup are addressed and the model is solved numerically. The results show that the second best policy in form of soil protection scores is highly inefficient, while another second best policy in form of land-use taxes is nearly as efficient as the first best policy.