We develop an optimal control model for cost-effective management of pollution, including two state variables, pollution stock and ecosystem quality. We apply it to Baltic Sea pollution by nitrogen leachates from agriculture. We present a sophisticated, non-linear model of leaching abatement costs, and a simple model of nitrogen stocks. We find that significant abatement is achievable at reasonable cost, despite the countervailing effects of existing agricultural policies such as price supports. Successful abatement should lead to lower nitrogen stocks in the sea in 5 years or less. However, the rate of ecosystem recovery is less certain. The results are highly dependent on the rate of self-cleaning of the Baltic Sea, and less so on the discount rate. Choice of target has a radical effect on the abatement path chosen. Cost-effectiveness demands such a choice, and should therefore be used with care when stock effects are present.