The paper deals with a means to improve nitrogen removal in an alternating activated sludge waste water treatment process by automatic control. A control strategy is formulated involving the on demand increase in denitrification rate by the addition of a readily metabolizable carbon source to the point in the process where denitrification occurs. An algorithm is presented which automatically adjusts the carbon addition rate to match current process conditions. Two models are employed : a prediction model based on mass balances to determine the needed denitrification rate; and an experimentally determined, empirical relational model describing the denitrification rate as a function of carbon addition rate. An adaptive technique is employed to continually update estimates for the slowly changing parameters of the relational model. The algorithm is illustrated with a simulated example and two pilot plant demonstrations.