In pipe repair maintenance operations for leaky water distribution networks in cities, it is usually the practice that whenever pipe leaks are detected, repair actions are undertaken until the funds assigned are exhausted. The allocation of limited repair funds to different pipe sections is normally done without careful quantitative analysis but according to an intuitive rule, whereby the section of the larger leak volume receives the larger share of funds. This paper proposes a new modeling approach to determine optimal repair fund allocation to different sections of a city. A basic periodic-review stochastic demand inventory model is first applied to ascertain the optimal funds to be assigned to a city under an assumption of no fund limitation. The model is then extended to determine the optimal allocation of the limited funds to different sections of a city. The models minimize the expected total system cost composed of repair cost, storage cost (of spare parts and materials), and penalty cost for water loss due to lack of funds. Numerical experiments are then conducted to illustrate the applicability of the proposed models, and the characteristics of the optimal solutions with regard to parameter variations are investigated. Data used in the numerical example are taken from actual data of the secondary network of Ho Chi Minh City in Vietnam. ¿ 2000 American Geophysical Union