The expansion and operation of water supply systems under growing demands, hydrologic variability, and water scarcity requires strategic decisions on supply sources for reducing and improving reliability and flexibility. The design and operation of such supply portfolio merits decisions of what and when to expand, and how much of each source to use considering interest rates, economies of scale and hydrologic variability. This paper provides an integrated framework to optimize water supply system expansions using dynamic programming and combining short and long term water supply source optimization using quadratic programming. The approach represents water supply infrastructure features and hydrologic conditions under increasing water demands. Results allow (a) identification of tradeoffs between cost and reliability of various expansion paths and supply source decisions, and (b) evaluation of potential gains from water system loss reductions, an often neglected supply augmenting alternative in developing countries. Results show the need for supply infrastructure expansions can be dampened by investments in reduction of systemwide losses. A losses reduction program cutting losses in half will cost R$ 532 million annually with a minimum system reliability of 60 percent. This program would yield net savings of 2.7 percent with respect to the status quo annual costs. Decreasing system losses from fifty to five percent, will cost R$ 547 million per year with a minimum system reliability of 87 percent. Earlier investments in system loss reduction programs will greatly improve reliability and system expansion costs. The framework proves its usefulness for planning its transferability to similarly urbanized systems.