Around the world, long-term changes in the timing and magnitude of streamflow are testing the ability of large managed water resource systems constructed in the 20th century to continue to meet objectives in the 21st century. Streamflow records for unregulated rivers upstream of reservoirs can be combined with records downstream of reservoirs using a paired-watershed framework and concepts of water resource system performance to assess how reservoir management has responded to long-term change. Using publicly available data, this study quantified how the intra-annual timing of inflows and outflows of 25 major reservoirs has shifted, how management has responded, and how this has influenced reliability and vulnerability of the water resource system in the 668,000 km² Columbia River basin from 1950 to 2012. Reservoir inflows increased slightly in early spring and declined in late spring to early fall, but reservoir outflows increased in late summer from 1950 to 2012. Average inflows to reservoirs in the low flow period exceeded outflows in the1950s, but inflows are now less than outflows. Reservoirs have increased hedging, that is, they have stored more water during the spring, in order to meet the widening gap between inflows and outflows during the summer low flow period. For a given level of reliability (the fraction of time flow targets were met), vulnerability (the maximum departure from the flow target) was greater during periods with lower than average inflows. Thus, the water management system in this large river basin has adjusted to multi-decade trends of declining inflows, but vulnerability, that is, the potential for excess releases in spring and shortfalls in summer, has increased. This study demonstrates the value of combining publicly available historical data on streamflow with concepts from paired-watershed analyses and metrics of water resource performance to detect, evaluate, and manage water resource systems in large river basins.