This dataset includes model projections of seasonal temperature (T), precipitation (P), and runoff (R) from 214 climate simulations from coupled model intercomparison project (CMIP) 3 and CMIP5 scenarios for 19-year periods centered on 2030, 2060, and 2090. The summaries of the climate model projections are presented as percentiles (5th, 25th, 50th, 75th, and 95th) of seasonal (October through March, January through March, April through June, and July through September) changes in T, P, and R for the 214 climate models. The metrics are calculated from variables previously summarized across the conterminous United States for hydrologic response units of the Geospatial Fabric for National Hydrologic Modeling (Viger and Bock, 2014). T, P, and R were previously derived using a monthly water balance model (Bock and others, 2016; 2017). Names, sources, and references of the climate inputs are described in Bock and others (2017).