High-resolution historical climate grids are readily available and frequently used as inputs for a wide range of regional management and risk assessments, including water supply, ecological processes, and as baseline for climate change impact studies that compare them to future projected conditions. Because historical gridded climates are produced using various methods, their portrayal of landscape conditions differ, which becomes a source of uncertainty when they are applied to subsequent analyses. Here we tested the range of values from five gridded climate datasets. We compared their values to observations from 1231 weather stations, first using each dataset’s native scale, and then after each was rescaled to 270-m resolution. We inputted the downscaled grids to a mechanistic hydrology model and assessed the spatial results of six hydrological variables across California, in 10 ecoregions and 11 large watersheds in the Sierra Nevada. PRISM was most accurate for precipitation, ClimateNA for maximum temperature, and TopoWx for minimum temperature. The single most accurate dataset overall was PRISM due to the best performance for precipitation and low air temperature errors. Hydrological differences ranged up to 70% of the average monthly streamflow with an average of 35% disagreement for all months derived from different historical climate maps. Large differences in minimum air temperature data produced differences in modeled actual evapotranspiration, snowpack, and streamflow. Areas with the highest variability in climate data, including the Sierra Nevada and Klamath Mountains ecoregions, also had the largest spread for snow water equivalent, recharge, and runoff.