Abstract. Global climate models predict relative humidity (RH) in the western US will decrease at a rate of about 0.1–0.6% per decade, although with seasonal differences (most drying in spring and summer), geographical variability (greater declines in the interior), stronger reductions for greater anthropogenic radiative forcing, and notable spread among the models. Although atmospheric moisture content increases, this is more than compensated for by higher air temperatures, leading to declining RH. Fine-scale hydrological simulations driven by the global model results should reproduce these trends. It is shown that the meteorological algorithms used by the Variable Infiltration Capacity (VIC) hydrological model, when driven by daily Tmin, Tmax, and precipitation (a configuration used in numerous published studies), do not preserve the original global model's humidity trends. Trends are biased positive in the interior western US, so that strong RH decreases are changed to weak decreases, and weak decreases are changed to increases. This happens because the meteorlogical algorithms VIC incorporates infer an overly large positive trend in atmospheric moisture content in this region. The result could downplay the effects of decreasing RH on plants and wildfire. RH trends along the coast have a weak negative bias due to neglect of the ocean's moderating influence. A numerical experiment where VIC's values of Tdew are altered to compensate for the RH error suggests that eliminating the atmospheric moisture bias could, in and of itself, decrease runoff up to 14% in high-altitude regions east of the Sierra Nevada and Cascades, and reduce estimated Colorado River runoff at Lees Ferry up to 4 % by the end of the century.