We investigate two sources of uncertainty in observational estimates of the relationship between changes in Lees Ferry naturalized flow and precipitation and temperature in the basin: observational uncertainty and meteorological variability. Uncertainty in estimating trends and multi-decadal variability in naturalized flow records and in the gridded estimates of precipitation and temperature are first considered. While temperature trends are largely consistent among datasets, precipitation trend estimates differ significantly, resulting in a range of estimates that is as large as that fr Meteorological variability is investigated in two ways: the first is a standard block-bootstrapping resampling of the historic record, the second is using a large ensemble of GCM simulations as a proxy for natural variability. Both methods indicate a large range of possible temperature sensitivity consistent with the historic record. Finally we present a suite of ensemble AMIP-style simulations from 3 different GCMs (NCAR CAM5, ECHAM5 and MRI), for both historic and counterfactual “1880s climate” scenarios at approximately 50-70 km resolution. These models exhibit a temperature sensitivity at the low observational uncertainty.