Technical Abstract: Statistical approaches have been widely used to downscale global climate model (GCM) projections to finer spatiotemporal resolution for impact assessment of climate change. However, a major concern of the approaches is whether a statistical relationship derived from historical climate will hold for changed climate. The objective is to verify that a statistical method, which downscales monthly precipitation to daily series using a stochastic weather generator, is fully applicable to any climate state, using historical station data that have experienced non-stationary changes. Daily precipitation records of about 100 years from five meteorological stations, with mean annual precipitation ranging from 450 to 1330 mm, were split into calibration and validation periods based on abrupt changes in mean annual precipitation. Linear relationships between transition probabilities of wet-follow-wet (Pw/w) and wet-follow-dry (Pw/d) and mean monthly precipitation amounts were established between 30 driest months and 30 wettest months of the calibration periods, and were used to interpolate Pw/w and Pw/d for the validation periods. Mean and standard deviation of daily precipitation amounts were algebraically derived using mean and standard deviation of monthly precipitation values of the validation periods as well as the interpolated Pw/w and Pw/d. Statistics of downscaled daily precipitation amounts resembled those of observed daily amounts of the validation periods reasonably well. The downscaling method preserved statistics of monthly precipitation amounts extremely well, demonstrating the validity of the method for temporal downscaling of non-stationary climate. The linear relationships that transcend the driest and wettest hypothetical climate states are suitable for interpolating Pw/w and Pw/d with mean monthly precipitation to any intermediate climate state.