For many hydrologic applications information from seasonal climate forecasts must be disaggregated to daily weather sequences at individual stations. An attractive disaggregation method is through the use of a weather generator. Weather generators generate new weather sequences, which can be biased to reflect forecasted climate conditions. This paper introduces a simple method for generating weather. The method re-samples data sequences from the historical record nens times (nens=number of ensemble members), and re-orders the ensemble members to reconstruct the spatial correlation between neighboring stations, the correlation between variables (e.g., between precipitation and temperature), and the temporal correlation between subsequent days in the generated weather sequence. This paper also introduces methods for conditioning the weather generator on climate indices and probabilistic climate forecasts in which data are re-sampled from a biased set of years. The conditionded weather generator model is used to produce probabilistic forecasts of streamflow for three river basins in the contiguous USA.