In previous research we described a prototype system for medium range (up to two week lead) flood prediction intended for use in large rivers, primarily outside the developed world where in situ data are sparse. The vision is to rely heavily on weather prediction model output, and satellite remote sensing, so as to reduce the need for in situ precipitation and other observations in parts of the world where surface networks are sparse, and where a hydrologic forecast capability arguably would have the greatest value. The hydrologic component of the system is the Variable Infiltration Capacity (VIC) macroscale hydrology model. In the prototype, VIC generates hydrologic states for forecast initialization (nowcasts) using daily ECMWF analysis field wind, surface temperature and 24-hour ECMWF deterministic forecast precipitation. In hindcast mode, VIC is driven by global ECMWF Ensemble Prediction System 10-day forecasts. Previously, the ensemble precipitation forecasts were bias corrected with respect to the analysis fields and then downscaled using a resampling of the higher spatial resolution Tropical Rainfall Measuring System (TRMM) 3B42 precipitation. Although the mean errors of the forecasts were reduced by this procedure, the reliability of the forecasts was not improved, and the space-rank correlation between hydrologic model grid cells was not conserved. Here, we evaluate alternative approaches for the precipitation forecasts downscaling; one is based on a conditional regression and the so-called Schaake Shuffle and the other on the Hamill and Whitaker analog method. The performance is improved relative to the original method described above (bias correction followed by resampling). The procedure is evaluated through forecast verification over the Ohio Basin at one-quarter degree spatial resolution for the period 2002-2007.