Canadian Precipitation Analysis (CaPA) has been developed by Environment Canada to produce the most accurate near-real-time gridded precipitation estimates. It uses the Global Environmental Multiscale model (GEM) as a background and assimilates the synoptic network of weather stations through Optimal Interpolation. Accurate estimation of gridded precipitation is useful for hydrological modeling, stream ow forecasting, and climate change studies. However, the calibration and validation of hydrologic models requires long temporal coverage of data for a better performance. Since GEM/CaPA data are available only for the recent past (2002-present), the development of historical data sets starting earlier than 2002 becomes important. Using alternative models for producing the atmospheric gridded background is one solution to overcome the short temporal coverage of archived GEM data. This thesis evaluates and analyzes two candidate data sets. ERA-Interim and NARR were selected as potential alternatives to GEM background. The general conclusion of the study is that the use of ERA-Interim and NARR as background elds leads to performance results that are not signi cantly inferior to GEM after assimilation with stations in the CaPA framework. While result with the GEM background remains the best, one can cautiously conclude that for most practical applications, ERA-Interim and/or NARR may be used for the period that predates archived GEM data. The thesis presents a more detailed evaluation of ERA-Interim and NARR for di erent seasons and di erent regions of Canada.