A systematic comparison of the surface wind speed records extracted from the North American Regional Reanalysis (NARR) database and the network of automated surface observation stations of the Mexican National Weather Service has been conducted. The work presented here is based on a multi-step procedure that initiates with a point-to-point comparison of surface observations and NARR grid point, followed by the reconstruction of missing data from the quality-controlled observational data set using an iterative technique based on principal component analysis (PCA). Subsequently, the empirical orthogonal function (EOF) sets for NARR and the reconstructed observational data set were compared globally using a group comparison criterion that is also based on PCA. Different observational periods were evaluated, leading to the highest similarity between the two sets for the case in the year 2010. A wind speed map for Mexico was created by joining the quality-controlled, filtered and reconstructed observational data with the NARR map using a weighting technique based on distance weighting, local similarity of the EOFs and data recovery.