In this study, the impact of various rescaling approaches in the framework of data fusion is explored. Four different soil moisture products (Advanced Scatterometer; Advanced Microwave Scanning Radiometer for EOS, AMSR-E; Antecedent Precipitation Index; and Global Land Data Assimilation System-NOAH) are fused. The systematic differences between products are removed before the fusion utilizing various rescaling approaches focusing on different methods (regression, variance/cumulative distribution function (CDF) matching, multivariate adaptive regression splines, and support vector machines based), stationarity assumptions (constant or time-varying rescaling coefficients), and time-frequency techniques (periodic or nonperiodic high- and low-frequency components). Given that statistical descriptions (e.g., standard deviation and correlation coefficient) of reference data sets are utilized in rescaling approaches, the precision of the selected reference data set also impacts the final fused product precision. Experiments are validated over 542 soil moisture monitoring sites selected from the International Soil Moisture Network data sets between 2007 and 2011. Overall, results highlight the importance of reference data set selection-particularly that a more precise reference product yields a higher precision fused soil moisture product. This conclusion is sensitive neither to the number of fused products nor the rescaling procedure. Among rescaling approaches, the precision of fused products is most affected by the choice of rescaling stationary assumption and time-frequency decomposition technique. Variations in rescaling methods have only a small impact on the precision of pair fused products. In contrast, utilizing a time-varying stationary assumption and nonperiodic decomposition technique produces correlation improvements of 0.07 [-] and 0.02 [-], respectively, versus the other widely implemented rescaling approaches.