Precision yield data is commonly recorded by modern combine harvesters and can be used to help growers optimize their operations. However, there have been very few attempts to predict variation in yield within a given field using multispectral satellite data. We used a precision yield dataset gathered in canola (Brassica napus L.) crops in central Alberta, Canada, and a time series of medium-resolution Sentinel-2 data collected over the growing season. Using two mapping methods, random forest regression and functional data analysis, we were able to predict crop yield to within 12-16% accuracy of actual yield, and to capture within-field variation. Our results demonstrate that time series of medium-resolution multispectral imagery is capable of mapping small-scale variation in crop yields, presenting new research and management applications for these techniques.