Wastewater sampling has been shown to be an effective tool for monitoring the dynamics of an infectious disease. During the COVID-19 pandemic, many sampling sites were opened in order to capture as much information as possible. However, with the pandemic waning, not all sampling sites need to continue operating.In this work, we investigate a method for evaluating sampling sites for which sampling can stop. We apply machine learning methods to predict the mutation frequencies from wastewater sites on the next day in one location based on the frequencies on previous days in other locations, then record the prediction error. The sites with the lowest prediction error are the ones that contain the least amount of unique information, and sampling can cease at those locations. We demonstrate a systematic approach to evaluating prediction errors and several interpretations of the error. We demonstrate this method on five locations in Switzerland, finding two locations that could be removed with minimal information loss.