The rise of automated weather stations, which are both affordable and easy to set up, has led to an influx of meteorological data from members of the public. Websites such as the Met Office's Weather Observations Website (WOW) act as hubs for sharing such data, making it freely and widely available. This user-contributed data could be useful in many applications. Feeding the data into data assimilation schemes to construct the initial weather conditions in numerical weather prediction models could be one such use. Alternatively it could help post-process weather model output, or aid in urban climate studies. However, as there is no guarantee that these amateur networks conform to the same strict quality guidelines as professional networks it is crucial to accurately quantify our uncertainty about the data before we can begin to use it. We introduce our proposed approach of automatically learning this uncertainty, an approach that must also identify and correct for any biases in the data. In particular we'll describe the Gaussian Process model being used to interpolate observations from high quality Met Office weather stations to the locations of the amateur stations, providing prior information on the weather conditions at these locations.