In the climate change era, it is fundamental to monitor the availability of water resources. One of the possible causes for a change in the water availability is related to variations in the meteorological conditions. To track this change, ground-based observations are one of the commonly used measurements. However, these datasets might include both extreme but realistic values and erroneous information. A necessary but not trivial preliminary process for exploiting the observations is to filter the former while retailing the latter. The Station Observation Outlier fiNder (SOON) is a highly innovative algorithm, that identifies errors in large dataflows. SOON can be used on historical datasets as well as in real-time dataflows. A first prototype has been tested on 8 years (2007-2014) of hourly data recorded by about 10000 stations around Europe, which includes 7 meteorological variables: temperature, dewpoint temperature, pressure, precipitation, wind speed, wind gusts, and cloudiness. The dataset belongs to the Ubimet archive and has been provided within the EDI incubator programme.
