Functional boxplots for summarizing and detecting changes in environmental data coming from sensors

Nowadays, environmental sensor networks produce a large amount of streaming time series whose storage, manipulation and indexing is impractical. In this work, we propose a new strategy for summarizing and describing this kind of data based on functional data representation. It discovers trends and potential anomalies by using an informative exploratory tool: the functional boxplot. Functional boxplots are introduced for conveying location and variability information. In addition, for detecting and illustrating variation a distance among functional boxplots is used.