In much of atmospheric science, primary scientific data are gathered during intensive campaigns lasting from a few days to weeks or months and often involving multiple platforms. Each sortie in a campaign is typically planned a few hours ahead by a mission core group and overseen throughout its duration by a mission scientist charged with making real-time decisions on platform routes or sensor operating modes. Modern instruments often provide scientifically useful raw data streams that are multidimensional (i.e., multiple chemical species, state parameters, and/or heights) and output at order 0.1 10 Hz. Platform payloads often consist of 10-20 such instruments. The platforms themselves e.g., the NASA ”Global Hawk” can have endurance many times the duty cycle of a typical mission scientist. Mission Scientists, are, therefore, increasingly faced with a potentially overwhelming data torrent from which they are required to identify and use mission critical information. In the face of this torrent, mission scientists often focus on a very small subset of the data stream, limiting the depth of the analysis which can be carried out. Newly developed techniques can help mission scientists through on-line cluster analysis. The data stream can be separated in real-time to reveal different data groups and hence isolate specific regions of interest. We demonstrate the utility of these new techniques by applying them in a simulated real-time environment, using data gathered during the SAMBBA campaign. The resulting output shows how on-line cluster analysis can help differentiate between actual and apparent anomalies, supporting the mission scientist in their task.