A k-means clustering algorithm applied to the height-time data on 2000 January 01–03 (from Fig. 4), in an effort to distinguish the multiple CME profiles that were detected as a single event due to their close proximity in space and time. Top plot: The height-time measurements of the CME detections from the automated CORIMP catalog. (Note, this dataset has been put through a cleaning algorithm, discussed in Sect. 2.1, that removes a lot of inner-core and trailing-material datapoints, thus making it easier to distinguish their separate profiles.) Middle plots: The resulting clusters for the cases of k = 3 (left) and k = 4 (right), applied to the normalised parameters of the slope of a linear fit to, and the mean time of, the height-time profile at each position angle. The clusters are distinguished by different plot symbols and colours, with black asterisks to indicate the mean of each cluster. Bottom plots: The resulting effort at separately distinguishing the CME height-time profiles.