Authors: Manzhu Yu*, Pennsylvania State University
Topics: Geographic Information Science and Systems, Climatology and Meteorology, Cyberinfrastructure
Keywords: abnormal movement patterns, multi-spatial-temporal-scale, natural phenomena, dust events
Session Type: Paper
Presentation File: No File Uploaded
Natural phenomena constantly evolve in both space and time in a complex way. The increasingly available observation and simulation datasets have provided various opportunities to understand these complex evolvement patterns. Existing researches have investigated the generalized movement patterns of natural phenomena, but it is also necessary to identify the abnormal movement patterns in a certain spatial and temporal range to better predict the occurrence of severe weather events, e.g., thunderstorm, dust storm, and tropical cyclones. This research proposes a multi-spatial-temporal scale approach for detecting abnormal movement patterns of natural phenomena. The detection considers the spatial (direction and speed) and non-spatial (volume and intensity) attributes to examine global movement as well as the local intensity change of natural phenomena. The outliers are determined by the degree to which each attribute deviates from the average value of each spatiotemporal unit. Different outliers are detected under different spatiotemporal scales. The outliers detected consistently from different scales are considered as the major outliers. The outliers detected in a finer spatiotemporal scale but are left out in a coarser spatiotemporal scale might also be of interest as they might represent local movements or short-lasting and fast-moving events. The proposed approach is applied and demonstrated using a dust simulation dataset but can be easily applied to observational datasets or other types of natural phenomena events, such as thunderstorms and ocean eddies.
To access contact information login