Issue link: https://read.dmtmag.com/i/41362
It's Not a Party, It's an Event WHERE IT'S ABOUT TIME T BY ERIK SHEPARD he real power of geospatial technology shines in its integrative role to bring together disparate sources of information into the decision- support process. GIS is the system in the enterprise stack that provides data on land use, socioeconomics, demographics, weather and environmental quality as well as asset and facility information. These data—some of them external to an organization—then can be layered with financial, performance, engineering and other internal data. Whereas internal data have an obvious temporal component (i.e., the business activity that the data are tracking happened at some reference point in time), external data have a temporal frame of reference, but they're also inherently dynamic. Performance data often are tracked at discrete points in time: weekly, monthly or quarterly—sum- marizing trends. The ubiquity of sensors (another technology with a geospatial foundation) made available vast quantities of dynamic and rapidly changing data associated with events. Erik Shepard is principal of Waterbridge Consulting and principal consultant with SSP Innovations; e-mail: erik@ waterbridge.biz or erik.shepard@ sspinnovations.com. 30 Truly Dynamic Events are different from the types of spatiotem- poral data this column has previously discussed. Spatiotemporal data typically are thought of in terms of a baseline representation of features and discrete changes over time (e.g., an asset database that's initially populated with assets as they were designed and then modified in forward-looking snapshots as changes are made). The modifications exist in a temporal framework, but the data still are relatively constant through time. Events happen at specific points in time—tornadoes, snow and ice storms, accidents (clearly, no positive events are interesting, as the news always shows us). These data truly are dynamic. They don't consist of a base representation with changes; they exist completely in the temporal frame of reference and change continuously. Inferring Causality A deluge of data generated by the environment's multitude of sensors creates a need to analyze, filter GEO W ORLD / AUGUST 2O11 and parse event data into information. Data describ- ing events may be contained in multiple information sources, including a GIS. The challenge then is plac- ing event data into a context that allows information to be correlated. For example, a utility's outage-management sys- tem may detect a faulted device, while a GIS reports below-freezing temperatures, and SCADA reports increased line tension. Correlating these items together may allow a system to infer that a complex event has occurred; in this case, below-freezing temperatures led to ice on the lines, which led to sagging and ultimately to a faulted device. It's the ability to correlate these items together, in a number of ways, that allows determinations to be made of this complex event. Event representation is key to facilitate other types of decision support, including complex event detection. Such correlation can be accomplished by detect- ing causality and timing; freezing temperatures were detected first, followed by sag on the line and finally the faulted device. Without the ability to store the temporal context, such linearity can't be detected, and causality can't be inferred. Complex Processors Of course, filtering through the large amounts of data that led to the complex event is impossible without the assistance of automation and engines. A complex event processor integrates the multiple data sources, and performs the correlation and inference that allows connections to be made between events that may seem unrelated. Temporal representation of events, and an appropriate timestamp, is essential for the complex event processor to be able to utilize the information from source systems. For external event data from the GIS to be incorporated into the complex event proces- sor, it must be appropriately framed, and that means spatiotemporal representation in the form of events. Although spatiotemporal representation in terms of base layers and deltas to those layers allows for change detection over time (and several applications described in previous columns), event representation is key to facilitate other types of decision support, including complex event detection.