Towards a discipline of geospatial distributed event based systems

Abstract

A geospatial system is one in which the state space includes one, two or three-dimensional space and time. A geospatial event is one in which an event impacts points in space over time. Examples of geospatial events include floods, tsunamis, earthquakes, and emission of toxic plumes. This paper discusses aspects of the theory of geospatial distributed event based systems (GDEBS). The paper describes algorithms for rapid detection of geospatial events which can be used on Cloud computing architectures, in which many servers collaborate to detect events by analyzing data streams from large numbers of sensors. Sensor noise and timing errors may result in false detection or missed detection as well as incorrect identification of event attributes such as the location of the event source. The paper presents mathematical analyses and simulations dealing with rapid event detection for geospatial events of varying speeds in the presence of substantial sensor noise and timing error. The paper also describes some of the algorithmic and machine-learning techniques for improving event detection in the Cloud with large numbers of noisy sensors. Experience with GDEBS using a seismic network is described.