Computer Science and Engineering, Department of


Date of this Version

Spring 5-11-2014


Hart, K. DECAF: A New Event Detection Logic for the Purpose of Fusing Delineated-Continuous Spatial Information. (2014). Master's thesis, University of Nebraska-Lincoln.


A THESIS Presented to the Faculty of The Graduate College at the University of Nebraska In Partial Fulfillment of Requirements For the Degree of Master of Science, Major: Computer Science, Under the Supervision of Professor Ashok Samal. Lincoln, Nebraska: May, 2014

Copyright (c) 2014 Kerry Q. Hart II


Geospatial information fusion is the process of synthesizing information from complementary data sources located at different points in space and time. Spatial phenomena are often measured at discrete locations by sensor networks, technicians, and volunteers; yet decisions often require information about locations where direct measurements do not exist. Traditional methods assume the spatial phenomena to be either discrete or continuous, an assumption that underlies and informs all subsequent analysis. Yet certain phenomena defy this dichotomy, alternating as they move across spatial and temporal scales. Precipitation, for example, appears continuous at large scales, but it can be temporally decomposed into discrete spatial events (e.g., storms, fronts) inside which rainfall is continuous. We describe such phenomena as behaving discretely-continuous.

This thesis presents an event detection framework that both leverages existing techniques such as indicator kriging as well as a novel embedded-graph based algorithms. This event detection framework is applied to the spatial information fusion problem in order to more intelligently parameterize existing interpolation methods according to local spatial structure (i.e., presence or absence of events). Tests on simulated data demonstrate the efficacy of the event detection logic as well as help inform how to better apply this event detection logic to the fusion problem. Results of tests on real-world precipitation data compare favorably, and in some cases outperform, traditional interpolation methods.

Adviser: Ashok Samal