Multiple representations of elevation for dynamic process modeling
Namikawa, Laercio Massaru
MetadataShow full item record
A major benefit from the existence of widely available geographic information is the capability to simulate events for prediction, analysis, and decision-making. A reliable simulation of a geographic event is essential for the search of solutions for problems with environmental and socio-economic impacts from the local to the global level. However, the question of how to select the best among the available geographic information has not been answered. This dissertation's objective is to develop an integrated framework that allows simulations of dynamic geographic phenomena by using multiple representations of geographic information. These simulations will be calibrated and tested against actual geographical events. Therefore, reliable simulations will be enabled by using available multiple sources of information in an effective way. Since simulations of a geographic phenomenon require the existence of models constructed from entities of the real world, an analysis of modeling should use notions from Ontology. Ontology provides the framework that accounts for all existing entities. An information systems ontology that distinguishes entities by their modes of existence in time is used to classify geographic entities in process and objects. Therefore, the distinction between models of process and model of objects is defined. A representation of elevation is a model of an object. Modeling, the procedure to create a model from the real world entity, produces multiple representations of elevation. The framework proposed in this dissertation handles multiple representations of elevation and provides a linkage to a model of process. The main feature of the linkage is the use of information about elevation modeling to select the best representation of elevation based on the spatial and temporal setting of the simulated event. The framework is validated through simulations of geophysical mass flow events. Simulations are executed by a process model of geophysical mass flow linked to multiple representations of elevation. Simulation results are compared to another simulation that uses only one representation of elevation. Since quantitative methods for comparison of geophysical mass flow simulations do not exist, a method using logistic regression was developed and used in the validation. Simulated events are the block-and ash flow event, which occurred in April, 1991, at the Colima Volcano, Mexico, and the debris flow event, which occurred in December, 2003, in San Bernardino County, California. Comparisons of simulation results indicate that the use of multiple representations of elevation yields better results than the use of one representation only; therefore, the integrated framework proposed in this dissertation has the potential to provide reliable simulations of geographic phenomena when multiple representations of elevation are available.