Thesis METODOLOG´IA PARA EL DISEÑO DE UNA RED DE SENSORES DE TSUNAMI
Loading...
Date
2016
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Recientes eventos tsunamig´enicos han demostrado la necesidad de mejorar la confiabilidad en lossistemas de alerta para este tipo de fen´omenos. Los m´etodos tradicionales han consistido en basesde datos de tsunamis con escenarios pre-calculados (Ozer et al., 2016 ) debido al escaso tiempo derespuesta entre la generaci´on y el arribo del tsunami a la costa, donde la elevaci´on inicial de la superficielibre del tsunami se estima a partir de una gama arbitraria de magnitudes y ubicaciones del terremoto.T´ecnicas de inversi´on de tsunami podr´ian proporcionar una alternativa sin la necesidad de establecersupuestos respecto a la geometr´ia de la falla y magnitud del sismo, mediante una r´apidaestimaci´on de la deformaci´on inicial de la superficie libre. Esta ventaja es considerable respecto a lasm´etodos convencionales, ya que obtener los datos s´ismicos no es un proceso instant´aneo y a menudoes traducido de manera inexacta a los datos del tsunami (Wei et al., 2014 ). Es probable que unametodolog´ia robusta comprender´a una combinaci´on de las anteriormente mencionadas.En relaci´on al ´ultimo m´etodo, s´olo unos pocos pa´ises tienen la densa red de sensores necesariospara estimar las deformaciones iniciales de la superficie libre de un tsunami. Sin embargo, podr´ia serde gran inter´es establecer el n´umero m´inimo de sensores necesarios para proporcionar una estimaci´onrazonable, teniendo en cuenta la precisi´on de los resultados y costos generales. Lo anterior es cr´iticopara contar con dicha herramienta en pa´ises en vias de desarrollo, como es el caso de Chile. Debidoa esto ´ultimo, en el presente trabajo se propone una metodolog´ia sistem´atica para determinar elposicionamiento ´optimo de los sensores en el norte de Chile debido a las ventanas sismicas de la zona,minimizando el n´umero de estos y la redundancia de los datos.El primer paso de esta metodolog´ia consiste en la creaci´on de una base de datos de las funcionesde Green (Tsushima et al., 2009 ). Esto procedimiento se basa en la propagaci´on de tsunamis“elementales” utilizando las ecuaciones lineales de onda larga (Satake et al., 1995 ) desde un conjuntode fuentes unitarias de tsunami a una serie de puntos de pron´ostico (i.e. funciones de Green).Luego, se seleccionan como escenarios extremos los terremotos esperados en el norte de Chile, zonacon alto potencial tsunamig´enico (Cienfuegos et al., 2014 ). Una amplia gama de arreglo de sensores,establecidos con base en criterios t´ecnicos, se prueban para realizar la inversi´on. Posteriormente, laspredicciones obtenidas con los diferentes arreglos de sensores son comparadas mediante una funci´onde coste que incluye varios par´ametros para cuantificar la capacidad de predicci´on. Entre ellos, se considerael tiempo de arribo, la amplitud m´axima y un ajuste global. La red de sensores que proporcionala configuraci´on ´optima es seleccionada como candidata. Finalmente, dicha red se prueba utilizandoescenarios intermedios y reales para garantizar as´i la calidad de la predicci´on en todos los puntos depron´ostico.Los resultados muestran que una configuraci´on que contiene s´olo tres sensores es capaz de proporcionar estimaciones precisas de los tiempos de arribo y amplitudes de la primera onda. En t´erminosgenerales, es posible observar una fuerte correlaci´on entre la ubicaci´on de los sensores y los estimadoresde error. A modo de ejemplo, la estimaci´on de los tiempos de arribo es de mayor calidad cuando almenos un sensor se encuentra situado en frente de los puntos de pron´ostico. En cambio, con respectoa la amplitud m´axima y el ajuste global, mejores pron´osticos son obtenidos cuando los sensores est´ansituados en la zona de solevantamiento o frente a ´esta.Esta metodolog´ia ofrece un gran potencial para ser una herramienta que permita definir la ubicaci´onde posibles sensores de presi´on, de esta forma se asegura una buena calidad de las prediccionesde tsunami al utilizar la t´ecnica de inversi´on. Ademas, la metodolog´ia aqu´i propuesta tiene un granpotencial para mejorar la evaluaci´on de riesgo en tiempo real, apoyando a herramientas de toma dedecisiones en sistemas de alerta temprana.
Recent tsunamis have shown the necessity of improve the reliability of tsunami warning systems.Traditional methods are usually based on data sets of scenarios calculated a priori (Ozer et al.,2016 ), where the initial tsunami free surface elevation is estimated from arbitrary range of earthquakemagnitudes and source locations. More recently, there has been promising advances in the possibilityof establishing near real time simulations, owing to improvements in rapid assessment of earthquakeparameters. Finally, tsunami inversion techniques could provide an alternative without the need ofestablishing assumptions of the fault geometry and the earthquake magnitude, by quickly estimatingthe initial sea surface deformation. It is likely that a robust methodology will comprise a combinationof the above.Regarding the latter method, only a few countries have the dense network of sensors required toestimate the tsunami initial deformations. However, it might be of interest to establish what couldbe the minimum number of sensors required to provide a reasonable estimate, thus trading precisionand overall costs. This might be of interest for developing countries, for example. To date, the locationof these sensors is based primarily on the good judgment and intuition. Here we propose asystematic methodology to determinate the optimal location of sensors, minimizing the number andthe redundancy of data.The first step is to place a database of different tsunami sources (Tsushima et al., 2009 ). Tocompute the Green’s functions, we calculated the finite difference approximation of the linear longwaveequations (Satake et al., 1995 ) from a set of unit tsunami sources to a set of forecast points(I.e. Green’s functions). Next, expected earthquakes in northern Chile (Cienfuegos et al., 2014 ) areselected as end scenarios, providing the most extreme condition. A wide range of arrays of sensors arecompared by introducing a cost function involving several parameters to quantify predictive skill. Thenetwork providing the optimal configuration is selected as candidate. Finally, the candidate networkis tested using intermediate (between the end scenaries) and real scenarios, to ensure the quality ofthe prediction at all forecasting points.Results show that a configuration comprising just/only three sensors is capable of providing accurateestimations of the wave’s arrival times and peak amplitudes of the first wave. In general terms,it is possible to note a correlation between the location of the sensors and error estimators. As anexample, arrival times are better predicted with sensors located opposite to the tide gauges and betterresults on wave amplitudes are assessed when sensors are in front of or above the uplift zone.This methodology provides a tool that will allow to define the location of the minimum amountof sensors to get a good quality of predictions using the inversion technique. Thus, it allows for abetter modeling of tsunami event, which appears particularly promising to improve real-time danger assessment and decision making tools
Recent tsunamis have shown the necessity of improve the reliability of tsunami warning systems.Traditional methods are usually based on data sets of scenarios calculated a priori (Ozer et al.,2016 ), where the initial tsunami free surface elevation is estimated from arbitrary range of earthquakemagnitudes and source locations. More recently, there has been promising advances in the possibilityof establishing near real time simulations, owing to improvements in rapid assessment of earthquakeparameters. Finally, tsunami inversion techniques could provide an alternative without the need ofestablishing assumptions of the fault geometry and the earthquake magnitude, by quickly estimatingthe initial sea surface deformation. It is likely that a robust methodology will comprise a combinationof the above.Regarding the latter method, only a few countries have the dense network of sensors required toestimate the tsunami initial deformations. However, it might be of interest to establish what couldbe the minimum number of sensors required to provide a reasonable estimate, thus trading precisionand overall costs. This might be of interest for developing countries, for example. To date, the locationof these sensors is based primarily on the good judgment and intuition. Here we propose asystematic methodology to determinate the optimal location of sensors, minimizing the number andthe redundancy of data.The first step is to place a database of different tsunami sources (Tsushima et al., 2009 ). Tocompute the Green’s functions, we calculated the finite difference approximation of the linear longwaveequations (Satake et al., 1995 ) from a set of unit tsunami sources to a set of forecast points(I.e. Green’s functions). Next, expected earthquakes in northern Chile (Cienfuegos et al., 2014 ) areselected as end scenarios, providing the most extreme condition. A wide range of arrays of sensors arecompared by introducing a cost function involving several parameters to quantify predictive skill. Thenetwork providing the optimal configuration is selected as candidate. Finally, the candidate networkis tested using intermediate (between the end scenaries) and real scenarios, to ensure the quality ofthe prediction at all forecasting points.Results show that a configuration comprising just/only three sensors is capable of providing accurateestimations of the wave’s arrival times and peak amplitudes of the first wave. In general terms,it is possible to note a correlation between the location of the sensors and error estimators. As anexample, arrival times are better predicted with sensors located opposite to the tide gauges and betterresults on wave amplitudes are assessed when sensors are in front of or above the uplift zone.This methodology provides a tool that will allow to define the location of the minimum amountof sensors to get a good quality of predictions using the inversion technique. Thus, it allows for abetter modeling of tsunami event, which appears particularly promising to improve real-time danger assessment and decision making tools
Description
Catalogado desde la version PDF de la tesis.
Keywords
INVERSION DE TSUNAMIS, MODELAMIENTO DE TSUNAMIS, SENSORES DE PRESION, SISTEMA DE ALERTA TEMPRANA DE TSUNAMIS
Citation
Campus
Universidad Técnica Federico Santa María UTFSM. Casa Central Valparaíso