Estimativa de velocidades a partir de séries temporais posicionais para atualização temporal de coordenadas: avaliação integrada de métodos e tratamento de dados
Loading...
Date
Journal Title
Journal ISSN
Volume Title
Publisher
Universidade Federal de Viçosa
Abstract
Esta pesquisa teve como objetivo analisar a influência que outliers presentes em séries temporais posicionais exercem nas velocidades estimadas a partir destas séries, assim como realizar a otimização do processo de estimativa de velocidades, indo desde a obtenção das séries com um software desenvolvido (GPTS Downloader) até a interpolação espacial das velocidades para pontos de interesse. A partir de séries de 20 anos de dados de 113 estações dos Estados Unidos, estimou- se a velocidade dessas estações, atualizaram-se suas coordenadas e calcularam-se as discrepâncias planimétricas em relação às coordenadas de referência. Empregando séries temporais de dados reais, que refletem a complexidade e a variação natural intrínseca da Terra, verificou-se que a remoção de outliers das séries não apresentou impacto significativo nas discrepâncias planimétricas finais. Ou seja, as velocidades não se mostraram sensíveis aos outliers das séries das quais são estimadas. Analisando técnicas de estimativa de velocidades, verificou-se que os métodos baseados nos mínimos quadrados são intercambiáveis, uma vez que seus resultados mostraram diferenças desprezíveis para esta finalidade. As velocidades estimadas com a regressão linear simples e o MIDAS (Median Interannual Difference Adjusted for Skewness) também foram similares para a maioria das estações, mas o MIDAS apresentou resultados significativamente piores em algumas séries com saltos. No que se refere às multivelocidades, não foi possível prever quando ou qual multivelocidade de período uniforme apresentará melhores resultados que a monovelocidade. No entanto, para séries com saltos ou mudanças de tendência, a multivelocidade com períodos distintos permite uma melhor adequação de cada segmento ao comportamento das séries, reduzindo a discrepância após a atualização das coordenadas. Em relação à interpolação, por apresentar os resultados mais consistentes entre os métodos analisados, a interpolação por Funções de Bases Radiais se mostrou o método mais confiável para ser empregado em velocidades estimadas a partir de séries temporais. Dessa forma, mesmo analisando regiões de alta atividade sísmica, foi possível atualizar coordenadas por 20 anos com baixas discrepâncias planimétricas, apresentando resultados superiores aos dos modelos de velocidades convencionais, mesmo após interpolação.Palavras-chave: geodésia; GPS; outlier; velocidades geodésicas; série temporal posicional
This research aimed to analyze the influence of outliers present in positional time series on the velocities estimated from these series, as well as to optimize the velocity estimation process, from obtaining the series with a developed software (GPTS Downloader) to the spatial interpolation of velocities for points of interest. Based on 20 years of data from 113 stations in the United States, the velocities of these stations were estimated, their coordinates were updated, and the planimetric discrepancies in relation to the reference coordinates were calculated. Using real data time series, which reflect the complexity and intrinsic variation of nature, it was found that removing outliers from the series did not have a significant impact on the final planimetric discrepancies. In other words, the velocities were not sensitive to the outliers of the series from which they were estimated. Analyzing velocity estimation techniques, it was found that least squares-based methods are interchangeable, as their results showed negligible differences for this purpose. The velocities estimated with simple linear regression and MIDAS (Median Interannual Difference Adjusted for Skewness) were also similar for most stations, but MIDAS showed significantly worse results in some series with jumps. Regarding multivelocities, it was not possible to predict when or which multivelocity with a homogeneous period would yield better results than the monovelocity. However, for series with jumps or trend changes, the multivelocities with distinct periods allows for better adjustment of each segment to the series' behavior, reducing discrepancies after coordinate updates. As for interpolation, due to presenting the most consistent results among the methods analyzed, interpolation using Radial Basis Functions proved to be the most reliable method to be employed for estimating velocities from time series. In this way, even when analyzing regions of high seismic activity, it was possible to update coordinates over 20 years with low planimetric discrepancies, showing results superior to those of conventional velocity models, even after interpolation. Keywords: geodesy; GPS; outlier; geodetic velocities; positional time series
This research aimed to analyze the influence of outliers present in positional time series on the velocities estimated from these series, as well as to optimize the velocity estimation process, from obtaining the series with a developed software (GPTS Downloader) to the spatial interpolation of velocities for points of interest. Based on 20 years of data from 113 stations in the United States, the velocities of these stations were estimated, their coordinates were updated, and the planimetric discrepancies in relation to the reference coordinates were calculated. Using real data time series, which reflect the complexity and intrinsic variation of nature, it was found that removing outliers from the series did not have a significant impact on the final planimetric discrepancies. In other words, the velocities were not sensitive to the outliers of the series from which they were estimated. Analyzing velocity estimation techniques, it was found that least squares-based methods are interchangeable, as their results showed negligible differences for this purpose. The velocities estimated with simple linear regression and MIDAS (Median Interannual Difference Adjusted for Skewness) were also similar for most stations, but MIDAS showed significantly worse results in some series with jumps. Regarding multivelocities, it was not possible to predict when or which multivelocity with a homogeneous period would yield better results than the monovelocity. However, for series with jumps or trend changes, the multivelocities with distinct periods allows for better adjustment of each segment to the series' behavior, reducing discrepancies after coordinate updates. As for interpolation, due to presenting the most consistent results among the methods analyzed, interpolation using Radial Basis Functions proved to be the most reliable method to be employed for estimating velocities from time series. In this way, even when analyzing regions of high seismic activity, it was possible to update coordinates over 20 years with low planimetric discrepancies, showing results superior to those of conventional velocity models, even after interpolation. Keywords: geodesy; GPS; outlier; geodetic velocities; positional time series
Description
Citation
FREITAS, Krisley Xavier Soares de. Estimativa de velocidades a partir de séries temporais posicionais para atualização temporal de coordenadas: avaliação integrada de métodos e tratamento de dados. 2025. 149 f. Tese (Doutorado em Engenharia Civil) - Universidade Federal de Viçosa, Viçosa. 2025.
