Detecção de rastros de tornados por meio de imagens de sensoriamento remoto orbital nos municípios de Água Doce a Tangará (SC) e Nova Candelária a Campos Novos (RS)
Loading...
Date
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Universidade Estadual de Ponta Grossa
Abstract
Os tornados são responsáveis por uma série de danos que resultam em prejuízos econômicos e perdas humanas, podendo resultar desta forma, em um desastre. Embora o fenômeno fosse historicamente documentado no Brasil, o mesmo é relativamente desconhecido por uma parcela significativa da população. O sudeste da América do Sul abriga um corredor de tornados que inclui o sul e sudeste brasileiro, com a região Sul respondendo a uma fração significativa das ocorrências reportadas. Diante da perspectiva que o uso de sensoriamento remoto, SIG e as geotecnologias podem ser mais exploradas e a relevância do tema, essa pesquisa investigou dois estudos de caso de tornados retrógrados, detectando seu percurso e a largura máxima encontrada. Para a execução desse propósito, foram utilizadas imagens de satélite MSI/Sentinel-2 de 2017 a 2023, sendo adquiridas cenas pré-evento e pós-evento. O ano inicial foi definido tendo como base o lançamento global do satélite. Para a seleção preliminar das ocorrências, foi utilizado a base de dados de Almeida (2023) e Piotrovski (2025), selecionando eventos de maiores categorias dentro do contexto brasileiro e do período disponível. Com as imagens de satélites adquiridas, foram utilizados procedimentos que pudessem realçar através de processamento digital de imagens, o rastro de tornados. Para isso, foram usados o NDVI pós-evento, a mudança de NDVI e o PCA, métodos aplicados em estudos anteriores e que foram eficazes em identificar a trajetória dos tornados. O NDVI é um indicador da saúde e o estado da vegetação, sendo derivado do balanço espectral entre o infravermelho próximo e o vermelho, enquanto que, a mudança de NDVI reflete as mudanças na superfície para períodos distintos (antes e após o distúrbio causado pelo tornado). O PCA, entretanto, trata-se de um procedimento estatístico. O método exibe a máxima variância em componentes principais, sendo a quantidade total de componentes principais determinada pelo número de bandas espectrais empilhadas no arquivo raster de entrada. Cada componente explica diferentes mudanças na superfície, dentre elas o rastro de tornado e são computadas em ordem decrescente da variância total explicada. Os resultados demonstraram que principalmente a mudança de NDVI e PCA foram eficazes em detectar rastros de tornados estimados entre EF2 e EF3, embora a variabilidade espacial do uso do solo e variações na intensidade do fenômeno podem resultar em descontinuidades espaciais no percurso detectado. Algumas evidências corroboram alegações de trabalhos anteriores, dentre eles, de que o PCA identifica melhor o rastro em ambientes urbanos e a mudança de NDVI em áreas de perda de vegetação, mas que o substrato não foi exposto. No entanto, em linhas gerais, tanto a mudança de NDVI quanto o PCA apresentam comportamento similar na identificação de rastro de tornados para os eventos analisados, destacando a importância do recurso multitemporal de análise. Os métodos aplicados são úteis para tornados extensos, auxiliando o mapeamento de danos e subsidiam análises futuras para correlacionar com as características da superfície terrestre.
Tornadoes are responsible for a series of damages that result in economic losses and human casualties, which can lead to disaster. Although the phenomenon has been historically documented in Brazil, it is relatively unknown to a significant portion of the population. Southeastern South America is home to a tornado corridor that includes southern and southeastern Brazil, with the southern region accounting for a significant fraction of reported occurrences. Given the prospect that the use of remote sensing, GIS, and geotechnologies can be further explored and the relevance of the topic, this research aimed to investigate retrograde tornadoes, detecting their path and maximum width. To achieve this purpose, MSI/Sentinel-2 satellite images from 2017 to 2023 were used, with pre-event and post-event scenes being acquired. The initial year was defined based on the global launch of the satellite. For the preliminary selection of occurrences, the database of Almeida (2023) and Piotrovski (2025) was used, and as necessary for the search for more significant events, the list on Wikipedia containing independent external sources was consulted, filtered by EF2-EF3 occurrences. With the satellite images acquired, procedures were used that could highlight the trail of tornadoes through digital image processing. For this, post-event NDVI, NDVI change, and PCA were used, methods applied in previous studies that were effective in identifying the trajectory of tornadoes. NDVI is an indicator of the health and condition of vegetation, derived from the spectral balance between near infrared and red, while NDVI change reflects changes in the surface for different periods (before and after the disturbance caused by the tornado). PCA, however, is a statistical procedure. The method displays the maximum variance in principal components, with the total number of principal components determined by the number of spectral bands stacked in the input raster file. Each component explains different changes in the surface, including the tornado track, and is computed in descending order of total variance explained. The results showed that NDVI change and PCA were particularly effective in detecting tornado tracks estimated between EF2 and EF3, although spatial variability in land use and variations in the intensity of the phenomenon can result in spatial discontinuities in the detected path. Some evidence corroborates claims from previous studies, among them that PCA better identifies trails in urban environments and NDVI change in areas of vegetation loss, but that the substrate was not exposed. However, in general, both NDVI change and PCA show similar behavior in identifying tornado tracks for the events analyzed, highlighting the importance of multitemporal analysis. The methods applied are useful for extensive tornadoes, assisting in damage mapping and supporting future analyses to correlate with land surface characteristics.
Tornadoes are responsible for a series of damages that result in economic losses and human casualties, which can lead to disaster. Although the phenomenon has been historically documented in Brazil, it is relatively unknown to a significant portion of the population. Southeastern South America is home to a tornado corridor that includes southern and southeastern Brazil, with the southern region accounting for a significant fraction of reported occurrences. Given the prospect that the use of remote sensing, GIS, and geotechnologies can be further explored and the relevance of the topic, this research aimed to investigate retrograde tornadoes, detecting their path and maximum width. To achieve this purpose, MSI/Sentinel-2 satellite images from 2017 to 2023 were used, with pre-event and post-event scenes being acquired. The initial year was defined based on the global launch of the satellite. For the preliminary selection of occurrences, the database of Almeida (2023) and Piotrovski (2025) was used, and as necessary for the search for more significant events, the list on Wikipedia containing independent external sources was consulted, filtered by EF2-EF3 occurrences. With the satellite images acquired, procedures were used that could highlight the trail of tornadoes through digital image processing. For this, post-event NDVI, NDVI change, and PCA were used, methods applied in previous studies that were effective in identifying the trajectory of tornadoes. NDVI is an indicator of the health and condition of vegetation, derived from the spectral balance between near infrared and red, while NDVI change reflects changes in the surface for different periods (before and after the disturbance caused by the tornado). PCA, however, is a statistical procedure. The method displays the maximum variance in principal components, with the total number of principal components determined by the number of spectral bands stacked in the input raster file. Each component explains different changes in the surface, including the tornado track, and is computed in descending order of total variance explained. The results showed that NDVI change and PCA were particularly effective in detecting tornado tracks estimated between EF2 and EF3, although spatial variability in land use and variations in the intensity of the phenomenon can result in spatial discontinuities in the detected path. Some evidence corroborates claims from previous studies, among them that PCA better identifies trails in urban environments and NDVI change in areas of vegetation loss, but that the substrate was not exposed. However, in general, both NDVI change and PCA show similar behavior in identifying tornado tracks for the events analyzed, highlighting the importance of multitemporal analysis. The methods applied are useful for extensive tornadoes, assisting in damage mapping and supporting future analyses to correlate with land surface characteristics.
Description
Citation
KUBASKI, Kauan Mateus. Detecção de rastros de tornados por meio de imagens de sensoriamento remoto orbital nos municípios de Água Doce a Tangará (SC) e Nova Candelária a Campos Novos (RS). 2026. Dissertação (Mestrado em Gestão do Território) - Universidade Estadual de Ponta Grossa, Ponta Grossa, 2026.
Endorsement
Review
Supplemented By
Referenced By
Creative Commons license
Except where otherwised noted, this item's license is described as http://purl.org/coar/access_right/c_abf2
