Analyzing Spatio-temporal Land Cover Dynamics in an Atlantic Forest Portion Using Unsupervised Change Detection Techniques

• Autor(a) principal: Sapucci, Gabriela Ribeiro [UNESP]
• Data de Publicação: 2021
• Outros Autores: Negri, Rogério Galante [UNESP], Casaca, Wallace [UNESP], Massi, Klécia Gili [UNESP]
• Tipo de documento: Artigo
• Idioma: eng
• Título da fonte: Repositório Institucional da UNESP
• Texto Completo: http://dx.doi.org/10.1007/s10666-021-09758-6; http://hdl.handle.net/11449/206045
• topic: Change detection; Landsat imagery; Multitemporal analysis; Remote sensing; Unsupervised

Over the past decades, the Southeast Atlantic Forest in Paraíba do Sul River Valley has suffered intense deforestation and human disturbances. Due to the Atlantic Forest biodiversity and the economic relevance of such a region in Brazil, spatial-temporal analyses are of crucial importance to protect the forest, as well as to support economic decision-making of public and private agents. In this context, the use of change detection techniques applied to remote sensing imagery arises as a powerful tool to track and map the Earth’s surface transformations. Therefore, this work investigates the effectiveness and practical feasibility of distinct unsupervised change detection approaches when they are applied to reveal the spatial-temporal dynamics in Paraíba do Sul River Valley across the last four decades. Different change detection approaches such as Change Vector Analysis (CVA), a K-Means and Principal Component Analysis (PCA-KM) framework, and a Alternating Sequential Filtering (ASF) based process were taken and properly tuned to cope with Landsat image series. The analysis of the results revealed a permanent land cover change rate over the last decades. Moreover, these changes do not necessary occur in the same locations, as it was confirmed the existence of successive modifications in original coverage of the study area. Another observed aspect is that the simplest technique for detecting changes, CVA, turned out to be the best approach to map the changes in the examined region.