Portal de Periódicos da CAPES

Artigo Produção Nacional Revisado por pares

BIBTEX RIS

Analysis of the wetland classification using optical satellite imagery in the environmental protection area of Guaraqueçaba, PR, Brazil

2021; Elsevier BV; Volume: 112; Linguagem: Inglês

10.1016/j.jsames.2021.103615

1873-0647

Raoni Wainer Duarte Bosquilia, Frank Müller‐Karger,

Soil and Land Suitability Analysis

Located at an interface between water and land, wetlands are complex ecosystems, which are important for the maintenance of biodiversity. The objective of this work was to evaluate five different optical satellite imagery with different spatial resolution (original: RapidEye – 5 m; Sentinel-2B – 10 m; Landsat 8–30 m; resampled: RapidEye – 30 m; Sentinel-2B – 30 m) in order to obtain the classification of Land Use/Land Cover (LULC) types of a coastal ecosystem, including its wetlands, in the Environmental Protection Area of Guaraqueçaba, PR, Brazil. For this, a supervised classification was made using the algorithm of maximum likelihood to obtain the LULC of each satellite image, using a mosaic of high-resolution spatial imagery from DigitalGlobe as ground-truth verification. A SRTM relief image was used to check wetlands classification against their location in flat and low altitude areas. Analyzing the spatial resolutions of the studied imagery, no significant differences were observed between RapidEye and Sentinel-2B imagery in obtaining the wetland areas, differently of the Landsat 8 image, which showed a classification accuracy a little lower than the others, which may be related to its lower spatial resolution. Using multivariate statistics, three groups were formed with two of them very close to each other (RapidEye and Sentinel-2B), corroborating the conclusion obtained by the kappa index and the overall accuracy. Analyzing the slope and altitude classes, all imagery classified wetlands into flat and low-altitude regions. It is recommended that the Sentinel-2B imagery are used by governments and research agencies for wetland-related policies and research, since this image is distributed free of charge and has a temporal resolution of only 5 days, against payed access of RapidEye.