Rodrigues, Thanan Walesza Pequeno [UNESP]Guimarães, Ulisses Silva [UNESP]Rotta, Luiz Henrique Da Silva [UNESP]Watanabe, Fernanda Sayuri Yoshino [UNESP]Alcântara, Enner [UNESP]Imai, Nilton Nobuhiro [UNESP]2018-12-112018-12-112016-01-01Boletim de Ciencias Geodesicas, v. 22, n. 2, p. 303-323, 2016.1982-21701413-4853http://hdl.handle.net/11449/173158The use of remote sensing focused on the determination of field samples is of great value to environmental studies, since the satellite images have attributes able to assess the spectral variability of water surface considering a wide area. Thus, this approach aims to define a stratified working method of selecting samples based on images variability in the visible and infrared wavelengths derived from Landsat-8/OLI sensor. The method relies on the use of raster data representing the standard deviation of a time series of images Landsat-8/OLI and then the automatic setting field points supported by the stratified random sampling. The choice of the image that yielded the selection of the samples was based on the greatest spectral variation component by means of Principal Component technique. As a result were obtained twenty representative points of a total of six spectrally similar classes created by using a Geographic Information System.303-323porGeoprocessingRemote sensingSamplingArtigo10.1590/S1982-21702016000200017S1982-21702016000200303Acesso aberto2-s2.0-84976646042S1982-21702016000200303.pdf298577110250533066913103944104900000-0003-0516-05670000-0002-8077-2865