samples_cerrado_lc8: Samples of Cerrado biome used in the paper "Satellite Image...

samples_cerrado_lc8R Documentation

Samples of Cerrado biome used in the paper "Satellite Image Time Series Analysis for Big Earth ObservationData"

Description

A sits tibble with time series samples from Brazilian Cerrado biome. The samples is based on a systematic sampling design of 5-kilometers over all Cerrado extension.

Labels were automatically attributed from a cross-comparison among land cover products of the year of 2018 from the following sources: Pastagem.org (LAPIG-UFG), IBGE Land use and land cover map (IBGE), and MapBiomas (MapBiomas).

The classification labels were: "Annual_Crop", "Cerradao", "Cerrado", "Nat_NonVeg", "Open_Cerrado", "Pasture", "Perennial_Crop", "Silviculture", "Sugarcane", and "Water". All those points not consistent with these classes as well as those with conflicting labels among input maps were discarded.

The time series were extracted from the Landsat 8 BDC data cube (LC8_30_16D_STK-1). These time series comprehends a period of 12 months (24 observations) from "2017-09-01" to "2018-08-31". The extracted attributes were: NDVI, EVI, B1, B2, B3, B4, B5, B6, and B7. Cloudy values were removed and interpolated.

The validation data set used in the referred paper is in the external data folder of this package that can be accessed using:

system.file("extdata/cerrado/validation.csv", package = "sitsdata")

This validation samples as well as the training data set were collected as part of the Brazil Data Cube project (http://brazildatacube.org). More details on how these data sets were obtained can be found in the referred paper.

Usage

data("samples_cerrado_lc8")

Format

A sits tibble with ~48K samples and 10 classes ("Annual_Crop", "Cerradao", "Cerrado", "Nat_NonVeg", "Open_Cerrado", "Pasture", "Perennial_Crop", "Silviculture", "Sugarcane", "Water").


e-sensing/sitsdata documentation built on April 17, 2025, 5:32 a.m.