Lançamos a Coleção 5 (1985-2019)

Glossary

In this glossary are terms commonly used in describing Mapbiomas's methodology and products

Word Definition
Accuracy (Accuracy Analysis) Quantitative analysis of mapping accuracy. Indicates the allocation error and the area error.
Accuracy Samples Points collected over the anual mosaics and visually classified by the interpreter as belonging to a specific land use land cover class.
Algorithm Set of rules and procedures established to solve a task.
Asset Collection of maps, images or georeferenced data available for processing and analysis in Google Earth Engine.
ATBD (Algorithm Theoretical Basis Document) Document with methodological description and the algorithms used.
Band It refers to each layer of information of an Asset - either maps or images.
Chart or Millionth Chart Mapping division of the Brazilian territory is defined by IBGE and integrated to the International Map of the World (IMW) or Millionth Map. This division is used to organize the work of processing MapBiomas maps. Each chart covers an area of approximately 18,700 square kilometers and about 20 million pixels.
Classification Distribution of pixels in classes of a given biome or theme.
Classifier Generic name for an automated classification method (example of a classifier is Random Forest).
Cloud Computing Data processing performed on distributed processors available on the world wide computer network. In MapBiomas the cloud computing process runs through Google Earth Engine and Google Cloud Computing.
Code Editor Google Earth Engine programming tool with graphical interface for viewing the results.
Collection Each version of MapBiomas annual mapping data. The collections may vary in the period, methodology and legend.
Collect Mobile Mobile application developed by MapBiomas for the collection of reference data in the field.
Dashboard (Control Panel) Platform for visual presentation of information and consolidated data for easy tracking of information.
Empirical Decision Tree A cascade of parameters set to define the pixel classification. In the empirical decision trees the format and parameters of the tree are defined by the analysts, as well as the parameterization of each decision node.
Feature Space Set of spectral information, indices and metrics used in Random Forest classification.
Fusion Table Tabular data that connects with Google tools. Used to parameterize variables and processing rules (rules applied during the transition filter).
Google Cloud Storage Google's tool for storing lots of cloud information.
Google Earth Engine Platform for analysis and visualisation of scientific spatial data on the Earth's surface, in cloud computing. All image processing and production of MapBiomas maps is done on this platform.
Image Mosaic Set of Landsat pixels with good quality (little cloud interference, for example) selected in a given period. The MapBiomas mosaics are constructed by individually analyzing each pixel of the Landsat images available for the period. In the mosaic, we try to represent the analysis area for the specified period in the best possible way. In MapBiomas image mosaics generally represent the period of one year.
Integration Overlap routine of classes in order to generate an integrated maps. Different classes of MapBiomas are made separately and then integrated using prevalence rules
Integration Map Final map consolidating maps of biomes and themes.
Landsat Image Image generated by a set of Landsat satellites - launched by NASA and operated by the American Geological Survey.
Pixel The smallest unit in a digital image. A satellite image is composed of an array of pixels, each pixel with a digital value. The pixel in MapBiomas corresponds to the pixel of Landsat images with 30m resolution. The area of the pixel undergoes variations according to its latitude. Further away from the equator the distorted will be the area.
Post-Classification Automated routines to improve the consistency of maps performed after classification and map integration. The temporal and spatial filters are examples of post classification.
Random Forest Supervised classification method that is based on decision trees.
Raster Digital image, composed of an array of values (pixel).
Satellite Sensor Satellite instrument responsible for the remote sensing of electromagnetic energy. A satellite may have multiple sensors for picking up different spectral ranges.
Scene Refers to the image generated by the sensor of a satellite. To cover the Brazilian territory, 380 Landsat scenes are required.
Scripts Set of instructions written in a programming language for a function to be executed.
Shapefile File format with a set of spatial data in vector format.
Spatial Consistency Distribution of pixels of a certain class in space must be consistent with landscape characteristics of the place. For example, in the middle of a hillside forest area several pixels appear as water indicating a spatial inconsistency.
Spatial Filter Post-classification analysis used to correct errors of spatial consistency in a class.
Spatial resolution Describes the level of detail of an image. Landsat (TM) images have an average spatial resolution of 30m.
Spectral Band Interval between two wavelengths in the electromagnetic spectrum. Landsat has several spectral bands each one covering a range of the electromagnetic spectrum.
Spectral Index A spectral index is the result of mathematical operations between numerical values of pixels from the spectral bands of a sensor. For example, the Normalized Difference Vegetation Index (NDVI) is calculated by: (NIR - R) / (NIR + R) - NIR being the near infrared band and R is the Red band.
Temporal Consistency Classification history of a pixel in time is consistent with possible or probable transitions of land use and land cover. For example, a pixel that is classified as forest for 20 years but in a year in the middle of the series appears as non-forest. This is likely to be a misclassification.
Temporal Filter Post-classification analysis to correct temporal consistency errors between classes and years.
Training Samples Points or polygons used to train the classifier
Transition Map Map showing the main transitions of land use and land cover. It is produced from a comparison of a pair of maps (eg 2000 x 2016). In these maps each pixel can be classified as change or no change. For each change, it receives a code that represents the class in t1 and the class in t2.
Transversal Theme Land use land cover classes that occur transpassing the limits of different biomes. The cross-cutting themes of MapBiomas include agriculture, pasture, planted forest, urban area, mining, mangrove, apincum, aquaculture, beach and dunes.
WebCollect Platform used to collect points for the training of the classifier or accuracy analysis.
Workspace Web platform developed by MapBiomas for parameterization and classification of land use and land cover maps. The platform serves as an interface between analyst work and the cloud processing on Google Earth Engine.