|
Welcome to the GOFC-GOLD Reference Data Portal
last update on October 2015
Spatial distribution of the reference datasets available on the portal
Background and objectives
Observation of land cover at global scale is useful for many scientific and managerial applications. Several global land cover (GLC) maps have been produced for such purposes using remote sensing data, and GLC mapping is progressing towards higher spatial resolution.
GLC maps are commonly validated using higher-quality reference data such as independent validation datasets and regional maps. Currently, there are several independently validated GLC datasets which generation required significant efforts to analyze a large number of satellite images and interpret the land cover type.
These reference datasets are not always easily accessible to the community and are not used to their full potential despite the scarcity of such datasets. Therefore, the GOFC-GOLD Land Cover Office in coordination with reference data producers is taking the initiative to make a series of datasets available to the GLC mapping community.
This portal aims to:
provide the most appropriate databases among the existing ones based on internal quality criteria. For instance the GLC2000 and the GlobCover 2005 datasets are provided in a consolidated version.
foster the use of recommended practices for land cover validation in the GLC mapping community,
direct users to the most appropriate dataset(s) according to their needs. Each dataset is provided along with detailed information and recommendations for an appropriate use. A guiding tool will be made available when a larger number of datasets will be available from the portal.
The portal provides access to a first series of reference datasets to which new datasets will be appended in the coming months.
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License.
Reference datasets
We provide below the list of reference datasets currently available on this portal. Map 1 illustrates the spatial distribution of all these datasets
Quick links to the reference datasets
- GLC 2000 dataset
- GlobCover 2005 dataset
- STEP dataset
- VIIRS dataset
- GLCNMO 2008 dataset
- Urban dataset (U. Tokyo)
- Upcoming datasets and updates
- GLC 2000 database
This dataset is the result of a consolidation work realised on the original GLC 2000 dataset.
The original GLC 2000 classes were aggregated following the Aggregated Generalised Legend (AGL) resulting in a set of 11 land cover classes.
Access to the dataset
- GlobCover 2005 database
This dataset is the result of a consolidation work realised on the original ESA-GlobCover 2005 dataset.
A total of 186 points, associated with a pure land cover class, were randomly selected and interpreted according to the GlobCover legend (which counts 22 land cover classes).
The purpose of this dataset was to validate the global GlobCover 2005 land cover map, derived from 300-m MERIS time series.
Each sample is a 5*5 MERIS pixel window (~225 ha) drawn on GoogleEarth that was interpreted in terms of FAO-LCCS classifiers by independent experts.
The consolidation of the entire dataset is still going on. More samples will addded in the coming months.
Access to the dataset
- System for Terrestrial Ecosystem Parameterization (STEP) database
The STEP database was developed and maintained by Boston University between 2000-2013. The major purpose of this database is to train decision tree classifiers that are used by the MODIS Land Cover (MCD12Q1) product. Each site is a polygon (~4-sq km) drawn on GoogleEarth that is considered a stable example of a specific land cover type. Recently STEP attributes were revised to be compliant with the FAO-LCCS and the database was filtered for sites that had been disturbed through time.
Access to the dataset
- Visible Infrared Imaging Radiometer Suite (VIIRS) database
This database was derived directly from the Boston VHSR database (under development) in order to meet the validation requirements for the VIIRS Surface Type (ST) product in the summer of 2013. Each VIIRS (~1-km) pixel within each VHSR site was manually interpreted according to the 17-class International Geosphere-Biosphere Programme (IGBP) legend. The database contains primary and secondary IGBP labels at each pixel, a confidence for the first label, and records any land cover changes that may have occurred between the date of the imagery and the year 2012. Because of the similarity between the MODIS and VIIRS grids, this database can be used to compare these two land cover products.
Access to the dataset
- Global Land Cover National Mapping Organizations (GLCNMO) 2008 training dataset
This training dataset was developed to produce the GLCNMO Global Land Cover Map product for year 2008.
Access to the dataset
- Urban dataset from the University of Tokyo
A ground truth database specialised on urban areas was developed by the University of Tokyo. The dataset is based on the visual interpretation of ASTER/VNIR images acquired between 2000 and 2008.
Access to the dataset
- More datasets and updates to come in the next months
Follow us on social media to be informed on the next dataset releases:
or subscribe to our newsletter (released every 3-4 months):
GOFC-GOLD LC newsletter
| 
