Summary

Description

This dataset classifies the land cover class present on the land surface at 10 m spatial resolution and 2-5 day temporal resolution, with global coverage from 2015 onwards. Probabilities are delivered for 9 land cover classes, alongside a label index indicating the land cover class with highest probability of occurrence.

Usage notes

Actual temporal resolution varies and may be lower than expected due to Sentinel-2 revisit frequency, overlap of Sentinel-2 scenes, and cloud cover. Data is delivered as daily composites to ensure a maximum of 1 pixel observation per day. Composites are generated by taking daily means of class probabilities then reassigning the label index as the class with highest mean probability.


Licence

Open

Version

1

Release date

June 2022


CRS

Tile-specific (UTM-based)

Update schedule

One day

Provider cost

$0.00/ha


Variables

Water

Description

Water quantifies the relative probability (range 0-1) that a pixel is completely covered by water. Water is defined as a pixel containing water but little to no vegetation, rock, outcrops, or infrastructure. Probabilities are calculated using a Fully Convolutional Neural Network model trained to predict >20000 manually annotated land cover classifications from Sentinel-2 imagery.

Usage notes

Occasionally, unmasked shadows may be misclassified as Water. Overall prediction accuracy for this class compared to expert annotations is 90.6%.

Variable category

Land cover class

Variable type

Continuous


Trees

Description

Trees quantifies the relative probability (range 0-1) that a pixel is completely covered by trees. Trees are defined as a clustering of dense vegetation with a closed canopy. Probabilities are calculated using a Fully Convolutional Neural Network model trained to predict >20000 manually annotated land cover classifications from Sentinel-2 imagery.

Usage notes

Overall prediction accuracy for this class compared to expert annotations is 70.2%.

Variable category

Land cover class

Variable type

Continuous