Hansen Global Forest Change

Global Forest Change is a set of world maps of forest extent and loss, made from time series of Landsat satellite images.¹ Its makers are the Global Land Analysis and Discovery (GLAD) laboratory at the University of Maryland, working with Global Forest Watch, now Global Nature Watch.¹ The method is described in "High-Resolution Global Maps of 21st-Century Forest Cover Change" by Matthew Hansen and colleagues, in Science, November 2013.^1,3 Version 1.13, the current one, covers 2000 to 2025.^1,2

The maps come in 10-by-10-degree tiles at 1 arc-second, about 30 m at the equator.¹ The layers are:^1,2

• tree canopy cover in 2000, the share of each cell covered by vegetation taller than 5 m¹
• forest gain from 2000 to 2012, not updated since²
• the year of forest loss, where loss means a change from forest to non-forest¹
• a mask of land and water¹
• cloud-free Landsat composites from around 2000 and from the last year¹

Only the loss year and the last composite are updated each year.¹

The makers describe the limits. Changes in the Landsat sensors, in the number of images each year and in the algorithm mean that years are not strictly comparable.¹ Loss from 2011 onward has been reprocessed with newer methods, while earlier years have not.¹ A full validation with Landsat 8 data has not been done.¹ They call the maps a relative indicator of trends, suggest smoothing them over time, for example with a three-year moving average, and say forest area should not be estimated by counting pixels; instead the maps can guide a statistical sample.¹ A version 2.0 reprocessing of the whole record is planned with no date set.¹

The data are licensed CC BY 4.0, free for any use including commercial, with the display credit "Source: Hansen/UMD/Google/USGS/NASA".¹ Tiles can be downloaded as GeoTIFFs, and the whole set is also an asset in the Google Earth Engine catalogue, where it can be analysed in place; the makers provide an introductory tutorial.^1,2 They ask users to register so they can keep them informed of updates.¹

For someone modelling a place, the maps give a year-by-year record since 2001 of where tree cover was lost, at about 30 m, starting from the tree cover of 2000.¹ The updated method detects some kinds of loss better: fire in boreal forests, smallholder rotation clearing in the tropics, selective logging and the clearing of short-cycle plantations.¹ Comparing one year with another needs their warnings in mind.¹