This map follows the WRI map from 1997 assessing intact forest landscapes, but for the first time uses the latest satellite imagery for the global assessment. It is based on strict rules of interpretation in order to make the results globally as comparable as possible. Please, read The Natural Value of Large Intact Forest Landscapes before going into details.

The mapping method
Mapping algorithm
Important remarks

The mapping method

The method of mapping is based on a subtractive approach to identifying intact forest landscapes. All sources of information were analyzed to detect disturbed areas or infrastructure lines dividing natural landscape into separate isolated parts. All areas remaining after the identification of disturbed areas and infrastructure were classified as intact forest landscapes, so long as they met the size criteria (see definitions). This method follows the precautionary approach: unless an area is clearly identified as disturbed, according to the available sources of information, it is considered to be an intact forest area.

The second important aspect of the mapping method is the stepwise use of materials of different precision and accuracy.

Mapping algorithm

1) the forest zone was defined and identified according to the "Global Percent Tree Cover at a Spatial Resolution of 500 meters" map (Hansen et al., 2003). As this map shows a gradual transition from the forest zone to non-forest territories, the following algorithm was used to establish the linear borders of the forest zone. All parts of the forests with a treecanopy cover density of 20 percent or more were considered solid forest tracts if the distance between them is less than two kilometers. All forest plots combined in this way were considered to be the parts of the forest zone if their area was 50 thousand hectares or more. All non-forest plots within the forest zone (i.e. fully surrounded by forests) were regarded as a part of the forest zone. The borders of the forest zone were slightly changed in some areas of the world, but only in cases when they obviously did not fit with the existing vegetation maps or higher resolution imagery (this happened only in areas of open forests with tree cover of 20 to 39%).

2) The best and most up-to-date topographic maps available in every region were used. According to these maps, the global forests were divided by infrastructure (with 1 km buffer zones along it) into a great number of separate sections. Each of these was regarded as a potential intact forest landscape. Questionable objects of infrastructure in these maps (for example, temporary or seasonal roads and trails) were not considered. After that, the first sorting of potential intact forest landscapes was held according to their size. This considerably decreased both the area for further analysis of satellite images and the amount of work. So, before the beginning of work with satellite images, significant territories within the most densely populated and developed parts of the forest biomes were excluded from further analysis.

3) LANDSAT TM (global coverage of about 1990) and LANDSAT ETM+ (global coverage of about 2000) space images were used to identify disturbed landscapes and objects of infrastructure, which were not shown in reliable topographic maps. Both sets of images were used together: LANDSAT ETM+ images made it possible to detect the most recent disturbances, while LANDSAT TM - were used for older ones, whose traces have become less evident now. This is especially important in tropical forests, where the traces of disturbances disappear much faster than in temperate and boreal forests. 1 km buffer zones around human infrastructure (roads, waterways, settlements etc) and fire scars in the vicinity of human infrastructure where excluded from intact forest landscapes, because most natural fire cycles have been significantly altered in the vicinity of human infrastructure (eg increased fire frequency).

Disturbances: logging roads. Congo	Disturbances: logging roads. Gabon
Disturbed areas: agriculture clearings and pastures with fire dynamics. Democratic Republic of the Congo	Deforested landscape with human-induced pyrogenic dynamic. Congo

Using satellite images in the final stage allowed for the smoothing out of differences between regions that occurred at the stage of using topographical maps (as the quality, scale and age of the maps was significantly different for different areas). If any infrastructure was not shown in topographical maps or if it was of an "unreliable" type (like a temporary road), the decision regarding its exclusion from potentially intact forest landscapes depended on how it looked in the LANDSAT image.

Disturbances: different types of anthropogenic disturbances in boreal forests

4) Borders of intact forest landscapes were refined: long (where length was greater than width) ledges with a width of 2 meters or more were eliminated. Then the final sorting of the remaining intact forest landscapes was carried out according to their size, and those of less than 50 thousand hectares were excluded.

Intact Forest Landscape borders, Tropical forests. Papua-New Guinea	Intact Forest Landscape borders, Boreal forest region. Russia

The result of the project is a map of the borders of intact forest landscapes within the forest zone. The information about the limits of the forest zone itself, as well as limits of closed and open forests and non-forested areas within the forest zone and within intact forest landscapes is taken from the "Global Percent Tree Cover at a Spatial Resolution of 500 meters" map.

Important remarks

	The World Intact Forest Landscapes map is the first global assessment of intact forest landscapes based on high-resolution satellite imagery analysis. The world map of intact forest landscapes is based on a global set of criteria and approaches, the same for all countries and continents. This can lead to the neglection of some regional differences in understanding the human role in certain types of forest landscape disturbances (like fires or hunting). Also, the world map may contain inaccuracies caused by lack of available information about local land-use practices and the state of forests in some regions of the Earth. The authors plan for the map to be periodically updated as soon as new data and more sophisticated sources of information and technologies are available. Ideally it should become a "living map", which will periodically be improved and updated to reflect the results of intact forest landscape monitoring.
	The world map is based on the best available non-commercial space images for each region. Most of them are GLCF images of Landsat 7 1999-2002. This has two important implications: first, the map does not reflect changes of intact forest landscapes that occurred during last 3-5 years, and second, the map does not reflect some types of disturbances or fragmentation of the natural landscape which cannot be detected using Landsat images (for example, local forest roads, small temporary agricultural clearings or seismic lines that are not used as roads). These disadvantages of the world map can be corrected in the process of making more detailed regional or national maps, which could be based on more recent and detailed information.
	The world map may contain some inaccuracies caused by a lack of available information for mapping experts about local land use practices and the state of forests in each country or region of the Earth. This is an inevitable challenge for any group of experts struggling to secure the global compatibility of the mapping work. Inaccuracies can be corrected during the composition of regional and national maps, which can better take into account local knowledge and experience.