To reverse the trend of global climate change caused by greenhouse gases, the United Nations Framework Convention and the Kyoto Protocol have stipulated that it is necessary to take immediate policy actions to reduce the emissions. Some of important emission-cutting actions are within the forestry sector, such as sustainable forest management, forest ecosystem health, and woody plantations, which require consistent and periodic inventory and reporting to verify their status, progress, and credibility. The Food and Agriculture Organization (FAO) of the United Nations (UN) conducts a periodic global Forest Resources Assessment (FRA) program (FAO 1995). The basis and rationale of the program are provided through the UN Conference on Environment and Development held in Rio de Janeiro in June 1992, which required the observation and assessment of global forest resources as a key element in Agenda 21, the conference documentation. The FAO program has been conducted for every ten years since 1980, with the current FRA survey referenced for year 2000 (FRA2000).
Regular FRA program components in the previous and current surveys are comprised of a tally and report of the most current forest inventory from each country, change analysis based on a statistical sample of Landsat image interpretations, and other modeling and analysis techniques (FAO, 1995). In addition to these components, FRA2000 was also required to produce a global forest cover map showing spatial distribution of the world's existing forests for this survey. Expert panels convened by FRA in the early stages of the program (Nyssönen and Ahti, 1996, Päivinen and Gyde, 1997) noted that a new remote sensing-based global forest cover map should be produced using concurrent satellite data to represent the geographic distribution of the forest resources surveyed under FRA2000. A number of large-area land/forest cover mapping products (discussed below) were presented and their applications were discussed. However, it was apparent that none of them met concurrent needs of FRA2000 in terms of scope (global), theme (forest cover and forest density), and timing (data from mid 1990's). These existing products demonstrated the feasibility of a FAO global forest cover mapping effort for use by FRA2000 and its constituents.
Thematically, the proposed FRA2000 forest cover legend was not complex – only five land cover classes were required (Table 1): closed forest, open or fragmented forest, other wooded land, other land cover, and water. This map legend is a simplified version of a more comprehensive FRA classification system used for the ground survey and reporting components in the previous and the current surveys (FAO, 1995).
Table 1. FAO FRA 2000 global land cover map legend, definitions, and representative land cover types.
FRA2000 Class |
FAO Definition |
Representative land cover |
Closed forest |
Land covered by trees with a canopy cover of more than 40 percent and height exceeding 5 meters. Includes natural forests and forest plantations. |
- Tropical/subtropical moist forest - Temperate broadleaf mixed forest - Subtropical/temperate conifer plantation - Boreal conifer forest |
Open or fragmented forest |
Land covered by trees with a canopy cover between 10 and 40 percent and height exceeding 5 meters (open forest), or mosaics of forest and non-forest land (fragmented forest). Includes natural forests and forest plantations. |
- Northern boreal/taiga open conifer or mixed forest - Southern Africa woodland - Tropical fragmented/degraded forest |
Other wooded land |
Land either with a 5-10 percent canopy cover of trees exceeding 5 meters height, or with a shrub or bush cover of more than 10 percent and height less than 5 meters. |
- Mediterranean closed shrubland - Tropical woody savanna |
Other land cover |
All other land, including grassland, agricultural land, barren land, urban areas. |
- Grassland, cropland, non-woody wetland, desert, urban |
Water |
Inland water |
- Inland water |
This legend was developed considering several factors, including source data, time and scope requirements, previous project experience, and FRA forest cover needs. In the 1990’s, owing to its coarse resolution and daily repeat cycle, the Advanced Very High Resolution Radiometer (AVHRR) sensor was perhaps the only choice for a global-scale land cover mapping effort. Constraints of the coarse resolution, plus the requirement for a quick 2-3 year delivery time for the global scope, necessitated that the classification legend should be relatively simple. Mapping forest types, in addition to canopy density, may be desirable, but at the global level, consensus and consistency are lacking as to what and how many forest types should, and can be mapped within the scope of this project.
The public and institutional awareness of the scientific and policy issues surrounding forest resources exploration and conservation has led to an increased emphasis in the last two decades on land cover and forest cover mapping programs. Maturing technologies in space-borne sensor, computer, and image processing techniques are another reason for the recently increased use of remote sensing for monitoring and assessment of forest resources (Loveland et al. 1999). While many mapping programs were intended for specific locations and in response to specific local or regional needs, there are large-area (large countries, continents, global) land/forest cover mapping programs designed to provide data to meet global science and policy requirements (Tucker and Townshend, 2000).
In the 1980’s and 1990’s, a number of studies demonstrated the feasibility and effectiveness of large-area, wall-to-wall land cover or forest cover mapping using coarse or medium resolution satellite data such as AVHRR and Landsat MSS (Multi-Spectral Scanner). Studies of Africa land cover (Tucker et al. 1985), Canadian land cover (Cihlar et al. 1996), European land cover (Mücher et al. 2000) and the U.S. land cover (Loveland et al. 1991) relied on spectral and seasonal properties of land cover from AVHRR imagery for classification of general land cover patterns. Large-area forestry applications, conducted using a variety of techniques, include Amazon deforestation monitoring (Skole and Tucker, 1993), the TREES tropical forest mapping project (Malingreau et al. 1995), mapping of the U.S. forest types (Zhu and Evans, 1994) and classification of European forest ecosystems (Roy et al. 1997). At the global scale, Hansen et al. (2000) classified land cover at 1 km resolution using 41 metrics as input data into a classification tree. The U.S. Geological Survey (USGS) EROS Data Center (EDC) has developed a flexible, multi-theme land cover characteristics database (Loveland et al. 1999) for use by a broad range of applications. Depending on land cover complexity and land use history, the numbers of land cover classes in the database, derived primarily from their seasonal characteristics, ranged from 137 for Australia to 255 for Eurasia. The seasonal land cover classes were further summarized to a 17-class legend defined by the International Geosphere and Biosphere Programme (IGBP) (Belward, 1996). The IGBP classification legend includes five forested classes based on leaf type and longevity: evergreen broadleaf forest, evergreen needleleaf forest, deciduous broadleaf forest, deciduous needleleaf forest, and mixed forest1. In addition to coarse or medium resolution satellite data, 30-meter Landsat image data have been increasingly used to produce wall-to-wall land cover maps in large areas, (e.g., China (Liu and Buheaosier, 2000), U.S. (Vogelmann et al. 1998)). However, presently these mapping efforts remain prohibitively expensive in terms of time and effort to be applied to the global level.
Following the successful completion of the USGS global land cover effort, it would be logical to consider using the USGS/IGBP results for the FAO FRA2000 mapping needs. However, because the USGS seasonal land cover database was not intended to optimize forest canopy, no direct relationship exists to permit a simple conversion of the seasonal forest cover type classes to the FAO forest canopy classes. The closed versus open forest requirement in the FAO classification implied that forested classes in the USGS database, described on the basis of vegetation seasonality, productivity, and leaf type, were not entirely appropriate for an one-to-one conversion to FAO classes. A new approach was required for the FAO mapping effort.
1 Remaining 17 IGBP classes are: closed shrubland, open shrubland, woody savanna, savanna, grassland, wetland, cropland, urban, cropland and natural vegetation mosaic, barren, permanent snow and ice, water.
