Land Cover

IFPRI cultivated area (1992/93)

Any comprehensive assessment of the state and potential of agricultural development requires a clear understanding of where farming is practiced and what land resources are utilized. Without such knowledge it is difficult to gage the dynamics of the agricultural footprint, to see where specific policies and technologies are bringing about change, or where, for example, expansion of the agricultural frontier might involve tradeoffs with regard to conservation or ecosystem service provision. At a local scale, determining where and how farming is practiced is most effectively done by working with farmers. More broadly, however, to derive consistent knowledge at a national, regional or global scale this is clearly an impractical option. At this scale, satellite based (remotely-sensed) land cover data (along with agricultural statistics) offer practical insights into the geography, nature, extent and intensity of agricultural practices; particularly with regard to the delineation of rainfed and irrigated croplands and extensive pastures and grazing land.

Aggregated land cover—IFPRI cultivated area (1992/93)

A number of coarse resolution (1km) satellite-derived global land cover datasets exist, but the reliability with which agricultural areas - particularly cultivated¹ areas - can be distinguished varies widely. These datasets include: IFPRI’s extent of cultivated area , derived from the Global Land Cover Characterization (GLCCv2.0: USGS EDC 1999) that is based on 1992/93 AVHRR satellite data (for further explanation on how this dataset was created see Wood et al. 2000); JRC's GLC2000(JRC 2003) which was derived from 2000 SPOT satellite data; and Boston University’s Global Land Cover dataset(BU-MODIS) based on 2000 MODIS data. The three satellite-derived datasets (GLCC , GLC2000 and BU-MODIS ) characterize each 1x1km pixel in terms of its dominant land cover with a heavy focus on natural land covers. Each of these datasets includes cropland and cropland mosaic classes but each were derived using different criteria and thresholds, and none stands out as fully encompassing the areas across the globe that are cultivated, particularly those areas characterized by mosaics of cultivated and natural land covers. It remains a challenge, therefore, for individuals and organizations working with these datasets to deduce and maintain over time a reliable ‘picture’ of the geographical extent and intensity of cultivation.

In 2006 IFPRI/HarvestChoice undertook a detailed examination of the existing global land cover datasets in order to generate a best representation of cultivated areas. The resulting extent of cultivated area product is a re-interpretation of the detailed land cover classes of GLCC v2.0 with a focus on identifying cultivated and mosaic areas and their area intensity (the actual share of each grid cell occupied by cropland). For the moment, HarvestChoice considers this its best representation of the extent of cultivated area on a global basis (chiefly because existing datasets usually only identify areas as cropland if they occupy at least half the grid cell area). With this reinterpretation we are able to map areas with as little as 20% cultivated area. The maps on this page show two versions of the IFPRI dataset. The first provides the full list of land cover classes and shows that the integrity of the non-agricultural classes was maintained to the degree possible in the reinterpretation. The second map shows a class aggregation focused on best representing the area share of cropland in each grid cell. One drawback is that this database is derived from a 1992/93 AVHRR satellite data (albeit with over 900 land cover classes) whereas other datasets are derived from 2000 data (but with moreaggregated land cover classes). HarvestChoice continues to monitor land cover dataset products and to create and distribute new agriculture focused map and database derivative products as they become available. This will certainly be necessary after July 2008 when GlobCover, a new global 300m resolution land cover dataset based on FAO's hierarchical Land Cover Classification Scheme(LCCS) is scheduled to be released by JRC (European Space Agency).

The Crop Allocation Model uses aggregations of IFPRI's extent of cultivated area , GLC2000 , and BU-MODIS as inputs. The merger of three different satellite derived products allows the model to identify, and prioritize, areas that were identified as cultivated in any of the three input datasets. The shortcomings of individual datasets in certain areas can thus potentially be overcome by the strengths of the others.

Map data available for download at the HarvestChoice GeoNetwork Site.

Notes
¹ Most land cover datasets identify cultivated areas as those that encompass cropland and highly managed pasture. Extensive pasture and grazing lands are difficult to distinguish from natural grasslands and thus are usually not identified separately.