HR: 12:05h
AN: GC22A-07    [Abstracts]
TI: Land Cover of Northern Eurasia: Comparison and Assessment of Coarse Resolution Maps
AU: * Krankina, O N
EM: krankinao@fsl.orst.edu
AF: Oregon State University, 321 Richardson Hall, Corvallis, OR 97331, United States
AU: Pflugmacher, D
AF: Oregon State University, 321 Richardson Hall, Corvallis, OR 97331, United States
AU: Cohen, W
AF: Forest Service, 3200 SW Jefferson Way, Corvallis, OR 97331, United States
AU: Kennedy, R
AF: Forest Service, 3200 SW Jefferson Way, Corvallis, OR 97331, United States
AU: Nelson, P
AF: Oregon State University, 321 Richardson Hall, Corvallis, OR 97331, United States
AU: Loboda, T
AF: University of Maryland, 1104 LeFrak Hall, College Park, MD 20742, United States
AB: Consistent measurements of land cover are critical for addressing a range of important science questions, from quantifying the effects of vegetation on the carbon, energy, and water cycles, to understanding the social and economic causes and consequences of land-use and land-cover change. While multiple moderate and coarse- resolution land-cover products have been developed, they disagree significantly. Resolving discrepancies among maps is particularly challenging for boreal and temperate Northern Eurasia, where validation sites are sparse and processes of ecosystem disturbance and land-cover change are widespread. To identify specific needs and possibilities for improved mapping of land cover across boreal and temperate Northern Eurasia, we compared the performance of three recent land-cover products based on different sensors: MODIS (Global Land Cover Collection 4), AVHRR (DISCover v. 2.0), and SPOT VEGETATION (GLC2000 for Northern Eurasia v. 4.0). First, we examined the level of agreement among these data sets across the entire region. On a qualitative level, the assessment of general patterns indicates the highest degree of disagreement in transitional zones at the northern and southern fringes of boreal forest, in mountainous regions, and in areas of extensive wetlands, agricultural development, and urban land use. The quantitative analysis measured the level of disagreement between land-cover classes aggregated according to dominant type of vegetation (trees, shrubs, herbaceous, bare land, permanent snow/ice). Secondly, validation of these products was performed at two test sites where Landsat-based classifications were developed based on FAO Land Cover Classification System. Fractional land cover was calculated for each 1x1 km pixel and used to construct fractional error matrices. Most errors were associated with "mixed" coarse-resolution pixels (i.e. those having nearly equal percentage of multiple class types), while errors in "pure" (single class) pixels were low. In addition to actual differences in land-cover classifications, other sources of discrepancy among land cover products include differences in class definitions, map projections, and spatial resolution.
DE: 0480 Remote sensing
DE: 1632 Land cover change
DE: 1640 Remote sensing (1855)
SC: Global Environmental Change [GC]
MN: 2007 Fall Meeting