Incorporating reference classification uncertainty into the analysis of land cover accuracy P. SARMENTO†‡, H. CARRÃO†‡, M. CAETANO*†‡, and S. V. STEHMAN§ †Portuguese Geographic Institute (IGP), Remote Sensing Unit (RSU), Rua Artilharia Um, 107, 1099-052 Lisboa, Portugal ‡CEGI, Instituto Superior de Estatística e Gestão de Informação, ISEGI, Universidade Nova de Lisboa, 1070- 312 Lisboa, Lisboa, Portugal §SUNY, College of Environmental Science and Forestry, 320 Bray Hall, Syracuse, NY 13210, USA *Corresponding author. Email: mario.caetano@igeo.pt To accommodate the difficulty of identifying a single ‘true’ or ‘reference’ class, the reference data protocol of an accuracy assessment may include identifying both a primary and alternate reference land cover label along with a rating of the interpreter’s confidence in the reference classification obtained for each sample location. This additional reference information is used to construct a nominal variable (called CONF) in which the categories represent the ‘confidence’ in the correctness of the map land cover classification at a given location. An accuracy measure that incorporates uncertainty in the reference classification is then derived by assigning partial credit weights to each CONF class. Further, the accuracy reporting format can be organized by CONF classes to provide additional understanding of the relationship between accuracy and uncertainty in the reference classification. The analysis is illustrated using an accuracy assessment of a land cover map of Portugal. These analyses incorporating uncertainty in the reference classification are intended to supplement traditional analyses to further enhance understanding of the accuracy of land cover maps. Keywords: Land cover maps, accuracy assessment, reference database uncertainty. 1 Introduction Land cover maps play a significant role in many environmental science studies and natural resource policy processes. If land cover maps are used for decision-making, then the quality of these data would certainly affect the type of the decisions. Traditionally, accuracy assessment of land cover maps is conducted by comparing the produced maps with a reference sample database (Foody 2002) that represents the ‘real’ land cover on the Earth’s surface. For land cover maps derived from automatic classification of satellite images, the sample locations are pixels and the comparison of the map and reference land cover classes is on a per- pixel basis. This comparison is summarized by a confusion matrix, where the reference and map classifications are represented by the columns and the rows of the matrix, respectively. This approach assumes that only one reference land cover class is assigned to each sample location. Often the most suitable reference land cover class at a given sample location may be difficult to identify, and thus some uncertainty exists in this component of the accuracy assessment protocol. The uncertainty associated with the reference class designation may be attributable to characteristics of land cover, for example, existence of more than one land cover class at the sample location (i.e. mixed pixel), landscape fragmentation, and the natural continuum between land cover types. In this sense, Goodchild (2003) states that land cover maps are vague and full of