ABSTRACT
Assessing the thematic accuracy of maps or other spatial data requires sampling, because it is not economically feasible or time effective to visit every place on the ground. Thematic map accuracy assessment assumes that the information contained within the error matrix is a true characterization of the map being assessed. The level of detail in the scheme strongly influences the time and effort needed to make the map and to conduct the accuracy assessment. The more complex the scheme, the more expensive the map and its assessment. Normal theory statistical techniques that assume a continuous normal distribution may be inappropriate for map accuracy assessment. The existence of spatial autocorrelation can violate the assumption of sample independence, which, in turn, can affect the sample size and especially the sampling scheme used in the accuracy assessment. A large number of accuracy assessments have been conducted using a single pixel as the sampling unit.
