The most commonly assessed errors in maps are characterized as
classification or thematic errors. However, the effects of positional
errors, which includes both rectification errors and GPS errors, on
classification accuracy and inferences based on classifications are
often ignored. As the label suggests, rectification error results from
imperfect rectification of imagery to the coordinate system. Congalton
(2007) asserts that topography is the primary factor affecting
rectification accuracy and that rectification accuracy is an integral
component of classification accuracy. Rectification error is typically
quantified as the root mean square distance between known ground
point locations and their locations in the rectified imagery. Congalton
(2007) further indicates that rectification errors on order of a halfpixel
width are sufficiently small for most applications using Landsat
or SPOT imagery. Cooke (2000) and Patterson and Williams (2003),
however, indicate that rectification error is typically as great as one
pixel width, whereas Goodchild (1994) suggests that the pixels in
which points appear in a map may be as many as two pixels away
from the actual point location and may be anywhere within a 3 × 3
pixel array. Carmel et al. (2001) note that few studies have explored
the combined effects of positional and classification error. Of the few,
Arbia et al. (1998) found that rectification error had large effects on a
wide range of map parameters.