2.5 Accuracy Assessment
An accuracy assessment using an error matrix (Congalton and
Green, 1999) was performed on each of the object-based and
pixel-based classifications to determine which produced the
most accurate identification of LULC. The error matrix
measures an overall accuracy for the classification, as well as a
producer and user’s accuracy for each feature class.
To determine if two classifications were significantly different
(α = 0.05), a Kappa analysis and pair-wise Z-test were
calculated (Congalton and Green, 1999; Dwivedi et al., 2004;
Zar, 2007):