ABSTRACT
Measures of map accuracy do not explicitly consider generalization effects, although excessive line generalization and spatial filtering can greatly reduce the predictive value of a map. In the following sections, some experiments using vegetation transects are described in order to study the effect of generalization on the attribute accuracy of a small-scale vegetation map. The accuracy assessment of a vegetation map needs to be made within the framework of a particular classification system that will play the role of a filter through which space is partitioned. In order to determine the accuracy of a map such as the one described, one would need to compare it directly with the truth, i.e. with the vegetation existing on the ground. In this second case there is a need to study only each classification system but also how the systems can be cross-checked with minimum accuracy loss.
