Geoscience Reference
In-Depth Information
Figure 14.3
Map of fuzzy set reference data.
A fuzzy set analysis provided more information about the agreement between the reference data
and the map and was less biased against a small sample size compared to number of cover types.
The M statistics were disturbing, but less so when the R statistics were considered. The R function
indicates that many cover types were more accurately classified to the life form level. Yet, even for
this statistic, accuracies did not reach the targeted 80% in most instances. This added information
allows the user and producer to judge the value of the LC map for different applications. For
example, for certain cover types, the map performed adequately to the life form level and could be
used in applications where this determination is all that is required. Fuzzy set theory was particularly
appropriate for LC classification systems that must be discrete but represent a continuum.
Adding the spatial location of accuracy to the accuracy ranks contributed additional accuracy
information to the LC map. Thematic map accuracy may vary spatially across a landscape in a
manner partially or totally unrelated to cover type. In other words, a cover type may be misclassified
more often when it occurs in certain contexts, such as on steep slopes. Also, cover types that were
