To the extent that measurement error can be
quantified and propagated through an analysis, an
expression of decision risk is thus not very difficult to
achieve. However, the recognition of factors in MCE
as fuzzy sets implies a very different form of
uncertainty from that of measurement error. The
suitability map that results from weighted linear
combination is a clear expression of uncertainty
about the suitability of any particular piece of land
for the objective under consideration. However, as an
expression of uncertainty, it has no relationship to the
frequentist notion of probability that underlies the
treatment of decision risk in the context of
measurement error. Thus a traditional treatment of
decision risk as the probability that the decision made
will be wrong cannot be developed. Eastman (1996)
has therefore suggested that decision risk for such
cases be expressed by the concept of relative risk.
A mapping of relative risk can be very simply
achieved by rank ordering the alternatives and
dividing the result by the maximum (i.e. worst)
rank that occurs. The outcome is a proportional
ranking that can directly be interpreted as relative
risk. Then in cases where no specific area
requirement for the decision set is being sought
(e.g. the best 10 hectares), the final decision set can
be established by selecting the alternatives where
the relative risk does not exceed a specific threshold
(e.g. the best 5 per cent of the areas under
consideration). Figure 6 illustrates such a mapping
of relative risk for the result of Plate 32 along with
a mapping of the best (least risky) 10 per cent of
cases outside the wildlife reserve.