t al., 2012).
The multiplication factor sensitivity analysis indicated that the
three most critical monitors were unaffected in their rank-order
when we varied the multiplication factors in the index, which
demonstrates for this system that the index is stable and not largely
driven by the choice of using multiplication factors to emphasize
the false-positives and false-negatives. We suspect that the index is
not susceptible to the multiplication factors because of the limited
number of hours that concentrations exceeded the AQS, which was
only 2.5% of the period when one or more monitors recorded a
concentration above the AQS. In regions with a lower AQS or higher
concentrations, the index will be more susceptible to the choice of
the multiplication factors, and various values should be tested to
determine the magnitude of any effect. This process will provide a
Pareto front of index values by varying the multiplication factors,
from which an average rank or index value can be calculated. Pareto
fronts have been applied when many potential near-optimal solutions
are computed in network design problems (Kornelsen and
Coulibaly, 2014