Information Technology Reference
In-Depth Information
low inventory cost. Evacuation centres with high inventory cost, e.g. at location
(193,948), are established only due to insucient capacities of the low-inventory-
cost centres and the remoteness of the communities served.
By examining the solutions for the cases with
D
=3
,
6 and 9, we can see
how the optimal strategy changes as the anticipated level of damage increases.
When there are a small number of evacuation centres destroyed (Figure 1.b), the
location-allocation is the same as for the case of
D
= 0, since there are enough
evacuation centres with low post-disaster inventory-delivery cost, e.g. the cen-
tres at (217,168), (321,96), (407,609) and (459,594), to support the communities
assigned. When the disaster damage is moderately high (Figure 1.c), the best
strategy established some centres with low inventory-delivery cost post-disaster
but a higher inventory pre-positioning cost. For example, the centre at (853,20)
is replaced by the centre at (935,105). It indicates that, when a significant num-
ber of centres are destroyed by the disaster, there is a trade-off between the costs
of pre-positioning inventory and of delivering inventory post-disaster.
Another interesting finding is shown in Figure 1.d. When a large number of
evacuation centres are expected to be destroyed by the disaster, it is helpful to
establish
more evacuation centres
with low post-disaster inventory-delivery costs,
even if some of them may not have any inventory pre-positioned, e.g. the centres
at (120,179) and (459,594). The number of evacuation centres increases from 12
to 14 in this case. Again, there is also a trade-off between the pre-positioning and
post-disaster costs, as illustrated in the replacement for the centre at (978,486)
by the centre at (833,572). A solution that does not anticipate post-disaster costs
would be much more expensive when a severe disaster strikes.
This computational study shows that the optimal location-allocation strat-
egy that accounts for recovery cost post-disaster may be very different from one
that does not consider the costs after a worst-case scenario. Therefore, in disas-
ter preparedness planning, it is very important that post-disaster scenarios are
considered when determining the location of evacuation centres.
6.4 Case Study
We investigated a case study on hurricane preparedness in southeast USA, based
on the study of Rawls and Turnquist (2011). There are 30 cities, each being
a possible location for an evacuation centre; see Figure 2. We use the actual
distances and populations for our computations. Other parameter values are
adapted from Rawls and Turnquist (2011) as follows. The size and fixed cost of
the centre considered for each city is pre-selected (with an equal chance) to be
either
Small (F
i
=58,800, C
i
=109,200)
,
Medium (F
i
=565,200, C
i
=1,224,600)
or
Large (F
i
=900,000, C
i
=2,340,000)
. For each evacuation centre,
h
i
is randomly
generated from a discrete uniform distribution
U
[20
,
50] with the post-disaster
inventory delivery cost
H
i
being 40 times higher.
We solve the problem for
D
= 0, 3, 6 and 9; the optimal set of centres es-
tablished and pre-positioned inventory levels are shown in Table 3. Similar to
results reported in Section 6.3, when no disaster strikes (
D
= 0), the optimal so-
lution establishes fewer evacuation centres and pre-positions inventory to levels
