From Preparedness to Recovery: A Tri-Level Programming Model for Disaster Relief Planning - Computational Logistics - page 226

Fig. 2. Cities in the case study - Southeast USA. (Map data: Google, INEGI)

close to capacity to reduce fixed costs. When the disaster damage is higher, it
is beneficial to establish more centres that are low cost (e.g., centre 23 for D =
3, and centres 4 and 23 for D = 6). When a large number of evacuation centres
are expected to be destroyed by the disaster ( D = 9), the optimal solution relin-
quishes some high capacity (but high-cost) centres (e.g., centres 13 and 18) and
replaces them by more lower-cost centres. We also note that if the damage is
severe so that more evacuation centres are needed, the pre-positioned inventory
levels may not fully use the centres' capacities.

Tabl e 3. Pre-positioned inventory level (thousand units)
Centres established Size h i D =0 D =3D=6 D =9

2. Corpus Christi Large 25 2105 2105 1875 1875

3. San Antonio Large 20 2161 2161 2161 2161

4. Dallas Medium 21 — — 249 249

7. Memphis Medium 26 — — — 1207

8. Jackson Large 29 — — — 607

9. El Dorado Large 27 2306 2306 2287 2034

13. New Orleans Medium 30 725 725 725 —

18. Atlanta Medium 38 1110 1110 1110 —

19. Columbia Large 32 2289 2289 2289 1016

23. Savannah Small 22 — 20 20 20

24. Tallahassee Medium 29 — — — 1121

26. Jacksonville Medium 28 — — — 1211

27. Orlando Medium 30 1218 1198 1198 413

29. Miami Small 27 25 25 25 25

— indicates that the centre is not established.

We summarize the insights for disaster relief planning from our computational
study as follows: