Research in Coastal Infrastructure Reliability: Rerouting Intercity Flows in the Wake of a Port Outage

Research in Coastal Infrastructure Reliability: Rerouting Intercity Flows in the Wake of a Port Outage Megan S. Ryerson, Ph.D Department of City and Regional Planning Department of Electrical and Systems Engineering University of Pennsylvania mryerson@design.upenn.edu meganryerson.com

Introduction Port and airport outages present a major disruption to intercity flows: passengers, freight, and possibly life-saving goods In the immediate aftermath of an outage: Flights/ships are diverted to the nearest safe port, stranding travelers/goods In the short and medium-term after an outage (hours to days): Closed hub airports and seaports impede transfer flows Airports and seaports are located in climate vulnerable locations Establish ad-hoc hubs in an outage scenario and reroute ships and aircraft to these hubs to maximize passenger and goods throughput or minimize passenger/goods delay

Airports, Seaports & Sea Level Rise Boston SFO, OAK NY Metro Area Airports, Philadelphia Miami Major hub airports in low-lying areas that are subject to flooding today at high tide. 3

Extreme Weather Events LGA during Hurricane Sandy 2012 Northeastern Snowstorm 2011 BOS during Winter Storm Juno, 2015 4

Modeling diversions to ad-hoc hubs immediately after an outage occurs: A Large Neighborhood Search Heuristic to Establish an Optimal Ad-hoc Hubbing Strategy in the Wake of a Large-Scale Airport Outage Modeling diversions to ad-hoc hubs hours to days after an outage occurs: The q-ad Hoc Assignment Problem in the Aftermath of a Seaport Outage 5

Research Question: How could flights be diverted in the immediate aftermath of an airport outage such that passengers can be accommodated on existing flights departing non-impacted airports Objective: Design a heuristic that identifies an ad-hoc hubbing strategy: a strategy to reroute flights bound for a disrupted airport to a hub airport that is not disrupted, with the goal of accommodating passengers on existing flights departing the non-disrupted hub with minimal travel and wait time while maintaining physical feasibility (including constraints on on-board fuel and diversion airport capacity) 11

Ad-hoc Diversion Algorithm Schematic Set of en route flights i Set of diversion airports d No Can flight i reach diversion airport d? Yes Calculate ground cost from closest traditional alt airport Passenger Wait Time Algorithm

Ad-hoc Diversion Algorithm Schematic Passenger Wait Time Algorithm Set of true destinations r Does div airport d serve true destination r? No Set of secondary airports s Yes Calculate pax wait time costs from d to r Select s with minimum wait time costs Wait Time Costs

Case Study Approach: Design an ad-hoc diversion algorithm to test the optimal diversion scenario if San Francisco International Airport experiences an outage and other coastal airports are not available Outage scenario SFO (outage airport) 11/24/2014 at 2pm local time (outage date and time) Search distance = 1300 miles Identify flights en route to SFO 29 en route flights at the time of outage Potential diversion airports These airports must be physically equipped to handle large aircraft that are not in low lying flood zones (OAK and SAN are removed) 18 potential diversion airports (OAK, PHX, LAX, SLC, SEA, LGB, TUS, BUR, PDX, SAN, ONT, SJC, DEN, SMF, PSP, SNA, LAS, ABQ) BUR is the selected diversion airport with minimum wait costs 18

19 Identify en route flights

20 Divert to BUR for passenger optimization

22 Three transport levels: FTAs, Seaport hubs, and Inland demand/supply nodes

23 Status Quo Model Formulation: Classic Hub Location Problem

24 Ad-hoc Hub Location Problem with Single and Multiple Allocation (Case of 2 Hub Outages)

Data Sources Data from the Department of Transportation Bureau of Transportation Statistics, Freight Analysis Framework The data includes 2015 estimated freight flows between the seven FTAs (Canada, Mexico, Europe, Africa, Central Asia, East Asia, and Oceania) and the 17 U.S. major seaports nodes plus 19 inland destinations 25

Criticality of seaport hubs (top 10) Criticality Ranking AHLP-SA, α=0.3 AHLP-MA, α=0.3 AHLP-SA, α=0.9 AHLP-MA, α=0.9 1 Boston 2 Seattle 3 Houston 4 Miami 5 Los Angeles/Long Beach 6 Hampton Roads Savannah Hampton Roads New Orleans 7 Savannah Hampton Roads Savannah Hampton Roads 8 New Orleans New York New Orleans Savannah 9 Baltimore Philadelphia Baltimore Portland 10 Philadelphia Portland New York 26

Recuperability of the AHLP to outages with the increase of ad-hoc hubs 27 The recovery scenario is tested to the worst scenario of hubs outage: Anchorage (8) for R =1, and Anchorage (8), Boston (12), and Houston (23) for R =3, respectively, q=0 to 6

Critical research gaps in network planning for coastal infrastructure reliability Day to day port/airport self-reflection: What are my vulnerabilities today? Collaborative Decision Making for Ad-hoc Hubbing The trade-off between carrying fuel on board (buying options) and burning additional fuel 28

29 Megan S. Ryerson, Ph.D. Assistant Professor Department of City and Regional Planning Department of Electrical and Systems Engineering University of Pennsylvania mryerson@design.upenn.edu meganryerson.com

Additional Slides 30 Megan S. Ryerson, Ph.D. Assistant Professor Department of City and Regional Planning Department of Electrical and Systems Engineering University of Pennsylvania mryerson@design.upenn.edu meganryerson.com

Algorithm output table Airport DivPAX SecPAX Flight Time to Div Total Wait Time (Div) Total Wait Time (Div to Sec) Flight Time Div to Sec Total Wait Time (sec) Flight Time Sec to Div NoFuelCos t BUR Burbank, CA 3417.20 33.98 40.81 58.85 13.21 24.35 1.36 100.60 1393.11 1632.28 LGB Long Beach, CA 3297.08 26.45 40.40 26.09 0.72 64.84 10.51 61.09 1501.67 1705.34 PSP Palm Spring, CA 3205.29 65.34 38.42 12.17 0.81 140.92 1.26 306.08 1501.67 2001.33 LAS Las Vegas, NV 3454.13 19.29 39.77 744.22 33.05 15.33 0.45 77.64 1362.66 2273.12 SLC Salt Lake City, UT 2442.76 15.73 23.00 90.27 6.35 16.73 0.35 70.91 2234.63 2442.22 PHX Phoenix, AZ 2085.22 1.08 22.13 54.12 0.42 0.55 0.01 1.29 2513.98 2592.50 TUS Tucson, AZ 1941.93 1.12 22.58 0.30 0.02 0.72 0.02 3.66 2643.49 2670.79 SJC San Jose, CA 5116.32 2.95 46.06 1276.09 0.91 1.11 0.91 1.11 1501.67 2827.87 LAX Los Angeles, CA 3417.20 3.44 41.48 1480.10 9.52 2.56 0.11 1.42 1393.11 2928.31 ABQ Albuquerque, NM 1596.86 3.80 13.25 0.30 0.16 8.52 0.06 12.77 2911.01 2946.07 DEN Denver CO 1493.42 1.08 9.83 12.41 0.46 1.95 0.04 0.44 3004.50 3029.63 SNA Santa Ana, CA 3248.63 6.56 39.39 1.63E+06 0.02 3.64 0.02 3.64 1501.67 1.63E+06 PDX Portland, OR 2062.49 1.08 37.34 4.68E+06 0.20 1.03 0.04 0.44 2533.97 4.68E+06 ONT Ontario, CA 3297.08 67.75 39.75 7.62E+13 67.49 52.78 1.37 316.95 1501.67 7.62E+13 SMF 31 Sacramento, CA 4570.41 6.02 44.46 4.16E+15 0.20 2.49 0.21 2.49 493.40 4.16E+15 SCORE

33 Ad-hoc Hub Location Problem with Single Allocation

Ad-hoc Hub Location Problem with Multiple Allocation flows from i to j travel through ad-hoc hub(s) k and m allowing for a more flexible transport prevent the flow from i to j being routed via k and m if any of hubs k and m remains disrupted. 34