Spatial Economic Impact Models: Applications to Terrorism Events and Natural Disasters

1
Spatial Economic Impact Models Applications to
Terrorism Events and Natural Disasters

• SAE 599 Special Topics, Modeling and Simulation
  for
• Systems Architecting and Engineering
• October 31, 2007

Prof. Jim Moore, ISE Prof. Qisheng Pan,
TSU Prof. Peter Gordon, SPPD Jiyoung Park, PhD,
CREATE Sungbin Cho, PhD, ImageCat Prof. Harry
Ward Richardson, SPPD
2
Scenario Development

• Challenging parts of this work involve developing
  plausible scenarios that are
• compatible with the models, and
• interesting to policy makers.
• Recognizing scenario limits requires recognizing
  model limits.
• How far into the future can modeling be useful?
• Ensure that users appreciate the limits.

3
Objective Develop Operational Models that
Include Spatial and Economic Detail

• Avoid errors created by spatial aggregation.
• Integrate economic and infrastructure impacts.
• The Southern California Planning Model (SCPM).
• The National Interstate Economic Model (NIEMO).
• Risks from Earthquake Damage to Roadway Systems
  (REDARS).

4
Southern California Planning Model (SCPM)

• An integrated highway network-economic-spatial
  allocation model of the Los Angels metropolitan
  area.
• 47 economic sectors (USC Sectors ) ,
  translatable into other U.S. industrial and
  commodity codes.
• 3,191 Traffic Analysis Zones (TAZs) and 89,356
  highway links including 647 HOV lane-miles.

5
(No Transcript)
6
What is Endogenous in SCPM?
7
SCPM (Cont.)

• Hypothetical or anticipated scenarios determine
  changes to infrastructure and/or jobs and
  population by location.
• Results are based on transportation network
  equilibrium costs and trip production and
  attraction vectors determined in the model,
  calibrated via 38 separate spatial interaction
  models (9 flows involving people, and 29 classes
  of commodity flows).

8
Baseline Calculations

• Begin with an empirical (or otherwise estimated)
  set of travel requirements for freight and people
  and a network.
• Adjust each of the matrices of inter-zonal flows
  separately in response to a common measure of
  network equilibrium costs. The structure of
  inter-zonal requirements in each of the matrices
  influences network equilibrium costs. The
  equilibrium network costs influences inter-zonal
  requirements.

9
Baseline Calculations (cont.)

• Baseline calibration requires iteration between
  the network assignment model and the set of
  gravity models (one for each type of flow) to
  find gravity parameters that match flows to the
  empirical travel time distribution.
• Result is a matrix of equilibrium link costs and
  volumes consistent with a corresponding set of
  equilibrium trip-interchange matrices and gravity
  model parameters for each flow.

10
Allocating Spatial Impacts

• Resulting passenger and freight flows are used to
  compute Garin-Lowry style matrices that would
  otherwise be exogenous, and which are used to
  estimate the spatial allocation of indirect and
  induced economic impacts (jobs and dollars) by
  sector.  

11
Allocating Spatial Impacts (Cont.)

• The current version of SCPM is able to loop
  travel delays from traffic assignment back to
  trip generation and distribution, which allows
  the model to smoothly bridge the gaps between
  upper- and lower-stream modeling.
• Impacts on economic structure and the
  transportation system performance are captured in
  TAZ-level spatial detail. They can also be
  reported at more aggregated level, such as city
  or county level.

12
Baseline Inputs The Special Case of Freight

• SCPM captures micro-level commodity freight
  flows. Identifies major freight facilities, such
  as ports, airports, rail yards, and highway
  entry-exit points in a metropolitan region and
  examines freight movement in and out of these
  facilities.
• These data are fed into a freight module with
  intra-regional commodity flows to generate a
  freight origin-destination (OD) matrix in
  passenger-car-equivalent (PCE) values, which
  allows SCPM to load freight trip ODs together
  with personal trip ODs onto the regional highway
  network and allows these two kinds of trips be
  assigned simultaneously.

13
Recent Applications

• Detailed spatial and sectoral impacts of a dirty
  bomb attack on the Los Angeles/Long Beach ports.
• Detailed spatial and sectoral impacts of a dirty
  bomb attack on downtown Los Angeles.

14
Dirty Bomb Attacks on the Los Angeles / Long
Beach Ports

• Assume the explosion of one or two radiological
  dispersal devices (RDDs) at the ports, which
  results in the closure of one or both ports on
  both health and security grounds.

15
Scenarios for Impact Analysis

• Local impact Ports close down imposing an
  economic shock directly on the port areas.
• Regional and national impacts interruption of
  trade flows to and from the ports
• Two alternative closure times 15 days with no
  bridge damage and 120 days with bridge damage
• Closure of Port of Los Angeles, Port of Long
  Beach, or both.
• Six scenarios for regional and national impacts
• 15-day Port Closure, No Bridge Damage (Port of
  Long Beach, Port of Los Angeles, or both).
• 120-day Port Closure with Bridge Damage (Port of
  Long Beach, Port of Los Angeles, or both).

16
Modeling Results

• The maximum regional and national impacts (the
  consequences of the interruption in exports and
  imports with bridge damage) would close the ports
  for at least 120 days, resulting in 34 billion
  of lost output -- 212,165 person years of
  employment (PYE) and 648 million in local travel
  cost delays.

17
Spatial Distribution of Job Losses 120-Day Port
Closure, Ports of Long Beach-Los Angeles, Bridge
Damage
18
Good Computational Behavior Rapidly Achieves
Stopping Conditions
19
Summary of Findings

• The local impacts of this type of attack are
  modest because direct loss is limited to the port
  areas.
• About 2/3 occur within the five-county region
• gt50 within LA County.
• It is relatively easy to disrupt port access, and
  the costs of trade flow interruption are very
  high.
• In the case of trade flow interruptions, about
  2/3 of the impacts are felt outside Southern
  California.
• The high economic impact costs justify
  considerable resource expenditures on prevention
  -- especially on freeway access routes
• The methodology used in this study is adaptable
  to almost any kind of terrorist attack and is
  also transferable to other large metropolitan
  areas (e.g. New York, Washington D.C., San
  Francisco, Houston) if a similar model were
  created for these areas.

20
Dirty Bomb Attack in Downtown Los Angeles

• Assume a 50 lb. radiological bomb explodes in a
  large office building in downtown Los Angeles
  (Financial District).
• Creates a radiological plume over many km2 in
  Downtown Los Angeles and its surrounding areas.
  The business interruption effects of evacuation
  are estimated.

21
Scenarios

• Exit Scenario All firms and households leave the
  region (or close down)
• Relocation Scenario All firms and households
  relocate elsewhere in the region for the
  evacuation period
• Hybrid Scenario Firms in the Inner Zone exit
  (there are no households) Firms and households
  in the Outer Zone relocate

22
Economic Impact of a Terrorist Attack on Downtown
LA, Exit Scenario, All Businesses and Households
Moving out of the Inner and Outer Zones
Source Authors calculations
23
NIEMO, An Operational Multi-Regional Input-Output
Model (MIRO)

• Uses IMPLAN I/O models for the 50 states and DC.
• Aggregates to 47 USC sectors.
• Uses Commodity Flow Survey (and other) data to
  estimate interstate commodity flows.
• Results in (47x47) x (51x51) MRIO.
• Demand-side and supply-side versions have been
  applied and tested.

24
Schematic Diagram of NIEMO Port Closure Scenario
25
Recent Applications

• One-month shutdowns of three major U.S. ports.
• Terrorist attack on theme parks.
• One-year border closure in response to avian flu
  epidemic.

26
One-Month Shutdowns of Three Major U.S. ports
(M)

• Los Angeles / Long Beach - 23,258.21
• New York / New Jersey - 16,824.25
• Houston - 10,049.93

27
Terrorist Attack on Theme Parks (M).
28
One-Year Border Closure, Avian Flu Epidemic (M)

• International Air Travel 113,429
• International Trade 2,223,037
• Legal Immigration 10,122
• Illegal Immigration 2,039
• Cross-Border Shopping 9,941
• Total 2,358,568

29
FlexNIEMO (Preliminary)

• Study post-event (Katrina) value added and final
  demand changes
• Apply RAS methods to estimate adjustments of
  input-output coefficients.

30
FlexNIEMO (Cont.)
31
Monthly Multipliers Changes for FlexNIEMO
32
TransNIEMO Placing Interstate Truck Trade on the
Interstate Highway Network

• Apply FHWA Freight Analysis Framework data to
  estimate truck trip OD data between sub-state
  areas.
• Network Centroids 10 sampled intersections
  within a specified boundary of economically
  weighted FHWA Freight Analysis Framework
  Centroids of each sub-state
• TransNIEMO baseline Estimation of shortest
  paths between network centroids and distribution
  of OD data onto the paths
• Calculate freight costs for each USC sector.
• Scenario-based simulations Alternative shortest
  paths imply additional freight costs, which
  increase costs of product and hence induce losses
  in final demand. In turn, price-type IO model can
  address any increased prices and NIEMO will
  estimate their total economic losses.

33
Quantifying Economic Loses from Travel Forgone
Following a Large Metropolitan Earthquake

• SAE 599 Special Topics, Modeling and Simulation
• for Systems Architecting and Engineering
• October 31, 2007

Prof. Jim Moore, USC Prof. YueYue Fan,
UCD Sungbin Cho, PhD, ImageCat Stuart D. Werner,
Seismic Systems and Engineering Consultants
34
Acknowledgements and Disclaimer

• This work was supported primarily by the
  Earthquake Engineering Research Centers Program
  of the National Science Foundation under award
  number EEC-9701568 through the Pacific Earthquake
  Engineering Research Center (PEER).
• This work made use of the Earthquake Engineering
  Research Centers Shared Facilities supported by
  the National Science Foundation under award
  number EEC-9701471 and by the Federal Highway
  Administration through the Multidisciplinary
  Center for Earthquake Engineering Research
  (MCEER).
• Any opinions, findings, and conclusion or
  recommendations expressed in this material are
  those of the authors and do not necessarily
  reflect those of the National Science Foundation
  or the Federal Highway Administration or the
  California Department of Transportation.

35
Key References

• Moore, II, J.E, S. Cho, YY. Fan, and S. Werner
  (2006) Quantifying Economic Loses from Travel
  Forgone Following a Large Metropolitan
  Earthquake, PEER Report 2007/, Berkeley, CA
  Pacific Earthquake Engineering Research Center,
  forthcoming.
• Kiremidjian, A., J. E. Moore, II, YY. Fan, N.
  Basoz, O. Yazali, and M. Williams (2006) Pacific
  Earthquake Engineering Research Center Highway
  Demonstration Project, PEER Report 2006/02,
  Berkeley, CA Pacific Earthquake Engineering
  Research Center, forthcoming.
• Werner, S. D., S. Cho, C. E. Taylor, J. P.
  Lavoie, C. K. Huyck (2006) Seismic Risk Analysis
  of a Roadway System in the Los Angeles,
  California Area, Proceedings of the of National
  Seismic Conference on Bridges and Highways, San
  Francisco, CA.
• Werner, S. D., C. E. Taylor, S. Cho, J. P.
  Lavoie, C. K. Huyck, C. Eitzel, R. T. Eguchi, and
  J. E. Moore, II (2004) New Developments in
  Seismic Risk Analysis of Highway Systems, Paper
  2189, Proceedings of the 13th World Conference on
  Earthquake Engineering, Vancouver, BC.
• Cho, S.B., YY. Fan, and J. E. Moore, II (2003)
  Modeling Transportation Network Flows as a
  Simultaneous Function of Travel Demand,
  Earthquake Damage, and Network Level of Service,
  Advancing Mitigation Technologies and Disaster
  Response for Lifeline Systems, Proceedings of the
  6th US Conference, Long Beach, CA.

36
Motivation

• Caltrans District 7 was immediately attacked in
  the press following the Northridge Earthquake.
• Some facilities had failed.
• The media has a tendency to equate bad outcomes
  with bad decisions.
• Repair of the I-10 bridges following the
  Northridge Earthquake in 1994 produced two
  controversies.
• Bonuses paid to C. C Meyers to accelerate the
  work were thought by some to be a poor use of
  public resources.
• Some prominent earthquake engineers criticized
  the design standards of the new bridges as not
  sufficiently earthquake resistant.

37
Insights

• There is more to a facilitys importance than
  Average Daily Traffic.
• Available redundant capacity in the network can
  and should be accounted for.
• Prioritizing bridge retrofits (or
  reconstructions) is an exercise in network
  design.
• Resources are scarce.
• We cannot afford to design every transportation
  structure in the inventory to withstand a maximum
  credible earthquake.
• District 7 still did a commendable job deploying
  innovative, low cost retrofits prior to the
  Northridge Earthquake.
• There is considerable serious work to be done at
  the interface of transportation engineering and
  earthquake engineering.

38
Stages of Interdisciplinary Work

• Your fields problems must be trivial, otherwise
  my own fields methodologies would have already
  addressed them.
• Your field focuses on substantive problems, but
  these must be intractable, otherwise my fields
  methodologies would have already addressed them.
• Denial, Anger, Bargaining, and Acceptance.
• Your field includes methodologies that might be
  relevant to standing problems in my own field.
• Understanding your fields methods helps define
  new problems and opportunities in my own field.

39
Imposing Pre-earthquake Travel Demand on a
Post-earthquake Network

• Fails to account for
• Movement along the travel demand curve, or
• Shifts in the travel demand curve.
• Overestimates post-earthquake travel volumes.
• Generates unrealistic volume/capacity ratios.
• Generates unrealistic travel delays.
• Is a source of embarrassment for transportation
  engineers who are attempting to persuade
  earthquake engineers of the importance of
  transportation engineering.

40
Applying Standard Transportation Planning Models
to Earthquakes
41
Treating Post-earthquake Travel Demand as a
Function of Network Level of Service

• Adds considerable economic and behavioral realism
  by allowing equilibria in the market for
  transportation services to shift along a
  conventional demand curve.
• Better estimates post-earthquake travel volumes.
• Generates wholly realistic volume/capacity
  ratios.
• Generates wholly realistic, yet elevated
  zone-to-zone travel delays.

42
Stepping Back Recognizing that Travel Demand is
a Function of Level of Service
43
Treating Travel Demand as a Function of Network
Level of Service

• Substantially complicates network assignment
  calculations intended to identify user
  equilibrium flows.
• Is outside standard practice, but almost within
  the grasp of standard computational tools, and
  should likely become standard practice.
• Generates an apparent reduction in total travel
  delay due to reduced travel demand, thereby
• Making it appear that earthquakes improve
  transportation system performance, and
• becoming a source of embarrassment for
  transportation engineers who are attempting to
  persuade earthquake engineers of the importance
  of transportation engineering.

44
Cumulative Distribution of Post-earthquake
Volume/Capacity Ratios
45
REDARS (Risks from Earthquake Damage to Roadway
Systems)

• Software package supplied by the Federal Highway
  Administration (FHWA).
• An advanced seismic risk analysis (SRA) tool that
  enable users to better plan for and respond to
  earthquake emergencies.
• Methodologys risk-based framework uses
• models for seismology and geology, engineering
  (structural, geotechnical, and transportation),
  repair and reconstruction, system analysis, and
  risk analysis to
• estimate economic losses due to
  earthquake-induced repair costs, increased travel
  times, and reduced trip demands due to earthquake
  damage to the highway system .

46
REDARS (cont.)

• Developed by FHWA and the Multi-Disciplinary
  Earthquake Engineering Research Center (MCEER) as
  a future public-domain software package, and
  ?-tested by Caltrans as the REDARS Demonstration
  Project.
• REDARS 2.0 incorporates a version of the Variable
  Demand Model operationalized in the PEER Highway
  Demonstration Project.
• Successfully applied to the
• Memphis, TN highway network, a location that is
  vulnerable to a repeat of the 1812 New Madrid
  zone earthquakes,
• The northern, central and southern sections of
  the Los Angeles highway network, and to a
• limited portion of Caltrans highway network
  extending from Fairfield to Oakland.
• The California project was intended to transfer
  technical expertise from the developer community
  within FHWA and MCEER to Caltrans.

47
REDARS Methodology
48
REDARS Seismic Risk Analysis (SRA) Modules
49
The PEER Variable Demand Model is incorporated
into REDARS 2.0

• For a given earthquake scenario and network data,
  REDARS 2.0 sequentially analyzes
• ground motion, liquefaction, and surface fault
  rupture hazards
• bridge / tunnel / roadway damage states
• network configurations
• executes a VDM analysis of network level of
  service
• Reports results for any four time periods
  following the earthquake, in this application
• 7 days
• 60 days
• 150 days following the event.

50
Endogenizing Travel Demand

• Requires parameterization of travel demand
  curves,
• Which can be done on a zone-to-zone basis
• Based on baseline travel demands and costs, and
• A gravity model calibration, or equivalent
  calculation.
• But which ideally would be based on a model of
  the urban activity system
• Makes it possible to determine
• The total increased delay experienced by
  travelers who remain on the network, and
• The number of trips eliminated from the network,
  and their value to the people who were previously
  making them, thereby
• Providing a long sought after source of
  credibility for transportation engineers who are
  traveling in the intellectual company of
  earthquake engineers.

51
Economic Losses Linked to Network Level of
Service Following an Earthquake
52
Obtaining Empirical Estimates of Coefficients for
Monotone Travel Demand Functions
53
Empirical Travel Demand Curves are Non-monotonic
54
REDARS 2.0 Import Wizard

• Combines federal, state, and local data from
  public sources to generate transportation network
  data for study area.
• Public data sources used to compile the network
  database consist of
• National Highway Planning Network (NHPN) from the
  Federal Highway Administration (FHWA),
• FHWA Highway Performance Monitoring System (HPMS)
• FHWA National Bridge Inventory (NBI),
• Bay Area transportation analysis zone map from
  the Metropolitan Transportation Commission (MTC),
  and
• MTC 1998 Bay Area (passenger) trip table (Peak 4
  hours).

55
The Bay Area Highway Network Model Characterized
by the Import Wizard Includes

• 10,154 directional links
• 3,288 nodes,
• 1,136 Travel Analysis Zone (TAZ) centroids,
• 1,475 bridges, and
• eight tunnels.

56
The San Francisco Bay Area Roadway Network
Characterized by the REDARS 2.0 Import Wizard
57
Hayward Fault Scenario Earthquake (Early Results,
Since Revised)

• Moment magnitude 7.1 event along the Hayward
  fault.
• Epicenter at -122.0866 o / 37.7266 o in decimal
  longitude and latitude.
• REDARS 2.0 estimates
• bridge collapses
• damaged bridges
• links subject to pavement failures due to
  liquefaction
• Full reconstruction or repair in 231 days,
  assuming no constraints on resources
• Note The most recent revision to REDARS
  predicts fewer bridge collapses and more
  extensive damage to roadway links due
  liquefaction and surface fault rupture.

58
Bridge and Link Damage States Associated with the
Hayward Fault Scenario Earthquake
59
Variable Demand Model Algorithm Performance

• Four minutes of calculations using desktop
  computing resources.
• Travel demands associated with only 20 of the
  origin-destination zone pairs have converged to
  values consistent with the associated set of
  empirically estimated travel demand functions.
• The flows associated with these zone pairs
  account for 95 percent of the total trips in the
  system.
• The remaining 80 of the zone pairs account for
  only about 5 percent of the trips.

60
Variable Demand Model is Effective for Most
travel, but not Most Zone Pairs
61
(No Transcript)
62
Total Household Transportation Impacts
63
Extensions

• Most recent revisions to REDARS.
• Treating freight flows.
• Accounting for demand shifts, as opposed to
  movements along a demand curve.
• Decision support and network design.

64
Sample Results from REDARS 2
65
Freight Trip Generation

• MTC does have freight origin-destination tables
  available.
• Alternatively, employment data from the 2000
  Census Transportation Planning Package (CTPP) for
  the San Francisco Bay Area can be used to
  construct intra-regional freight trip generation
  estimates.
• The CTPP includes employment data by economic
  sector and by place of employment (by Traffic
  Analysis Zone).
• Commodity flows between industries can be used to
  estimate freight trip productions and
  attractions.
• To convert this aspatial information to spatial
  flows, disaggregate and assign these interactions
  to each TAZ based on 2000 CTPP employment by TAZ
  and by sector.
• Interregional flows are estimated by
• Identifying network locations associated with
  inter-regional freight movement, including
  seaports, airports, rail yards, and highway
  network entry points, and
• Assembling freight tonnage data for inbound and
  outbound freight for each of these sites,

66
Movement Along a Demand Curve versus a Shift in
Demand
67
Network Design Problem

• Broadly stated, our research goal is to find,
  subject to certain resource constraints, which
  network components should be retrofitted, and
  where new components should be added so that the
  overall performance of any metropolitan
  transportation system is maximally improved.
• This well-defined network design problem is
  important in the transportation network
  literature (Yang and Bell 1998).

68
Deterministic Network Design is Reasonably
Difficult

• Individual users and network planners do not have
  the same objectives. Consequently, the network
  design problem involves multiple levels of
  optimization.
• At the upper level, the system planner makes
  decision on resource allocation to achieve the
  best system performance.
• At the lower level, the network users make their
  travel decision based on their individual travel
  preferences.
• For a large network, this kind of network design
  problem is computationally challenging.

69
Stochastic Network Design is Even More Difficult

• Uncertainty makes the pre-event network design
  problem very challenging. The problem has been
  formulated (Yang and Bell 1998), but never
  treated at a realistic scale.
• Subject to budget constraints, the objective is
  to find the transportation network configuration
  on which user equilibrium flows produce the
  minimum expected total congestion.
• This stochastic version of the problem is an
  embedded optimization problem with a tri-level
  structure.
• The upper level is the decision by the network
  authority, in this case a pre-event retrofit or
  reconstruction decision.
• The intermediate level outcome, a function of the
  upper level decision, is a random result of
  nature.
• The lower level, a function of the upper level
  decision and the intermediate outcome, is the
  decision by the network user.

70
Complexity

• Assuming that retrofitting transportation
  structures is not a matter of degree, but rather
  a binary decision (itself an over
  simplification)
• then a network with M transportation structures
  supporting its links presents 2M retrofit
  options.
• A random act of nature converts the network to a
  collection of L lt M links.
• The total number of possible networks to be
  considered is thus an impossibly large value,
• Explicit enumeration of options is out of the
  question, so now what?