Decision Models for Emergency Response Planning

DECISION MODELS FOR EMERGENCY

RESPONSE PLANNING

Larson, R., Metzger, M., & Cahn, M.

CREATE REPORT

Under FEMA Grant EMW-2004-GR-0112

September 28, 2004

Center for Risk and Economic Analysis of Terrorism Events

University of Southern California

Los Angeles, California

3710 McClintock Avenue, RTH 314 Los Angeles, CA 90089-2902 ~ (213) 740-5514 ~ www.usc.edu/create

Report #04-002

DRAFT

Decision Models for Emergency Response Planning

Richard C. Larson, Massachusetts Institute of Technology

Engineering Systems Division and

Department of Civil and Environmental Engineering Room 1-170

Cambridge, MA 02139

and

Structured Decisions Corporation

1105 Washington Street

West Newton, MA 02465

This research was supported by the United States Department of Homeland Security through the Center for

Risk and Economic Analysis of Terrorism Events (CREATE), grant number EMW-2004-GR-0112.

However, any opinions, findings, and conclusions or recommendations in this document are those of the

authors and do not necessarily reflect views of the U.S. Department of Homeland Security.

A chapter submitted to

The McGraw-Hill Handbook of Homeland Security, David Kamien, editor.

September 28, 2004

Operations Research (O.R.), born in World War II (WWII), has for 65 years proved

invaluable as a decision-planning tool. Known as the science and technology of decision-

aiding, O.R. is an empirical science that uses the scientific method to assess the

consequences of alternative decisions, be they long-term strategic planning decisions or

shorter range tactical or operational decisions. Since a decision can be viewed as an

allocation of resources, Operations Research is the science of resource allocation. In

WWII O.R. helped guide the allocation of scarce resources against the enemy.

Successful applications ranged from finding optimal locations for new and expensive

radar installations in Great Britain (in order to detect incoming enemy aircraft and

missiles), to the invention of ‘optimal search theory,’ used to deploy aircraft and ships in

search of enemy submarines [32]. The search theory results were deemed so important

that the original papers by Bernard Koopman remained classified for 15 years. Today

O.R. is ideally suited for evaluating and guiding our operational strategies and actions

with regard to large scale emergency incidents, be they acts of terrorism, acts of Mother

Nature (e.g., earthquakes, floods, tornadoes, hurricanes) or industrial accidents.

Following WWII, Operations Research found widespread applications in civilian sectors,

both in private companies and in the nonmilitary government sectors. The collective

result has been savings of billions of dollars from costs of operation and significant

increases in the quality of services provided – in both the public and private sectors. The

majority of Fortune 500 companies have utilized O.R. inside to help them in their

decision making, long, medium and short term. The O.R. civilian sector includes the US

Postal Service, which for decades has used O.R. extensively for designing routes,

scheduling personnel and designing its national distribution network. It also includes the

City of New York, which as a result of 30 years of successful O.R. experience, created its

DRAFT

own permanent O.R. group within the City’s Office of Management and Budget. And

the military has its own Military Operations Research Society, whose 3,000 members

must have a security clearance to attend its meetings.

O.R. is in many places where you least expect it. New England Patriots football coach

Bill Belichick uses O.R. in the execution of a football game. In a famous incident in

2003, with the ball on the Patriots’ side of the 50-yard line, on a fourth down and one-

yard-to-go situation, he went for first-down. His decision was viewed as contrary to what

90 percent of NFL football coaches would have done, but was supported by 30+ pages of

sophisticated “Bellman equation” O.R. analysis of a Stanford University professor

. The

decision proved correct, the first down was achieved, the team continued to a touchdown

and later in the season to another Super Bowl victory. O.R. is not just an academic

discipline, any more than electrical or mechanical engineering are just academic

disciplines. O.R. is an academic discipline that continues to have numerous substantial

beneficial impacts on organizations, large and small. While football and response to

terrorist attacks are very different operations, we cite the football example to show that

O.R. can be used to inform and transform decisions in the real world, with all its gritty

complexity. One needs that type of down-to-earth practical science to confront the many

challenges of emergency response planning and operation. As explained below, that is

why O.R. proved invaluable to New York City firefighters on September 11, 2001.

What are the methods of O.R.? The answer is: any aspect of the scientific method that

sheds light on the problem at hand. Usually but not always mathematics is involved.

Usually but not always a mathematical model of the system or problem being studied is

created, tested, refined and then implemented on a computer. Once computer-

implemented, the O.R. analyst runs the model under alternative assumptions, leading to

optimal or near-optimal system configuration or decision policies. But some very famous

O.R. studies have involved no mathematics at all, just clever application of ‘common

sense,’ often leading to new and insightful re-definitions of the problem.

Response to a major emergency incident requires careful planning and professional

execution of plans, when and if an emergency occurs. Decisions involve the movement

(‘deployment’) of people, equipment and supplies. They also include the development of

policies with regard to operation, once men, women and materiel are in place. This

setting is nearly perfect for the application of O.R. – to create better emergency response

plans and to design better operational policies once the rescue efforts are underway.

In this chapter we review briefly the major O.R. work done to date in emergency

response. Some of this work is quite recent and aimed directly at what we now call

Homeland Security issues. Most of it has evolved over the past 50 years, motivated by

other emergency applications, especially operation of first responders in municipalities –

police, fire and emergency medical. The good news is that we are now building from a

rich legacy of 50 years of research in emergency response. The bad news is that only

recently have we collectively refocused our energies towards catastrophic events

New York Times, David Lionhardt, “Incremental Analysis, with Two Yards to Go,” February 1, 2004.

Also see

http://espn.go.com/nfl/columns/garber_greg/1453717.html.

DRAFT

associated with Homeland Security. But our 50-year’s of research and implementation

results provide us today with solid platforms for moving forward. Space limitations do

not allow us to provide a comprehensive literature review of the field of “O.R. and

emergency response.” So, we have selected the works that we feel are most important for

progressing with emergency response and Homeland Security. The new threats posed by

terrorists present myriad new problems for O.R. analysts. In some ways, today we stand

at a place analogous to the place that Drs. Philip M. Morse, George Kimball, Bernard

Koopman and other O.R. pioneers stood near the beginning of WWII. There are

numerous new O.R.-related problems to identify, to frame, to formulate and to solve.

Hopefully the end result will be the most rational deployment of our scarce national

resources, resulting in maximally many lives saved and injuries averted in the instance of

another terrorist attack. But since the methods we describe also apply to emergencies

created by Mother Nature and by human accident, let us hope that the vast majority of

incidents in which these methods are applied are from these latter two categories.

For purposes of this chapter, a major emergency is one in which local first responder

resources are overwhelmed. There simply are not enough resources to do the many jobs

at hand. To read a history of such an event, one that took place across the entire United

States, we suggest studying the ‘great influenza epidemic’ of 1918-1919, a pandemic that

took more American lives than all wars of the 20

Century [1]. Nurses, doctors and other

resources were simply too few to care for the sick and dying. But more critically, history

shows that lack of timely response to events as they unfolded and lack of disciplined

management strategies led to many unnecessary deaths. One of our goals here is to show

that careful and systematic approaches to the myriad operational problems associated

with emergency response can lead to policies and procedures that maximize the

effectiveness of the resources available. As a result, lives should be saved, and lifelong

debilitating injuries should be minimized.

1. First Responders: Police, Fire and Emergency Medical Services

Perhaps most relevant to emergency response planning is the 40 years of O.R. work

focused on urban and municipal first responders, i.e., police, fire and emergency medical.

This work started with the Science and Technology Task Force of the President’s

Commission on Law Enforcement and Administration of Justice in 1966 [6]. It led

directly to the national implementation of the three digit emergency number, ‘911,’ and it

sparked a generation of important O.R. emergency services research. When New York

City implemented its 911 system in 1970, managers there discovered how useful and

necessary queueing theory is in the scheduling of 911 call takers. Their original

scheduling of personnel, without the benefit of O.R. analysis, yielded intolerable 30+

minute telephone queue delays on weekend evenings, with the caller hearing only the

familiar and annoying ‘electronic ringing sound.’ An O.R. analysis quickly showed how

rescheduling current personnel – without additions -- brought the delays to within

acceptable limits. The recommendations were fully implemented within one month of

the study’s completion [21].

Queues occur when the available resources are not adequate to handle real time demand

for those resources. Queueing is a type of rationing of resources. Sometimes the

DRAFT

rationing and delays are deliberate, as with some private sector call-in complaint centers.

In a major emergency, queues are endemic and must be managed aggressively by using

techniques such as prioritization and triaging. Triage classifies those who are injured into

various priority categories, and acts with urgency on the highest priority categories first,

trying to save as many lives as possible with the limited resources at hand. Without

triage, queues would grow without bound, and few would be treated in a timely manner.

Sometimes triaging requires very difficult deferral decisions, such as occurred on

December 7, 1941 at Pearl Harbor: triage nurses deciding against medical treatment

other than morphine for those who had been so severely injured that near-term death is a

certainty regardless of medical intervention. Such difficult decisions may be required to

save scarce medical expertise for treatment of those whose lives can be saved. Modeling

work on ‘cut-off priority queues” provides a methodology of setting priorities and

predicting system performance under alternative triaging schemes [35, 36]. In [33] we

show how this would work with data from the Hartford Connecticut Police Department.

The author of this chapter was a member of the Science and Technology Task Force of

the President’s Commission, and as a result of that work, wrote his Ph.D. thesis on urban

police patrol allocations. This culminated in a book, Urban Police Patrol Analysis, MIT

Press, 1972 [22]. This book offered a variety of O.R. models to examine police response

times, patrolling patterns, impact of new technologies (such as automatic vehicle

locations systems), personnel scheduling and more. This effort led to a major four-year

NSF-funded research program at the Massachusetts Institute of Technology, the “IRP

Project,” Innovative Resource Planning in Urban Public Safety Systems. That project led

to many graduate theses and computer-implemented models related to police and

emergency medical operations. It started ‘public safety’ O.R. careers of at least five

doctoral students [27].

The key model from the IRP project was the “Hypercube Queueing Model.” This

mathematical model depicts the detailed spatial operation of urban police departments

and emergency medical services [23, 25]. The model is an equation-based model that

uses various analysis tools from the technical field known as ‘stochastic processes.’ It is

a multi-server queueing model that reflects the unscheduled nature of 911 calls by

modeling them as a Poisson process. The service times of different servers (i.e., police

patrol cars or ambulances) are uncertain (i.e., probabilistic) and have different average

values, reflecting differing workloads and travel times in their areas of responsibility.

The Hypercube model has found application is police beat design, dispatcher car-picking

strategies, allocation of patrolling time, evaluating the response time reduction value of

automatic vehicle location systems, and more [28]. By combining the spatial and

temporal aspects of police and ambulance operations within one unified probabilistic

framework, the Hypercube model was the first to be able to predict the operational

consequences of alternative police deployments over space and time. It illustrated the

inadequacies of some deeply rooted beliefs that had become folklore in police

deployment [24]. In contrast to popular wisdom at the time, it predicted a large number

of inter-beat dispatches of police cars, a forecast later verified in the field [20]. It showed

that a police car may have an above-average workload even if its ‘own beat’ had virtually

no internally generated workload. The Hypercube Queueing model has been

DRAFT

implemented in many cites, including New York City [30]; Boston, Massachusetts [7];

Hartford, Connecticut; Orlando, Florida; Dallas, Texas and Cambridge, Massachusetts.

The Orlando Police Department, for instance, in 1992 essentially redesigned the police

beat layout of the entire city, in order to create a new central city precinct [34]. Without

the model, the city’s police planners would have had no scientific basis for making such a

dramatic change in deployments. But with the model, they could be confident that the

new allocation scheme would satisfy all performance standards set by the department

[26].

Various vendors have commercialized the Hypercube model, and its full impact is

impossible to determine since not all implementations have been documented in the open

literature. From the perspective of Homeland Security, the analytical structure of the

Hypercube model offers promise in guiding response resources depleted in the event of a

major emergency. But the model needs to be generalized in order to include the impact

of second and third tier responders, from regional, state and federal agencies, and of

specialized responders such as HAZMAT and bio-terrorism units. It also needs a time-

dependent solution structure, fed with (potentially massive) data from the field. The

author, with research colleagues at Structured Decisions Corporation of West Newton,

Massachusetts, is building from the Hypercube model a new deployment model for

response to terrorist attacks and other large emergencies. The desired result is a model

that will guide event managers in the dispatch and routing of heterogeneous responders

on a regional interagency basis. The effort will also help local emergency planners to

identify and correct weaknesses in their response plans for major emergencies. This

effort is part of the CREATE project at the University of Southern California, funded by

the U. S. Department of Homeland Security.

In 1969 New York City commissioned the RAND Corporation of California to open the

New York City Rand Institute (NYCRI). The NYCRI assembled a team of analysts to

examine a wide variety of operational problems of the City [13, 17, 18]. Over the years,

their award-winning O.R. work on emergency services has stood the test of time. Some

of this work is still the best available today. An example is the NYCRI’s fire department

relocation model [19, 41]. When there is one large fire or a collection of smaller fires in

geographic proximity, the fire-fighting resources near the vicinity of the fire (or fires)

become depleted. Most fire departments try to re-balance protection by moving some of

the still-available fire companies from more distant firehouses to occupy temporarily

some of the firehouses left vacant by the busy companies. But this in turn creates new

relative vacancies at the more distant firehouses, which in turn require reassignment of

even more distant fire fighters into the newly vacated firehouses. This wave-like

cascading process, if not carefully managed, can create conditions in the city in which

certain neighborhoods are left uncovered, should a new fire occur there. The number of

ways to implement relocations is literally in the hundreds of billions, and no human can

contemplate the consequences of each option and pick the best. But computer-

implemented mathematical models such as the NYCRI relocation model are perfect for

the job at hand. The NYCRI was shut down in 1975 due to New York City’s severe

budget crisis. Remarkably, 30 years later, the NYCRI fire relocation model lives on in

the NYFD. In fact, it proved invaluable on September 11, 2001, in managing the

DRAFT

relocations of NYC fire fighters on that infamous day. With the help of that model plus

implementation of a “Fallback 3” response strategy (meaning far less than usual number

of units initially dispatched to an incident), the NYFD managed to keep its average

response times to other more routine fire incidents to an average of 5.5 minutes, only

about one minute above the usual average

The NYCRI relocation methodology is most relevant in planning response to a terrorist

attack. The New York City 9/11 case is an ‘existence proof.” Any other terrorist attack is

also likely to overwhelm nearby first responders, thereby putting the entire city or region

at risk, if resources are not managed carefully. According to Dr. Peter Kolesar of

Columbia University, co-inventor of the NYCRI relocation model, in the event of

terrorist attack,

"Several core principles underlying the NYFD version would probably be appropriate.

First, solve the problem as it occurs rather than trying to plan in advance since you

probably cannot anticipate the dimensions of the attack and following crisis. Second, use

some politically acceptable mathematical measure to define when coverage is inadequate

and to evaluate alternative relocation options. Third, employ a computer driven

optimization algorithm to generate actual solutions. Fourth, allow the actual decision

makers to modify or override the algorithm's suggestions.”

University students are becoming increasingly interested in research on emergency

response. For example MIT doctoral student Michael Metzger has recently completed a

Masters thesis on deployment of rescue and recovery resources in response to an

earthquake [31]. He used data from the world’s most earthquake-prone country, Iran, to

illustrate the results of his O.R. analysis. Among other measures, his model predicts the

number of hospital admissions, and fatalities over the hours and days following the

earthquake. The results depend on the response strategy selected. Particularly important

are the strategies to select when more than one populated community is affected by

earthquake damage. The results are important, controversial and counter-intuitive. For

example, he shows that to save the maximum number of lives in all affected

communities, it may be necessary to dispatch ‘local’ responders from one of the affected

communities to travel to a more distant damaged community, in order to save lives there.

The result might be the saving of 100 or more lives in the more distant city ‘at a cost’ of

not saving some fewer number of lives in the ‘home city.’ Such sharing of resources

may be politically difficult and would require considerable citizen education prior to any

event in which it is implemented. Metzger’s methodology, which will appear soon in a

published article, demonstrates how one can model the temporal allocation of resources

following a catastrophic event, thereby finding ways to deploy personnel to save lives

and reduce occurrence of debilitating injuries.

We have given a very brief sampling of the many O.R. contributions to analysis of first

responders. For those who want additional detail, there are review papers available [13,

17, 18, 27], as well as summary books [29, 41]. And, there is prize-winning work in the

McKinsey Report, Increasing FDNY’s Preparedness.

http://www.ci.nyc.ny.us/html/fdny/html/mck_report/index.shtml

Peter Kolesar, private communications, August 10, 17, 2004.

DRAFT

scheduling of police personnel [37]. We feel that the entire 40-year body of emergency

response O.R. work will prove invaluable in building the next generation of emergency

response models and methods, ones that are applicable to response to terrorist attacks and

to other major emergencies.

2. Hazardous Materials

The transportation of hazardous materials on trains, trucks and vessels exposes the public

to risks of environmental catastrophes, even in the absence of terrorist threats. The

possibility of a terrorist attack on hazardous materials in transit only increases the risk.

As one example, currently there is much debate about using deep caves at Yucca

Mountain in the Nevada desert for long-term storage of radioactive waste from nuclear

power plants. Should that or another location be selected and operations started, there

would be a massive transportation effort throughout the United States, hauling spent fuel

rods and other radioactive wastes to the selected location. Each city, town, village, or

farm that is passed by the train or other conveyance carrying the hazardous materials is at

risk of an accident and severe contamination.

Because of these threats, there have been O.R. studies focusing on the routing and

scheduling of hazardous materials, point to point on a transportation network such as the

national railway system, in ways that mitigate the risk and/or spread it equitably. The

work has shown that there are tradeoffs between efficiency and equity [2, 3, 5, 10, 14].

The lowest total system risk routes trains (or other conveyances) along the same path

each time. A more equitable policy employs various routes, with more people sharing the

risk, at a modest increase in total risk exposure. For the nuclear waste problem, the

selected routes are of course yet to be decided, but analyses point to the ways in which

efficiency and equity can be addressed in an integrated fashion.

Risk reducing routing of hazardous materials is an example of pre-event O.R. analysis.

By making improved decisions based on such analysis before a major emergency event

occurs, one can reduce the damage caused -- should the event occur -- and on occasion

even reduce the chance of the event ever happening. Of course, the ‘nonevent’ is the best

one for saving lives and property!

3. Bio-Terrorism

Carefully planned detection of and response to any bio-terrorism attack is crucial in terms

of saving lives. This new area of concern has only recently been the focus of O.R.

analyses. But the work has been widely reported and has had major national impact. The

developed models provide a consistent framework for considering operations following a

bio-attack. The work has changed our national policies with regard to immunizations and

medications following a bio-terrorist attack.

With regard to a possible anthrax attack, the co-authors Lawrence Wein and Edward

Kaplan state,

Two pounds of weapons-grade anthrax dropped on a large

American city could result in more than 100,000 deaths,

even if early cases were successfully diagnosed,

DRAFT

antibiotics were distributed broadly and drug adherence was

high. The reason for the catastrophic death toll: Not

enough people would receive antibiotics quickly enough to

prevent symptoms from developing, and those who developed

symptoms would overwhelm the medical facilities.

Any plan to cope with this scenario must include (1)

immediate intervention, (2) rapid distribution of

antibiotics to everyone in the affected region, (3)

aggressive education to ensure adherence to the full course

of treatment and (4) creation of "surge capacity" to treat

the sudden influx of patients. [43]

Their conclusions, together with their colleague David Craft, were based on a highly

sophisticated set of mathematical models that included an airborne anthrax dispersion

model, an age-dependent dose-response model, a disease progression model, and a set of

spatially distributed two-stage queueing systems consisting of antibiotic distribution and

hospital care [42]. One of their most controversial recommendations is to have non-

professionals disperse antibiotics very soon after an attack and/or have those antibiotics

in the hands of citizens at all times – pre-positioned at the points of need in case of such

an attack [15]. Based on these recommendations, the US Postal Service has announced

that its mail carriers will help to distribute antibiotics if a large attack occurs in the

Washington D.C. area

The same three co-authors also used O.R. methods to study response to smallpox attack

[16]. The initial federal policy had been to isolate the symptomatic victims, trace and

vaccinate their contacts, quarantine others, and hope that the spread of disease could be

limited by these measures. The O.R. analysis, again based on a highly complex but

compelling set of models, indicated that the initially selected policy would result in many

deaths. Instead, the analysis suggested a different response: as soon as the attack is

recognized, undertake mass vaccination across the entire population. This

recommendation caused quite a stir nationally, in the press, among physicians and with

policy makers, but now has been adopted as official US policy.

O.R. is playing major roles in other aspects of medical response to major emergencies as

well. For instance, Linda Green has shown how usual efficiency measures defined in

terms of bed occupancy in hospitals cause large queueing delays for beds even in the

presence of routine demand; demands caused by major emergency events would

overwhelm such hospitals [11, 12]. Bravata et. al. extend the policy conclusions of the

anthrax and smallpox work described above to examine regionalized or local stockpiling

of drugs and response to bio-terrorism events [8, 44].

One can see here the need for additional O.R. research on optimal locations of drug and

equipment stockpiling. Traditional location theory seeks global optimal solutions that

minimize some measure of total system travel time or distance [29, Chapter 6]. Usually

one or a small number of carefully positioned facilities accomplish this travel time

United States Postal Service. U.S. Postal Service may deliver medicine in the event of a catastrophic

incident. News release no. 04-015, February 18, 2004.

DRAFT

minimization goal. Within an environment of a major emergency, the traditional

formulation of the facility location problem may be highly inappropriate. Instead, one

has to consider that one or more of the stockpiled facilities may be destroyed by the

emergency event and/or travel paths leading from them may be damaged or inaccessible.

In such cases, one may want to position more than the usual number of facilities, each

containing fewer medications and supplies, in order to increase the probability of

survivability of the drug and supply distribution system. This version of the problem is

somewhat similar to the so-called the p-dispersion location problem, where p is the

number of facilities being dispersed. These issues are addressed in new papers by Gong

et. al. [9] and Berman et. al. [4].

Should a major bio-terrorism event occur at one identified location or limited region,

getting timely appropriate medical care to those exposed is critical for their survival. One

can imagine scenarios in which victims are first triaged, those identified as needing

immediate transport are taken to nearby hospitals or other medical facilities, initial

treatments are administered, and then many patients at the nearby hospital are moved out

to more distant locations. For if such outward movements are not done, the nearby

hospitals become queueing choke points in the system, with their own limited resources

totally overwhelmed. The cascading wave-like movement of patients out of nearby

facilities to more distant ones reminds one of the reverse of NYCRI’s fire relocation

model. Creating such hospital “surge capacity” (in the words of Kaplan and Wein)

certainly warrants further research.

Effective responses of healthcare systems are essential to the total societal response to

major events, be they terrorist attacks, acts of nature or man-caused accidents. The

number of components of these systems can be large, the relevant factors many, and their

interactions complex. Mathematical models are essential in order to understand all of the

complexities and tradeoffs, ultimately leading to more informed decisions and allocations

of scarce societal resources.

4. Private Sector Response

Emergency response is not limited to public sector agencies. In the event of a major

emergency such as a terrorist attack, it is important that private firms whose operations

have been interrupted by the emergency resume normal operation as soon as possible.

Operations Research can play a role in that normalization process.

There are few companies whose operations are more complex than airlines. With

thousands of flights scheduled each day, the efficient matching of planes and crews to

schedules and airports is an intricate, carefully choreographed optimization problem.

When unplanned events occur, myriad decisions must be made. “Usual events” in the

airline industry are “Chicago O’Hare closed due to snow” or “Miami closed due to a

hurricane.” These Mother-Nature-caused events are difficult enough to handle, as

hundreds of flights and thousands of passengers may be affected. But imagine what

happens when all planes are unexpectedly grounded. And, on September 11, 2001 all

civilian airlines were grounded. Planes that had been in the air at the time of the 9/11

emergencies were directed to nearby airports for landing. At the end of the day, the

DRAFT

airlines and their passengers found themselves literally all over the country and even

outside of the country, often at locations far from the intended destinations. The state of

each airline was very far from what had been carefully planned. Yet, as described in

award-winning work [38, 39, 40], O.R. optimization resulted in Continental Airlines

having the "best" recovery of any major airline in terms of percentage of

delays/cancellations during the restart phase that followed the nationwide grounding of

commercial aircraft. The computer-implemented O.R. methodology determined the

least-cost sequence of decisions to get the airline up and flying again, consistent with the

thousands of constraints dealing with matching crews to planes to which they are

qualified, getting each plane back on schedule, adhering to maintenance schedules,

obeying FAA rules with regard to maximum allowable flying times of crews, etc. Since

that time, many other airlines have adopted this proprietary O.R. methodology to assure

their swift recovery from major events to resuming regular operations.

The events of 9/11 affected many industries in addition to airlines. For instance the

mantra of ‘just in time’ supply chain management was tossed out the window on 9/11 due

to a lack of redundancy and slack in ‘just-in-time” systems. Hundreds of trucks were

lined up on the Canadian border on September 12, 2001, awaiting customs and

immigration clearance into the USA. As a result of these delays, factories in the US

began experiencing parts and supplies shortages. This has led to a new type of supply

chain analysis, one that requires robustness in case of emergencies, one that trades of

just-in-time efficiencies with redundancies needed to maintain normal operations in case

of interruption from a major event. This is another area for the analytical, model-based

approaches.

5. Implementation

Many of the models and methods discussed herein are being used on a daily basis by first

responder emergency services throughout the United States. As discussed, the NYCRI

‘fire department relocation model’ was used extensively and successfully by the New

York Fire Department on that fateful day, September 11, 2001. Private firms have

implemented and often extended the methodologies discussed here into real-time

command and control systems, such as computer-aided dispatch systems and regional

emergency management systems. The end user in such circumstances probably does not

even know that “O.R. is inside” the computer programs she is using. This is as it should

be, just as the user of an Internet search engine such as Google does not care about the

mathematical or logical details of Google’s search engine, only in the usefulness of the

results. The final proof of the value of O.R. is in the quality of decisions made from those

who benefit from its use.

With computer computation and storage being exceedingly inexpensive these days,

relative to the past, we are seeing more and more databases being assembled that will

assist the O.R. planner in preparation for emergency response. One of these is New York

City’s Citywide Assets and Logistics Management System

(CALMS). CALMS, set up

for disaster response, is the only New York City system that cuts across jurisdictional

http://www.nyc.gov/html/oem/html/response/calms.html

DRAFT

lines and retains knowledge of the whereabouts of supplies, equipment and personnel

from many different agencies. It is organized according to six asset types: fleet,

equipment and supplies, facilities, contracts, personnel and donated goods. CALMS

automatically gets periodic uploads, to refresh its databases. And, its spatially oriented

data can be displayed on maps of the appropriate parts of the City via a Geographic

Information System (GIS) mapping tool. Eventually we see systems such as CALMS

instilled with intelligent O.R.-based models and algorithms that would recommend the

best movements of men, women and materiel in response to an emergency event. A

similar inclusion of O.R. may be expected in now widely-implemented Emergency

Incident Management Systems, computer-based systems to coordinate the management of

resources in response to an emergency and one in which there is now an effort to

standardize nationally

The need for O.R. talent in the domain of Homeland Security is apparent. The U.S.

Department of Homeland Security has job openings for professionals with O.R. training.

The City of New York has hired such people. Even the taxicab system of New York City

has sought such professionals, as needed talent to interact with consultants who are

examining the consequences of using GPS vehicle positioning technology on New York

City taxicabs. O.R. professionals often have undergraduate degrees in electrical

engineering, computer science or mechanical engineering, so they can integrate technical

engineering knowledge into the systems framework of Operations Research.

A growing new professional career is that of emergency services manager. Emergency

services managers may also be called emergency program managers or directors,

operations center chiefs and risk management experts. They are professionals who

coordinate equipment, emergency workers and volunteers who move into action

following any disaster. They make sure that government, volunteer and medical

personnel work together cohesively and effectively during an emergency. Knowledge of

operations research is increasingly a job requirement.

Universities such as Virginia’s

University of Richmond offer certificates and degrees in Emergency Services

Management.

The public, in weather forecasts, has been exposed for many years to the types of

probabilistic analyses we have discussed. We are all accustomed to hearing that, “…the

probability of rain tomorrow is 0.4.” Quantified uncertainty has become a part of our

daily lives. More critical to Homeland Security, the public is also aware of probabilities

associated with hurricanes, particularly where and when they will achieve landfall.

Public preparedness for and response to hurricanes are important components of

Homeland Security, as a Category 3, 4 or 5 hurricane making landfall of the United

States certainly conforms to our definition of major emergency. The types of analyses

we have been discussing are used today to create the ‘probability risk profile’ of

approaching hurricanes. They are also used to make decisions on evacuations. Here

additional O.R. analyses may be valuable on the detailed planning of evacuations, to

minimize false alarms and the public apathy that may result and also to minimize the

See http://www.dhs.gov/dhspublic/display?content=3259.

http://www3.ccps.virginia.edu/career_prospects/briefs/E-J/EmergencyManage.html

DRAFT

chance that evacuees could end up in congested roadways subject to local flooding.

There is still much to do.

6. Summary and Conclusions

Operations Research, the science and technology of decision aiding, helped immensely in

WWII. Today we face a different set of threats, a new type of warfare labeled

asymmetrical. This type of threat creates the possibility for large-scale devastation

similar to that caused by Mother Nature and by man-made accidents. Planning

appropriate societal response to such large-scale emergencies, should they occur, can

save many lives and reduce extent of injuries. Hardware technology alone, without

careful systems planning is not enough. And there is not enough money in the public

coffers to think that simply ‘throwing money at the problem’ will solve it. Operations

research offers a scientifically valid, integrated framework for considering all aspects of

the problem and for assessing the consequences and tradeoffs associated with alternative

decisions. We expect to see many more new results from Operations Research in the

years ahead, as the nation comes to grip with the new threats from terrorists and the old

threats from Mother Nature and industrial accidents.

Acknowledgement. This research was supported by the United States Department of Homeland Security

through the Center for Risk and Economic Analysis of Terrorism Events (CREATE). The work was

conducted through a subcontract from CREATE at the University of Southern California to Structured

Decisions Corporation, West Newton, Massachusetts. For helpful comments on an earlier draft, the author

thanks Andrea Allocca (New York Fire Department [NYFD]), Linda Green, Peter Kolesar, Lawrence

Wein, Gang Yu and the editor of this book, David Kamien.

References

1. Barry, John M., The Great Influenza: The Epic Story of the 1918 Pandemic. Viking Penguin, NY, 2004.

2. Batta, Rajan and Samuel S. Chiu, “Optimal Obnoxious Paths on a Network: Transportation of

Hazardous Materials,” Operations Research, 36 (1), 1988, 84-92.

3. Berman, Oded, Z. Drezner and G. Wesolowsky, “A Routing and Location on a Network with

Hazardous Threats,” Journal of the Operational Research Society, 51, 2000, pp. 1093-1099.

4. Berman, Oded and A. Gavious, “Location of Bioterror Response Facilities: A Game Between State and

Terrorist,” working paper, Rotman School of Management, University of Toronto, 2004

5. Berman, Oded, V. Verter, Y. Bahar, “Designing Emergency Response Networks for Hazardous

Materials Transportation,” working paper, Rotman School of Management, University of Toronto, 2004

6. Blumstein, Alfred, et. al. Task Force Report: Science and Technology, A Report to

The President's Commission on Law Enforcement and Administration of Justice, Prepared by The Institute

for Defense Analyses, U.S. Government Printing Office, Washington, D.C,. 1967.

7. Brandeau, M. and R. C. Larson. 1986. "Extending and Applying the Hypercube Queueing Model to

Deploy Ambulances in Boston." A chapter in Delivery of Urban Services, A. Swersey and E. Ignall, (eds.).

New York: North Holland.

DRAFT

8. Bravata, D.M., McDonald, K., Owens, D.K., Wilhelm, E., Brandeau, M.L., Zaric, G.S., Holty, J-E.C.,

Liu, H., and V. Sundaram, Regionalization of Bioterrorism Preparedness and Response (Evidence

Report/Technology Assessment). Agency for Healthcare Research and Quality, Rockville, MD, 2003.

9. Gong, Qiang, Arun Jotshi and Rajan Batta, “Dispatching/Routing of Emergency Vehicles in a Disaster

Environment using Data Fusion Concepts, proceedings for the International Data Fusion Conference,

Stockholm, June 2004.

10. Gopalan, Ram, Krishna S. Kolluri, Rajan Batta and Mark H. Karwin, “Modeling Equity of Risk in the

Transportation of Hazardous Materials,” Operations Research, 38 (6), 1990, 961-973.

11. Green, L.V., How Many Hospital Beds? Inquiry, Winter 2002/2003, 400-412.

12. Green, L.V. Capacity Planning in Hospitals, in Handbook of Operations Research/Management

Science Applications in Health Care, Kluwer Academic Publishers, 2004.

13. Green, L.V. and P.J. Kolesar, Applying Management Science to emergency response systems: lessons

from the past, Management Science, August 2004, 50 (8) 1001-1014

14. Jin, Honghua, Rajan Batta and Mark H. Karwin, “On the Analysis of Two Models for Transporting

Hazardous Materials,” Operations Research, 44 (5), 1996, 710-723.

15. Kaplan, Edward H., William J. Bicknell, and Lawrence M. Wein, The Citizen as First Responder,

August 13, 2003. Fox News

http://www.foxnews.com/story/0,2933,94665,00.html.

16. Kaplan, Edward H., David L. Craft and Lawrence M. Wein, “Emergency Response to a Smallpox

Attack: the Case for Mass Vaccination,” PNAS, August 6, 2002, 99:16, pp. 10935-10940.

17. Kolesar, P. 1982. "Ten Years of Research on the Logistics of Urban Emergency Services" in

Operational Research '81, Brans, J. P. (ed.), North Holland, N.Y.

18. Kolesar, P. and A. Swersey. 1985. “The Deployment of Urban Emergency Units: A Survey,” in The

TIMS Studies in Management Science, Volume 22.

19. Kolesar, P. and W.E. Walker, “An Algorithm for the Dynamic Relocation of Fire Companies,”

Operations Research, Vol. 22, No. 2 (Mar. – Apr., 1974), 249-274.

20. Larson, Richard C. 1971a. "Measuring the Response Patterns of New York City Police Patrol Cars."

New York City Rand Institute R-673-NYC/HUD.

21. Larson, Richard C. 1972a. "Improving the Effectiveness of New York City's 911." A chapter in

Analysis of Public Systems, A. W. Drake, R. L. Keeney, and P. M. Morse (eds.). Cambridge, MA: MIT

Press.

22. Larson, Richard C. 1972b. Urban Police Patrol Analysis. Cambridge, MA: MIT Press.

23. Larson, Richard C. 1974a. "A Hypercube Queueing Modeling for Facility Location and Redistricting

in Urban Emergency Services." Journal of Computers and Operations Research 1(1): 67-95.

24. Larson, Richard C. 1974b. "Illustrative Police Sector Redesign in District 4 in Boston." Urban Analysis

2(1): 51-91.

25. Larson, Richard C.. 1975. "Approximating the Performance of Urban Emergency Service Systems."

Operations Research 23(5) pp. 845-868.

DRAFT

26. Larson, Richard C. 2001. “Hypercube Queueing Model,” in Encyclopedia of Operations Research

and Management Science, Saul I. Gass and Carl M. Harris, eds., Kluwer, Boston, Mass., pp. 373-377.

27. Larson, Richard C. 2002 “Public Sector Operations Research: A Personal Perspective,” Operations

Research (50) 1 pp. 135-145.

28. Larson, Richard C. and E. Franck 1978. "Evaluating Dispatching Consequences of Automatic Vehicle

Location in Emergency Services." Journal of Computers and Operations Research 5 pp.11-30.

29. Larson, Richard C. and A. R. Odoni 1981. Urban Operations Research. Englewood Cliffs, NJ:

Prentice-Hall.

30. Larson, Richard C. and T. Rich. 1987. "Travel Time Analysis of New York City Police Patrol Cars."

Interfaces 17(2) pp. 15-20.

31. Metzger, Michael. Formulating Earthquake Response Models in Iran, Masters thesis in Electrical

Engineering and Computer Science, Massachusetts Institute of Technology. May 2004.

32. Morse, Philip M. and George E. Kimball, Methods of Operations Research, Technology Press,

Cambridge, Mass. 1951. Republished, Dover Publications, Saul I. Gass, ed., 2003.

33. Sacks, Stephen R., Richard C. Larson and Christian Schaack. 1993. "Minimizing the Cost of Dispatch

Delays by Holding Patrol Cars in Reserve," Journal of Quantitative Criminology, 9 (2) pp. 203-224.

34. Sacks, Stephen R., Grief, Shirley, "Orlando Magic: Efficient Design of Police Patrol Districts," OR/MS

Today, 21 (1) February 1994.

35. Schaack, C., R. C. Larson. 1986. "An N Server Cutoff Priority Queue." Operations Research 34(2) pp.

257-266.

36. Schaack, C., R. C. Larson. 1989. "An N Server Cutoff Priority Queue Where Arriving Customers

Request a Random Number of Servers." Management Science 35(5) pp. 614-634.

37. Taylor, P.E. & Huxley, S.J. 1989. “A break from tradition for San Francisco police: patrol officer

scheduling using an optimization-based decision support system.” Interfaces, 19 (1), 4-24.

38. Thengvall, B., J. Bard, and G. Yu. “Multiple fleet Aircraft Schedule Recovery Following Hub

Closures,” Transportation Research, Part A 35 (2001) 289-308.

39. Thengvall, B., J.F. Bard, and G. Yu. “Balancing User Preferences for Aircraft Recovery during Airline

Irregular Operations,” IIE Transactions on Operations Engineering, 32, 181-193, 2000.

40. Yu, G., M. Arguello, M. Song, S. McCowan, and A. White, “A New Era for Crew Recovery at

Continental Airlines,” Interfaces, Vol. 33, No. 1, 5-22, 2003.

41. W. Walker, J. Chaiken, E. Ignall (eds.). 1979. Fire Department Deployment Analysis, North Holland

Press, N.Y.

42. Wein, Lawrence M., David L. Craft and Edward H. Kaplan, “Emergency Response to an Anthrax

Attack,” PNAS, April 1, 2003, 100:7, pp. 4347-4351.

43. Wein, Lawrence M., and Edward H. Kaplan, Unready for Anthrax, Washington Post, Monday, July 28,

2003, p. A21.

DRAFT

44. Zaric, G.S., Brandeau, M.L., Bravata, D.M., Holty, J-E.C., Wilhelm, E.R., McDonald, K.M., and D.K.

Owens, “Modeling the Logistics of Response to Anthrax Bioterrorism”, Working Paper, Stanford

University, 2004.

DRAFT