Evacuation Planning and Engineering for Hurricane Katrina

The evacuation of New Orleans had some unprecedented successes . . . and glaring failures.

The hurricane season of 2005 will go down in the record books. It included 14 hurricanes (a new record), three of which were among the most powerful and costly in the 154-year history of record keeping in the Atlantic Basin. In addition, there were 27 named systems—the latest, Tropical Storm Zeta, continued to move through the Atlantic even as this article was being written in January 2006. Some meteorologists suggest that this pattern may last for at least another decade.

Although little can be done to alter the weather, we can prepare for the eventuality of hurricanes and other natural and man-made hazards. For decades, engineers and scientists have been developing techniques, strategies, and materials to help the built environment withstand the effects of hurricanes. In addition, building and zoning codes have been changed to keep critical infrastructure away from hazardous areas to minimize the risks of flood and wind damage. The only way to protect people, however, is to evacuate them when threats arise, but this is often easier said than done.

At the fundamental level, the concept of evacuation is simple—move people away from danger. In reality, evacuations, particularly evacuations on a mass scale, are complex undertakings. As the nation clearly saw during Hurricanes Katrina and Rita, it is not always possible to evacuate everyone who is in danger. The most obvious problem is the sheer scope of the event. Hurricane evacuations may involve millions of people over hundreds of thousands of square miles. In addition, because evacuations are inconvenient and disruptive, evacuees often delay travel decisions until the threat appears imminent, thus compressing the enormous travel demand into shorter time periods.
______________________________________
Transportation infrastructure
is not designed to
accommodate evacuation-
level demand.
_______________________________________

One complicating factor is that transportation infrastructure is neither planned nor designed to accommodate evacuation-level demand; building enough capacity to move the population of an entire city in a matter of hours is simply not economically, environmentally, or socially feasible. Roadways are not even designed to be delay-free under routine peak-period conditions. The effectiveness of an evacuation is also greatly affected by human behavior and socioeconomics. No matter how threatening the conditions, some people refuse or are unable to leave.

Despite these difficulties, the evacuation of New Orleans for Hurricane Katrina was widely viewed as a success; data show that more people were able to leave the city in a shorter time than had been thought possible. There were also apparent failures, however, particularly in the evacuation of low-mobility groups.

This article highlights the development of the evacuation management plan for Hurricane Katrina and summarizes some of the facts, findings, and unresolved issues. The discussion is presented from the perspective of a transportation engineer and centers primarily on the highway-based aspects of the evacuation, including demand, capacity, and issues related to the non-evacuees. This article also presents some lessons learned and how they may be applied to other locations and other threat scenarios and identifies unanswered questions and research needs that should be addressed in the future.

The Katrina Evacuation Plan
The city of New Orleans has long been considered “a disaster waiting to happen.” For those who prepare for, respond to, and study such events, the level of death and destruction wrought by Katrina was not outside the realm of possibility. Although a complete evacuation of the city has been the cornerstone of hurricane preparedness planning for the region, the highway evacuation plan used for Katrina evolved over a period of many years based on valuable lessons learned from prior storms in Louisiana and elsewhere.

Fortunately for New Orleans, officials in Louisiana were able to evaluate and refine their evacuation plan based on two “practice runs.” In 1998, Hurricane Georges appeared to be heading directly for the city, leading to the first major evacuation in some 20 years. From that experience, it was apparent that making conventional use of available routes in the region was not an adequate strategy. As a result, the Louisiana State Police (LSP) developed a plan to implement two short segments of contraflow (LSP, 2000).

Six years later, Hurricane Ivan threatened another direct hit on the city, triggering an implementation of the new plan and the first-ever implementation of evacuation contraflow in Louisiana. Like Georges, Ivan tracked east prior to landfall and largely missed Louisiana. The evacuation that it precipitated, however, revealed numerous deficiencies in the plan that resulted in monumental congestion and delays on several key evacuation routes. After a period of considerable public criticism, the Louisiana Department of Transportation and Development (LA DOTD) and LSP formed a Louisiana Evacuation Task Force with input from consultants in industry and academia to identify where and how the congestion occurred and to develop and test ways to reduce it.

After a review of traffic volume and speed data, traffic videos, media accounts, and interviews of evacuees, the task force identified four primary issues that had hampered the evacuation: (1) over-reliance on westward traffic movement; (2) inefficient loading of the contraflow freeway segment out of New Orleans; (3) extreme congestion resulting from the confluence of multiple regional evacuation routes in Baton Rouge, Hammond, Lafayette, Covington, and Slidell; and (4) the lack of real-time, accurate traffic information.

New Orleans Analyses
In New Orleans, the goal was to get as many people as possible out of the threat zone as quickly as possible by making full use of the available traffic lanes. The movement of traffic out of the city can be pictured as sand moving through an hourglass. The outflow of sand (evacuation demand) is controlled by the neck of the glass (over-water elevated freeway segments). Given the limitations and the infeasibility of constructing additional lanes on outbound bridges in the foreseeable future, efforts were focused on making the most effective use of available capacity by using engineering and operations techniques to open the neck of the hourglass as wide as possible.

The analysis after Ivan showed that the movement of traffic was further hindered by the configuration of the contraflow initiation point. As shown in Figure 1, contraflow was initiated by crossing the left two lanes of the outbound side of I-10 into the inbound lanes. The traffic control strategy at this location resulted in a capacity restriction that effectively regulated flow into the downstream section thereby preventing full use of the contraflow lanes. While traffic upstream of this location was heavily congested, traffic downstream moved much more freely.

Traffic modeling was used to assess loading techniques and identify a method that would allow for optimal use of the contraflow lanes without limiting the capacity of the normal lanes or vice versa. It was hypothesized that direct access to the contraflow lanes, similar to normal entrance ramps to a freeway, would avoid the friction and, therefore, lost capacity of a crossover/decision point for drivers. It was thought that, if decision points were located on the surface-street system, the interstate would not experience the loss of capacity. It was also believed that by using multiple loading points with direct access to the contraflow lanes, the demand would be spread spatially throughout the surface-street system rather than being limited by the capacity of a single loading point (as happened during the Hurricane Ivan evacuation).

Based on these assumptions, simulation was used to evaluate several alternatives. One of the first ideas was to move the first loading point of the contraflow upstream and to add two contraflow loading points downstream. This configuration would prevent the normal outbound and contraflowing traffic streams from ever merging. The additional loading points would provide direct access by using existing exit ramps as entrance ramps to the contraflow. Overall, this configuration was expected to accomplish several objectives. First and foremost, it would spatially spread the loading of the demand. Next, it would not only give evacuees routing choices, but would also locate decision points on the surface-street system instead of the interstate system, where maximizing capacity was a necessity.

This plan was thought to be ideal, not only from the standpoint of efficiency, but also for balancing traffic on the separate sides of the freeway, eliminating a major merge point, removing the decision point from the interstate, and providing access from both Orleans and Jefferson Parishes to both evacuation routes. To spread traffic demand spatially, vehicles would enter freeway lanes via six downstream interchanges, rather than at a single point, as in the earlier plan. The anticipated benefits of this plan were clearly supported by the results of traffic simulation modeling, which showed a potential increase in outbound volume of more than 60 percent, or 30,000 vehicles, over a 12-hour period.

Ultimately, this concept was rejected, however, because of the cost and time required to build the crossover, but the knowledge gained about the benefits of using multiple loading points with direct access was used to develop a plan with three interchanges and a single lane crossover. Simulation modeling showed that with this plan outflow volumes would be only marginally (about 15 percent) lower than for the original plan. The final access points (interchanges) were selected based on input from both LSP and LA DOTD, as well as other factors, such as interchange configurations, construction requirements, emergency access, and population distribution.

Efforts were also made to evaluate alternatives on the east side of the city. Although eastbound routes were not affected by Hurricane Ivan to the same degree as westbound routes, significant congestion was evident on eastbound routes as evacuees moved east to avoid congestion in the city. To remedy the situation, a contraflow alternative was developed and tested to route all westbound traffic entering from Mississippi north back into Mississippi, thus precluding entry into Louisiana. Interstate contraflow had never been implemented before, and the state of Mississippi would have to accommodate hundreds of thousands of evacuees from Louisiana. Fortunately, an agreement was reached between officials in Louisiana and Mississippi just months before Katrina. Realizing the enormous threat posed to New Orleans, the state of Mississippi also permitted Louisiana to reroute traffic back into Mississippi and initiate contraflow on another freeway route.

Baton Rouge Analyses
Because of the level of demand generated by mass evacuations, impacts to the transportation network often occur far away from the source of the demand. In Louisiana, significant increases in volume were clearly apparent in downstream locations, such as Baton Rouge, Lake Charles, and even Shreveport, nearly 300 miles from New Orleans. The focus of regional improvements following Ivan was centered in Baton Rouge where the merge point of two major freeway evacuation routes (I-10 and I-12) had created a bottleneck that restricted flow and caused traffic congestion and delay over many miles. The most radical and interesting proposal to address this problem involved the use of an eight-mile-long segment of freeway contraflow through the heart of Baton Rouge to eliminate the merging of intersecting evacuation flows. As diagrammed in Figure 2, the contraflow segment would keep the I-10 and I-12 traffic streams separate, sending them onto two different freeways and eliminating congestion associated with merging. S
imulation modeling showed that such a plan would eliminate all congestion from the area (where delays as long as 14 hours had previously been reported).

Despite its potential benefits, it was decided that contraflow through Baton Rouge was not a viable alternative. The main objection was that it would have required closure of the sole interstate freeway route through the city, raising concerns about the movement of commercial and hazardous cargo outside of the I-10 corridor, local traffic from the west side of the Mississippi River, loss of access to property or family east of Baton Rouge, and loss of access for emergency and service vehicles.

Interestingly, the plan that was ultimately decided upon and refined by LSP was potentially even more radical. In fact, in terms of its scale and level of strategic control, it was unprecedented. As shown in the schematic diagram in Fig-ure 3, the final plan eliminated confluence traffic by forcing evacuees to use northward evacuation routes and prohibiting cross-state westbound traffic into the Baton Rouge area. The plan featured the reversal of approximately 100 miles of interstate freeway from Louisiana into Mississippi, forcing the northbound movement of traffic instead of the usual congestion-causing westbound movement. It also restricted access to nearly 100 additional miles of freeway to prevent a regional bottleneck in Baton Rouge. Combined with the improved loading plan in New Orleans and the coordination of transportation assets on a regional basis, the plan was expected to enable a rapid, orderly exodus of more people than was ever possible before.

Successes
Soon after the storm, efforts were initiated to assess and evaluate traffic impacts associated with the evacuation, including how, when, where, and how long it took evacuees to move along the road network of the Gulf Coast, how contraflow impacted the efficiency of the evacuation, and, in particular, how some of these impacts were related to the revisions in the plan. Early indications of these analyses have revealed the number of vehicles involved, when evacuations started and ended at key locations, the geographic extent of the evacuation in Louisiana, the directions of travel, the capacity gains made through contraflow, and some of the implications of the modifications to the Louisiana evacuation highway management plan.

Traffic-count data recorded as part of the LA DOTD traffic monitoring network showed elevated volumes at every station throughout the state, even on lightly traveled roads through sparsely populated areas hundreds of miles from the storm landfall location. Increases in volume were evident as early as Friday evening as some travelers moved into the New Orleans area and as late as early Monday morning as traffic was winding down through more distant areas like Monroe and Shreveport. During the peak 48-hour period of evacuation, more than 430,000 outbound vehicles were recorded on the six freeway and major arterial roadways in southeast Louisiana. Although this total does not include several other primary highways, it has been assumed to mean that about 80 to 90 percent of the population evacuated the area.

Figure 4 graphically illustrates the progression of the New Orleans evacuation over time. In the figure, the heavier line represents the average traffic volume between Thursday and Monday in the two westbound lanes of I-10 during the three weeks prior to Katrina. The lighter line shows the volume at the same location during the four-day period preceding the storm landfall. The graph clearly shows that the volume increased significantly in the late morning hours of Saturday, August 27 and that traffic volume decreased for a brief period as contraflow was implemented some 20 miles upstream at around 4:00 p.m. Volumes increased again and remained far above average throughout the night. The flow peaked at more than 2,500 vehicles per hour and remained at that level throughout the day. It dropped dramatically the evening before the storm made landfall.

Temporal analyses supported by Figure 4 also show that the duration of the evacuation required only about half of the 72-hour clearance time that had historically been assumed prior to the 2005 modifications. Given the limited number of roadways, this was a tremendous achievement. The accomplishment was even more significant because of the small number of traffic-related injuries or deaths directly resulting from the evacuation. Preliminary analyses also reveal other indications of the effectiveness of the plan, including better use of low-volume routes and significant gains from contraflow.

Failures
Not all of the evacuation news was positive, however. Images of thousands of desperate people being plucked from rooftops by helicopter, stranded at the New Orleans Convention Center and Superdome, and awaiting rescue on freeways have overshadowed the successes of the highway-based evacuation plan. It has been estimated that between 100,000 and 300,000 people did not or could not be evacuated from the city.

The most serious questions, however, relate to the city’s poor populations. Local governments have been blamed for poor planning and not providing adequate transportation to shelters of last resort. For example, it was widely known that some 112,000 people did not have access to personal vehicles at the time of the storm (Russell, 2005). Given these numbers and the limited capability of moving this enormous number of people quickly, public officials have long advocated “neighbor helping neighbor” policies, urging low-mobility individuals to arrange for transportation with friends, family, neighbors, and church members. Local plans also included using Regional Transit Authority buses to carry people to the Superdome from 12 locations around the city (Russell, 2005).

A major failure of the plans for evacuating the low-mobility population was the lack of communication. Evacuation plans can only be effective if people are aware of them, and evacuation orders can only be heeded if they are received in time.¹ Thus, the of problem evacuating low-mobility populations will be one of the most important issues for all levels of government in future evacuation plans.

Lesson Learned and Future Needs
In Louisiana, three major traffic modifications significantly improved the evacuation. First, a staged evacuation plan identified the order of evacuees, starting with the lowest lying areas, and suggested a time line for the initiation of contraflow. Second, multiple interchanges and a crossover were used for the contraflow loading plan to minimize congestion and maximize the capacity of bridges; this spread the traffic demand to several critical roadway segments. Third, the access management plan made maximum use of major arterial roadways; this helped spread demand to many highways instead of concentrating demand on the freeways. Although these changes may have limited flexibility and caused some inconveniences to evacuees and pass-through traffic, they accomplished their primary goal of saving lives.

Coastal populations are expected to continue to grow while available roadway capacity remains relatively unchanged. As a result, future efforts must also focus on controlling evacuation travel demand. This may be accomplished through better public information and education programs, including educating the public about which areas are truly at risk and working with the news media to provide more accurate descriptions of threat levels.

During Hurricane Floyd, the Florida Division of Emergency Management estimated that about 35 percent of the approximate 2.0 million evacuees on the road in that state did not have to leave (Collins, 2002). This so-called “shadow evacuation” also contributed to the enormous congestion in Houston. Phased evacuations may help, although they may be difficult to implement, and measures can be taken to strengthen building codes and encourage in-place sheltering practices to reduce the need for people to leave.

Emergency management and transportation officials involved in evacuations most often cite a need for better communication and coordination among emergency management, transportation, and law enforcement agencies and the public. Many states are working to combine emergency management personnel into single facilities and to establish coordinated evacuation policies. Evacuation plans can be communicated throughout the year through public information campaigns, public service announcements, and tourist information centers. In addition, the collection and transfer of traffic information during an evacuation is critical to the management of evacuation routes and the allocation of transportation resources.

LA DOTD and LSP are already working to communicate evacuation plans via news media and public service announcements urging people to prepare for future evacuations by (1) developing personal evacuation plans before the need for an evacuation arises and knowing which routes will be available; (2) being prepared for traffic delays; (3) leaving as early as possible; and (4) encouraging people on the Gulf Coast to make northward travel plans and shelter arrangements (e.g., Jackson, Mississippi; Memphis, Tennessee; and Birmingham, Alabama), rather than westward plans (Houston or Dallas) (LA DOTD, 2005a). Overall, people must be encouraged to take a more active role in planning for their own safety.

Conclusion
The experiences of the 2005 hurricane season have shown that overall evacuation readiness in the United States has improved in the past decade, although a great deal remains to be done. We now have evidence of the significant gains that can be made through large-scale, aggressive, proactive regional traffic management. Policy makers and plan developers must understand that evacuations are not business as usual and cannot be considered local events. Evacuations are life-and-death scenarios, and planning decisions must be made with preserving lives in mind. This may mean that convenience, accessibility, route choices, and perhaps even safety, may have to be compromised somewhat to maintain the most efficient traffic flow.

As we move forward and prepare for the next hurricane season, the search for solutions will extend far beyond the realm of engineering. The lessons learned, information exchanged, and new knowledge that will be created in this field will have implications far beyond hurricane evacuations. Temporary reversible lanes, unconventional spatial and temporal loading strategies for freeways, and the use of mass transit under emergency conditions have applications for homeland-security scenarios, major-event traffic management, and traffic congestion during peak commuter periods.

References
Collins, R. 2002. Florida State Perspective, Session 14: Hurricane Partners—Avoiding the Perfect Storm. Presented at Institute of Transportation Engineers Spring Conference and Exhibit, Palm Harbor, Florida, March 2002.
LA DOTD (Louisiana Department of Transportation and Development). 2005a. “Prepare Ahead of Time,” Coalition Urges. Baton Rouge, Louisiana, June 2005. Available online at: http://www.dotd.louisiana.gov/press/pressrelease.asp? nReleas e=489.
LA DOTD. 2005b. Metropolitan New Orleans Contraflow Plan. Available online at: http://www.dotd.state.la.us/maps/Web_ ContraFlow2.jpg.
Louisiana Recovery Authority. 2005. Louisiana Recovery Initiatives. Baton Rouge, Louisiana, November 2005. Available online at: http://www.lra.louisiana.gov/assets/initiatives.pdf.
LSP (Louisiana State Police). 2000. New Orleans Emergency Evacuation Plan. Troop B, Kenner, Louisiana, May 2000.
Russell, G. 2005. Nagin Orders First-Ever Mandatory Evacuation of New Orleans. The Times-Picayune, New Orleans, Louisiana, August 28, 2005. Available online at: http://www.nola.com/newslogs/breakingtp/index.ssf?/mtlogs/ nola_Times-Picayune/archives/2005_08_28.html.

FOOTNOTES

1 “Many said that mandatory evacuation orders came too late, or that leaving, even with transportation, was not a simple matter for older residents. LeShawn Hains could not find a special-needs shelter for her mother, Gilda, who was on oxygen and had heart and lung trouble. Eddie Cherrie Jr. stayed behind with his mother, Onelia, who relied on a walker and blood pressure medication. ‘It’s true nothing stopped us from leaving,’ he said. ‘But also, it’s not that easy to leave with a 91-year-old woman’” (Russell, 2005).