The overarching topic of evacuation practice is well represented throughout the literature. Several related topics areas were reviewed given their relevancy to the issues being addressed as part of the research being presented herein. To fully understand research needs, the existing literature as related to the following topics has been reviewed:
- University Evacuation Plans
- Transportation Management & Weather Emergencies
- General Evacuation Considerations
- Pedestrian Crowd Modeling
- Evacuation Simulation Models
UNIVERSITY EVACUATION PLANS
Existing campus evacuation plans differ in various accounts. Some of the selected plans are for universities in either rural or urban areas and they pertain to universities of varying sizes. Regardless of these differences, these campus evacuation plans relate in their purposes and priorities. The reviewed campus plans include the University of Massachusetts Amherst, (2016) Hampshire College, Mount Holyoke College, Smith College, (2016). Boston University, (2006). University of Nebraska Kearney, (2016) and the University of North Carolina Charlotte (2014). Throughout the university plans, the most communicated priority at each of the campuses is community safety, second to efficiency, and the return of services to the community. Additionally, suggested by the University of Massachusetts Amherst has been minimizing environmental and property damage. For the purpose of this research the University of Massachusetts Amherst’s existing priorities will be followed as they align similarly with universities nation wide.
University of Massachusetts Amherst (2016) Emergency Management Plan Priorities:
- Protect lives,
- Stabilize the incident,
- Minimize environmental and property damage, and
- Continue critical services to customers.
TRANSPORTATION MANAGEMENT & WEATHER EMERGENCIES
To come up with a disaster plan various scenarios are simulated. An example can be an earthquake where we can find out which bridge has fallen down, which roadways are crumbled, where motorists are trapped, etc,, Information should be posted on message boards wherever possible with the coordination between various agencies like law enforcement, transportation, fire and rescue, EMS.
Transportation Management Centers(TMCs) should form an Emergency Preparedness Working Group with agencies that regularly work with the TMC to discuss, develop, and review topics and initiatives related to emergency preparedness.
The working groups who are to be prepared for emergencies are:
- Law enforcement (Federal, state, local), Fire departments, Emergency medical services, Emergency management agencies (Federal, state, local), Towing and recovery providers, Transportation agencies, including non highway agencies such as public transit (large metro systems and local bus services), rail, airport, or maritime;
- Private or not-for-profit organizations such as the Red Cross, towing associations and AAA; and
- Other Federal, state, or local agencies such as Department of Homeland Security and state/local environmental agencies.
According to the FHWA (2015), the group should establish regularly scheduled and published meetings and develop the activities, including the following:
- Conduct a needs assessment and planning activities;
- Perform training and drill exercises;
- Introduce and evaluate new technologies that can benefit both transportation and emergency operations which may offer opportunities to pool funds; tabletop exercises to build trust and good working relationships; and
- Perform formal debriefs or after-action meetings after critical incidents
Planners should identify the operational tasks like what action is required, who is responsible for the action, when should the action take place, how long should the action take and how much time is actually available, what has to happen before and after, what resources does the person/entity need to perform the action.
GENERAL EVACUATION CONSIDERATIONS
The fire departments evacuation protocol on typical college campuses is explored in order to highlight the most significant areas of safety during varying levels of fire emergency. Tracy (2001) suggests that supplying water to the area and access in and out of the building are the two most vital areas of focus. Both of which, rely on the accountability of the institution. A reliable and consistent water supply relies typically on standpipe operations and otherwise resorts to hydrants or supply lines placed by the institution. Furthermore, access must be distinguished by staircase for those exiting and entering buildings in order to minimize the delay of the evacuation. The institution is responsible for providing platform equipment, including ladders, to aid in the evacuation of students. This equipment may also be crucial for access. The paper concludes that proper training is essential for this protocol to be carried out.
An evacuation model, by Daganzo, influenced by previous models Daganzo and Lin(1993) and Daganzo and Lovell (2000) explores the scenarios in which an evacuation route may render ineffective if left unattended for. However, by incorporating simple adjustments the strategy optimizes the number of people evacuating at all times and favors the groups of people most at risk during the evacuation. The relevancy to this project being that the greatest focus is the areas on campus where the students are most at risk of the impending emergency.
The goal of the model is to minimize the time at each nest in order to minimize the time of the entire system. Additionally, the model aims to maximize the evacuees by prioritizing upstream residents, who are more at risk. The model works by a ramp controller, likely a police officer controlling the flow from a ramp onto a freeway by observing incoming flow and allowing out the appropriate number. In order to further optimize this traffic, downstream queues must be prevented from pouring back into the exit. Findings showed that the freeway with limited options gave people incentives to access the upstream ramps if they were being used less. However, since the network has many routing options there is a worry that this strategy would not be used optimally by people because they can change routes so easily. (Daganzo 2010)
EVACUATION SIMULATION MODELS
In order to effectively plan for traffic operation during an evacuation process, a microsimulation model will be the most suitable framework to encapsulate all externalities involved in the process.
Reviews of research show that there have been a significant amount of work done if the field of evacuation simulation and modeling. A study done by Wojtowicz & Wallace, (2010) proposed a methodology that uses Transmodeler, a microsimulation software, in conjunction with tabletop exercise to evaluate traffic operation under several special events namely. Relevant to this project was the case study for the evacuation of a minor league baseball stadium located in New York. The overall goal was to use various traffic control strategies to improve traffic management during and emergency requiring evacuation. The methodology adopted in this paper made use of a tabletop exercise which was adopted with the transmodeler simulation software to identify and review possible high impact incident scenarios for the case study of interest. It started with the detection of an event and ended when traffic flow was restored to the pre-incident condition.
Some characteristics of the case study included:
- Total capacity: 6,600 spectators
- Time: event begins at 7pm (Well after peak hour of traffic)
The run the model, Traffic analysis Zones (TAZ) were developed to model OD matrix flows for the area of interest and also an event OD matrix for the venue, which included parking lots information and an estimate of person/vehicle etc. Findings showed that the use of the microsimulation model was the best course of action to take when evaluating scenarios involving non routine events and evacuation incidents that cannot be handled by routine traffic management. Solutions adopted included the use of barricades and police at intersections to redirect event traffic way from the accident location, the use of a one-way street structed to move vehicles around the incident. Under normal traffic management the simulation illustrated that if the incident occurred 5 minutes after the end of the event, it would take 8–10 min for the queue to extend back to the local street network. This is useful for traffic managers because this is the amount of time police officers have to redirect vehicles away from the incident before it affects the event traffic on the local roads (Wojtowicz, 2010, 115).
MICROSIMULATION BASE MODEL
The base model simulation was created to reflect upon traffic on campus and some relevant surrounding roadways at 9PM. We chose this time because although in the evening, there are still significant occupants on the roadways heading out to local places or getting home. Choosing our volumes was challenging because they vary so much from peak hour traffic but we chose 55% as our percentage (multiplier) of the regular peak hour. This though likely over estimated, allows us to plan for the 'worst case scenario' in the overnight hours in terms of traffic volumes so that if similar evacuation were needed later in the evening, the proposed evacuation plan would still be suitable. Volumes were calculated by summing the peak hour for each intersection's turning movements and multiplied by .55 resulting in our 9PM volumes. In cases where data was not provided we found AADT from MassDOT's Transportation Management System and took 11% to give us relative peak hour data, then proceeded with the 9PM calculation above (2016). We also added the 'emergency volumes' to our current volumes which includes the extra buses and cars. The extra cars include 890 vehicles based on our data from parking services exiting lot 22 with 30% turning left and traveling toward 116 and 70% turning right. Of the left turning vehicles, half go toward Rt 9 for 91 access and half go toward the center via Fearing for 90 access. For purpose of simulation, as to account for lag in some residents evacuating and getting to their cars we have simulated 2000 vph exiting lot 22 for a 26 min period at which point all cars are evacuated from lot 22.
After our assumption that all 890 residents with vehicles would want to leave campus, we also had to account for the remaining 4,942 residents who must be transported safely and efficiently to the Mullins Center. We assumed that 75 passengers would load onto each bus at capacity in such emergency conditions. We also assumed all the buses would be used in the evacuation, if necessary, which is a fleet of 35. Fernandez states that boarding times assuming no platform with a wide door in the back and a narrow door in the front are 1.36 secs/pax and alighting .975 secs/pax (2014). Our calculation of boarding time is ~1.70 mins/bus and ~1.22 mins/bus alighting. There is plenty of curb space both in front of Southwest residential area and the Mullins Center, but in order to strategize to drop students as close to door to door as possible while not sacrificing efficiency we have assumed curb space at each of these origins and destinations as 5 buses. Total travel time for each bus including the 8 mins of actual roadway travel time on the loop, acquired using GoogleMaps, comes to ~11 mins per bus. Considering curb space, that number actuates to 11 mins per 5 circulating buses. To account for delay the buses might run into if boarding and alighting isn't as efficient as possible and with startup loss time, we've chosen to circulate 4 fleets of 5 buses, that is 20 buses in circulation for the duration of the evacuation. For purpose of simulation, we've simulated 100 buses per hour, as 20 buses circulate in an 11 min increment. This leaves us with 5 mins of time per hour to account for slow boarding and alighting, as we anticipate this might happen. If the residents respond slowly to instruction and don't keep up with the timing estimated for the plan, a fleet could be removed, but for purpose of our calculations we have used 4 fleets of 5 buses. 4,942 residents filling each bus to capacity will take exactly 66 bus loads to evacuate. Assuming not every bus can hold exactly 75 occupants, we have rounded up to 70 buses to be certain this evacuates all residents. Since we are simulating 100 buses per hour, that part of the simulation will only occur for 42 mins at which point, evacuation bus traffic will be mitigated.
We simulated, randomly, 890 pedestrians crossing the crosswalk most northbound of the parking lot from Southwest to lot 22. We assumed no pedestrians would be crossing otherwise due to the shelter-in-place transportation system set up. We didn't simulate any additional pedestrians assuming the residents directed to shelter-in-place will do as directed. If they do not follow instruction, there will be so few vehicles on the other areas of campus that they will not impact the model or risk their own safety.
These volumes were used to populate various OD matrices including typical 9PM traffic, passenger vehicles exiting lot 22, circulating buses picking up students at the Southwest loop and dropping them at the Mullins Center bus stop and pedestrians. Besides normal intersection controls that exist on campus, the base model does not exhibit any traffic management.
THE BASE MODEL NETWORK
The final base model network is pictured below. At 9PM we made the assumption that few people would be traveling inside of the campus loop where the academic buildings are located. Our network generally includes the campus loop including Massachusetts Ave, N. Pleasant St, Commonwealth Ave, and Governors Dr. Also, Triangle and Main St. for access toward Rt. 90, N. Hadley Rd and Rt. 116 for access to the North or Rt. 9, Fearing St , Lincoln Ave and Sunset Ave, as well as Amity St.
University evacuation plan network
The existing network is very busy at node 77 which is the lot 22 parking exit but not tremendously busy throughout the whole network which is good for holding our emergency plan to the highest standard of rapid dissipation. To further our discussion, we took a look at what existing resources at the university might assist our plan to evacuate. Some of these resources include 4 portable LED traffic signs used for campus event instruction which are owned by the University, emergency texts and emails, and outdoor warning sirens all suitable for relaying information to the endangered residents and even to the residents and faculty out of harms way to keep them in place and safe. In addition the resident assistants and resident directors, campus transit services, the office of emergency management, and both campus and Town of Amherst police and firefighters will all be viable resources to move the students and faculty while simultaneously keeping them safe.
We will need to evacuate 5,832 students from Southwest according to the provided data on housing in Fall 2016 and the assumption that 500 people may be using the various Southwest dining locations. We looked at the data about shelter locations provided by Robert Laford, the Assistant Director of Emergency Management, which explained that in circumstances of a staging area rather than an actual shelter, there is space for up to 9,000 in Mullins Center. Although there is sufficient space for all 5,832 southwest residents to relocate to Mullins Center we understand many residents may chose to get off campus and stay at friend's or family's homes. Additionally, its important to realize that if the threat is not located or determined safe after a few hours, residents would be relocated to shelter-in-place locations across campus which is an evacuation practice that is not discussed in our research.
To rapidly remove all Southwest residents and relocate them to their perspective location, either sheltering-in-place at the Mullins Center or traveling by vehicle to a safe off campus destination, traffic management is key. Initial assumptions about our mitigation strategies were centered around road closures, simple detouring, critical instruction to faculty and residents, police details, and most importantly bus prioritization. For simplification, we will propose ideas for the existing network and further detour plans extending beyond our network should be put in the hands of Amherst and Hadley law enforcement but supported by the university.
The purpose of our road closures is to keep uninvolved residents of other dormitories, apartments, and locals away from the incident. Reducing conflict points will produce the fastest possible evacuation time and keep the roadway users safe. The proposed alterations and road closures are as follows:
- At the roundabout by the Lederle Graduate Research Center, close N. Pleasant Street southbound and Governor's Drive so vehicles entering the roundabout from N. Pleasant Street in the north or Eastman Lane cannot get further access to the network beyond the roundabout.
- A short segment and boundary node at Butterfield Terrace was added to the network creating a new destination for all the unassociated vehicles in the evacuation who enter the network at node 7.
- Finally the node 3 and 5 segments are extended and Stadium Drive is added to the network. Vehicles entering at nodes 3 and 5 are redirected to Stadium Drive emptying these vehicles at the new node at Rocky Hill Road.
- No vehicles are allowed into Stadium Drive from Rock Hill Road so this node will be a destination but not an origin.
- Detectors are used to prioritize buses at all intersections where they exhibit conflicts but signify police details in actual evacuation implementation.
- The roadway into lot 22 is created into a one-way two lane roadway that has all right-turners in the right lane and all left-turners in the left lane.
The proposed design is exhibited with related signage in the image below which shows the minimizing of conflict points needing only 3 police detail locations to optimize the system while keeping all users safe.
THE IMPORTANCE OF TIME
Time is crucial in terms of emergency evacuation procedures. In the circumstances of a bomb threat in Southwest there are two procedures in which time should be measured. First and foremost is getting people out of the immediate threat being the buildings. There are two manners of building evacuation utilized based on the specific circumstances. Evacuating in phases is more efficient when feasible in circumstances like a fire where the immediate threat location is known and temporarily maintained in place. In the case of a bomb threat, the evacuation method must be simultaneous for all occupants in order to evacuate everyone as quickly as possible since the exact supposed location and size impact of the threat is assumed unknown (Milke, 2007). After 911, many have studied the event giving us viable information about building evacuation times. In terms of Southwest, there are 5 22-story towers and 11 low rise residential halls. Considering that the towers would take the most time, they are the evacuation times that have been analyzed. All buildings will evacuate simultaneously, so the time it takes to evacuate a single tower, is considered to be the maximum building evacuation time for all Southwest residents. Milke makes the below assumptions in his analysis:
- All persons start to evacuate at the same time and hence no pre-movement time is considered (e.g., talking to others, turning off electronic devices, putting on coats).
- Occupant travel is not interrupted to make decisions or communicate with other individuals involved.
- The persons involved are free of any disabilities that would significantly impede their ability to keep up with the movement of the group. This includes any temporary disabilities as a result of fatigue.
- Rescue team coming into the stairway do not impose a significant impact on the flow rate of occupants traveling down the stairs.
- The controlling feature of the flow rate of people from the building is the door at the bottom of the exit stairway. This assumes that people develop a queue in the stairway that ends at the doorway at the base of the stairway. Also, the time for the first people to form the queue is assumed to be much less than the total evacuation time.
- The density of the people traveling through the doorway is in the range of observed values (i.e., 6-10 ft2/person). As such, the flow rate per foot of effective width for each doorway would be anticipated to be in the range of 18 to 24 persons/min.
- Assuming to exit stairways may take up to 15 minutes to evacuate the building (2007).
Estimated evacuation times for high-rise buildings
With 550-600 residents living in the Southwest towers, the analysis is for the worst case scenario of 600 students. From analyzing the graph, we can estimate 15 minutes to evacuate a fully occupied tower. Unlike the study used, campus residents both younger and less mature than typical residents of a tower evacuation may take the situation's validity less serious and take additional time preparing to evacuate. For our purposes, we have added 5 minutes to the evacuation time for such reasons. Such has lead us to an estimated evacuation of 20 minutes. Additionally, we must consider the time for the students to travel from various locations in Southwest to the curbside bus load zones along the Southwest loop.
Travel times from Southwest residential buildings to curbside bus pick up.
In the worst case scenario of traveling from the farthest Southwest point to the curbside pickup we found a 6 minute travel time. Accounting for all aspects, students should take 26 minutes to evacuate the buildings and arrive at the loading areas or begin traveling toward their vehicles.
The two models were compared based on the following measurements of effectiveness for the entire network as well as for the users exiting Lot 22 where most of the queuing was found in the base model:
- Average Delay
- Average Travel Time
The results show an improved network throughout where in the entire network, average delay improves by 1.3 mins and average travel time improves by 19.7 mins. For lot 22 origin users, the average delay improves by .2 mins and the average travel time improves by 28.9 mins.
The entire evacuation time frame from emergency alert to dissipation is demonstrated in the Gantt Chart below. The model considered human factors within reason but actual human behavior can never be fully and accurately predicted. Having a dynamic plan in the first place is a step in the right direction where changes can be made with behavior.
In further research the following circumstances should be analyzed:
- Psychosocial impact of the bomb threat on students and their compliance with instruction; one might consider instruction for individuals who chose to walk or run from the threat so they end up in a safe place and get there safely.
- Consider disabled individuals or individuals with special transportation needs as well as their abilities to exit the building safely.
- Extension of the lot 22 roadway so the queue does not extend off the roadway in the base model. (This will improve the accuracy of measures of effectiveness)
- Consider safe ways to allow students to leave in other areas of campus if they don't comply with instruction.
- Consider a couple of parent pick-up location in the North side of campus and simulate such trips.
Boston University, (2006). Campus-wide Evacuation Plan. http://www.bu.edu/ehs/files/2010/06/BU-Evac-Plan1sm11-1.pdf
Campus Police Department Hampshire, Mount Holyoke, and Smith Colleges, (2016). Hampshire College Evacuation Plan. https://www.mtholyoke.edu/campuspolice/hampshire-college-evacuation-plan
University of Massachusetts Amherst, (2016). Emergency Management Plan. https://www.umass.edu/emergency/campus-plans/emergency-management-plan
University of Nebraska Kearney, (2016). Campus Evacuation Plan. https://www.unk.edu/offices/emergency_management/_files/campus-evacuation-plan.pdf
University of North Carolina at Charlotte, (2015). Campus Evacuation Plan. http://emergency.uncc.edu/sites/emergency.uncc.edu/files/media/2015%20Campus%20Evacuation%20Plan_1.pdf
University of Massachusetts Amherst, (2016). Living at UMass Amherst. https://www.umass.edu/living/residence/southwest
Milke, (2014). A Overview of Fire Protection in Buildings. http://911research.wtc7.net/mirrors/guardian2/wtc/WTC_apndxA.htm
Fernandez, (2011). Experimental Study of Bus Boarding and Alighting Times. https://www.researchgate.net/publication/263890456_EXPERIMENTAL_STUDY_OF_BUS_BOARDING_AND_ALIGHTING_TIMES