Hypothetical Greenhaven Vehicle Evacuation
Matthew Trevor | Spring 2017 | Geography 190
Introduction & Site
When creating an evacuation model for an area there
are a few general aspects needing to be addressed. They
range from vital details directly affecting the design of the
evacuation model to other aspects that should be considered
in the event of an actual evacuation. While the latter may not
be affected by the model, it’s still important to understand
them and use the output of the study to help address them.
However, in some cases the output of the model can help
give insight into planning an evacuation.
The problem attempting to be solved is the balance of
flow through a network, including such aspects as traffic flow,
density, and speed (Sheffi et al, 1982). Through the usage of
GIS software, keeping track of these metrics is easier and
more effective. More importantly, being able to understand
this information ahead of time allows for planning instead of
waiting for a crisis event to establish evacuation zones or
other safe areas (Murray-Tuite et al 2013; Ma et al, 2012).
The area of focus is the Pocket area of Sacramento as
well as the road network leading into and out of the area. The
study site was limited to a couple hundred feet west of
Interstate 5 allowing estimation of traffic flow out of the area.
The natural border of the river levee provides the remaining
boundary of the study area. The Sacramento River levee
contains the neighborhood like a bowl. This forces all
evacuation in an eastward direction towards Interstate 5 and
the area of south Sacramento. For the flood model, a break
on the northern edge of the levee, about halfway across, will
be taken into account. The lack of elevation change in the
Pocket area allows flooding to cover a large amount of area in
a short amount of time.
There are 10,945 individual residential locations and
27,975 vehicles. This is significantly larger due to apartment
complexes accounting for hundreds of vehicles at a
single “residential location”.
Residential Locations Unable to Evacuate
6
Street Congestion
As seen in Figure 1, most of the residential locations
were successful in evacuating. However, 508 residential
locations accounting for 7,132 total vehicles were unable.
Many of these were located on the western side of the
neighborhood, though many apartment complexes, even
some located relatively close to the safe zones were still
unable to completely evacuate in time.
Most of the congested streets shown in Figure 2 were
near the eastern border of the neighborhood. Also, many
roads close to apartment complexes were backed up due to
the number of vehicles attempting to exit at one time. Also,
there is a low amount of congestion on Interstate 5 leading
northward and southward out of the neighborhood.
Figure 3 shows a breakdown for each of the nine “Safe
Zones” defined earlier in the study for how many locations
and vehicles were allocated to said zone. It is easy to see how
not all residential locations and their vehicles were allocated
evenly to a safe zone. Some areas, such as zone five received a
small number of evacuating vehicles, while others like zone
four and six received a majority of them.
5
4
7
8
Discussion & Conclusion
3
9
Figure 2: Map of street congestion during the evacuation event
2
Methods
To create and run an evacuation model, ESRI’s ArcGIS
software was utilized along with CASPER’s evacuation model
software (Shahabi et al, 2014). To model the hypothetical
evacuation in the Pocket Area – the decision to focus on a
vehicular type of transportation was made since this
neighborhood is mostly residential. Also, the study assumes
all families will follow the “one car per family” rule (Dow and
Cutter, 1998). To track how many vehicles will be utilizing the
network, a recent dataset showing each parcel of residential
property was collected from the Sacramento County Tax
Assessor. Within this parcel data is a field called “units” which
details the amount of individual living spaces, this was
utilized to calculate the number of vehicles at each parcel.
To measure the amount of time it takes flooding to
spread a Sacramento County hypothetical flood depths map
was utilized and is shown in Figure 4. This map contains three
separate “stages” of flooding across certain intervals of time
showing the spread of water. By digitizing these, one can
compare the amount of time it takes to evacuate against how
long it takes flooding to reach them.
To keep the road network as accurate as possible, an
ESRI generated United States Streets dataset was obtained.
There were nine total safe zones created for the study:
1) Southbound I-5 from the southern edge of the Pocket area,
2) Pocket Road, 3) Florin Road, 4) 43rd Avenue, 5) Northbound
I-5 from the northern edge of the Pocket area, 6) Riverside
Boulevard, 7) Gloria Drive, 8) 56th Avenue, and 9) South Land
Park Drive. These routes were chosen because they are all
through streets leading directly out of the Pocket area.
After running the evacuation model, consideration of
“cost" is achieved through a spatial join between the cost of
each route taken and the overall time it takes for flood
inundation to occur. After this, subtracting the evacuation
time from flooding time shows which locations are able to
evacuate. If this result is negative, then the location would
not be able to leave the area in the allotted time.
Results
Flood Inundation Time
1
Levee
Break
Figure 1: Evacuation model output showing residential locations unable to evacuate and the location of Safe Zones
Safe Zone Allocation
In the event of an evacuation many residential locations
would be able to evacuate in time. However, there are a few
different aspects that could have resulted in a more even
distribution of vehicles to the safe zones, or in a quicker
evacuation time overall.
Although the amount of residential areas allocated to
each safe zone is pretty evenly balanced, there are large
discrepancies in the number of vehicles allocated to each
zone. This could be due to the geographic layout of the road
network along with the amount of time it takes to balance
traffic in an area. Also, due to a large cluster of residential
locations unable to evacuate on the western edge of the
pocket area, it would be beneficial to increase the amount of
effort to educate that population about sheltering-in-place
on top of the evacuation plan.
When observing patterns on the map, one notices how
the homes deemed “unable to evacuate” aren’t right next to
each other, they appear to happen every other location. This
could be due to the nature of how CASPER “releases” vehicles
into the road network.
Part of the flood analysis included creating a shapefile of
flood inundation times in the area. However, this data came
from a map with only three intervals of time covering the
whole geographical area. If there were more intervals of time.
A more accurate representation of the relationship between
flood inundation and evacuation times would be more
accurate and useful.
CASPER has a function allowing creation of interactive
polygons to act as “barriers” in the road network. If the layers
dictating the flood times were input as these dynamic
polygons, the modeler could update the road network as it
flooded which would allow a much more accurate
representation of the changing conditions as event time
continued.
One final aspect that could greatly affect the outcome of
another iteration of this study is the implementation of a
strategy called contraflow (Tuydes et al, 2006). When utilizing
this, both directions of a road are used for outward traffic. In
theory, twice the number of vehicles could travel the same
route helping locations farther to the west and and any
complexes bottlenecked during the evacuation.
References:
Figure 3: Breakdown of the amount of residential locations and associated vehicles to the various Safe Zones
Figure 4: Map of flood inundation times and the location of the hypothetical levee break
Dow, K., & Cutter, S. L. (1998). Crying wolf: Repeat responses to hurricane evacuation orders.
Ma, J., Liu, S., Wang, W., Lin, P., & Lo, S. (2012). A GIS-based micro-simulation queue model for vehicle evacuation. Journal of Risk Analysis
and Crisis Response, 2(3), 178-187.
Murray-Tuite, P., & Wolshon, B. (2013). Evacuation transportation modeling: An overview of research, development, and practice.
Transportation Research Part C: Emerging Technologies, 27, 25-45.
Sheffi, Y., Mahmassani, H., & Powell, W. B. (1982). A transportation network evacuation model. Transportation research part A: general,
16(3), 209-218.
Shahabi, K & Wilson, J.P., “CASPER: Intelligent capacity-aware evacuation routing,”Computers, Environment and Urban Systems, vol. 46
pp.12-24, Apr. 2014
Tuydes, H., & Ziliaskopoulos, A. (2006). Tabu-based heuristic approach for optimization of network evacuation contraflow. Transportation
Research Record: Journal of the Transportation Research Board, (1964), 157-168.