Professional Documents
Culture Documents
NRE 509 Lab #5 –STELLA Modeling of a Zombie Invasion
Background
The Zombie Invasion scenario has been a common trope in American popular culture for
decades. In one version, the outbreak begins with a mutant virus introduced to a small
population of humans. The virus kills those whom it infects, and after a period of time, causes
them to rise from the dead with an insatiable hunger for human flesh. Decaying mockeries of
human life, these walking undead are insensible to cold, heat, fatigue, or fear. Mindless, they
possess only the most basic instinct to feed upon the living.
The infection spreads by fluid contact, typically a bite. An infected human who manages to
escape being consumed entirely will shortly, die, only to rise again as a fully contagious member
of the walking dead. The most often‐cited way to kill a zombie for good is to destroy its brain.
The Zombie Invasion scenario is often used as a vehicle for critique of modern man’s utter
unpreparedness for survival in the event of the collapse of social order. In books and films, the
highest occupants of the modern socioeconomic ladder, whose position is wholly dependent on
modern economic and technological infrastructure, are typically the least equipped for survival
in a post‐apocalyptic world. Mnual laborers, farmers, survivalists, and others whose
accumulated life skills are less dependent on functioning modern infrastructure—as might be
the case with, say, a hedge fund manager—are the most likely to survive and promote the
survival of others. The collapse of social order, followed by the outright reversal of its
smoldering remains, is part and parcel of the horror of the zombie apocalypse.
Past Work
In their seminal work on mathematical modeling of a zombie scenario, Professor R.J. Smith?
(the ? is part of his legal name) and his team applied a modified epidemiological model to the
problem. i However, I observe a small deficiency in the model. In the most basic form, there are
three stocks.
S – susceptible humans
Z – zombies
R – removed
Figure 1. Modified epidemiological model for zombie invasion
The term S represents the death of humans by natural causes. However, the model clearly
shows that all humans that die from natural causes are directed to a common pool of corpses
(R) from which zombies may resurrect. Humans which die of natural causes are not expected to
rise from the dead. This oversight appears to persist in more sophisticated models presented in
the paper.
Jarett Diamond
A careful observer might also note that there is no allowance for zombie destruction in this
model. However, in a later section of the paper, Smith?, et al implement an “impulsive
eradication” mechanism, in which the humans would attempt to control the zombie population
by “strategically destroying them at such times that... resources permit.”
Overview of STELLA Model
In my STELLA model, I attempted to expand on the past work by Munz, Hudea, Imad, and Smith
by incorporating the following elements:
A hybrid Lotka‐Volterra/epidemiological model to simulate a multistage disease spread
by predation, with a 1:1 efficiency of conversion of prey to ‘infected prey’
A persistent, rather than impulsive, zombie eradication mechanism which more closely
resembles prey defense systems in the natural world
A ‘panic factor’ coupled to the ratio of infection vs. zombie elimination rate, plus the
zombie feed rate which gives a rough approximate value for the general feel of which
side is winning. The panic factor causes indirect human casualties (i.e. not directly
caused by zombie infection or consumption).
A ‘learning curve’ which simulates adaptive human behavioral changes over time
A ‘zombie feed rate’ term which simulates destruction, but not infection and
conversion, of living humans.
The STELLA model omits human birth rate, and death by natural causes. For small populations
(under 10,000) and short modeling times, these factors were felt to be of negligible
consequence.
Detailed Description of STELLA Model
In my STELLA model, there are four stocks:
S – susceptible humans
I – infected humans
Z – zombies
X – removed
and two conveyors:
IFREE– “free infected” infected humans not quarantined
Q – quarantined humans
A conveyor is similar to a stock, but the primary output is controlled by a transit‐time factor.
A susceptible human (S) may be infected at a given rate proportional to the number of humans
and the number of active zombies (Z). Alternatively, a susceptible human may die (via s elim
rate) from 1) the mass panic caused by the zombie invasion or 2) be wholly consumed by a
zombie. The zombie feed rate is proportional to the number of zombies. The panic factor was
described previously.
A susceptible human unlucky enough to be infected moves to the infected stock (I). At this
stage, the infected human may be detected. If the infected human is detected, he may be sent
Jarett Diamond
to the quarantine conveyor (Q), in futile hopes that a cure may be found. Since no cure is
possible, all infected humans are removed (X) after the latency period of the disease expires.
However, there is also a leakage rate (quar escape rate) of infected humans who escape
quarantine, and join the free infected conveyor. The model assumes that successful escapes are
made immediately upon arrival in quarantine (i.e. at the onset of the disease latency period,
which begins again in the IFREE conveyor). In time, the humans realize the futility of the
quarantine strategy. This is simulated by the learning curve factor, which causes the proportion
of infected humans sent to the free infected conveyor to increase over time.
This leaves two categories of infected humans.
1) non‐quarantined infected humans
2) undetected infected humans
Both groups proceed directly from the infected stock (I) to the free infected (IFREE) conveyor.
Thus, the IFREE conveyor consists of these two populations, plus the population of escaped
infected quarantined humans. Unless they are destroyed first, the disease latency period
expires and members of the IFREE conveyor resurrect as zombies. The leakage rate from IFREE
represents the active detection and elimination of the IFREE population.
Resurrected zombies are capable of creating more zombies by infecting humans, or destroying
humans outright by feeding on them. The zombie feed rate is also an important term in the
panic factor equation, which results in indirect human casualties. The zombie population is
controlled by active retaliation by humans (z elim rate), which is proportional to the product of
the number of humans, the number of zombies, and a variable zombie kill factor. The zombie
kill factor increases over time with the learning curve, as humans learn better methods for
controlling the undead.
Jarett Diamond
Figure 2: Complete STELLA model for zombie invasion
Analysis of Results
The model itself is inherently highly unstable. A small change in the infection rate, or a small
change in the intrinsic zombie kill factor can result in rapid population crash of either humans
or zombies. Parameters were finely tuned to demonstrate a long, drawn‐out contest between
the living and undead.
The three‐day disease latency is clearly apparent in Figures 3 and 4, as the graph of flows and
stocks indicate a sharp jump in activity/population after the 3rd day of the invasion.
As might be expected, widespread panic can result in catastrophic loss of life. Panic has
elements of both positive‐ and negative‐feedback. As panic casualties increase, there are fewer
humans to fight the zombies, which increases the rate of undetected zombie resurrection. This
increases the panic factor. At the same time, there is also a smaller human population for the
zombies to infect and feed upon. This decreases the panic factor. The relative weight of the rate
Jarett Diamond
of zombie feeding, as compared to the ratio of infection/zombie elimination may be adjusted to
reflect a real‐life scenario when empirical data becomes available.
Figure 5 illustrates that the learning curve is the key to human persistence. The steadily
increasing zombie population reaches its peak at after approximately 10.5 days, at which point
the battle swings to favor the humans.
1: S 2: I 3: Q 4: Z 5: X
1: 10000
2: 7 1
3: 0 5
4: 10 3
5: 1000 3 2
2
2 3
3 5 2
1
1: 9500 4
2: 4 4 4
3: 0 5
4: 6 1
5: 500
4
5 1
1: 9000
2: 0
3: 0
4: 2
5: 0
0.00 7.50 15.00 22.50 30.00
Page 1 Days 9:58 PM Fri, Dec 11, 2009
Untitled
Figure 3: Stocks (S, I, Z, X) and conveyor (Q). Conveyor IFREE is not shown.
1: inf rate 2: z elim rate 3: z feed rate 4: Panic Casualties
1: 10
2: 6
3: 0
4: 250
2
2
2
1
3
1: 6 3
2: 4
3: 0 1
4: 150
3
4 1 1
4
2 4
3 4
1: 2
2: 1
3: 0
4: 50
0.00 7.50 15.00 22.50 30.00
1: Learning Curve 2: z elim rate 3: ifree elim rate 4: inf not quar rate 5: Z
1: 3 3
2: 6 4
3: 1
4: 6 4
5: 10
3 2 3
4 4
2
2
1
1: 2 1
5 5
2: 4
3: 0 5
4: 3 1
5: 6 3
5
2
1: 1 1
2: 1
3: 0
4: 0
5: 2
0.00 7.50 15.00 22.50 30.00
Untitled
Figure 5: The effect of the Learning Curve on infected human elimination, zombie elimination,
use of quarantine, and overall zombie population
Conclusion
The keys to human survival in the face of a zombie invasion rest in the ability of citizens to
mount an immediate, ruthless, and above all, adaptive campaign of resistance. A strategy of
quarantine of known infected humans is impractical. A large‐scale security breakdown (not
included in this simulation) could result in thousands of quarantined, infected humans escaping
into the general population.
It is worth noting that the zombie population never exceeded seven. After 30 days, the human
population of 10,000 lost nearly 10% of its total membership. Of this loss of life, almost 8% is
attributed to indirect casualties caused by panic. Therefore, all possible measures must be
undertaken to counteract panic, be they educational or coercive.
As this model clearly demonstrates, the destructive potential of the walking undead cannot be
underestimated.
i
P. Munz, I. Hudea, J. Imad and R.J. Smith? When zombies attack!: Mathematical modelling of an outbreak of
zombie infection (Infectious Disease Modelling Research Progress 2009, in: J.M. Tchuenche and C. Chiyaka, eds,
pp133‐150).