Please visit our new site!


Undead ringer

What can nuking zombies tell us about infectious disease control?

Bill Hanage 21 August 2006

Stealth operator: a zombie outbreak (green) spreads before any human lives are taken (screenshot courtesy of

A single zombie is highly soon as one emerges you have to give up on the entire surrounding block

Zombies are a major challenge for public health authorities. The walking dead exert their damaging effect through two clearly delineated mechanisms. Firstly, they are wantonly destructive of public property in an attempt to obtain their primary source of nutrition, namely human brains. Secondly, once these brains have been harvested, their donor rises through infernal power as a new zombie, hungry for more of the same, and with the potential to create additional zombies. Strategies for the effective control of zombie outbreaks have varied. However, a new simulation model provides a simple illustration of the likely effects of different policies, which may be explored here.

The model contains three classes of entity: humans, zombies, and obliterated corpses (humans and zombies). These move around a simulated cityscape at pre-determined rates. Zombies maraud according to a random walk, and upon encountering a human, eat its brains to produce a new zombie. Zombie generation is hence an infectious process. A particularly nice feature is the simulation of panic in the human population, which leads them to move at a greater rate in the presence of a zombie, which may well of course drive them into another zombie.

The most effective countermeasure against zombie outbreaks has previously been shown to be aerial bombardment. Troops on the ground are not favoured because they are ineffective at the dismemberment – a prerequisite for zombie inactivation (for a simulation study of this, visit this link). However, carefully targeted high explosives successfully achieve this. A regrettable corollary is collateral damage, as nearby human life is inevitably destroyed. However, these innocent bystanders would, in the absence of intervention, would likely have become zombies – and a threat to their neighbours. This contingency is introduced into the model as a ‘user-directed interaction’ which destroys humans and zombies, making them functionally equivalent: in other words, non-infectious.

OK, enough. I can’t keep this up any longer. As you may already have guessed, the ‘model’ to which I’m referring is actually a game where you try to nuke the zombies from orbit in order to save as many humans as possible. Besides the fact that it’s kind of fun (and the name of the website, ‘hardcorepawn’, is brilliant), why am I writing about it?

Because the tongue-in-cheek scientific writing style I’ve employed above is not as far from the mark as you might think. The key word is ‘infectious’. The underlying model is similar to simple simulations of infectious disease spread, and we can use it to illustrate several important concepts.

Firstly, have a go at the game. If you’re not gifted with super-quick reactions, you’ll find that the zombies rapidly overwhelm your ability to bomb them back to hell. A single zombie is highly infectious, and given the high population density in the city, as soon as one emerges you have to give up on the entire surrounding block.

Which illustrates why, in the control of hugely infectious diseases like influenza, early detection is essential. If by the time you are aware of the problem the virus has already spread to multiple locations, it is very difficult to control it. Modeling studies not entirely dissimilar to our zombie game in spirit have shown that influenza could only be stopped if a very high (approximately 90%) proportion of the population around a point outbreak was administered antivirals. But such mass dosing is obviously logistically very difficult, especially in the parts of the world where any pandemic strain is most likely to emerge, such as Southeast Asia.

Another problem is how many times outbreaks emerge. One of the most difficult things about the zombie game is that it starts with four randomly placed zombies. By the time you have identified and bombed one or even two nests, the third and fourth are already well advanced. A zombie in the middle of a densely populated area is a far more immediate threat than one with few potential humans to infect, so you quickly learn to deal with these first.

One effective strategy is the firebreak. Here, you start the game by identifying a region without zombies, preferably at the edge of the map, and then create a cordon sanitaire by bombing between this area and the nearest zombies. This is not, however, enough, because zombies can wander across the bombed region. So you must maintain it and gradually bomb outwards, increasing your area of coverage and obliterating the remaining zombies left. By this means you can save over 3000 of the original population of 4000. The downside, of course, is that the humans outside the cordon are condemned – and those who in the vicinity of bombs are destroyed by your own actions.

When we face an outbreak of disease, there are several important differences from the zombie simulation. In the first place, people can react to the virus in two ways: they can recover from disease, thereby becoming immune to reinfection, or they can die from it. Both mean that after a period there is no further transmission. Secondly, we have drugs and vaccines that we can use to deny the disease onward transmission, as well as to treat sufferers. This is obviously better than bombing.

But the UK foot-and-mouth virus epidemic of 2001 is still a useful case study. The modeling work, done in my own department at Imperial College, showed how difficult it would be to stop an extraordinarily infectious disease after it had been introduced to the vulnerable population and not detected until it had already spread to several regions of the country. Following its remorseless spread, in which livelihoods and animals alike were destroyed, all options were on the table, but the decision was finally taken for a contiguous cull. This measure denies onward transmission to the virus by killing all potentially infected and susceptible hosts for a given radius around a case. In doing so, you will inevitably kill many healthy animals. However, you also hope to kill those animals either already infected or soon to be which would have spread to disease to the rest. Controversy still rages over whether a vaccination policy would have been more effective, but I do not wish to comment on this; instead, I merely wanted to point out how difficult it can be to stop a highly transmissible agent in its tracks. Just as some of the people in the simulated zombie city had to be sacrificed, so may some people in the face of an infectious disease outbreak.

Of course, what we really need is a vaccine against zombies. But that still eludes the best minds of science.

Related links

Visit Bill Hanage’s webpage to find out more about his epidemiological research.