Hurricane Irene evacuation naysayers point out some fundamental human problems with understanding riskPosted 29 Aug 2011 / 0
A downed tree in East River Park in Manhattan’s East Village after Hurricane Irene. Photo and caption text by David Shankbone.
The east coast of the United States woke up this Monday morning to begin cleanup following the passing of Hurricane Irene. In preparation for the hurricane, over half a million New York and New Jersey residents were evacuated from coastal areas. In addition, the entire New York City, Long Island, New Jersey, and Westchester County rapid transit systems were shut down. While some of the damage–particularly in suburban inland areas–was pretty serious, overall the New York City region is breathing a sigh of relief because Hurricane Irene’s storm surge effect on urban areas was relatively minimal.
This, of course, has brought out many naysayers. People who were evacuated grumble that they were inconvenienced for nothing, and even people fortunate enough not to be in the evacuation zone questioned the expenditure of resources required to facilitate such a large-scale evacuation.
To me, these naysayers epitomize a more general problem faced by human beings: most of the time, we are very bad at assessing and understanding risk. When scientists attempt to communicate their findings to the general public, they run smack into this collective risk-comprehension disability, often with disastrous results. One can see this phenomenon in microcosm with this most recent hurricane: the general public expects that government officials will only evacuate them when there is certainty about the dangers involved in a particular unfolding natural disaster. If precautions are taken and the natural disaster in question turns out to be less dangerous, for some reason the public is upset. This highlights how poorly most people understand probability and risk.
Here is how one ought to view the recent evacuations in response to the approach of Hurricane Irene. Meteorologists monitoring the storm were able to make some assessment of how dangerous the storm might be based on direct observations of wind speed, satellite images, and other fairly-reliable sources of data. This is the “mostly-certain” component of predicting how dangerous a particular storm will be, because our observational instruments have become so good that there is minimal error in the measurements used to make predictions. I think that a lot of people know that observational technology is quite accurate, and make the erroneous leap to assume that if observations are accurate then the predictions made based on those observations should be equally accurate. This, unfortunately, is not the way the prediction works. In the process of predicting a storm or the trajectory of other natural systems, all sorts of uncertainties enter into the process. We may know the wind speed of a particular storm, but that information alone may not be sufficient to fully predict where the storm is going and how devastating its effects will be. Not being a meteorologist, I cannot report on all the uncertainties faced in making these predictions, but it should be pretty easy to imagine that some relevant information may not be obtainable as the prediction is made. This leads to uncertainty. As we try to predict farther into the future, that uncertainty gets larger because of a process called propagation of error: predictions farther into the future are made based on already-uncertain predictions made about the near future. This is why weather reports are only good for several days.
So what did the predictions of this uncertain process look like? In the end, they actually are presented in very simple fashion: as probabilities. Usually prediction models are run as simulations, which means that the starting conditions of a particular natural phenomenon (in this case a hurricane) are entered into a model used to predict the future dynamics of that phenomenon. Hundreds of simulations are run, allowing any uncertainty to propagate through the predictions. Some simulations go to one extreme, some simulations go to another extreme, but most center on the most-likely outcome. This allows scientists to produce a risk profile that depicts how likely different outcomes may be.
I cannot tell you exactly what that risk profile looked like for Hurricane Irene, but let us just imagine for a second what it could have looked like. The hurricane was never strong enough to be a category five, so it is likely that the models predicted low to no probability of the storm developing to this intensity. But based on the news reports I have heard, there was some decent probability of the storm getting up to category three. Let’s say that this probability was 10%, just to have a number to work with. Now maybe the probability of the storm getting up to category two was 20%, remaining a category one was 30%, and the storm was likely to be downgraded to a tropical depression with a probability of 40%. How does one deal with these numbers?
The naysayers suggest that government officials “overreacted” to the storm because the storm actually turned out to be a tropical depression by the time it hit the New York City area. They might even say something like “this was the most-likely outcome, so the city should have allowed people to stay in their neighborhoods because the chances were that evacuation would not be necessary”. This gives us a good sense of the kind of logic that leads to poor decision-making in the face of uncertain future results. Simply choosing to prepare for the most likely outcome does not make sense, because the probabilities lay out possibilities rather than predicting exact futures. Those who claim that we should not have evacuated vulnerable areas are–whether they realize it or not–advocating for disaster on a regular basis. In the made up example above, I posited a 10% chance of having a category three hurricane hit the metropolitan area. Anyone who was willing to say that this probability is too low to merit evacuation is basically saying that he is happy to have one out of every ten hurricanes of this type cause a major disaster. What government officials like Mayor Bloomberg and Governor Christie are saying when they decide to evacuate is that even one out of ten hurricanes turning into a disaster is unacceptable. These government officials realized that the chances were that the evacuations would be unnecessary, because they are well-counseled and probably well-educated in probability. I doubt that the naysayers are equally aware of the implied values in their complaints about “unnecessary evacuations”.
I should also point out that failure to understand the meaning of uncertainty and the difficulties in making decisions when predictions are uncertain is only part of this story. As the above paragraph implies, it is possible that well-educated people who understand the nature of probability will still come to different decisions when faced with the same information. Maybe for some having one out of ten hurricanes lead to a complete disaster is acceptable. There are no values implied in scientific predictions: people need to bring their values to bear on these scientific predictions.
Humans are notoriously bad at assessing risk, particularly when that risk is long-term. In the case of a hurricane, we tend to make better decisions collectively because uncertainties are smaller (once the hurricane has formed, we know it will hit somewhere) and potential risks are more immediate. That Governor Christie and Mayor Bloomberg took very similar approaches to the hurricane does not surprise me, because for anyone who values human life the costs associated with the evacuation are dwarfed by the potential (albeit low-probability) costs of a major disaster. Clearly Governor Christie and Mayor Bloomberg part ways when it comes to assessing longer-term risks. Although they both have large areas of their constituency in zones that are vulnerable to storm surge, they are not taking the same actions to prevent long-term risk to these coastal areas. While Mayor Bloomberg has assessed the scientific predictions that suggest that continued global warming will put New York City at higher risk for storm surge flooding and decided that the city ought to actively participate in efforts to slow climate change, Governor Christie has looked at the same predictions and made the opposite decision. Obviously, this has to do with the politics of each executive, but I would suggest that these politics reflect values.Belief, Climate Change, Ethics, Political Science, Prediction, Sociology, Stochasticity