Science of Cities, Part 2
In last week’s blog I mentioned the amazing work that has emerged from looking for scaling laws for cities and which shows strong predictive power across a wide range of sizes of city for attributes from crimes rates to wages to the number of Thai restaurants.
One question it does not answer is whether there is an “optimum size” for a city. Not too big, and not too small. The scaling laws show monotonic increases in the attributes studied, although these include both positive (healthcare) and negative (crime rates)
attributes. Many organizations have developed indices for cities that combine multiple metrics into a single number. I could imagine that for those that combine both positive and negative attributes, these would tend to cancel out in the index even for very large
Nonetheless, the question deserves some thought. Starting from the small end, there are clearly benefits to larger cities – greater range of employment opportunities, greater social mixing, greater productivity, and so forth. As I noted last week, this “super-linear” scaling applies to the distribution and not necessarily to the development of a given city. Nonetheless, to misquote our former CEO Lou Gerstner, “I never met a small city that did not want to be a bigger city”. But is there some size beyond which the gains start to be outweighed by the disadvantages, in ways that may not be revealed by the scaling laws?
One area of concern that I am studying these days is the resilience of cities against Natural Catastrophes such as earthquakes, floods, hurricanes, and so forth. One aspect of this is the ability to evacuate the inhabitants when a disaster threatens. Evacuating
the City of Dubuque, Iowa (pop. 60,000) seems entirely feasible. Evacuating the Metro Tokyo region (pop. some 30 million) seems impossible. In fact it was this prospect, following the failures at the Fukushima Dai’ichi reactors in March 2011 and the real
possibility of clouds of radiation being released, that had the Government of Japan panicked. Fortunately it was not necessary, but it gives pause for thought.
In the past fifty years, the frequency and impact of natural disasters has increased dramatically; see The International Disaster Database . This trend is hard to explain. The recent IPCC report on violent weather events concludes that there has been no significant increase in the numbers of such events, although their intensity has been rising. On the other hand, it is difficult to see why the incidence of earthquakes would be driven by climate change. But the continued growth of mega-cities and regions makes me worry whether we are presenting ever bigger targets by concentrating more people and assets. Not only are the potential impacts on mega-cities larger, but the costs of mitigating these risks are disproportionately largely.
It is conceivable that the evacuation of Dubuque could be completed within one hour, since the city has an effective radius of less than 10 km and is not densely populated. Metro Tokyo on the other hand is very densely populated and has an effective radius of at
least 50 km. Both are bounded on one side by water. Almost every family in Dubuque has at least one car, whereas most people in Tokyo have to rely on public transportation, since there is no space for everyone to own a car.
It is conceivable that one could employ building codes to ensure that the city infrastructure would provide protection against these natural hazards, indeed earthquake engineering in Tokyo has gone a long way towards that goal. But this starts to introduce
new costs for living in a major city that are not imposed on smaller cities. Moreover, even if the city is better able to withstand the immediate impact of the event, it may still be uninhabitable if the event has shutdown the utilities and the supply chains of food,
water, and energy into the city. Hence an evacuation may still be needed.
So as the Urban Systems Collaborative, we would ask: How would better use of information change this problem? Past experience of attempted evacuations, e.g. Hurricane Andrew in 1991, suggests that managing large-scale evacuations is almost impossible as they quickly degenerate into gridlocked roads or impossibly overloaded rail services. Basic challenges in these processes are the scale and completeness of view. In 1991 it was impossible for a regional emergency management authority to have an integrated view of a situation that extended more than a few blocks from the operations center. Other qualitative information could be received via radios from police or other public safety forces, but no mechanisms existed to integrate these into a whole. The evacuees were essentially left to heir own devices to work out how to escape from the blocked roads and stations. Today the information from surveillance cameras and other traffic sensors can be integrated to provide a detailed, quantitative, complete, real- time view of the situation that can extend over an entire metro region. However tragic anecdotes circulate about emergency managers being prevented during evacuations from exploiting surveillance cameras that were installed for other purposes.
While it is not my real point, I would point to this scenario as an example of why some aspects of urban systems need centralised management systems. I still wonder about such limits to scaling of cities. I am willing to believe that the scaling laws are true under “normal circumstances”. But striving for resilience requires us also to think about the exceptional circumstances and how urban systems behave in the limit. I can find arguments such as that presented above that suggest that information engineering can help to overcome what seem to be practical limits to scalability, but I wonder if these hypotheses can be verified. IBM’s own work on massive agent- based simulations () of traffic flow at the scale on millions of agents is still an order of magnitude or two short of representing mega-cities. Although they can model driver behavior to some extent, they still leave me wondering about whether there is – if not an optimum size of cities – a maximum size for a city. And if there is, to what extent can we extend that limit through better information flows within the city?