Black Swans, Sitting Ducks, and Geospatial Analytics
Find out how data scientists can use geospatial analytics to help mitigate risks from natural disasters
Many natural disasters are foreseeable, if only because they’ve occurred in the past, and the conditions for them to strike in roughly the same areas—perhaps with greater ferocity—rarely vanish entirely. Not to be a downer, but clearly the advance of global warming accelerates the rate at which humanity confronts unforeseeable disasters of a meteorological nature.
Populations that have grown accustomed to living perched just above sea level are likely to encounter inundations with increasing frequency. Those who live in formerly arid climates may experience flash floods more frequently than before. And people at low latitudes may have to dig themselves out from “once in a lifetime” blizzards more often. Even formerly well-watered temperate regions are likely to suffer paralyzing droughts.
Try as they may to predict how global warming impacts the earth as a whole or in particular regions, data scientists still face the methodological restrictions enumerated in my blog, “Data Science’s Limitations in Addressing Global Warming.”1 Historically shallow data sets, spatiotemporal scale mixing, and other challenges can complicate efforts to predict which regions will be hit when and where by meteorological black swans—high-consequence events of low predictability that often become retrospectively predictable.2 And even if data science models could pinpoint exactly when Category 5 hurricanes will strike Florida (fat chance), they probably wouldn’t be effective in predicting when superstorm Sandy 2.0 flattens the entire mid-Atlantic seaboard.
From a statistical analysis standpoint, modeling natural disasters as black swan risk-mitigation scenarios is better than modeling them as events that can be predicted, much less averted, with any degree of actionable precision. For example, earthquakes remain the classic example of these kinds of catastrophes: they can’t be predicted with confidence, but their impact can be mitigated through various proactive measures. That risk profile also describes the threats that regions face from many meteorological events. In such scenarios, impacted regions are often little more than sitting ducks whose best hope is to mitigate the severity of the inevitable effects.
Geospatial analytics is pivotal to risk mitigation in such scenarios. Considering the unlikelihood that humans will ever be able to control the planet’s caprices to the nth degree, geospatial risk assessments can drive community efforts in proactive planning and reactive retrofitting. Activist data scientists can perform a valuable public service by wielding location intelligence models populated with deep geospatial data on all affected roads, bridges, utilities, buildings, and other infrastructure. Geospatial risk profiling can guide governments, developers, and other community stakeholders in prioritizing highly feasible, cost-effective responses. For example, communities may leverage geospatial prescriptive analyses in mandating that all future urban development conform to risk-mitigating standards. And retrofitting can mitigate risk for the majority of built-out infrastructure components that remain vulnerable.
Where seismological risks are concerned, there are high-profile success stories of proactive community initiatives that have saved lives and property. Decades ago, for example, Japan and the US—along its West Coast—began to construct earthquake-resistant buildings that can sway with the most foreseeable land movements from earthquakes and aftershocks. Indeed, deep historical data on these and other disasters enables urban planners to build fairly precise maps of likely fault zones to minimize risks from potential future temblors.
Though historical data may be an unreliable guide for predicting meteorological catastrophes that strike far afield from their usual zones, this data can at least help climate data scientists frame a reference impact profile for sounding the necessary alarms. For example, historical profiles of the effects of hurricanes striking the Gulf Coast and the efficacy of levees and seawalls can help assess the extent to which New York City may benefit from similar measures in the face of rising sea levels.
Natural disasters are rarely tidy events in the sense of being just one type of disaster. Future natural disasters are likely to trigger black swan catastrophes in which geological, meteorological, ecological, and technological factors converge in unforeseen combinations. These events may be regarded as the proverbial perfect storm that is far more than, say, just a bout of especially nasty weather.
For example, in the recent article, “Is Data the Best Preparation Against Natural Disasters?” a US Geological Survey (USGS) seismologist notes that Southern California may be vulnerable to a powerful earthquake that could seriously damage the aqueducts that supply its water.3 One worst-case scenario from such an event is the risk that the already parched region could go as long as 18 months without water. Clearly, such a catastrophe would dwarf any loss of life and property from the quakes themselves, and it could conceivably trigger disease, famine, pestilence, and other latent disaster scenarios.
The consequences from this chain-reaction mega-disaster could be avoided if, for example, Southern California took measures to retrofit the aqueducts most vulnerable to the San Andreas Fault. Geospatial analytics could help identify exactly where to apply the most urgent retrofits in such an effort.
Making these retrofits is intrinsically a political matter. To help societies overcome the apathy and inertia that keeps them from taking necessary action, data scientists need to refine their ability to tell the disaster-preparedness story with all relevant data, analytics, and visualizations. Check out Fern Halper’s recent article, “Present the Data Story Persuasively to Make the Point,” in IBM Data magazine; it offers an excellent perspective on the practical art of data scientific storytelling.4 When unthinkable scenarios become part of preparedness planning, data scientists’ ability to generate compelling geospatial visualizations will be critical to catalyze consensus within the community on the need for urgent action.
Limitations of analytical risk mitigation
Can geospatial analytics help humanity mitigate the risks from asteroids, comets, and solar flares? That level of risk mitigation is not likely, unless the earth stops turning on its axis and entire hemispheres become sitting ducks for extraterrestrially generated events and other cosmic black swans. Let’s leave that particular scenario in the unthinkable category for the time being.
Please share any thoughts or question in the comments.
1 “Data Science’s Limitations in Addressing Global Warming,” by James Kobielus, IBM Big Data & Analytics Hub blog, September 2014.
2 “Nassim Nicholas Taleb – What Is a ‘Black Swan’?” ForaTV excerpt presentation, courtesy of Long Now Foundation, San Francisco, February 2008.
3 “Is Data the Best Preparation Against Natural Disasters?” Israel’s Homeland Security newsdesk, October 2014.
4 “Present the Data Story Persuasively to Make the Point,” by Fern Halper, IBM Data magazine, January 2015.