NSF's Weather or Not!
Human forecasting? NSF-funded Paul Torrens uses Big Data techniques to better predict how people react to hazardous weather and the impact on infrastructure
National Science Foundation
Human forecasting? Paul Torrens uses Big Data techniques to better predict how people react to hazardous weather and the impact that can cause on infrastructure.
Interviewer: Charlie Heck
Interviewee: Paul Torrens
CHARLIE: Traffic jams, delayed openings, telework optional…winter is here.
Forecasters: We’re going to join our eye in the sky in a just a minute… but to repeat, we’re expecting several more inches of snow and the weather is playing havoc with the morning commute… many people may decide to stay home… even dedicated science geek Charlie Heck is considering it…stay tuned for the full forecasts
CHARLIE: And while forecasters will be busy trying to figure out the next move of Old Man Winter, a team of researchers at the University of Maryland will be busy trying to figure out our next move. That’s right -- you and me. Geographer Paul Torrens is interested in what we do after we hear about that foot of new snow or the ice storm coming tomorrow. Essentially, he’s trying to forecast “us”
PAUL: The human side of things is almost perennially difficult to study because humans are uncertain a lot of the time. They're imperfect, they're irrational.
CHARLIE: Torrens studies how snow and winter weather conditions affect human decisions about transportation and how those decisions ripple through other infrastructure, such as mass transit systems, bridges, tunnels and roads. Ultimately, what he finds out could be helpful to community planners and government agencies…
PAUL: People have been studying the physical nature of these infrastructures for a long time. And to a certain extent they're reasonably well understood. But the human side of things is almost perennially difficult to study because humans are uncertain a lot of the time. They're imperfect, they're irrational. They're often surprising. We also do things that are quite novel. And this makes the human side of the infrastructure equation quite difficult to plan and manage. So what we have been doing is looking at a couple of different areas primarily in trying to examine these systems using these data resources and data sources.
CHARLIE: In particular… Big Data
PAUL: So these are data that come from social media messaging, data that comes from transactions that people produce if they engage with each other and with the city. We say they're big because they're huge in size. But they're also big because for some of the data sets they cover entire populations.
PAUL: They're also big in that we get a lot of detail across lots of different activities. We get a lot of detail about individual things. But they're mostly not well organized data. So there's no official campaign to collect them. They tend to be misleading, they tend to be messy, they tend to be noisy, they tend to be raw. So we have to develop data mining techniques to swift through the largeness of the data but also to try and extract trustworthy, meaningful certain information from the data.
CHARLIE: So what’s the team doing with all this data? They’re working on two main projects: a near-real time atlas and a census of the current population. Think about it, the US government collects census data every 10 years…if you’re like me, living in a metro area like Washington, DC, chances are you don’t live in the same neighborhood you did 10 years ago.
PAUL: you think about a classic atlas as something you open in a book and it's static. And it doesn't change that much, of course. The world is much different, and realistically these things are changing on a second by second basis. You might have very, very fleeting and short run dynamics through to very marked and long run dynamics. So we want to build a dynamic atlas of that. At the same time we'd like to build a near real-time census so that we know where people are and we can guess what they're doing in a very, very high level of detail, both in terms of space and time.
PAUL: So most cities will collect data about transport, populations and interdependencies between the two of them on the order of about a year. Of course, the federal government collects a census of the entire population once every ten years. But a lot of things change at timescales that are much faster than that. So what we want to do is to try to supplement those official data sources with a trustworthy data from unofficial sources.
CHARLIE: After moving to the Washington, DC area from Arizona, Torrens saw firsthand just how snow and ice can upset even the most careful commute planning and…while not always crippling those disruptions have real economic costs and they can add up, quickly.
PAUL: when the snow falls, they need to get trucks out to plow and to salt the roads, and that costs a lot of money. And so they want to make that investment to right places and the right time with the greatest efficiencies.
PAUL: But also it starts to intermingle with just the general pattern of commerce. So lots of trucks making deliveries to offices and to supermarkets and so on. So when you get these kind of misconnections and interdependencies, it can propagate into lots of different systems. Also, people need to get to work, and so there's a delay of one hour for everybody who's starting work within an entire region. That can have quite dramatic consequences.
CHARLIE: That was Paul Torrens, a geographer at the University of Maryland. His research focuses on understanding and predicting how a not-so-winter-wonderland-weather situation can change people’s routines and in turn impact cities infrastructures and transportation systems.
CHARLIE: If you have any questions about this story or suggestions for interviews about super cool NSF-funded science, you have exactly two weeks before I hibernate for the winter. Email me at firstname.lastname@example.org.