When a blizzard rolls into town, cities usually bring out the big boys. By Monday night, facing a winter storm that dropped more than 2 feet of snow in some parts of the Northeast U.S., thousands of snow plows were lined up across the region. The next morning, long before the sun rose, many were already out on highways and on residential streets, shoving the glistening white powder out of the way.
But where they’re deployed can sometimes be a haphazard process, and depending on the city (say, Washington, D.C.), it’s not uncommon for the army of some hundred trucks to miss a spot—or, you know, an entire neighborhood. So, to ensure that the snow plows are reaching every corner of the city, some offices are relying on the power of big data.
In the northeastern region, it’s perhaps no surprise that Boston, which averages 45 inches of snow each year, is leading the effort. Since 2014, the city has monitored up to 700 trucks using an analytic tool called SnowCOP (for Snow Common Operating Picture). Developed by the city and the Virginia-based consulting company Qlarion, and with the help of GIS technology from Esri, SnowCOP tracks where each truck is in real time.
Every minute, GPS trackers on each truck ping its location to the central command center—the public works office—and that information gets recorded and visualized on a map of the city’s 30,000 streets. Streets that have been visited by a snow plow appear green, while those that the trucks may have missed appear red. Meanwhile, 311 calls to the mayor’s offices also get fed into the dashboard so that, over the course of the day, the city can keep track of the city to determine which streets or neighborhoods need another round of plowing or other services. “The goal is really to [help Boston] make decisions based on real-time data, and not necessarily on history," says Jake Bittner, CEO of Qlarion.
In 2015, as an effort to increase public engagement, Mayor Marty Walsh launched Snow Stats, an online tool that lets residents monitor snow removal activity in their own neighborhoods. By entering an address, the map will pull up information about the percentage of streets and number of miles plowed in that neighborhood, with extra information about how much time a truck has been on duty there. (The platform doesn’t appear to be active at the time of writing, though.)
Many other cities have launched their own public snow plow tracking websites, too. The D.C. area, for example, has at least four covering downtown and nearby suburbs of Maryland and Virginia. But Bittner says a tracking site only goes so far. “Tracking the location of a snow plow and putting it on a map is not very informational,” he tells CityLab. “To optimize the response, you need a lot more data.” Data like where the snow plows are supposed to be and which areas need more attention. The goal may be to ease concerns among the public, but absent that extra information, it could instill panic instead, and result in more 311 calls.
Kentucky’s transportation department also recently launched a program using sensors on 400 of its plow trucks to collect data about location, road conditions, temperature, and salt application. All that information then is combined with data gathered through social media and traffic monitors to allow operators to give their drivers directions in real time.
Elsewhere, researchers are helping their state and city officials harness big data as a predictive tool. In Iowa, for example, business researchers at Iowa State University analyzed 10 years’ worth of data to predict the best time to replace old snow plows. Iowa’s department of transportation generally replaces trucks every 15 years, but in analyzing maintenance records with purchase price and resale estimates, researchers concluded that they state can save between $2 million and $5 million in maintenance cost annually if they bought new trucks after six or seven years of use.
At the University of Maryland, the geographer Paul Torrens is focused on studying how humans make transportation decisions during major disruptions like snowstorms. Unlike conventional models, Torrens’s isn’t so much focused on data about land use or socioeconomics, which ignore the fact that actions are largely made based on “moment-to-moment” decisions and interactions: whether a person has internet access, the option to telework, or access to an Uber, for example. Instead, he and his team extract that sort of behavioral information from what people expressed through social media, feeding it into a machine learning program that will, in turn, help researchers predict how humans interact with the infrastructure around them. With enough information, their predictive tool can help cities respond and better communicate with their citizens under emergency events.
Bittner says that for many cities that have yet to harness big data, the first step is to change their way of thinking. “The challenge is that some cities are stuck in an old mentality, where they are more comfortable investing millions of dollars in new plows than they are in technology,” he says. “They haven't wrapped their heads around the idea that a few hundred thousands dollars in technology can have a larger impact than $10 million in equipment.”