Data-Directed Road Repairs Could Save Money and Lives
The United Nations Global Pulse, an innovation initiative on big data and data science, and Western Digital recently announced the winners of the Data for Climate Action Challenge (D4CA) at the Data Innovation: Generating Climate Solutions event during the United Nations climate change conference (COP23) in Bonn, Germany. This post highlights one of the D4CA thematic winning teams, from Georgia Institute of Technology, which created a framework to optimize data-directed road repairs.
An unprecedented open innovation challenge to harness data science and big data from the private sector to fight climate change, D4CA was launched earlier this year and called on innovators, scientists, and climate experts to use data to accelerate climate solutions. Access to large amounts of data – anonymized and aggregated to protect privacy – accelerates the ability to spot connections, gain insight and develop predictive algorithms that can provide more precise direction and decisions. The Data for Climate Action Challenge demonstrates what’s possible when public and private sector organizations partner for social good.
WIRED Brand Lab: Hi Caleb, can you walk us through your project?
Caleb: Our project is called Predicting And Alleviating Road Flooding For Climate Mitigation. We used optimisation and modelling techniques to help alleviate the impacts of flooding.
As climate change continues, flooding is going to become more severe and more frequent. For places where infrastructure isn’t as good, in particular parts in Africa for example, there’s a big challenge with how to maintain road networks and how to allocate budget.
What we’re doing is we’re trying to do is work out how best to fortify the road network from flooding in order to maximise accessibility during flood events.
We want to make the road network as resilient as possible to improve accessibility. This has implications on things like sustainable development goals, ensuring people have access to services like hospitals and schools.
WBL: What data set did you use?
Caleb: We were given datasets from Orange, mobile phone network provider. The company provided us details on how people move around Senegal.
WBL: How did you deduce that from phone data?
Caleb: When you make a phone call from one place, Orange knows who you are, and what cell phone tower you used to connect the call. The next time you make a call, it might connect to a different cell tower, indicating movement.
We aggregate the activity in one cell tower zone and the other cell tower zone to be able to calculate how many trips were taken. We then paired that with the road network to see what routes people would have used to travel between two places.
We obtained flooding data from a company called Fathom, they gave us a high resolution, gridded map, that tells us how much and how likely each grid cell on that map will flood under different scenarios. For example, in a 1 in 10 year flooding event, how many metres underwater is this grid cell going to be. From there we can see which roads will become impassable.
WBL: What did that approach tell you about flood prevention policy in Senegal?
Caleb: Our main finding showed that you needed sophisticated computational techniques to understand the trade-offs when it comes to allocating budget to road network improvements.
Under flooding conditions, you might lose some roads, but depending on how much budget you allocate, you may only be able to get back some of that mobility, or a lot. It’s a non-linear relationship, so we have to try and illustrate how that relationship is quite complex.
WBL: Can you explain that a bit more?
Caleb: If you have 100km of road to repair, and only X amount to spend on improvements, you might get a lot of improvement if you repair in an area where there is a high concentration of people, but almost no improvement in another area.
But, if you invest more money, the impact on the areas where previously improvements in mobility were low, can suddenly become a lot more effective. If you’re a policy maker, you need to understand what that trade-offs are between investment and where to spend it.
WBL: How was the data anonymised to prevent privacy issues from arising?
Caleb: We were given a sample of all phone records in Senegal from 2013. Orange broke that dataset down into two-week periods. By removing the personal identifying parts of the dataset, we were unable to tell whether we were tracking the same people or different ones. We were more interested in the movement, rather than who was moving.
WBL: What are some of the implications of this work?
Caleb: The way we built it was in three parts: how are the roads being used, how are they damaged and how can they be repaired. The first two parts of that are interchangeable, so they can be used for other incidents apart from flooding. It could be forest fires, or earthquake damage.
WBL: Could this sort of dataset be used as a proxy for development, trade, population movements, things like that?
Caleb: If you know one area of the country is responsible for producing a lot of food, losing access to that part of the country would be devastating. So you could encode that into the definition of accessibility to include movement of food, or critical goods rather than just the movement of people.
WBL: What’s next?
Caleb: Because this framework is general, we’d like to apply it to different settings. Using the computational models we’ve built, we think this can be applied to lots of areas to help make better decisions when it comes to protecting road networks against natural disasters and climate change.
This content is produced by WIRED Brand Lab in collaboration with Western Digital Corporation
Data Makes Possible will be following the winners as they work to implement their solutions and bring real change to our world, and we’ll be publishing interviews with the thematic and data visualization winners throughout January and February.