“Smart Refugee Camps”: applying the best of IoT and ICT for better camp management

UNICEF Jordan has been applying best practice from ‘smart city’ innovations in refugee camps, including an Uber-inspired system to optimize waste water management.

by Eva Kaplan, Innovations Specialist

Don’t be fooled by the name — refugee camps are no campsites, but more like pop-up municipalities. Housing tens of thousands of people (sometimes for decades), they have the same infrastructure requirements as any city, just with fewer resources and less time to develop them.

The humanitarian system has gotten fairly good at setting up and managing these camps, while constantly striving to do better. Perhaps this is best demonstrated in Jordan’s Za’atari camp, home to 80,000 Syrian refugees. In Za’atari, UNICEF Jordan takes the lead on water and sanitation, education, nutrition, and child protection. We use a “smart city” approach to all our operations.

Caution: nerd alert. This blog is about technology and data science, and there is no way to describe it without getting into the weeds.

But before we do, compare and contrast the above two pictures: On the left is my hometown of Portland Maine. Portland is a north-eastern US, a 400-year old city with 66,000 inhabitants. On the right is the Za’atari refugee camp in Jordan, population 80,000, established in 2012.

Despite very different histories and appearances, the same infrastructure makes each liveable, including a water and sewage system, law enforcement, schools, power grid, etc. Refugee camps could better be thought of as “pop-up municipalities” because all of the services a city needs, a refugee camp also needs. In Za’atari, we don’t have the luxury of 400 years of evolving service delivery systems, but we do have decades of experience from other crises. And we constantly strive to do better.

Using technology for “smart” refugee camp management

Smart city approaches essentially involve leveraging technology — including ICT and the Internet of Things (in particular sensors) to make cities more efficient. At UNICEF Jordan, we have three goals for using a smart city approach in the refugee camp:

· Using data to drive operations to improve efficiency. It will also strengthen the evidence base that supports our planning, for example to do contingency planning for anomalous events (like managing water and sanitation during severe storms).

· People-centred services to empower camp residents and improve programme delivery, for example using social networking sites to engage directly with refugees.

· Monitoring technology to improve equity in our delivery of services, for example tracking school attendance with RFID technologies.

In any smart-city initiative, designers have to be aware of the political and social context as well as the technical aspects. Humanitarians are well used to understanding the political and social context, but thinking through the technical constraints and opportunities is often less familiar. And the technicalities of working in a refugee camp are very different than they are in older, better established urban centres.

When assessing a situation in the field, I find it helpful to break things down into three interlinked components shown below: available infrastructure, available hardware, and available software. Once these three are understood, you can start designing solutions.

In our work, we have also found that where the ideal infrastructure, hardware, and software are not available, we can fill some of these gaps with data science, in particular with predictive analytics.

So far, this has all been very abstract — let’s get concrete.

“Uber for Waste”: using predictive analytics to streamline waste collection in the camps

Everyday, between 20 and 30 trucks collect 2.1 million liters of waste water from 2,500 tanks across Za’atari refugee camp, and take almost all of it to be processed in an internal treatment plant. UNICEF has managed this enormous operation for the five years since Za’atari was established. Today, all 80,000 residents of the camp have private waste facilities connected to these 2500 tanks, and over the next 18 months, we are rolling out a sewer network across the camp.

Yet challenges remain. An operation of this size requires massive oversight, coordination, and data management. When mistakes are made, refugees’ lives are impacted immediately. A waste tank nearing capacity smells bad; an overflowing one has health implications. The operation is also very expensive.

So far, data collection has mostly been done with pen and paper. Digitizing the operations is an obvious next step, but simply using tablets for the administration part would not help the logistical challenge of waste water collection.

Streamlining waste collection is exactly the type of project where smart-city approaches can make a real impact. There is a wealth of experience from civic tech data science projects, like NYC’s to optimize the locations where ambulance drivers should rest, as well as from innovative logistics companies like Uber. Learning from these, we decided to not only change the way we collect data, but to change the way we use data. Put another way, this is not an information management project, this is a data science project.

Our first idea was to essentially replicate the Uber system — refugees could report on a full tank via their phones, and trucks would be dispatched to do the collection. However, the infrastructure is not on our side. Since there is no Internet in the camp, there is no possibility for a truly real-time system. We then thought of equipping each tank with sensors that would connect to each other (without needing to reach the Internet), form a mesh network and let us know waste levels in real time. However, this was ruled out by the costs of procurement, cost, and maintenance. (But if you disagree, please do write in with any ideas!)

With the hardware solution ruled out, we developed software to predict when tanks will be full, and then to optimize trucking routes to maximize efficiency. Of course, this required real-world data on tank fill rates to feed into the system, and this is how it works:

When a truck arrives at the waste tank, the staff take a measurement of the waste level using a laser meter (picture below). They take the measurement again once the waste is fully connected. The information is entered into a tablet, and the next time the truck visits that tank, the information is collected again.

Photo: unicefjordan/dotmedia

The software’s algorithm uses this information to calculate a “fill rate” and predict when the tank will be full again. Based on this, the software produces a daily operational plan that dispatches trucks across the camps in an optimal manner.

Of course, things change. The size of a household may increase or decrease as people move around the camp; families may change their water use, for example if they acquire a washing machine. In such cases the truck will pick up the change in the fill rate at its next visit, and the algorithm will update accordingly.

The software’s algorithm uses the laser meter reading to calculate a “fill rate” and predict when the tank will be full again. Based on this, the software produces a daily operational plan that dispatches trucks across the camps in an optimal manner. Photo: unicefjordan/dotmedia

This sounds relatively simple — perhaps it is predictive analytics at its most basic. But in addition to the dynamic data (the changing fill rate), a hugely detailed amount of static data ensures that all of the factors that influence the optimal dispatch are captured.

The software can also account for another limitation: Za’atari’s waste water treatment plant can only handle a certain volume at a time. The system keeps track of how much waste water is reaching the plant. When that rate approaches maximum capacity, it sends trucks to places that are harder to reach and take longer, buying the plant a bit of time and respite.

But the question remains: Does this algorithm assistance actually increase the efficiency and equity of the service delivery? Well, we can’t quite say yet, but the initial results are fantastic. So far in September, reports of full tanks from refugees have dropped from 149 last year to 12 this year, a drop of 92%. So far we are optimistic not only that this will streamline operations here, but that it will provide lessons for any future camp instance. We’ll keep reporting as well, so watch this space.

What’s next?

We are excited about several ideas to strengthen the infrastructure, hardware, and software of the camp. As can be seen from the below, these initiatives are interlinked, and the progress on one front will depend on the progress of another.

For infrastructure, we are hoping to set up a local area network (LAN). The LAN will not connect to the global Internet, but will facilitate real-time feedback and communications on camp operations. While technically this is not difficult, setting up and managing such a network can present other problems, in particular as new devices in the camp are understandably treated with some suspicion by camp residents, and a LAN would require setting up routers of some description.

For hardware, we are excited about the possibilities that lie in what we are calling “Internet of Things without the Internet”, i.e. connected devices that can talk to each other even if not connected to the world wide web. This will also allow us to replicate the functionality of a LAN if necessary but without the actual routers. In particular, we would love to see better hardware (at low cost) that would allow us to put connected devices in the hands of refugees (ie. Phones with very strong Bluetooth). Such hardware would greatly expand what we are able to accomplish with software.

For software, the sky is the limit. However, our current explorations are around local instances of social networks. This would allow us, for example, to digitize our Time Bank initiatives, where refugees exchange services using time as currency instead of money. This would, however, require connected devices. Our assessment of existing available hardware is that the connections between them are not strong enough for these purposes.

Conclusion

Those are our thoughts — and we really welcome yours! If we are going to accomplish our smart city goals, it will require concentrated effort from a variety of partners. If you have ideas, or feel your work could contribute to these aims, feel free to get in touch.