Conversations with Data: #40
Do you want to receive Conversations with Data? Subscribe
Data is all around us -- be it the sounds of birds tweeting, or the levels of water pollution in a community -- it’s just a matter of capturing it. And while many of us rely on our trusty sources to analyse existing data, there’s an emerging opportunity to use sensors for the generation of new datasets on all types of physical phenomena.
In this 40th edition of Conversation with Data we’ll be looking at the weird and wonderful ways that journalists have used sensors, with some technical tips and tricks along the way.
What you said
We’ll start with some foundational advice from africanDRONE’s Frederick Mbuya: “It's the data, not the sensor that is of importance and, along the same line, it’s the story not the sensor.”
With that in mind, let’s take a look at some cool examples of story-based sensor work from our network. Over at the German public broadcaster, WDR, Thomas Hallet and his team created two web projects that played with the possibilities of sensor driven journalism: Superkühe and bienenlive.
“In Superkühe we followed three ordinary cows from three different farms (family farm, organic farm, factory farm) through the course of four weeks. We used different sensors and data streams to let the users observe how the cows were doing in a 24/7 mode: We tracked activity level, eating behaviour, health (body temperature and rumen pH) and -- of course -- milk production. The data were processed to feed a dashboard on the project website. Significant figures, on the other hand, were used to drive a chatbot on facebook messenger, where users could interact with simulations of each of the three cows,” Thomas told us.
“A bee hive is a blackbox. We installed sensors at three hives. A simple text engine transformed the sensor data into live diaries by the queen. We observed especially hard-working bees. But researchers found out: Most bees are not so active at all. We wanted to take a closer look -- and went on the trail of the chill bees. We found a paper by researcher Paul Tenczar who equipped bees with small transmitters. We got a bunch of transmitters. We tried to glue some transmitter backpacks on our bees.”
In the end, their sensors allowed WDR to report live from the lives of three bee colonies.
Frank Feulner, of AX Semantics, shared another example of air pollution reporting by Stuttgarter Zeitung. Using democratised sensors, the team was able to report on fine particle and nitrous oxide pollution in the German city of Stuttgart.
“We aggregated data over time so we could assess how bad the situation actually was in a certain area, and when relief was to be expected. We mingled sensor data with official data about air pollution alerts and other external factors to create a unique database,” he explained.
Sensors have also been a focus at Media Innovation Studio at the University of Central Lancashire, where John Mills has been leading research into different types, shapes, sizes, and use cases:
“Projects range from sensing biometric data to create data dashboard of Ranulph Fiennes’ attempt to run the Marathon des Sables a few years go, enabling school children to map pollution levels on their way to school in our DataMakers / UKKO project, to turning paper into an interactive surface that detects touch and creates media experiences. Our most recent project -- SenseMaker -- is working in Manchester in the UK. It asks: if journalists and communities could create their own sensors, what would they be? We’ve been working with data and editorial teams at the Manchester Evening News and with the local community to build a range of ‘sensing’ devices. Our early concepts -- which are currently being built or in the early stage of deployment -- span pollution in people’s homes, stress levels during the daily commute and image recognition what colour people wear the most!”
“For those getting started with sensor-based journalism, make sure to give yourself plenty of time to test out different sensors and their capabilities. There are a variety of manufacturers now that all have different capabilities and sensor data outputs, some are more accurate than others depending on cost. So having plenty of time for R&D before launching the sensor into the community is important. Doing a small pilot for a few weeks is highly recommended to get through any issues or problems before full launch. For our project, we gave ourselves six months to test out the sensors and determine the best ones and the overall plan,” she said.
Similarly, whilst at university, Reuters journalist Travis Hartman led a team to build a sensor-equipped tool that would collect data on noise pollution in Columbia, Missouri. His advice: “just dive in”.
“I started by learning to solder and building simple circuits with sensors that I could easily manipulate in the physical world. Thermometers, light sensors, etc, and learning how to write the code in the IDE to log and parse the data that was being logged. There are lots and lots of kits that come with the sensors and the code and wires all in the same package.”
And be sure to look for local initiatives and experts who might already be working on something, Jan Georg Plavec recommended. “You may then use your audience to further disseminate the sensors and work together with your readers.”
Now to the technical part: What sensors to use?
“The great Melexis MLX9064 makes a cheap but impressing 32x24 thermal camera with an ESP32 Microcontroller. With a Seed 4-Mic-Array we built a DIY-Voice-Assistant. An ordinary piezo-speaker serves as a great knock-sensor at my workshops front door, just knock the right rhythm and the door opens! My favourite sensor today is the cheap doppler radar sensor CDM324 (which is quite similar to HB100 or RSM2650). We needed to amplify its output signal quite a bit -- but with the proper preamp and an Arduino it is able to measure the speed of a passing vehicle -- so you can build a real radar gun for less than 20 Euros / 25$. I do wanna play more with ESP32-WHO which is already able to do machine learning-based face-recognition and identification on a sub 20$-device.”
Marcus lindemann added a few more tips:
“To get started I suggest buying a sensor that is easy to understand, e.g. we are all familiar with GPS since it has been built into our phones for a decade now. So buy a GPS-tracker for 100 Euros plus a SIM for the tracker (here you search for machine-to-machine-SIMs, short M2M, and buy a prepaid one). Then carry the tracker around in your office bag or in your car. Have a look at the data and try to make sense of it. Experiment with the settings -- how often should the GPS-sensor send its position? The one we use accepts any value from five seconds onward, the longer the interval the longer the battery will stay alive. What speed of movement do you expect from your target?”
Or, alternatively, why not build your own sensors. Jakob Vicari and Bertram Weiß often use the particle.io ecosystem for their sensors. “Particle offers a combined solution of hardware and cloud. It is easy out of the box and well documented. Find some quick start instructions in my blog.”
For more on sensors in journalism, be sure to follow future iterations of the Journalism of Things conference.
ICYMI: other happenings on DataJournalism.com
Editing investigations is challenging enough, but how can you succeed if the story involves data and you’re a numbers neophyte? In one of our latest Long Reads, Maud Beelman and Jennifer LaFleur share guidance for project editors on ways they can help their reporters shine, including best practices for workflows, bulletproofing, and writing techniques. Check it out here.
Our next conversation
It’s time for another AMA! Joining us in our next edition we have Steve Doig, renowned data journalist turned professor at the Walter Cronkite School of Journalism & Mass Communication of Arizona State University. You might’ve also seen his work in our Data journalism in disaster zones Long Read and our video course Doing Journalism with Data: First Steps, Skills and Tools. Comment to submit your questions.
As always, don’t forget to comment with what (or who!) you’d like us to feature in our future editions.
Until next time,
Madolyn from the EJC Data team