Learning About Radioactive Contamination
Working with Safecast data to understand the Chernobyl nuclear disaster and its consequences today
It probably does not come as a surprise, that we are easily fascinated by a dataset. This is especially true if the data is somehow related to a topic we are interested in to begin with. Not only have the two most severe nuclear disasters happened during our lifetime, but we have also used radioactive decay as a phenomenon to produce true random numbers.
So when one day, Ranga Yogeshwar, came back from a trip to Japan with a Safecast bGeigie Nano, we were hooked. This citizen science project is one of the most impressive initiatives “to empower people with data about their environments”. What started out as a grassroots initiative of some very talented creative technologists and activist is now the biggest publicly accessible resource of radiation data world-wide.
Fast-forward some months, this time approaching the 30th anniversary of the Chernobyl disaster, Ranga did pack the bGeigie again and took it to the Ukraine to shoot a documentary (Podcast link) for the German public broadcasting service WDR. And this time we wanted to get our hands on the data in order to understand the situation and put Safecast’s data into perspective. The result is this explanatory map, visualizing radiation exposure levels within the Chernobyl exclusion zone. The map accompanies not only the documentary on public German broadcast, but also countless other highly interesting stories, among them a stunning 360° video of the abandoned city of Pripyat, as well as this excellent storytelling website. So, what does the data say about Chernobyl today?
Safecast Data — An Overview
Safecast’s entire ~44 Mio (to date) readings are publicly accessible. Head over to their map to get an idea of the scope. Readings are available world-wide thanks to contributions from people who have purchased the bGeigie Nano and uploaded data. But unsurprisingly, data coverage is most dense across Japan’s main island Honshu. From the map we learn that near the reactor Fukushima Daiichi, radiation exposure peaks at around ~65µSv/h (micro-sievert per hour). But this does not reflect the actual levels. It reflects the fact that no Safecast volunteers where allowed closer to the buildings, particularly Unit 3. TEPCO’s official doserate maps of the site show readings as high as 1500 µSv/h.
The distribution of radiation levels around the Fukushima nuclear reactor is worth an entire post by itself and we will look at this another time. So let’s focus on Chernobyl instead.
Zooming in on the area around the Chernobyl nuclear power plant and the ghost town Pripyat, the density of readings is lower and maximum radiation levels near reactor 4—ground zero to the accident 30 years ago—seem to peak somewhere between 4 and 10 µSv/h. At this level though, some issues with the data also start to appear. First, GPS accuracy levels do vary of course, especially for readings inside the power plant, where the GPS signal strength is very low or there is no signal at all. Second, radiation seems to peak locally, often in a quite extreme way. That means, with relatively sparse readings in an area, we are only showing a small fraction of the entire picture. Third, it is hard to put the value of a reading into context. How high is 10 µSv/h really?
Massaging The Data
We cannot do a lot about the first two issues with the data. GPS accuracy varies between 5–30 meters, which in reality usually translates to a range of 10–50 meters depending on how many satellites are available to the sensor and its view of the sky. At the same time, radiation levels can be very specific to location, with often high fluctuation within meters. This depends heavily on where sources of radiation are located. In the case of Chernobyl, a lot of contaminated debris from the reactor has been buried in the area around the reactor. Sometimes these locations were documented, often they were not. Hence, to get the full picture, one would have to conduct measurements in a regular grid, something which is not easy given the size of the area. Based on this situation, we decided two things: First, we will have to aggregate locations to some extent to address the inaccuracy of the GPS. Second, we should not attempt to interpolate the values too much in regions where we do not have data. This would suggest that we know radiation exposure levels at locations where the situation could be entirely different.
Because of this situation, we have chosen the hex binning approach to come up with a just-fine-enough grid that evens out location inaccuracies but does not cover too much of areas we do not have data from. But instead of the usual approach to calculate the mean of each cell, we have chosen to keep the maximum value, because many times, the high number of (falsely located) low readings in one cell could average down local maxima too much. While initial experiments where done in QGIS, using a sample of the data in GeoJSON format and the MMQGIS Plugin, we settled with the immensely useful turf.js for the final aggregation. See this Gist for the code and the documentation of the data format.
What Do These Values Mean?
Finally, the biggest question: How high is 10 µSv/h really? In order to provide an answer for that, we needed more background information. So we have researched radiation exposure levels from many different sources, among which are the German Federal Office for Radiation Protection as well as the American Nuclear Society’s Radiation Dose Calculator. Initially, we were fascinated by the complexity of the issue and wanted to provide comparisons by duration, type and level of exposure.
In the end, we decided to include the most interesting of these comparisons directly into the legend of the map, which reveals annotated values if you expand its size. This decision was part of an overall shift to focus the map more strongly on explanation instead of exploration in order blend in more seamlessly with all the other excellent storytelling initiatives of this project.
A Simple Map
In the end, it took quite some steps to come up with a simple map that puts radiation exposure into perspective for such a broad audience. But with each step, we learned more about the actual topic and problems at hand, something which ultimately helped us to work side-by-side with the team at WDR on finding the right format for this topic.
At the same time it opened up many more questions and interesting aspects about the two most severe nuclear disasters of our life time. Things we look very much forward to exploring more in the future using Safecast data.
Thanks to Ranga Yogeshwar for introducing us to Safecast and to the team at WDR for their great support and collaboration: Lisa Weitemeier, Sami Skalli & Tobias Baum!
