Fukushima Fallout

Joe Honton


This web page was developed in April 2013 as an entry in a juried competition[1] sponsored by The Guardian, Google, and the Open Knowledge Foundation[2]. The contest was conceived as a way to draw attention to the freely available government data sources that exist around the world, and to encourage innovative ways to visualize it.

The data used for this entry was published by the Japanese organization responsible for disseminating radiation readings in the aftermath of the Fukushima meltdowns. The visualization uses a map of Japan overlaid with a Voronoi diagram colored using a new "SPEIRT" scale and animated using a time-lapse sequence of data from stations throughout the country.

Inspiration for this entry was drawn from previous work done by Joseph Honton,
"Timeline of Fukushima Meltdowns"[3] and "Can the Japanese Lead us to a Nuclear-free World?"[4].

About the data

The data shown here are collected, compiled, and published by 原子力規制委員会 [5], the Nuclear Regulation Authority (NRA) of Japan, the new agency established after the explosions and fires that occurred at the Fukushima Daiichi Nuclear Power Plant (FNPP) released radioactive material into the surrounding environment, on March 12 through 16, 2011[6].

These data are measurements of airborne ionizing radiation obtained from 3,813 locations throughout Japan. Samples are taken at a normalized height of one meter above the ground surface on monitoring posts established for this purpose.

Monitoring posts are located in cities, towns, and rural byways from Wakkanai in northernmost Hokkaido to Naha in southwestern Okinawa. There are data collection points in all of Japan's 47 administrative prefectures.

Measurements are collected in real-time, and published in ten minute intervals, providing the public with 144 readings per station per day. Over the course of a 24-hour period, more than a half million readings are taken and published online.

The data shown here are for the 24-hour period of February 28, 2013.


The methodology for processing and visualizing the NRA data was chosen with an eye towards simplicity and familiarity while maintaining fidelity to the underlying measurements.

The data shown here are relevant to a large body of people who need to make short-term and long-term decisions about housing, farming, and city planning in the aftermath of the meltdowns at Fukushima Daiichi Nuclear Power Plant.

Decision makers, including family members, business owners, and civic leaders have little specialized knowledge about the health effects of ionizing radiation. Furthermore, many of these decision makers need to make important choices without strong backgrounds in general science. This class of citizens has been inadequately served by the existing public websites that visualize and interpret these measurements. This is still a hot topic politically, even two years after the accident[7].

There are several challenges in presenting this data to such an audience: measurement units that are unfamiliar, sampling stations that are not evenly spaced, values that are fluctuating over time, and an overall unwieldy amount of data.

Measurement units

Modern dosimeters are very sensitive and can measure data values that span several orders of magnitude. This presents a special challenge because logarithmic scales, while being the most appropriate representation for such wide data ranges, are not familiar to the general public.

Ionizing radiation is variously measured in units of grays, sieverts, and rems. All three are special names for joules, which in this case is a derived unit for measuring heat (J = kg • m2/s2).

The data presented here have been published in sieverts, which medical practitioners use when representing radiation dosage (equivalent dose, effective dose, and committed dose). Although sieverts are familiar to nuclear experts and medical imaging specialists, they are not well understood by the average citizen. Compounding this unfamiliarity is the occasional confusion over scale, with some reports and analyses using microSieverts (µSv) and others using milliSieverts (mSv), potentially leading the casual observer awry.

Sieverts are expressed as an amount of radiation received over a given amount of time. Often the time duration is one hour, so a low amount of radiation might be expressed as something like 0.012 µSv/h. Oncology journals and nuclear regulatory agencies sometimes use a time duration of one year, so if a person were to be continuously exposed to the same amount of radiation as in the above example, the amount of radiation might be expressed as 0.105 mSv/a (0.012 µSv/h • 24 hrs • 365.242 days). Lay people not comfortable with algebra may not readily grasp the two implicit conversions just employed – from microSieverts to milliSieverts, and from per hour to per annum – and might be confused by the assertion that 0.012 µSv/h is identical to 0.105 mSv/a.

This project defines and uses a new SPEIRT scale, described below, in order to avoid these pitfalls.

Sampling stations

The monitoring posts used to measure airborne ionizing radiation are installed at 3,813 locations throughout Japan. Stations are heavily concentrated in Fukushima prefecture especially within the inland population belt of Shirakawa, Koriyama, Motomiya, Nihonmatsu and Fukushima City, all of which are outside the 20-kilometer exclusion zone. Stations are less concentrated in the other prefectures, except at locations near nuclear power plants, where there are more. Japan has 50 commissioned nuclear reactors clustered in 13 areas.

Data such as these are typically presented on thematic maps which aim to portray the geographic distribution and variation of values in a spatial context. Thematic maps come in a variety of types, oftentimes using different sized bubbles or various color schemes to represent data values. When carefully designed, thematic maps can facilitate intuitive comprehension. On the other hand, poorly designed thematic maps can distort the data, make it inaccessible, or lead the viewer to false conclusions. Recall the infamous maps of the Red Menace, which amplified the threat of communism's spread by plotting Eastern Bloc countries in red – without regard to population density – using the heavily distorted Mercator projection.

Bubbles would not work well for this data because the heavily concentrated areas would quickly become unreadable. Instead, this project uses a thematic map of Japan overlaid with colored Voronoi cells to represent the data. Voronoi cells are explained below.

Fluctuations over time

Measurement values fluctuate over time due to the meteorological effects of wind and precipitation. At the outset of this project it was not obvious whether these fluctuations would be visible over short time spans. In order to test this, the project was designed with a time-lapse technique for visualization.

The monitoring post network is continuously reading and transmitting data. Data are published by NRA in 10-minute intervals. Using a time-lapse interval of 1 second per sample, the project is able to show a 24-hour sequence of readings in about two and a half minutes.

Unwieldy data

There are approximately 400 million radiation data points in the NRA database. This presents computational challenges for storage, retrieval, and network transmission.

Projects of this scale are plentiful in the government sector. Consider, for example, the domains of medicine, transportation, earth sciences, communications, and others, where real time data could provide important benefits to the general public.

Every project with "big data" has explored ways to handle these computational problems. Some solutions use large storage devices, fast computers, and wide bandwidth networks. But these cost money to acquire, install and operate. Government agencies and NGOs seldom have enough money to cover these higher costs.

Many government projects, in order to fulfill their mandate to make the information publicly available, have resorted to placing the computational burden on the user. They publish the raw data but do not provide the tools to analyze, visualize or interpret the data. This omission becomes an insurmountable problem for the vast majority of potential users. The result is that much government data languishes, failing to reach the larger audience that could benefit.

The visualization tool created by this project removes the obstacle that stops most people from accessing and understanding the data, but it doesn't remove the computational challenge of transmitting large amounts of raw data over a limited bandwidth.

There are potentially many thousands of people who could use the NRA data presented here (Japan has a population of 128 million people). But if all of these potential users were to visit this website, the web server would quickly become choked by too many requests. In order to partially work around this problem, the project has artificially limited itself to a single day's collection of data.

A full solution would be to deploy this on a more robust web server platform.


SPEIRT scale

It is common knowledge that exposure to high amounts of radiation over a short period of time can lead to poisoning or even death. It is also commonly known that prolonged exposure to low amounts of radiation can eventually lead to the development of cancer, thus health care specialists and nuclear industry workers continuously monitor their exposure levels. This second case, of prolonged lifetime exposure, is the one that needs to be addressed now that the immediate crisis of the meltdown has subsided.

A study published in 2012 by NIEHS[8], examined the radiation-related risks of leukemia after Chernobyl. The study concluded that exposure to low doses and low dose-rates of radiation from post-Chernobyl cleanup work was associated with a significant increase in risk of leukemia. The study showed that an accumulated exposure to 200 milliSieverts of radiation was statistically correlated with an increase in leukemia. While no explicit threshold between "safe" and "unsafe" exposure to radiation was proposed in that study, other work on long term survivors of Hiroshima and Nagasaki have extrapolated data to suggest that the unsafe threshold begins at 90 or 100 milliSieverts.

Japanese officials have been correct in pointing out that there is no immediate health risk associated with living or working in areas outside the identified exclusion zone. But residents living in the fallout shadow, just beyond the exclusion zone, who are exposed to elevated radiation levels, need to have a way to decide whether prolonged exposure over years or a lifetime puts themselves and their families at risk. This is a particularly difficult decision if the radiation readings are only slightly elevated.

In order to interpret the data within the context of a lifetime, the data have been converted from hourly units to equivalent lifetime units. The life expectancy of a Japanese citizen born in 2012 is 83.91 years[9]. Thus, if the accumulated exposure to 100 mSv of ionizing radiation – over a lifetime – is accepted as the safety threshold, the same threshold per year would be 1.191753 mSv/a. Further conversion from annual units to hourly units (1.191753 / 365.242 days / 24 hours) results in a threshold of 0.0001359 mSv/h.

These two constants would be sufficient to define a general purpose threshold conversion factor, if the amount of gamma ray radiation were, on average, constant over time. But as the radionuclides decay, the probabilistic frequency of future gamma ray emissions is lowered: this is the "half life" of the material, and it necessitates a more sophisticated calculation based on the characterization of the isotopes.

The radionuclides dispersed by the explosions and fires have not been fully characterized, however isotopes of Iodine, Tellurium, and Caesium were measured in the air[10] and trace amounts of Lanthanum, Niobium, and Silver were additionally found in the soil[11].

The half-lives of these isotopes and the time it takes for 99% of the material to decay is shown in columns three and four of the table below. The relative ratios of radionuclides to the reference 137Cs – taken in soil samples at J Village, 20km south of FNPP, and corrected to the date of release – is shown in the fifth column. The percent of activity initially attributable to each of the longer-lived materials (those with half-lives greater than or equal to 131I) is calculated from the soil sample ratios and shown in the sixth column.

Element Isotope Half-life 99% decay Soil sample ratio % initial activity
Tellurium 129Te 69.6 minutes 7.7 hours
Iodine 132I 2.30 hours 15 hours
Technetium 99mTc 6.01 hours 40 hours
Tellurium 131mTe 30.0 hours 8.3 days
Lanthanum 140La 1.678 days 11 days trace
Tellurium 132Te 3.204 days 21 days
Iodine 131I 8.021 days 53 days 55 89.9869
Caesium 136Cs 13.16 days 87 days 0.22 0.3599
Tellurium 129mTe 33.60 days 223 days 4.0 6.5445
Niobium 95Nb 34.98 days 232 days trace
Silver 110mAg 249.8 days 4.5 years trace
Caesium 134Cs 2.065 years 13.7 years 0.90 1.4725
Caesium 137Cs 30.07 years 200 years ref 1.6361

Using the values for half-life, the percent of activity attributable to each material can be estimated at any future point in time. This is shown in the two tables on the accompanying page Decay of radionuclides from Fukushima Fallout.

Given a measurement on any day, an estimate of a full lifetime of prolonged exposure can be determined by using the decay factors from these tables. For example, for the measurements shown on this page, for February 28, 2013 (the 717th day after the accident), the decay factor of 2.3264 from table 1 is used as DF1. Then, for each of the next 84 years, table 2 is used to lookup DF2 and the annual contribution is calculated as: measured value • 24 • 365.242 • DF2 / DF1. These 84 annual estimates are summed to arrive at the estimated lifetime exposure. Units are converted from µSv to mSv by dividing by 1000, to prepare for the next step. Finally the estimated lifetime exposure is divided by 100, since we are using the accumulated exposure to 100 mSv of ionizing radiation over a lifetime as the safety threshold. This results in a value which is defined to be the Safe Prolonged Exposure to Ionizing Radiation Threshold (SPEIRT). Using this scale, lifetime exposure to values less than 1.0 are "safe" and lifetime exposure to values over 1.0 are "unsafe".

This scale can readily be interpreted under different scenarios. For example, a newborn child is safe growing up and living a full life in an area with a SPEIRT of less than 1.0; or, a middle aged couple in their 40s (with half of their expected life remaining) should not be living in an area with a SPEIRT of more than 2.0.

Since the SPEIRT scale is adjusted to the decay rates of the radionuclides released in the accident, it establishes a scale that can be interpreted the same way at any point in time. That is, a place with a current value of less than 1.0 is safe for living, irrespective of how much time has lapsed since the accident. This is not so when using readings that are expressed in microSieverts.

For example, if an individual – who was evacuated from his home shortly after the accident – kept a written log of measurements, he would easily notice a trend toward lower readings over time. Perhaps he recalls that in the early days a knowledgeable expert said that his town's evacuation was prompted by a reading of 4.000 µSv/h and that he and his neighbors were removed to a refuge shelter whose reading was only 1.000 µSv/h. Now, months later, the readings from his hometown are down to 0.500 µSv/h and he wants to know why the government won't let him return home. Was the government untruthful, at the time, about the safety of the refuge shelter, or is the government untruthful, now, about the current safety of his hometown, which is now only half as radioactive as the refuge shelter was at the time of the evacuation? This conundrum leaves the government expert with little recourse but to refer the individual to reading material that explains the physics of radioactive decay. By contrast, the SPEIRT scale is simpler and more intuitive.

Finally, a word about notation. The SPEIRT scale is dependent upon the initial makeup of the radionuclides dispersed in the accident as well as upon two constants: life expectancy and safe exposure threshold. The makeup of the radionuclides dispersed at Chernobyl and Fukushima are similar, but not exactly the same, so the decay factor will be slightly different in these two cases. And since life expectancy varies per country and per year, a different conversion factor should be used for similar projects in other locales. Likewise, since the safe exposure threshold is only an estimate, and since the policy objective for different projects will want to use a different threshold, we should establish some notional device to clearly distinguish scales that use different constants.

For this project the notation "SPEIRT jp2012" is used to mean life expectancy of 83.91 years and unsafe threshold of 100 mSv/lifetime. A project near Chernobyl in the Ukraine using different decay factors and constants might use the notation "SPEIRT ua1986", while a project near Three Mile Island might use the notation "SPEIRT us1979"

Voronoi cells

Data that is spatially oriented over an irregular set of grid points can be associated with irregularly sized polygons such that any place on a map is associated with the nearest neighboring point. Voronoi cells This is a Voronoi diagram. Every position on the map can be considered to possess a value whose range is somewhere between the cell's central node and its nearest edge. Edges can be considered to have a value that is halfway between the values of the two nodes on either side.

The beauty of this topology is that no point on the diagram is left uncovered. And the importance to thematic mapping is that the cell polygons can be computed and visualized rapidly. In this project, the entire set of 3,813 points is computed once every second.

This project uses a JavaScript implementation, by Raymond Hill[12], of Steven J. Fortune's algorithm for calculating the shape of the cells. It is drawn in the browser window using an HTML5 canvas.

Map projection

The outline of Japan is defined by a collection of latitude/longitude sequences that are rendered using HTML5 canvas commands. Each sequence represents one of Japan's 1,338 islands.

A high resolution shapefile for Japan was obtained from DIVA-GIS[13] and converted to a more compact form using Quantum GIS[14] by reducing the frequency of points in each sequence. The names and locations of 1,196 principal cities was obtained from the National Geospatial-Intelligence Agency[15].

In order to plot the Earth's spherical coordinates on the flat canvas of the browser, Lambert's Conformal Conic projection was applied, using the "gPoint" PHP class developed by Brenor Brophy and Hans Duedal[16], which they based on a C++ algorithm first developed by Chuck Gantz.

The centroid for the map projection is the Fukushima Daiichi Nuclear Power Plant located at 37.421601N, 141.032782E. First and second standard parallels three degrees north and south of the centroid were used for the projection. False northing and false easting were computed for the latitude/longitude coordinates at a 10-meter resolution. The distance from each station to the accident site was computed using the Haversine formula with a mean Earth radius of 6371.0 km.

The final collection of x/y coordinates defining the outline of Japan represents a good compromise between area distortion and directional skew, optimized for the project's feature of prominence.

Computational platform

The data from NRA has been saved to a MySQL database and optimized for retrieval by station and by time. Browser requests are handled by an Apache HTTP Web Server. Both of these are running on a Dell PowerEdge Model 1850 (vintage 2005) — inexpensive second-hand technology. The server operating system is Fedora Linux version 16.

A user's request for this page immediately invokes a response that sends the page's basic HTML layout to the user. This is accompanied by a small file of CSS for styling, a file of JavaScript that implements Fortune's algorithm and the algorithms for zooming, panning, labeling, and animating the map. When the server is under normal load, all of this can be returned to the user in about 200 milliseconds, allowing the initial page to be shown almost instantaneously.

Immediately after this page is loaded into the browser, four background requests are sent to the server to obtain data for the map: one request retrieves the coordinates of the map of Japan (625 kB); a second request retrieves the coordinates, names, and first time slice of values for the monitoring posts (537 kB); a third request retrieves the names of the principal cities of Japan (47 kB); and the last request retrieves the data for the remainder of the 144 times slices (6.6 MB).

Using this sequence of requests the user is able to see the initial plot of values in about two seconds. All zoom, pan, and hover features are available immediately. After the complete set of asynchronous data is downloaded (typically about 15 additional seconds), the browser will automatically initiate the 24-hour time-lapse loop.


Ionizing radiation readings are colored according to the SPEIRT scale, as follows:

SPEIRT jp2012 Interpretation
< 0.03125 Approaching normal background radiation.
0.03125 - 0.0625 No need to think about it.
0.0625 - 0.125 Probably safe unless a very large error in our research reveals otherwise.
0.125 - 0.25 Probably safe unless a moderate error in our research reveals otherwise.
0.25 - 0.5 Probably safe even accounting for a small margin of error in our research.
0.5 - 1.0 Safe for everyone, to the best of our current knowledge, but this may be subject to change with refined research.
1.0 - 2.0 Safe for middle aged adults living out the remainder of their lives, but not for people under the age of 42 intending to live here full time.
2.0 - 4.0 Safe for extended stays, but not for living full time.
4.0 - 8.0 Safe for cleanup workers, even over prolonged periods, but only if they leave the area at the end of each workday.
8.0 - 16 Safe for occasional long visits.
16 - 32 Safe for brief visits only.
32 - 64 Protective gear should be considered.
> 64 Protective gear recommended.

Using the map control panel

The map can be panned east-west or north-south using these controls.
The map can be zoomed in or out using these controls.
The map can be zoomed to Japan (日本) or to Fukushima Daiichi (福島) using these controls.
The time lapse sequence can be paused and restarted with the center button, or stepped forward and backward in 10-minute increments when paused.
City labels can be displayed in romaji or kanji using these controls.
Zoom The map can be zoomed to a specific area by drawing a rubber band around the location: press and hold the left mouse button, then drag down and to the right.
Identify Any point on the map can be identified by positioning the mouse over it. Use the readout at the bottom of the map to see the distance from the accident site; the current reading in SPEIRT and µSv/h; the station name; the name of the prefecture; and the name of the nearest city, ward, town, or village (in Japanese kanji only).

The location of the Nuclear Power Plant is shown on the map with the label 福島第一 (Fukushima Daiichi).


  1. The Guardian > News > Data visualization competition with Google, "Competition: visualise open government data and win $2,000" http://www.guardian.co.uk/news/2013/feb/12/government-data-free-our-data.
  2. Open Knowledge Foundation http://okfn.org/.
  3. Joseph Honton, "Timeline of Fukushima Meltdowns" http://fukushima-book.com/.
  4. Joseph Honton, "Can the Japanese Lead us to a Nuclear-free World?" http://nuclear-book.com/.
  5. 原子力規制委員会 (Nuclear Regulation Authority) measurement readings were obtained on March 6, 2013 from the NRA website at radioactivity.nsr.go.jp.
  6. Fukushima Daiichi Nuclear Power Plant explosions occurred at reactor unit 1 on March 12, reactor unit 3 on March 14, reactor units 2 and 4 on March 15. Fires occurred at reactor unit 4 on March 15 and 16.
  7. Reuters, March 10, 2013 "Thousands in Japan anti-nuclear protest two years after Fukushima" http://www.reuters.com/news/article/2013/03/10/us-japan-protest-idUSBRE92903Y20130310.
  8. Zablotska, Bazyka, Lubin, Gudzenko, et al., 2012, National Institute of Environmental Health Sciences (NIEHS), National Institutes of Health, U.S. Department of Health and Human Services. "Radiation and the Risk of Chronic Lymphocytic and Other Leukemias among Chornobyl Cleanup Workers " http://ehp.niehs.nih.gov/wp-content/uploads/2012/11/ehp.1204996.pdf.
  9. Central Intelligence Agency, "The World Factbook" https://www.cia.gov/library/publications/the-world-factbook/rankorder/2102rank.html.
  10. Stoehilker, Mikkinen, Gheddou, 2011, Preparatory Commission for the Comprehensive Nuclear-Test-Ban Treaty Organization. "Detection of Radionuclides Emitted during the Fukushima Nuclear Accident with the CTBT Radionuclide Network" http://www.rdss.info/librarybox/mrr/MRR2011/PAPERS/04-15.PDF.
  11. Tagami, Uchida, Uchihori, et al., 2011 Oct 15, The Science of the total environment, "Specific activity and activity ratios of radionuclides in soil collected about 20 km from the Fukushima Daiichi Nuclear Power Plant: Radionuclide release to the south and southwest." http://www.ncbi.nlm.nih.gov/pubmed/21906779?dopt=Abstract.
  12. Raymond Hill, "Javascript implementation of Steven J. Fortune's algorithm to compute Voronoi diagrams" version 0.99. http://www.raymondhill.net/voronoi/rhill-voronoi.php
  13. DIVA-GIS > download > Japan > Administrative areas http://www.diva-gis.org/datadown.
  14. Quantum GIS version 1.8 http://www.qgis.org/.
  15. National Geospatial-Intelligence Agency. "Complete Files of Geographic Names for Geopolitical Areas from GNS" http://earth-info.nga.mil/gns/html/namefiles.htm.
  16. Brenor Brophy and Hans Duedal, "gPoint PHP class". https://gist.github.com/840476.


© 2018 Joe Honton