Study: Modeling Fukushima NPP Radioactive Contamination Dispersion Utilizing Chino

Our goal involved developing contamination dispersion plots of the radionuclides emitted from the Fukushima Daichii Nuclear Power Plant; after which, we compare the simulation results to physical measurements taken at CTBTO monitoring stations located worldwide. Finally, we publish all data to the partMOM application for public analysis.


We took Initial I-131 source term estimations from Chino M., et al. [7]. We developed Cs-137 release rates from the I131/Cs137 ratios published by Chino[7]. All datetimes in the Chino publication were converted to UTC for comparison to other data sets within the partMOM application. After April 5, 2011, a release rate of 1 billion bq/hr (for both I131 and Cs137) was used for the remainder of the simulation. We derived this release rate from the I131/Cs137 ratios for the beginning of April (Chino[7]) and the published release rates of Cs137 through June (TEPCO [11]).

We utilized the open source software FLEXPART for all dispersion models. We used standard FLEXPART configuration. Our transport bounding box extended from pole to pole essentially including the entire Northern and Southern hemisphere. Simulations utilized 0.5 deg GFS weather data and a total particle population of 5 million.  Convection was not accounted for (lconv=0). FLEXPART accounts for I-131 half life in its species definition file. Similarly, standard FLEXPART installations omit the half life of Cs-137 – a reasonable assumption considering the half life of 30+ years.  Accordingly, We did not activate the agespectra feature.


In order to determine projected radionuclide concentrations,within the FLEXPART model, we defined >400 receptor points worldwide.  Our defined FLEXPART receptors correlate with locations for which we maintain physical, dose-rate or concentration, measurements. This allows a comprehensive analysis and validation of the FLEXPART model output.

Additionally, The ~8000 pFLEXPART plot maps establish an extensive visual diagram of the radioactive contamination dispersion.

In contrast to most original scientific studies, the aim of this project does not focus on publishing the consensus of a handful of scientists.  Rather, We aim to present the technical data, to a worldwide community, for the purpose of encouraging open commentary.

Please have a look at the published data and plot maps:

partMOM Application: FLEXPART model Utilizing Chino (20111) source terms.

We maintain a separate partMOM application into which we supply an ongoing analysis of this study. We encourage you to open a new partMOM application and begin an open analysis of your own.

Here are a few of the initial findings:

Initial review of the FLEXPART model output included comparison to published CTBTO concentration measurements.  FLEXPART output, in most cases, showed lower concentration levels than those published by the CTBTO.

Here in Takasaki, JP the levels of Cs137 and I131 were smaller,  by one order of magnitude, than CTBTO measured concentrations.

CTBTO publishes concentration measurements for Takasaki, JP when no concentrations are projected by the FLEXPART model output.

March 18, 2011 Sacramento, CA. FLEXPART output indicates concentrations of I131 (2) orders of magnitude lower than those published by the CTBTO while FLEXPART Cs137 level are consistent with those published by the CTBTO.

Upper altitudes (altitudes >32 meters above ground level) showed radionuclide concentrations several orders of magnitude higher than near surface concentrations.  This indicates physical measurements, taken at monitoring stations near ground level, may represent only a fraction of the total concentration of radionuclides dispersed over a location.  A more accurate assessment of radionuclide dispersion requires measurements at several different altitude levels.  Specifically, disclosure of helicopter or deploy-able monitor measurements would provide a much more detailed account of the contamination dispersion.

Sources of Error

Model Input Parameters. Small changes to input parameters can profoundly effect the FLEXPART model output.  For example,  increasing the lsync value by 6 times corresponds to an increase in concentration readings (at long range receptor locations) more than 4 times.  We used standard FLEXPART parameters (see COMMAND file above) but the importance of input parameters should not be overlooked.

Model errors.  FLEXPART uses probabilistic algorithms to predict particle transport.  It is possible the FLEXPART model contains some inaccuracies which would, in turn, produce errant concentration approximations.

Source Terms. The exact emissions from FNPP are not known.  Chino(2011) uses deductive algorithms to evaluate source terms – but the fact remains that no direct measurements, of the emissions at the FNPP reactors, were taken (or if they were they have not been publicly released). Any statistical or algorithmic errors in the evaluation could lead to errant source terms.  Furthermore, Chino(2011) does not consider the possibility of emissions from the spent fuel pools at reactors 2,3,and 4 – this possibility remains widely debated.  It should be pointed out that a “core melt on fresh air”, at spent fuel pool 4, has been specifically mentioned in an Areva report to the Japanese government.

Measurement Inaccuracies. It is possible the the physical measurements taken at monitoring stations contain some errors or that actual valid measurements have been withheld.

Weather Data.  We utilize GFS data at .5 degree resolution.  This resolution is sufficient for long range and meso scale dispersion forecasts; however, local weather data is preferable for short range transport and deposition modeling.


FLEXPART output consistently showed concentrations 1-2 orders of magnitude lower than those published by the CTBTO.  We cannot explain the conflict, between modeled and measured data, using model error solely.  We consider errors in the source term evaluation as one possible source of these conflicts. Our ongoing research includes utilizing source terms 2-3x greater than those published by Chino(2011).

Particle dispersion and deposition models, such as FLEXPART, currently do not exactly reproduce physical events.  That is to say, the model provides a statistical estimation of the movements of a particle in time and space which may differ from the actual physical movements of the particle.  If we maintained exact physical measurements of these particles, throughout their lifetime and travels, we would not need a dispersion model.

Models, utilizing statistical algorithms to approximate the properties of a population, present the potential for errors. It’s important to note, statistical models accurately reflect reality only if both the underlying algorithms and the parameters supplied to the model are correct.

The FLEXPART model (and others like it) provides a framework upon which we can both  develop ideas about both the travel and dispersion of radionuclides and forecast populations and locations with a potential for exposure to high levels of contamination. As we observed with the model of the Namie evacuation, FLEXPART revealed locations, along the existing evacuation route, subjected to high levels of contamination – this leads to very helpful inferences such as: “Evacuating [Namie town] to a location 15km Northwest may not be such a great idea – a better idea may be to visit your aunt, in Niigata prefecture, on the west side of the island”.

We contend that particle dispersion models, such as FLEXPART, should play an important role in both forecasting the dispersion of contamination and identifying “hot spot” locations exposed to high levels of contamination.

Have any Question or Comment?

Leave a Reply

Your email address will not be published. Required fields are marked *