Analysis of Radionuclide Releases from the Fukushima Dai-ichi Nuclear Power Plant Accident Part II
- First Online:
- Cite this article as:
- Achim, P., Monfort, M., Le Petit, G. et al. Pure Appl. Geophys. (2014) 171: 645. doi:10.1007/s00024-012-0578-1
- 4.1k Views
The present part of the publication (Part II) deals with long range dispersion of radionuclides emitted into the atmosphere during the Fukushima Dai-ichi accident that occurred after the March 11, 2011 tsunami. The first part (Part I) is dedicated to the accident features relying on radionuclide detections performed by monitoring stations of the Comprehensive Nuclear Test Ban Treaty Organization network. In this study, the emissions of the three fission products Cs-137, I-131 and Xe-133 are investigated. Regarding Xe-133, the total release is estimated to be of the order of 6 × 1018 Bq emitted during the explosions of units 1, 2 and 3. The total source term estimated gives a fraction of core inventory of about 8 × 1018 Bq at the time of reactors shutdown. This result suggests that at least 80 % of the core inventory has been released into the atmosphere and indicates a broad meltdown of reactor cores. Total atmospheric releases of Cs-137 and I-131 aerosols are estimated to be 1016 and 1017 Bq, respectively. By neglecting gas/particulate conversion phenomena, the total release of I-131 (gas + aerosol) could be estimated to be 4 × 1017 Bq. Atmospheric transport simulations suggest that the main air emissions have occurred during the events of March 14, 2011 (UTC) and that no major release occurred after March 23. The radioactivity emitted into the atmosphere could represent 10 % of the Chernobyl accident releases for I-131 and Cs-137.
KeywordsFukushima Dai-ichi accidentatmospheric transport modelingsource terms evaluationCs-137I-131Xe-133CTBTO
On March 11, 2011, the tsunami induced by the 9.0 magnitude earthquake that occurred east of Japan caused serious damage to the cooling systems of the Fukushima Dai-ichi Nuclear Power Plant (NPP). Due to the lack of cooling, hydrogen and vapor blasts, dewatering of spent fuel rod pools and fires led to the release of radioactive materials into the atmosphere. Although at the time this article was written, the comprehensive understanding of the accident was not fully established, it was accepted that units 1, 2 and 3 reached a fuel fusion state. The detection in the air of fission and activation products by monitoring stations belonging to the International Monitoring System (IMS) of the Comprehensive Nuclear Test Ban Treaty Organization (CTBTO) is likely to provide relevant information on the reactor core damage. CEA relied on these IMS data and on analysis performed by radionuclide laboratories supporting the CTBT network in order to estimate the worldwide distribution of the atmospheric releases of radioactive material and to better understand the accident features. The first part (Part I) of this study is dedicated to observations and interpretations of the accident characteristics relying on the chronology of the major detections and the nature of detected radionuclides (Le Petitet al., 2012). This linked part (Part II) deals with the Atmospheric Transport Modeling (ATM) at global scale (long range simulations). The objective is to assess the arrival time of radionuclides over IMS stations and to evaluate the quantities released into the atmosphere. We will mainly focus on Cs-137, I-131 and Xe-133 fission products. Hence, the two first radionuclides will be the main contributors to dose and worldwide industrial Xe-133 background could be modified by Fukushima radioxenon release affecting the performances of the CTBT radionuclide monitoring network.
2 Sequence of the Accident and Observations at Fukushima Dai-ichi Site
Sequence of main identified events that could have lead to atmospheric releases
Automatic shutdown of units 1, 2 and 3 due to earthquake
2) Hydrogen explosion in reactor building
6) Hydrogen explosion in reactor building
8) Explosion. Possible damage of pressure system. Damage of building wall of unit 4 reactor
9) Fire occurred in spent fuel cooling pool
10) Fire occurred in spent fuel cooling pool
11) White smoke generated
12) Rise of primary containment vessel pressure
13) Greyish smoke
14) Black smoke
2.1 Dose Rate Measurements
2.2 Assessment of Atmospheric Cs-137 and I-131 Leakage after March 22
Figure 2 (right) provides rough estimation of release rates required to obtain measurements presented in Fig. 2 (left). Release rates are calculated from a Gaussian dispersion formulation, assumed to be relevant due to the small distance between the reactors and the monitoring devices. It is assumed that sensors were under the direct influence of releases, which was not necessarily the case in reality. Building effects on dispersion are not taken into account and ground deposition and radioactive decay phenomena are neglected (both are negligible at short distances for considered radionuclides). Calculations are performed for a 20 m high source and for two meteorological conditions assumed to provide reasonable major and minor estimations of release rates: i.e., unstable atmosphere with a 2 m s−1 wind speed (Normal Diffusion 2) and a stable atmosphere with a 5 m s−1 wind speed (Weak Diffusion 5). By time integration of the release rates over the observation period (~2 months from March 22), these two conditions lead to releases ranging from 1 × 1013 to 1 × 1014 Bq for I-131 and from 1 × 1012 to 1 × 1013 Bq for Cs-137. The work should be continued to estimate the amounts released during leaks (i.e., between main releases) that may have occurred before March 22.
3 Elements on Atmospheric Transport Modeling (ATM)
ATM has been conducted at both the regional (short range simulations) and global scale (long range simulations). This section describes briefly the methodologies used and the objectives of the simulations.
3.1 Short Range Simulations
The regional scale simulations have been conducted using MM5 V3.7 (MM5, 2005) and WRF V3.3 (WRF website) mesoscale meteorological models. These well-known systems are parallelized, limited area, non hydrostatic, terrain following and sigma-coordinate models designed to simulate or predict mesoscale atmospheric circulation. NCEP’s Global Forecast System (GFS) meteorological data with 6 h and 0.5° resolutions have been used as initialization and boundary conditions (National Centers for Environmental Prediction/GFS website). Among the existing atmospheric dispersion models, the 3D lagrangian particle dispersion model (LPDM) FLEXPART was used (Stohlet al., 1998; FLEXPART homepage). The versions of the model used here are those specifically developed for MM5 (V.6.2) and WRF (Fast and Easter,2006). It should be noted that in the standard WRF FLEXPART version, wet deposition phenomena are not fully implemented. To correct this, the calculation of precipitation subgrid variability was made available in particular the calculation of the fraction of surface that undergoes precipitation. This fraction is a function of convective and large-scale precipitation and also depends on cloud cover. The method developed is based on the MM5 FLEXPART model, where cloud cover is estimated from the total water contained in the air column between ground and roof level of the calculation domain.
Several mesoscale simulation attempts were performed by varying the vertical calculation grid resolution and the relaxation coefficients towards the large-scale input data inside and outside the boundary layer for the coarser domain. A possible way to improve the quality of the simulations for this time period would be to proceed to observations assimilation. Satisfactory results have been obtained with the Japanese SPEED and WSPEEDI-II emergency response systems (Chinoet al., 2011; Katataet al., 2012a, b; Teradaet al., 2012). Large-scale data used in these models have higher temporal and spatial resolution than the GFS ones and a large amount of local meteorological observations are assimilated. As our regional simulations are still ongoing, only long range results are presented in this article.
3.2 Long Range Simulations
Atmospheric Transport Modeling at global scale has been carried out using the particle dispersion lagrangian model FLEXPART (V.8.2) and the NCEP/GFS meteorological data (http://weather.noaa.gov/pub/SL.us008001/ST.opnl) with 6 h, 0.5° × 0.5° and 1° × 1° resolutions. This ATM system is suitable regarding the spatio-temporal characteristic scales of the problem to be solved. The objective of long range simulations is to assess the arrival time of radionuclides at different monitoring stations located over the globe and evaluate the quantities released into the atmosphere. As previously quoted, we will mainly focus on Xe-133 (half life 5.244 days), I-131 (8.023 days), and Cs-137 (30.05 years) which are volatile fission products (see Part I of the publication).
3.3 Gas and Particulate Deposition, Particle Size and Emission Height
During events such as the Fukushima or Chernobyl accidents, gases and particles emitted in the atmosphere can be transported over very long distances and over long periods of time. Typically, the spatio-temporal scales of the problem are of the order of the circumference of the globe and several months. Because of these atmospheric dispersion and transportation scales, radioactive decay, gas-particulate conversion, and dry and wet deposition can drastically affect the behavior of emitted material. It should be noted that the noble gases (such as radioxenon) are not affected by deposition phenomena.
Wet deposition of gases or aerosols on the ground is due to the washout of the radionuclide plume by precipitation (rainfall, snow, etc.). It does not depend linearly on the precipitation rate. It is based on a scavenging coefficient, which depends on the precipitation rate and the considered radionuclide. To calculate the dry deposition of gases and aerosols, a classical approach in the dispersion models is to separate the gravity fallout (settling) and the interaction with soil and vegetation. The total deposition rate is the sum of these two contributions. The deposition due to interaction with the soil is calculated for altitudes between the ground and a reference height (e.g., 15 m). It is usually a function of aerodynamic drag terms produced by vegetation canopies and soil nature. Settling velocity is assumed to be zero for gases but depends on density and diameter for particles. This last point is important because as a FLEXPART particle can not represent several settling velocities, gases and aerosols trajectories must be calculated separately.
The height at which the radionuclides are emitted can have a significant influence on their dispersion into the atmosphere. Qiaoet al. (2011) used a climate model to estimate the long range dispersion of Fukushima releases over a period of 3 months. In particular, they studied the influence of the emission height of radionuclides by selecting a release close to the ground, in the 5,000 m layer and the 10,000 m layer. Particles are transported more rapidly towards North America, Europe and Asia especially when their emission height is high. In our simulations, since the releases were not energetic (excluding cooling pool fires) and the explosions that occurred were controlled, it is considered that releases take place in the 0–200 m layer. As mentioned in the following section, radionuclides can quickly reach the middle and upper troposphere in situations when an updraft occurs.
4 Meteorological Conditions
5 Radionuclides Considered in Dispersion Simulations
The analysis of the Chernobyl accident showed that large amounts of Cs-137 and especially of I-131 were emitted into the atmosphere (UNSCEAR, 2000). In order to compare these quantities with those released during the Fukushima accident, Cs-137 and I-131 are considered in the simulations. Because of the complexity of events that occur during such accidents, it is difficult to estimate what were the distributions of particle diameters and the gas/particulate ratios (I-131) that were emitted into the air. However, concerning long range simulations, it seems reasonable to assume that only small particles may be transported over such long distances. Without further information, a single 1 μm diameter particles distribution for Cs-137 and I-131 was considered. In addition to the Cs-137 and I-131, simulations were performed with Xe-133. Hence, as previously mentioned, in the framework of the Comprehensive Nuclear Test Ban Treaty (CTBT), monitoring of atmospheric concentration of radioxenon is relevant to provide evidence of atmospheric or underground nuclear weapon tests. However, during the couple of months after the Fukushima Dai-ichi accident, monitoring capabilities of the network could have been affected by the large amount of radioxenon released by the accident. The impact of Fukushima radioxenon releases on the worldwide Xe-133 background must also be investigated.
The gaseous form of I-131 is supposed to be deposited on the ground by the washout of the plume (wet deposition). Caputet al. (1993) have conducted experiments to determine the washout factor of elemental iodine in the gas form. The results show that molecular gaseous iodine, which is very reactive, seems to be irreversibly captured in rainwater. The average washout factor determined from experiments is Λ = 8.2 × 10−5 s−1 (this coefficient indicates the fraction of radioisotopes washed out from 1 m3 of air per second at a standard rain intensity of 1 mm h−1). This value is of the order of magnitude of those generally accepted in deposition models [e.g., Λ = 7 × 10−5 s−1 (Pittauerováet al., 2011)]. In the FLEXPART model, gaseous I-131 washout factor is set to Λ = 8 × 10−5 I0.62 s−1, where I is the intensity of the rain in mm h−1. Scavenging factor for particulates of I-131 is given by Λ = 1 × 10−4 I0.8 s−1. The dry deposition velocity of gaseous I-131 is calculated using the resistance method (Wesely and Hicks,1977). Calculated deposition velocities (interaction with soil) are approximately in the range 1 × 10−3–1 × 10−2 m s−1. These values are consistent with those generally accepted in impact assessment models (BIOMASS program, 2003). For the I-131 aerosol form, considering a 1 μm particle diameter class, the dry deposition velocity is of the order of 1 × 10−4 m s−1. This value is one order of magnitude smaller than those currently used in particle deposition models with diameters above 1 μm (about 1 × 10−3 m s−1) (BIOMASS program, 2003). In the case of particle diameters above 1 μm, the deposition velocity is increased by gravity settling (Slinn, 1982; Sehmel, 1980). Radioactive decay of I-131 is taken into account in the simulations.
The below cloud scavenging factor for Cs-137 is given by Λ = 1 × 10−4 I0.8 s−1. The calculated dry deposition velocity is about 1 × 10−4 m s−1 when considering 1 μm diameter particles. Regarding the simulated durations (several weeks), the influence of radioactive decay of Cs-137 will be negligible on calculated activity concentrations.
Radioxenons are highly volatile fission products. Noble gases are not affected by wet and dry processes. Radioactive decay of Xe-133 is taken into account in the simulations.
5.4 Iodine Gas/Particle Conversion
The behavior of I-131 in the atmosphere is known to be complex as the gaseous form could gradually evolve into the particulate form. This has been the subject of many studies, particularly following the accident at the Chernobyl plant in April 1986. Uematsuet al. (1988) used atmospheric measurements made in Japan and on a boat in the Pacific Ocean to estimate the characteristic conversion gas/aerosol time for I-131. With the assumption that about 60 % of the total I-131 was present in the gas form in the Chernobyl releases, they found an average conversion time of about 2–3 weeks, with a minimum conversion time of about 12 days. The relative uncertainty is estimated to be about a factor of 2. Massonet al. (2011) indicate that the I-131 gas/I-131 total (total = gas + particles) ratio measured at the site of the Fukushima Dai-ichi plant from March 22 to April 4, 2011 was 71 ± 11 %. The average ratio measured in Europe until April 12 on a station network is very close, 77 ± 14 %. According to the authors, the similarity of the measured ratios suggests that the conversion gas/particle is small.
To verify this hypothesis, a calculation of atmospheric dispersion was performed by considering a single release which occurred on March 14 from 18 to 24 h (UTC), linked to the explosion on unit 2. In this numerical experiment, the release is assumed to be made up of 70 % of I-131 gas and 30 % of I-131 aerosol and a diameter of 1 μm is chosen for particulate form. The wet and dry deposition processes for gas and particulate forms are those described above. The radioactive decay of I-131 is taken into account, but gas/particle form transfers are not simulated. Assuming that the deposition process is realistically reproduced, the transfer between the gas phase and particulate phase can be considered low if the gas/particle ratio calculated in Europe is close to the initial release one.
6 Simulations at the JPP38: Takasaki Station
Due to the relatively short distance between Fukushima Dai-ichi and JPP38 (~210 km), it was considered that the behavior of a tracer can be representative of the behavior of small sized aerosols of Cs-137 during the first few days of the accident. Results presented above have shown that the behavior of aerosols was insensitive to the particle diameter when it was smaller than 1 μm (Fig. 4). For this particle size, it was shown that dry deposition phenomena were weak. Previous results also highlight that wet deposition was not very effective for the period leading to higher activity concentrations of JPP38. Figure 7 shows that the calculated activity concentrations were almost the same on March 15 and for the period from March 21 to 23. During those days, the northeasterly wind was favorable to the transport of radionuclides towards the station. The figure shows that on March 15, the station was mainly sensitive to releases 7 and 8 (venting and explosion of unit 2) that occurred in the late afternoon of March 14. For the March 21 to 23 period, the station was sensitive to releases 12 and 13 (rise of primary containment vessel pressure and greyish smoke on unit 3) that occurred during March 20 and 21. From March 23, the contribution of other releases is more difficult to assess. From March 30, radionuclides emitted at the beginning of the event were able to travel around the globe and to contribute to the simulated signal.
Other releases, which have not reached the monitoring station due to meteorological conditions, might have to be added to the estimated I-131 and Cs-137 releases. Uncertainties on calculated source terms are significant and are related in part to the measurements. The I-131 and Cs-137 uncertainties have been assessed at ±25 % (3,700 ± 1,000 Bq m−3 and 400 ± 100 Bq m−3, respectively) on March 15 measurements. In addition, after March 15, the JPP38 radionuclide station had been contaminated by the radioactive cloud as shown in Part I of the publication. Uncertainties on source terms are also related to inaccuracies in the wind fields used and may also be due to the dispersion model itself. Hence, the 0.5° wind fields may not be sufficiently resolved to match the JPP38 observations, as this station is close to Fukushima and located in an area where the topography is complex (Appendix, Fig. 17).
7 Evaluation of Source Terms with Long Range Simulations
Based on source terms evaluated from JPP38 station measurements, long range simulations were performed using 1.0° GFS wind fields. Direct calculations were carried out considering the events listed in Table 1. It is assumed that all events can lead to Cs-137 releases. As unit 4 was already shutdown at the time of the tsunami, it is assumed that the reactor and spent fuel cooling pools can not lead to I-131 releases. Because Xe-133 is highly volatile, it is considered to be emitted into the atmosphere at the time of the hydrogen explosions of units 1, 2 and 3.
All simulations indicate that the material emitted into the atmosphere mostly remained over the northern hemisphere, showing that the inter-hemisphere exchanges are limited. Only a few minor detections were observed over the southern hemisphere (e.g., for I-131: 1 μBq m−3 at FJP26, Nadi, Fidji, from April 5 to 6 and 7 μBq m−3 at PGP51, Kavieng, Papua New Guinea, from April 11 to 12).
At stations where the aerosol behaviour is the most consistent with the observations, the source terms evaluated using the JPP38 detections lead to an overestimation of the long distance simulated detections. Total release estimated from the long range simulations is about 1 × 1016 Bq (~5–7 times lower than those obtained using the JPP38 measurements). The simulations show that the events that occurred in the afternoon of March 14 UTC on the unit 2 may have led to significant releases of Cs-137 in the atmosphere. For these events, a total release of the order of 6 × 1015 Bq is required to find an agreement with the measurements (~10 times lower than for the JPP38 calculation). The events on the unit 3 and that took place from March 20 to 23 are also found to be significant with a required total release of the order of 2 × 1015 Bq. Simulations suggest that the hydrogen explosion and/or venting operations of units 1 and 3 (March 12 and 13/14) may have resulted in a total release of 2 × 1015 Bq.
7.3 Discussion on Cs-137 and I-131 Results
Releases calculated for Cs-137 and I-131 appear to be much larger than those calculated for the continuous leakage between March 22 and mid-May 2011 (see Sect. 2.2). Continuous leakage calculated from March 22 may represent only 1/100 (Cs-137) and 1/1,000 (I-131) of total releases and they will be neglected.
Total source terms calculated from long range simulations for Cs-137 and I-131 (resp. 1 × 1016 Bq and 1 × 1017 to 4 × 1017 Bq) are in good agreement with those presented in the literature (Chinoet al., 2011; Mathieuet al., 2012; Winiarek, 2012) These results suggest that the source terms estimated from the station JPP38 are probably too high. Chinoet al. (2011) found also that the largest releases took place on early March 15 and that no major release occurred after March 24 which is also consistent with our results. The main releases could be linked to the explosion and the pressure suppression chamber damage of unit 2. However, our simulations suggest that the hydrogen explosions and/or venting operations of units 1 and 3 (March 12 and 13/14) may have also resulted in significant releases (Katata, 2012a). Because of the meteorological situation, these releases seem to have been rapidly blown towards the Pacific Ocean (especially the release of March 12) and may not have been significantly measured by the stations network in the vicinity of the NPP.
It was estimated that the total Cs-137 and I-131 releases emitted into the atmosphere during the Chernobyl accident were 8.5 × 1016 and 1.8 × 1018 Bq respectively. Concerning this event, the release of Cs-137 was estimated to be about 30 % of the core inventory and that of I-131 is estimated to be about 50 % (UNSCEAR). The source terms estimated in our study from long range simulations show that Fukushima releases could represent about 10 % of the Chernobyl emissions for these two isotopes.
The total estimated source term gives a fraction of core inventory of about 8 × 1018 Bq at the time of reactor shutdown. This result suggests that at least 80 % of the core inventory has been released into the atmosphere and indicates a broad meltdown of reactor core (see Part I of the publication). Total source term is in good agreement with literature (Mathieuet al., 2012). However our result is lower than in Stohlet al. (2012) where the authors found a total release from 12 × 1018 to 18 × 1018 Bq which is higher than the entire estimated Xe-133 inventory. According to the authors, a significant part of Xe-133 released could be due to I-133 decay.
This part of the publication (Part II) is dedicated to atmospheric transport modeling. Simulations were mainly carried out at global scale by considering Cs-137, I-131 and Xe-133 volatile fission products to assess the arrival time of radionuclides at different IMS stations located on the globe and to evaluate the quantities released into the atmosphere. These analyses are valuable to estimate the Fukushima reactor core damages (Part I of the publication) and to assess the monitoring capabilities of the CTBT network following the accident.
All simulations show that the cloud has been mainly dispersed in the northern hemisphere and air exchanges appear to be very low with the southern hemisphere. Regarding Xe-133, the total release is estimated to be of the order of 6 × 1018 Bq emitted during the explosions on units 1, 2 and 3. This result suggests that at least 80 % of the core inventory has been released into the atmosphere and indicates a broad meltdown of reactor core. Due to the large amount of radioxenon emitted into the atmosphere, levels of activity concentrations remained locally above 1 Bq m−3 for several weeks after the accident. Until early May, levels of activity concentrations remained significant throughout the whole northern hemisphere. Xe-133 concentrations due to the Fukushima Dai-ichi accident decreased following the half life of the radionuclide to be of the order of magnitude of industrial background from mid May 2011. Measurements by noble gas stations located in the northern hemisphere show that the situation had returned to near normal during June 2011. The evolution of measured Xe-131 m/Xe-133 ratios shows a clear signature of the Fukushima Dai-ichi accident over a long period of time (about 80 days). The knowledge of activity concentrations and of isotopic ratios over time were essential to maintain the monitoring capabilities of the CTBT network if another major event had arisen in the weeks following the accident.
Regarding Cs-137 and I-131, simulations were performed considering both JPP38 and distant IMS stations measurements. Atmospheric transport modeling results are in a reasonable agreement with measurements on most stations but appear poorer for stations located on the edge of the simulated plume and/or close to Fukushima NPP. Bias on transport may be due to the considered source terms and to the cumulative effect over long distances of inaccuracies in wind fields. Calculations suggest that the main air emissions have occurred on March 14 (explosion and pressure suppression chamber damage of unit 2) and that no major release occurred after March 23. The JPP38 station appeared to be mainly concerned by March 14 and March 20 and 21 releases (rise of pressure of unit 3). The hydrogen explosions and/or venting operations of units 1 and 3 (March 12 and 13/14) may have also resulted in significant releases. Because of the meteorological situation, these releases have been quickly blown towards the Pacific Ocean (especially the release of March 12) and may not have been significantly measured by the JPP38 station and the stations network located in the vicinity of the NPP. Total atmospheric releases of Cs-137 and I-131 aerosols estimated from long range simulation are found to be 1 × 1016 and 1 × 1017 Bq, respectively. By neglecting gas/particulate conversion phenomena, the total release of I-131 (gas + aerosol) could be 4 × 1017 Bq. Emissions estimated using JPP38 measurements are higher but uncertainties could be significant (total releases are ~5–7 larger regarding the long range results). The amounts of Cs-137 and I-131 emitted into the atmosphere during the Fukushima accident could represent 10 % of the Chernobyl accident releases (estimated to be 8.5 × 1016 Bq of Cs-137 and 1.8 × 1018 Bq of I-131).
The authors wish to thank the Comprehensive Nuclear-Test-Ban Treaty Organisation for fruitful discussions about the data collected by radionuclide stations of the International Monitoring Network during this event. A special thank you should be expressed to Harry Dupont of the Alten Company for his expertise and for the implementation of dispersion and mesoscale meteorological models.