Article pubs.acs.org/est
Episode Analysis of Deposition of Radiocesium from the Fukushima Daiichi Nuclear Power Plant Accident Yu Morino,* Toshimasa Ohara, Mirai Watanabe, Seiji Hayashi, and Masato Nishizawa Center for Regional Environment Research, National Institute for Environmental Studies, 16-2, Onogawa, Tsukuba, Ibaraki, 305-8506, Japan S Supporting Information *
ABSTRACT: Chemical transport models played key roles in understanding the atmospheric behaviors and deposition patterns of radioactive materials emitted from the Fukushima Daiichi nuclear power plant after the nuclear accident that accompanied the great Tohoku earthquake and tsunami on 11 March 2011. However, model results could not be sufficiently evaluated because of limited observational data. We assess the model performance to simulate the deposition patterns of radiocesium (137Cs) by making use of airborne monitoring survey data for the first time. We conducted ten sensitivity simulations to evaluate the atmospheric model uncertainties associated with key model settings including emission data and wet deposition modules. We found that simulation using emissions estimated with a regional-scale (∼500 km) model better reproduced the observed 137Cs deposition pattern in eastern Japan than simulation using emissions estimated with local-scale (∼50 km) or global-scale models. In addition, simulation using a process-based wet deposition module reproduced the observations well, whereas simulation using scavenging coefficients showed large uncertainties associated with empirical parameters. The best-available simulation reproduced the observed 137Cs deposition rates in high-deposition areas (≥10 kBq m−2) within 1 order of magnitude and showed that deposition of radiocesium over land occurred predominantly during 15−16, 20−23, and 30−31 March 2011.
■
INTRODUCTION Enormous quantities of radionuclides were released into the atmosphere after the nuclear accident at the Fukushima Daiichi nuclear power plant (FDNPP) on 11 March 2011.1−3 To estimate the atmospheric behavior of the radionuclides, particularly iodine-131 (131I), and cesium-137 (137Cs), atmospheric modeling studies were conducted over local,4 regional,5−7 and global2,8 scales. Observational data played a critical role in the assessment of model performance. Previous model simulations used data primarily from surface monitoring of atmospheric concentrations and deposition. However, monitoring of atmospheric deposition did not start until 18 March 2011, and thus model performance could not be evaluated for the period before 18 March, the period when the largest radiocesium emissions presumably occurred.1 Recently, data from an airborne monitoring survey conducted by the Ministry of Education, Culture, Sports, Science, and Technology became available.9 In this study, we made the first use of these data to assess the performance of models of radiocesium deposition patterns. The airborne survey data have two advantages for the evaluation of chemical transport models. One is the high spatial resolution. Combined analysis of this highly resolved observational data with modeling results helps to reveal the detailed mechanism of radiocesium deposition. The other advantage is the wide data © 2013 American Chemical Society
coverage over eastern Japan. Total radiocesium deposition over land in Japan can be estimated from the airborne monitoring data, and the fact that the simulated total deposition of radiocesium over land can be validated adds credibility to the 137 Cs budget analysis. In addition, by using the high-resolution observational data, we can assess the suitability of different model settings for critical modules. Because there was some uncertainty in selecting model settings (e.g., emission and wet deposition), we conducted nine sensitivity simulations and assessed the model performance. These simulations gave us insight into uncertainties and plausible model settings for simulation of radiocesium deposition, although a comprehensive uncertainty analysis is beyond the scope of this study. We then used the validated model to assess the radiocesium budget and deposition mechanisms. Budget analysis and episode analysis had previously been reported.5−7 However, the simulated deposition patterns of these analyses were not verified, owing to lack of observational data. Our analysis explains the mechanism by which the observed high-deposition Received: Revised: Accepted: Published: 2314
November 12, 2012 February 1, 2013 February 7, 2013 February 7, 2013 dx.doi.org/10.1021/es304620x | Environ. Sci. Technol. 2013, 47, 2314−2322
Environmental Science & Technology
Article
areas formed, and thus it may be the basis for future studies on radiocesium deposition.
although the horizontal resolution of the model used in this study (3 km) was finer than that in our previous study (6 km), and emission data were updated to the latest estimate from the Japan Atomic Energy Agency (JAEA).7 We conducted simulations with three sets of emission data: the estimates from JAEA,7 the Norwegian Institute for Air Research (NILU),2 and Tokyo Electric Power Company (TEPCO)3 (Figure S2 of the Supporting Information). All three emission estimates are based on inversion methods using simulation models and observational data. The JAEA analysis combined local- and regional-scale models, whereas NILU used a global-scale model, and TEPCO used a local-scale model. The model used for the JAEA estimate used a mesh size of 3 km, and that of TEPCO was 1 km. NILU used an objective inversion, whereas JAEA and TEPCO estimated release rates from a combination of observational data and atmospheric simulations under the assumption of a unit release rate. JAEA first used observed concentrations of radiocesium in air at 10 measurement sites1 and then modified their estimate by using the air dose rate at three monitoring sites in Fukushima Prefecture4 and surface deposition rates at 19 monitoring sites over eastern Japan.7 NILU used air concentrations at 45 monitoring stations (6 in Japan, 5 in the northern Pacific Ocean, 12 on the North American continent, and 12 in Western Europe) and surface deposition rates at 46 monitoring stations over Japan7 and in Tokai-mura in Ibaraki Prefecture. TEPCO used the air dose rate measured from a monitoring car that moved around the FDNPP, and estimated emission rates of 131I, 134Cs, 137Cs, and noble gases by assuming emission ratios for the respective nuclides. TEPCO estimated emission rates during 12−31 March 2011, and thus the simulation period for the EM3 simulation is 10−31 March. We also compared three wet deposition settings. The wet deposition modules are described in detail in section S1 of the Supporting Information and briefly described below. In CMAQ v4.6, wet deposition rates of accumulation-mode aerosols are calculated by considering washout time, which is calculated from the ratio of the water content of precipitation and that of clouds.11 The wet deposition module is process-based, and wet deposition amounts of aerosols calculated with CMAQ have been validated in several previous studies.12,13 We also conducted a simulation with the wet deposition module of the JAEA model (WD2 case).7 In that model, wet deposition rates are calculated using a scavenging coefficient (Λ), which is a function of the precipitation rate. This wet deposition module is an empirical module with fitting parameters included. Simulation with Λ of the JAEA model multiplied by a factor of 10 (WD3 case) was also conducted as shown later. Because the observed diameter of particulate radiocesium differs among studies,14 in our previous study we set the mean diameter and standard deviation to 1 μm and 1.1, respectively.5 Recently, observed size distributions of radiocesium have become available.15 Activity size distributions of 134Cs and 137 Cs in aerosols were measured at Tsukuba, a city 170 km southwest of the FDNPP, during 28 April−12 May and 12−26 May 2011. Means and standard deviations of 137Cs aerosol diameters during the two periods were derived after the data were fit to log-normal distributions. We used the derived mean diameter (0.65 μm) and standard deviation (1.35) in the sensitivity simulation (DD2 case). Aerosol size distributions change during transport because deposition rates differ by particle size. However, we did not consider this change in this simulation.
■
METHODOLOGY Simulation Models. We simulated distributions of 137Cs by using the Weather Forecast and Research Model version 3.110 and a 3D chemical transport model, Models-3 Community Multiscale Air Quality (CMAQ),11 for the period from 10 March to 20 April 2011. We conducted one standard-case simulation (STD) and nine sensitivity simulations as summarized in Table 1. The model domain is shown in Figure Table 1. Setup Parameters Used for Ten Model Simulations simulation STD EM2 EM3 WD2 E2W2 E3W2 WD3 E2W3 E3W3 DD2
emissionsa 7
JAEA NILU2 TEPCO3 JAEA NILU TEPCO JAEA NILU TEPCO JAEA
wet depositionb
particle diameter
11
1 μm 1 μm 1 μm 1 μm 1 μm 1 μm 1 μm 1 μm 1 μm Kaneyasu et al.15
CMAQ CMAQ CMAQ Scav. coeff.7 Scav. coeff. Scav. coeff. Scav. coeff. × 10 Scav. coeff. × 10 Scav. coeff. × 10 CMAQ
a
JAEA, Japan Atomic Energy Agency; NILU, the Norwegian Institute for Air Research; TEPCO, Tokyo Electric Power Company. bCMAQ, Community Multiscale Air Quality; Scav. coeff., scavenging coefficient.
1. The prefectures are indicated by numbers in Figure 1 (e.g., P1 for Iwate Prefecture and P5 for Fukushima Prefecture). The basic model settings are described in our previous article,5
Figure 1. Model domain used in the Community Multiscale Air Quality (CMAQ) simulation. Numbered prefectures: 1, Iwate; 2, Akita; 3, Yamagata; 4, Miyagi; 5, Fukushima; 6, Ibaraki; 7, Tochigi; 8, Gunma; 9, Chiba; 10, Saitama; 11, Tokyo; 12, Kanagawa; 13, Shizuoka; 14, Yamanashi; 15, Nagano; 16, Niigata. The white square indicates the site of the Fukushima Daiichi nuclear power plant (FDNPP), and the white cross indicates Mount Tsukuba. 2315
dx.doi.org/10.1021/es304620x | Environ. Sci. Technol. 2013, 47, 2314−2322
Environmental Science & Technology
Article
Figure 2. Observed (Obs) and simulated (Model) 137Cs deposition rates. Upper panels show deposition rates averaged over 18 March−20 April (18−31 March for EM3) at 15 surface monitoring sites (Figure 1). Both total and wet deposition rates are indicated because the deposition rates measured with bulk samplers were assumed to fall between the wet deposition rates and the total deposition rates, as noted in the text. Middle panels show daily deposition rates over the 15 surface monitoring sites. Lower panels show the comparison between 137Cs total deposition on model grids as determined from airborne monitoring and model simulation.
Observational Data. Daily deposition rates of 137Cs were monitored with bulk samplers over 46 Japanese prefectures starting on 18 March 2011.16 We assumed that the deposition rates measured with bulk samplers were between the wet deposition rates and the total deposition rates (i.e., dry plus wet) as in our previous study.5 As radiocesium deposition was dominated by wet processes, we believe that uncertainties due to this assumption are small. In addition, we used data from the airborne monitoring survey to evaluate model performance; details of the survey methodology are available elsewhere,9 and a brief description is given in section S2 of the Supporting Information. Note that
the airborne measurements over eastern Japan were conducted from June to November 2011, and thus these data cannot be directly compared with simulated deposition during March− April 2011. However, the measured amount of deposited radiocesium decreased by about 1.8% for reasons other than physical attenuation between 31 May−2 July and 22 October− 5 November 2011.9 In addition, radiocesium discharge was estimated to be small (0.3% of deposited 137Cs) in a forested catchment on Mount Tsukuba (Figure 1) over the year after the accident, as detailed elsewhere.17 Although no such budget studies have been conducted in other areas, the radiocesium discharge from a forest is not expected to be large. These 2316
dx.doi.org/10.1021/es304620x | Environ. Sci. Technol. 2013, 47, 2314−2322
Environmental Science & Technology
Article
Table 2. Comparison between Base-Case (STD) and Sensitivity Simulations; FA2 and FA10 Are the Proportions of Simulated Data That Reproduce the Observations within a Factor of 2 or 10, respectively; r and n Are Correlation Coefficient and Number of Data Points, Respectively (a) Comparison with surface monitoring (cutoffa: 1 kBq m−2 day−1) STD FA2 (%) FA10 (%) n STD FA2 (%) FA10 (%) r n STD FA2 (%) FA10 (%) r
FA2 (%) FA10 (%)
EM2
WD2
WD3
WD3
EM2 EM3 WD2 57.0 34.6 40.0 44.9 95.6 87.8 88.9 98.7 0.663 0.526 0.308 0.639 (d) Comparison with airborne monitoring (cutoffb: 1 kBq m−2, n = 8314)c
WD3
STD 28.6 65.6
EM2 23.8 64.7
8.79 2.21 2.16 1.81 1.75 4.75
EM3 24.4 60.7 (e) Budget analysis (PBq)
EM2 36.63 4.98 4.87 3.48 3.24 28.09
EM3 10.04 0.95 0.91 1.63 1.52 7.46
WD2 31.5 70.4 WD2
0.0 75.0 8 DD2
9.3 27.8 0.777 54
5.6 25.9 0.844 54 DD2
54.9 99.6 0.720
56.8 95.5 0.663
WD3 30.4 69.6 WD3
8.79 2.03 1.97 2.22 2.12 4.50
DD2 37.5 62.5 8
EM2 EM3 WD2 5.6 7.4 5.6 13.0 25.9 37.0 33.3 33.3 0.843 0.441 0.280 0.659 54 54 36 54 (c) Comparison with airborne monitoring (cutoffb: 10 kBq m−2, n = 2448)
STD emission total deposition over land wet deposition over land total deposition over ocean wet deposition over ocean outflow from the domain
EM3
0.0 0.0 25.0 37.5 75.0 37.5 75.0 62.5 8 8 8 8 (b) Comparison with surface monitoring (cutoffa: 0.1 kBq m−2 day−1)
DD2 28.3 65.5 Obsd
DD2 8.79 3.19 3.14 3.00 2.94 2.58
8.79 2.19 2.17 1.80 1.75 4.78
2.40
a
Minimum cutoff of observed daily deposition rates. bMinimum cutoff of model grid with observed deposition rates. cCorrelation coefficients are not shown as they are almost identical with those in part c of Table 2. dData from airborne monitoring survey.
EM3 reproduced the observations (≥1 kBq m−2 day−1) within a factor of 10 for 75%, 38%, and 75%, respectively (part a of Table 2). By contrast, for observed depositions higher than 0.1 kBq m−2 day−1, the proportion of observations reproduced by simulations within a factor of 2 (FA2) or 10 (FA10) were higher in the EM2 case than in the STD case (part b of Table 2), because the EM2 case reproduced the observations better in April suggesting that the JAEA analysis underestimated the 137 Cs emission rates in April. For low-deposition cases (
