Application of Bayesian machine learning for estimation of uncertainty in forecasted plume directions by atmospheric dispersion simulations

Kadowaki, Masanao; Nagai, Haruyasu; Yoshida, Toshiya*; Terada, Hiroaki; Tsuzuki, Katsunori; Sawa, Hiroki*

Journal of Nuclear Science and Technology, 60(10), p.1194 - 1207, 2023/10

https://doi.org/10.1080/00223131.2023.2177763

This study develops an estimation method using machine learning for uncertainty in forecasted plume directions. Bayesian machine learning was used in the machine learning approach. A three-day forecast simulation was conducted every day from 2015-2020, considering a hypothetical release of Cs from a nuclear facility to create training and test datasets for the machine learning. The findings reveal that the rate of good predictability was greater than 50% even in the forecast 36 h later when investigating the effectiveness of the Bayesian model on uncertainty estimation. Additionally, the frequency of miss prediction of higher uncertainty was low (0.9%-7.9%) throughout the forecast period. However, the rate of over-prediction of uncertainty increased with forecast time up to 31.2%, which is acceptable as a conservative estimation. These results show that the Bayesian model used in this study effectively estimates the uncertainty of plume directions predicted through atmospheric dispersion simulations.

Evaluation of uncertainties derived from meteorological forecast inputs in plume directions predicted by atmospheric dispersion simulations

Yoshida, Toshiya; Nagai, Haruyasu; Terada, Hiroaki; Tsuzuki, Katsunori; Sawa, Hiroki*

Journal of Nuclear Science and Technology, 59(1), p.55 - 66, 2022/01

https://doi.org/10.1080/00223131.2021.1951862

Atmospheric transport, dispersion, and deposition models (ATDMs) can support decision-making during nuclear emergencies; however, uncertainties in the ATDM results need to be carefully evaluated. To investigate the uncertainties derived from meteorological forecast inputs, we conducted three-day forecast simulations every day for one year with hypothetical releases of radionuclides (one-hour releases every 6 h) from a nuclear facility. The forecast outputs were compared with the analysis outputs during the same period. The difference between the outputs is treated as the uncertainty in the forecasts and is represented as an angle based on the discrepancy in the plume directions between the analysis and forecast outputs. Using meteorological inputs made by Japan Meteorological Agency, the discrepancy angle (Ang) increased by approximately 10 per day on an annual average basis. Meanwhile, the Ang values were occasionally 4-5 times higher than the annual average during short time periods. Since the Ang time series show seasonal and diurnal changes, the statistical characteristics likely depend on the geographical and meteorological conditions, as well as the types of meteorological inputs. Additionally, a main factor in the uncertainty is the wind-direction difference between the analysis and forecast outputs on scales of more than or less than 100 km.

Validity of the source term for the Fukushima Dai-ichi Nuclear Power Station accident estimated using local-scale atmospheric dispersion simulations to reproduce the large-scale atmospheric dispersion of Cs

Kadowaki, Masanao; Furuno, Akiko; Nagai, Haruyasu; Kawamura, Hideyuki; Terada, Hiroaki; Tsuzuki, Katsunori; El-Asaad, H.

Journal of Environmental Radioactivity, 237, p.106704_1 - 106704_18, 2021/10

https://doi.org/10.1016/j.jenvrad.2021.106704

The source term of Cs for the Fukushima Daiichi Nuclear Power Station (FDNPS) accident was estimated from the results of local-scale atmospheric dispersion simulations and measurements. To confirm the source term's validity for reproducing the large-scale atmospheric dispersion of Cs, this study conducted hemispheric-scale atmospheric and oceanic dispersion simulations. In the dispersion simulations, the atmospheric-dispersion database system Worldwide version of System for Prediction of Environmental Emergency Dose Information (WSPEEDI)-DB and oceanic dispersion model SEA-GEARN-FDM were used. Compared with the air concentrations of Cs measured by the Comprehensive Nuclear-Test-Ban Treaty Organization, overall, the WSPEEDI-DB simulation reproduced the measurements with some overestimation. Furthermore, the deposition amounts of Cs was investigated using concentrations of Cs in seawater. The simulated seawater concentrations of Cs were underestimated regionally in the North Pacific. The overestimation and underestimation could be improved without contradiction between the air and seawater concentrations of Cs using more realistic precipitation in atmospheric dispersion simulations. This shows that the source term validated in this study could reproduce the spatiotemporal distribution of Cs because of the FDNPS accident in both local and large-scale atmospheric dispersion simulations.

Atmospheric-dispersion database system that can immediately provide calculation results for various source term and meteorological conditions

Terada, Hiroaki; Nagai, Haruyasu; Tanaka, Atsunori*; Tsuzuki, Katsunori; Kadowaki, Masanao

Journal of Nuclear Science and Technology, 57(6), p.745 - 754, 2020/06

https://doi.org/10.1080/00223131.2019.1709994

We have estimated source term and analyzed processes of atmospheric dispersion of radioactive materials released during the Fukushima Daiichi Nuclear Power Station (FDNPS) accident by the Worldwide version of System for Environmental Emergency Dose Information. On the basis of this experience, we developed an dispersion calculation method that can respond to various needs in a nuclear emergency and provide useful information for emergency-response planning. By this method, if a release point is known, it is possible to immediately obtain the prediction results by applying provided source term to the database of dispersion-calculation results prepared in advance. With this function, it is easy to compare results by applying various source term with monitoring data, and to find out the optimum source term, which was applied for the source term estimation of the FDNPS accident. By performing this calculation with past meteorological-analysis data, it is possible to immediately get dispersion-calculation results for various source term and meteorological conditions. This database can be used for pre-accident planning, such as optimization of a monitoring plan and understanding of events to be supposed in considering emergency countermeasures.

Activities on predictions of atmospheric dispersion of radionuclides for nuclear tests by North Korea

Ishizaki, Shuhei; Hayakawa, Tsuyoshi; Tsuzuki, Katsunori; Terada, Hiroaki; Togawa, Orihiko

JAEA-Technology 2018-007, 43 Pages, 2018/10

JAEA-Technology-2018-007.pdf:5.67MB

When North Korea has carried out a nuclear test, by a request from Nuclear Regulation Authority (NRA), Nuclear Emergency Assistance and Training Center (NEAT) predicts atmospheric dispersion of radionuclides by WSPEEDI-II system in cooperation with Nuclear Science and Engineering Center (NSEC), and submits the predicted results to NRA as the activity to assist responses by the Japanese Government. This report explains frameworks of the Japanese Government and Japan Atomic Energy Agency (JAEA) to cope with nuclear tests by North Korea, and describes a series of activities by NEAT regarding predictions of atmospheric dispersion of radionuclides in response to the 5th and 6th nuclear tests carried out by North Korea in September 2016 and September 2017. Future plans and issues to be solved for responses to nuclear tests are also described in this report, together with an outline of a computer program system used in the predictions.

Sources of Cs fluvial export from a forest catchment evaluated by stable carbon and nitrogen isotopic characterization of organic matter

Muto, Kotomi; Atarashi-Andoh, Mariko; Koarashi, Jun; Takeuchi, Erina; Nishimura, Shusaku; Tsuzuki, Katsunori; Matsunaga, Takeshi*

Journal of Radioanalytical and Nuclear Chemistry, 314(1), p.403 - 411, 2017/10

https://doi.org/10.1007/s10967-017-5350-7

Fluvial export of particulate and dissolved Cs was investigated to reveal its sources and transfer mechanisms in a broadleaved forest catchment using a continuous collection system. The finest size fraction ( 75m), consisting of decomposed litter and surface mineral soil, was the dominant fraction in the particulate Cs load, although the contribution of coarser size fractions increased during high water discharge in 2014. The dissolved Cs originated from the decomposition of Cs-contaminated litter. Temporal changes in Cs distribution in the litter-mineral soil system indicated that the dissolved Cs load will be moderated in several years, while particulate Cs load has the potential to continue for a long time.

Year-round variations in the fluvial transport load of particulate Cs in a forested catchment affected by the Fukushima Daiichi Nuclear Power Plant accident

Matsunaga, Takeshi; Nakanishi, Takahiro; Atarashi-Andoh, Mariko; Takeuchi, Erina; Muto, Kotomi; Tsuzuki, Katsunori; Nishimura, Shusaku; Koarashi, Jun; Otosaka, Shigeyoshi; Sato, Tsutomu*; et al.

Journal of Radioanalytical and Nuclear Chemistry, 310(2), p.679 - 693, 2016/11

AA2015-0821.pdf:3.78MB

https://doi.org/10.1007/s10967-016-4840-3

Particulate Cs in stream water was collected continuously for two years in order to assess the long-term trend of the Cs discharge from the forest environment. Sampling was conducted from December 2011 to December 2013 in a mountainous stream, which received the Cs from the Fukushima Daiichi Nuclear Power Plant accident. A seasonal increase in fluvial transport load of particulate Cs associated with suspended solids (SS) was observed in August and September when rainfall was abundant. The particulate Cs concentration decreased at a faster rate than the rate due to radioactive decay. This decrease might be resulted from redistribution of the easily eroded and polluted soil surface due to heavy rain events such as typhoons. These findings indicate that the particulate Cs load was subject to the inter-annual variations in rainfalls, and decreased gradually over a long period of time due to a decrease in Cs concentration in SS.

Catchment-scale distribution of radiocesium air dose rate in a mountainous deciduous forest and its relation to topography

Atarashi-Andoh, Mariko; Koarashi, Jun; Takeuchi, Erina; Tsuzuki, Katsunori; Nishimura, Shusaku; Matsunaga, Takeshi

Journal of Environmental Radioactivity, 147, p.1 - 7, 2015/09

AA2014-0585.pdf:2.18MB

https://doi.org/10.1016/j.jenvrad.2015.05.004

We collected a large amount of radiocesium air dose rate data by mountain-walking with a small -ray survey system, KURAMA-II, to create an air dose rate map of a mountainous deciduous forest that received radiocesium from the Fukushima Dai-ichi Nuclear Power Plant accident. Measurements were conducted in a small stream catchment 0.6 km in size in August and September 2013, and the relationship between air dose rates and the mountainous topography was examined. Air dose rates increased with elevation, suggesting that more radiocesium was deposited on ridges, and that it had remained there for 2.5 years with no significant migration due to soil erosion or water drainage. Slope aspect also strongly affected air dose rates. By the continuous measurement using KURAMA-II, we describe the variation in air dose rates in a mountainous area and suggest that it is important to consider topography when selecting sampling points to estimating dose rates or contaminant deposition.

A Passive collection system for whole size fractions in river suspended solids

Matsunaga, Takeshi; Nakanishi, Takahiro; Atarashi-Andoh, Mariko; Takeuchi, Erina; Tsuzuki, Katsunori; Nishimura, Shusaku; Koarashi, Jun; Otosaka, Shigeyoshi; Sato, Tsutomu*; Nagao, Seiya*

Journal of Radioanalytical and Nuclear Chemistry, 303(2), p.1291 - 1295, 2015/02

https://doi.org/10.1007/s10967-014-3491-5

An innovative, yet simple method for the passive collection of radioactive materials in river water has been developed and validated. The method employes large filter vessels, containing multiple cartridge filters. River water is led to the system naturally using a drop of the riverbed by hose from upstream. This method makes long-term, unmanned monitoring possible. In addition to regular radioactivity analyses, this method provides an opportunity for the characterization of suspended materials based on its ample collection quantities (more than several tens of grams). This method may also be applicable to sediment-bound chemicals.

The Establishment of the framework and actual experience for the prediction of atmospheric dispersion of radionuclides against the nuclear test by North Korea

Nakanishi, Chika; Sato, Takeshi; Sato, Sohei; Nagai, Haruyasu; Kakefuda, Toyokazu; Katata, Genki; Tsuzuki, Katsunori; Ikeda, Takeshi; Okuno, Hiroshi; Yamamoto, Kazuya; et al.

JAEA-Technology 2013-030, 105 Pages, 2013/10

JAEA-Technology-2013-030.pdf:29.0MB

North Korea carried out the third nuclear test in February 2013. Due to the request of the Ministry of Education, Culture, Sports, Science and Technology (MEXT), Nuclear Emergency Assistance and Training Center (NEAT) and Nuclear Science and Engineering Directorate (NSED) of JAEA predicted the atmospheric dispersion of radionuclide by WSPEEDI-II for the purpose of contributing to the environmental monitoring plan. From February 12 to 22, they provided daily reports on the prediction to the MEXT and the Ministry of Defense. MEXT has published these reports on the website. Since April 2012, NEAT and NSED had prepared to predict by the framework for the prediction around the clock during 10months until February 2013. This report described this experience and pointed issues out on this system.