Inverse modeling of the Chernobyl source term using atmospheric concentration and deposition measurements

Abstract

Abstract. This paper describes the results of an inverse modeling study for the determination of the source term of the radionuclides 134 Cs, 137 Cs and 131 I released after the Chernobyl accident. The accident occurred on 26 April 1986 in the Former Soviet Union and released about 10 19 Bq of radioactive materials that were transported as far away as the USA and Japan. Thereafter, several attempts to assess the magnitude of the emissions were made that were based on the knowledge of the core inventory and the levels of the spent fuel. More recently, when modeling tools were further developed, inverse modeling techniques were applied to the Chernobyl case for source term quantification. However, because radioactivity is a sensitive topic for the public and attracts a lot of attention, high-quality measurements, which are essential for inverse modeling, were not made available except for a few sparse activity concentration measurements far from the source and far from the main direction of the radioactive fallout.For the first time, we apply Bayesian inversion of the Chernobyl source term using not only activity concentrations but also deposition measurements from the most recent public data set. These observations refer to a data rescue attempt that started more than 10 years ago, with a final goal to provide available measurements to anyone interested. In regards to our inverse modeling results, emissions of 134 Cs were estimated to be 80 PBq or 30-50 % higher than what was previously published. From the released amount of 134 Cs, about 70 PBq were deposited all over Europe. Similar to 134 Cs, emissions of 137 Cs were estimated as 86 PBq, on the same order as previously reported results. Finally, 131 I emissions of 1365 PBq were found, which are about 10 % less than the prior total releases.The inversion pushes the injection heights of the three radionuclides to higher altitudes (up to about 3 km) than previously assumed (≈ 2.2 km) in order to better match both concentration and deposition observations over Europe. The results of the present inversion were confirmed using an independent Eulerian model, for which deposition patterns were also improved when using the estimated posterior releases. Although the independent model tends to underestimate deposition in countries that are not in the main direction of the plume, it reproduces country levels of deposition very efficiently. The results were also tested for robustness against different setups of the inversion through sensitivity runs. The source term data from this study are publicly available.

Source term estimation of multi‐specie atmospheric release of radiation from gamma dose rates

“…Note that temporal profiles from the CO_LASSO and CO_Tikhonov methods are similar in the middle of the studied period while CO_LASSO estimates are higher in the beginning and at the end of the studied period. The proposed method (blue line with 95% uncertainty bounds using gray color in Figure ) estimated the total activity of 79 ± 14 PBq which is in agreement with the consensus as well as with recently published results by (Liu et al ) with range 70–130 PBq and with results by Evangeliou et al () with estimated total emission 86 ± 5 PBq. The source term was underestimated in the middle of the 10‐day period and overestimated at the end of the studied period.…”

“…Another significant difference from (Evangeliou et al ) is the use of the prior source term (also known as the first guess). In (Evangeliou et al ), the prior source term was calculated from six previously published Chernobyl source terms, (Brandt et al ), (Persson et al ), (Izrael et al ), (Abagyan et al ), (Talerko, ), and from (Talerko, ), as their mean emission estimate, and then used in inversion scheme. In the proposed method, we do not assume any prior information on the source term other than the ratios.…”

“…For this purpose, we use the recently published dataset (Evangeliou et al, 2016) that consists of 12,281 observations for Cs-137 in the form of both, deposition observations (10,682) and atmospheric concentrations (1,599). The same experimental setup (with exceptions described bellow) was used as by Evangeliou et al (2017).…”

“…(Tichý et al, 2016). The application area range from nuclear explosions (Issartel and Baverel, 2003) and accidents such as Chernobyl (Davoine and Bocquet, 2007;Evangeliou et al, 2017) or Fukushima (Winiarek et al, 2012;Stohl et al, 2012;Liu et al, 2017), accidental release of radioactive materials (Tichý et al, 2017), to emissions of volcanic ash during volcanic eruptions (Kristiansen et al, 2010;Stohl et al, 2011). These methods were mainly tested in scenarios where the measurements correspond to one specie in the source term, such as concentration measurements of individual nuclides.…”

“…Real data sets, such as the Chernobyl and the Fukushima accidents, do not provide ground truth of the source term (Aliyu et al, 2015), and most of the measurements is nuclide-specific (Evangeliou et al, 2016). However, the Chernobyl source term is also composed from many species, since the nuclides are attached to different particles sizes (Evangeliou et al, 2017). The fractions of different particle sizes in the total release (and thus their ratios) are approximately known from measurements (Malá et al, 2013).…”