Uncertainty quantification of pollutant source retrieval: comparison of Bayesian methods with application to the Chernobyl and Fukushima Daiichi accidental releases of radionuclides

Inverse modelling of the emissions of atmospheric species and pollutants has significantly progressed over the past 15 years. However, in spite of seemingly reliable estimates, the retrievals are rarely accompanied by an objective estimate of their uncertainty, except when Gaussian statistics are assumed for the errors, which is often an unrealistic assumption. Here, we assess rigorous techniques meant to compute this uncertainty in the context of the inverse modelling of the time emission rates – the so-called source term – of a point-wise atmospheric tracer. Log-normal statistics are used for the positive source term prior and possibly the observation errors; this precludes simple Gaussian statistics-based solutions. Firstly, through the so-called empirical Bayesian approach, parameters of the error statistics – the hyperparameters – are first estimated by maximizing their likelihood via an expectation–maximization algorithm. This enables a robust estimation of a source term. Then, the uncertainties attached to the retrieved source rates and total emission are estimated using four Monte Carlo techniques: (i) an importance sampling based on a Laplace proposal, (ii) a naive randomize-then-optimize (RTO) sampling approach, (iii) an unbiased RTO sampling approach, and (iv) a basic Markov chain Monte Carlo (MCMC) simulation. Secondly, these methods are compared to a more thorough hierarchical Bayesian approach, using an MCMC based on a transdimensional representation of the source term to reduce the computational cost. Those methods, and improvements thereof, are applied to the estimation of the atmospheric caesium-137 source terms from the Chernobyl nuclear power plant accident in April and May 1986 and Fukushima Daiichi nuclear power plant accident in March 2011. This study provides the first consistent and rigorous quantification of the uncertainty of these best estimates.

[1]  B. Welford.  Note on a Method for Calculating Corrected Sums of Squares and Products , 1962 .

[2]  J. F. Meirink,et al.  Inverse Modeling of Global and Regional CH4 Emissions Using SCIAMACHY Satellite Retrievals , 2009 .

[3]  Takuya Nagae,et al.  Estimation of release rate of iodine-131 and cesium-137 from the Fukushima Daiichi nuclear power plant , 2013 .

[4]  Marc Bocquet,et al.  An inverse modeling method to assess the source term of the Fukushima Nuclear Power Plant accident using gamma dose rate observations , 2013 .

[5]  M. Sambridge,et al.  Seismic tomography with the reversible jump algorithm , 2009 .

[6]  Janusz A. Pudykiewicz,et al.  Simulation of the Chernobyl dispersion with a 3-D hemispheric tracer model , 1989 .

[7]  Jens Mühle,et al.  Characterization of uncertainties in atmospheric trace gas inversions using hierarchical Bayesian methods , 2013 .

[8]  Yang Zhang,et al.  Real-time air quality forecasting, part II: State of the science, current research needs, and future prospects , 2012 .

[9]  G. C. Wei,et al.  A Monte Carlo Implementation of the EM Algorithm and the Poor Man's Data Augmentation Algorithms , 1990 .

[10]  A. Sterl,et al.  The ERA‐40 re‐analysis , 2005 .

[11]  G. Roberts,et al.  MCMC Methods for Functions: ModifyingOld Algorithms to Make Them Faster , 2012, 1202.0709.

[12]  Marc Bocquet,et al.  Ensemble Kalman filtering without the intrinsic need for inflation , 2011 .

[13]  Haruyasu Nagai,et al.  Source term estimation of atmospheric release due to the Fukushima Dai-ichi Nuclear Power Plant accident by atmospheric and oceanic dispersion simulations , 2013 .

[14]  Hendrik Elbern,et al.  Emission rate and chemical state estimation by 4-dimensional variational inversion , 2007 .

[15]  G. Ueno,et al.  Bayesian estimation of the observation‐error covariance matrix in ensemble‐based filters , 2016 .

[16]  Gerhard Wotawa,et al.  Xenon-133 and caesium-137 releases into the atmosphere from the Fukushima Dai-ichi nuclear power plant: determination of the source term, atmospheric dispersion, and deposition , 2011 .

[17]  Marc Bocquet,et al.  Estimation of the caesium-137 source term from the Fukushima Daiichi nuclear power plant using a consistent joint assimilation of air concentration and deposition observations , 2014 .

[18]  Haruyasu Nagai,et al.  Atmospheric discharge and dispersion of radionuclides during the Fukushima Dai-ichi Nuclear Power Plant accident. Part II: verification of the source term and analysis of regional-scale atmospheric dispersion. , 2012, Journal of environmental radioactivity.

[19]  N. Talerko.  Mesoscale modelling of radioactive contamination formation in Ukraine caused by the Chernobyl accident. , 2005, Journal of environmental radioactivity.

[20]  Eugene Yee,et al.  Theory for Reconstruction of an Unknown Number of Contaminant Sources using Probabilistic Inference , 2008 .

[21]  N. Metropolis,et al.  Equation of State Calculations by Fast Computing Machines , 1953, Resonance.

[22]  Vivien Mallet,et al.  Inverse modeling of NOx emissions at regional scale over northern France: Preliminary investigation of the second‐order sensitivity , 2005 .

[23]  Manuel Pulido,et al.  Offline parameter estimation using EnKF and maximum likelihood error covariance estimates: Application to a subgrid‐scale orography parametrization , 2013 .

[24]  Kevin R. Gurney,et al.  Maximum likelihood estimation of covariance parameters for Bayesian atmospheric trace gas surface flux inversions , 2005 .

[25]  P. Green.  Reversible jump Markov chain Monte Carlo computation and Bayesian model determination , 1995 .

[26]  Marc Bocquet,et al.  Accounting for representativeness errors in the inversion of atmospheric constituent emissions: application to the retrieval of regional carbon monoxide fluxes , 2012 .

[27]  Heikki Haario,et al.  Randomize-Then-Optimize: A Method for Sampling from Posterior Distributions in Nonlinear Inverse Problems , 2014, SIAM J. Sci. Comput..

[28]  Marc Bocquet,et al.  Parameter‐field estimation for atmospheric dispersion: application to the Chernobyl accident using 4D‐Var , 2012 .

[29]  A. Stohl,et al.  Reconstructing the Chernobyl Nuclear Power Plant (CNPP) accident 30 years after. A unique database of air concentration and deposition measurements over Europe. , 2016, Environmental pollution.

[30]  H. Yamazawa,et al.  Preliminary Estimation of Release Amounts of 131I and 137Cs Accidentally Discharged from the Fukushima Daiichi Nuclear Power Plant into the Atmosphere , 2011 .

[31]  Nagatomo Nakamura,et al.  Iterative algorithm for maximum‐likelihood estimation of the observation‐error covariance matrix for ensemble‐based filters , 2014 .

[32]  M. Bocquet,et al.  Estimation of volatile organic compound emissions for Europe using data assimilation , 2012 .

[33]  R. Lange,et al.  Chernobyl source term, atmospheric dispersion, and dose estimation. , 1988, Health physics.

[34]  T. Nakajima,et al.  First retrieval of hourly atmospheric radionuclides just after the Fukushima accident by analyzing filter-tapes of operational air pollution monitoring stations , 2014, Scientific Reports.

[35]  Marc Bocquet,et al.  Inverse modelling-based reconstruction of the Chernobyl source term available for long-range transport , 2007 .

[36]  A. Mathieu,et al.  Using the Wasserstein distance to compare fields of pollutants: application to the radionuclide atmospheric dispersion of the Fukushima-Daiichi accident , 2016 .

[37]  J. Moncrieff,et al.  Towards better error statistics for atmospheric inversions of methane surface fluxes , 2013 .

[38]  Haruyasu Nagai,et al.  Atmospheric discharge and dispersion of radionuclides during the Fukushima Dai-ichi Nuclear Power Plant accident. Part I: Source term estimation and local-scale atmospheric dispersion in early phase of the accident. , 2012, Journal of environmental radioactivity.

[39]  Thomas Kaminski,et al.  On aggregation errors in atmospheric transport inversions , 2001 .

[40]  K. Davis,et al.  Hyperparameter estimation for uncertainty quantification in mesoscale carbon dioxide inversions , 2013 .

[41]  Marc Bocquet,et al.  Assessment of the amount of cesium‐137 released into the Pacific Ocean after the Fukushima accident and analysis of its dispersion in Japanese coastal waters , 2012 .

[42]  Massimo Bonavita,et al.  On the use of EDA background error variances in the ECMWF 4D‐Var , 2012 .

[43]  A. Manning,et al.  Estimation of trace gas fluxes with objectively determined basis functions using reversible jump Markov Chain Monte Carlo , 2016 .

[44]  Alberto Malinverno,et al.  Expanded uncertainty quantification in inverse problems: Hierarchical Bayes and empirical Bayes , 2004 .

[45]  J. Louis.  A parametric model of vertical eddy fluxes in the atmosphere , 1979 .

[46]  Markus Reichstein,et al.  On the assignment of prior errors in Bayesian inversions of CO2 surface fluxes , 2006 .

[47]  Steven J. Fletcher,et al.  A data assimilation method for log‐normally distributed observational errors , 2006 .

[48]  Marc Bocquet,et al.  Source reconstruction of an accidental radionuclide release at European scale , 2007 .

[49]  Lin Wu,et al.  Technical Note: The air quality modeling system Polyphemus , 2007 .

[50]  Radek Hofman,et al.  LS-APC v1.0: a tuning-free method for the linear inverse problem and its application to source-term determination , 2016 .

[51]  Y. Roustan,et al.  Hints to discriminate the choice of wet deposition models applied to an accidental radioactive release , 2015 .

[52]  Taro Takahashi,et al.  Towards robust regional estimates of CO2 sources and sinks using atmospheric transport models , 2002, Nature.

[53]  Gérald Desroziers,et al.  An extended specification of flow‐dependent background error variances in the Météo‐France global 4D‐Var system , 2011 .

[54]  A. Mathieu,et al.  Atmospheric Dispersion and Deposition of Radionuclides from the Fukushima Daiichi Nuclear Power Plant Accident , 2012 .

[55]  T. Nakajima,et al.  Inverse modeling of the 137Cs source term of the Fukushima Dai-ichi Nuclear Power Plant accident constrained by a deposition map monitored by aircraft. , 2016, Journal of environmental radioactivity.

[56]  W. K. Hastings,et al.  Monte Carlo Sampling Methods Using Markov Chains and Their Applications , 1970 .

[57]  M. Sambridge,et al.  Transdimensional inference in the geosciences , 2013, Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences.

[58]  T. Nakajima,et al.  A Database of Hourly Atmospheric Concentrations of Radiocesium (134Cs and 137Cs) in Suspended Particulate Matter Collected in March 2011 at 99 Air Pollution Monitoring Stations in Eastern Japan , 2015 .

[59]  Luca Delle Monache,et al.  Bayesian Inference and Markov Chain Monte Carlo Sampling to Reconstruct a Contaminant Source on a Continental Scale , 2008 .

[60]  Marc Bocquet,et al.  Validation of the Polyphemus platform on the ETEX, Chernobyl and Algeciras cases , 2007 .

[61]  T. Nakajima,et al.  Model depiction of the atmospheric flows of radioactive cesium emitted from the Fukushima Daiichi Nuclear Power Station accident , 2017, Progress in Earth and Planetary Science.

[62]  Yukihisa Sanada,et al.  Detailed source term estimation of the atmospheric release for the Fukushima Daiichi Nuclear Power Station accident by coupling simulations of an atmospheric dispersion model with an improved deposition scheme and oceanic dispersion model , 2014 .