Objective inversion of the continuous atmospheric ¹³⁷Cs release following the Fukushima accident

Eleven years after the Fukushima accident, independent objective estimates of the atmospheric ¹³⁷Cs release still suffer from discontinuities such as negative release terms, oscillations, and temporal gaps, leading to noticeable differences from the subjective estimate. This paper describes an objective method that handles these artifacts and promotes the continuity of releases at fine resolutions. The proposed method uses the joint estimation model to reduce the oscillations induced by the model-observation discrepancies, and employs total variation regularization to recover the missing releases caused by insufficient observations. Adaptive parameterization is used to correct negative values. The application of this method to the Fukushima accident produces a source term that accurately approximates continuous releases at a fine temporal resolution of 1 h, providing a better match with the recognized subjective source term than nine published estimates, with a Pearson's correlation coefficient of 0.923 and an index of agreement of 0.872. This source term agrees with the timing of on-site gamma dose rate peaks, significantly improving the air concentration and deposition simulations, with FAC10 values of 0.564 and 0.990, respectively. The estimation error varies smoothly in a limited range with different regularization parameters, enabling automatic parameterization and demonstrating the potential for operational inversions.