Reproducibility of ocean dispersion simulations for ALPS-treated water release off Fukushima: comparison with one-year monitoring data

Validating the reproducibility of ocean dispersion models used in prior environmental impact assessments of ALPS-treated water release is essential for evaluating their applicability. In this study, we conducted reproduction simulations using actual release records together with observed meteorological and oceanographic conditions, and quantitatively compared the results with seawater monitoring data.

Model results were compared with tritium monitoring data collected by TEPCO, the Ministry of the Environment, the Nuclear Regulation Authority, and Fukushima Prefecture. The release was assumed to instantaneously disperse within a model grid (147 m × 186 m), with release scenarios prescribed for both the surface layer and the near-bottom layer at a depth of 10 m. As the actual discharged water is expected to rise upward from the seabed, results from the surface-release simulation are mainly discussed. Geometric means were used for model–observation comparisons to reduce the influence of outliers. Since background tritium concentration is not explicitly represented in the model, a constant background of 100 Bq m⁻³ was added to the modeled concentrations to ensure consistency with observations.

For the entire one-year period, the correlation coefficient between annual geometric means of modeled and observed concentrations was R = 0.30, indicating moderate reproducibility of temporal variability. In contrast, the mean log(Model/Obs) was −0.035, corresponding to a Model/Obs ratio of 0.92, demonstrating very good agreement in annual mean concentration levels. When the comparison was restricted to release periods, the correlation improved (R = 0.64), while the mean Model/Obs ratio increased to 1.37, suggesting a tendency toward overestimation associated with uncertainties in local release representation and model resolution near the outlet.

These results indicate that, although the model has limitations in reproducing short-term concentration variability, it reliably reproduces annual mean tritium concentrations that are critical for radiological dose assessment. The present validation demonstrates that the ocean dispersion model used in the prior environmental impact assessment has sufficient reliability for evaluating the dispersion behavior of ALPS-treated water, while highlighting the need for further improvements in the treatment of background concentrations and near-field processes.