The release rate of atmospheric radionuclide emissions is a critical factor in the emergency response to nuclear accidents. However, there are unavoidable biases in radionuclide transport models, leading to inaccurate estimates. In this study, a method that simultaneously corrects these biases and estimates the release rate is developed. Our approach provides a more complete measurement-by-measurement correction of the biases with a coefficient matrix that considers both deterministic and stochastic deviations. This matrix and the release rate are jointly solved by the alternating minimization algorithm. The proposed method is generic because it does not rely on specific features of transport models or scenarios. It is validated against wind tunnel experiments that simulate accidental releases in a heterogonous and densely built nuclear power plant site. The sensitivities to the position, number, and quality of measurements and extendibility of the method are also investigated. The results demonstrate that this method effectively corrects the model biases, and therefore outperforms Tikhonov’s method in both release rate estimation and model prediction. The proposed approach is robust to uncertainties and extendible with various center estimators, thus providing a flexible framework for robust source inversion in real accidents, even if large uncertainties exist in multiple factors.

中文翻译：

---

核事故中大气放射性核素排放的联合释放率估算和逐项测量模型校正：在风洞实验中的应用

大气放射性核素排放的释放速率是对核事故作出紧急反应的关键因素。但是，放射性核素迁移模型中不可避免存在偏差，导致估算不准确。在这项研究中，开发了一种同时纠正这些偏差并估计释放速率的方法。我们的方法通过考虑确定性和随机偏差的系数矩阵，提供了对偏差的更完整的逐次测量校正。该矩阵和释放速率通过交替最小化算法共同求解。所提出的方法是通用的，因为它不依赖于传输模型或方案的特定功能。它已针对风洞实验进行了验证，该实验模拟了异质密集核电站场址中的意外释放。还研究了对位置，数量和质量的敏感性以及方法的可扩展性。结果表明，该方法可以有效地校正模型偏差，因此在释放速率估计和模型预测方面均优于Tikhonov方法。所提出的方法对不确定性具有鲁棒性，并且可以通过各种中心估计器进行扩展，从而为实际事故中的鲁棒源反演提供了灵活的框架，即使在多个因素中都存在较大的不确定性也是如此。结果表明，该方法可以有效地校正模型偏差，因此在释放速率估计和模型预测方面均优于Tikhonov方法。所提出的方法对不确定性具有鲁棒性，并且可以通过各种中心估计器进行扩展，从而为实际事故中的鲁棒源反演提供了灵活的框架，即使在多个因素中都存在较大的不确定性也是如此。结果表明，该方法可以有效地校正模型偏差，因此在释放速率估计和模型预测方面均优于Tikhonov方法。所提出的方法对不确定性具有鲁棒性，并且可以通过各种中心估计器进行扩展，从而为实际事故中的鲁棒源反演提供了灵活的框架，即使在多个因素中都存在较大的不确定性也是如此。