Dynamic monitoring deployment for improved key-region radiological forecasts via ensemble-based forward-adjoint...

Accurate forecasting of atmospheric radiological pollutant concentrations is critical for emergency response to nuclear accidents, yet fixed monitoring networks are costly and often insufficiently sensitivity to unexpected releases. This study proposes an integrated framework for dynamic monitoring deployment and data assimilation to enhance early warning in key regions. The framework combines an Innovation-Corrected Regional Ensemble Transform Kalman Filter (ICRETKF) with a Dynamic Forward-Adjoint Strategy (DFAS). ICRETKF suppresses spurious background error correlations arising from small ensemble sizes and mitigates nonphysical negative-value truncation effects. DFAS incorporates adjoint-derived regional sensitivity to guide adaptive monitoring deployment toward locations exerting the greatest downstream influence on priority regions. Twin experiments demonstrate that ICRETKF reduces integrated concentration errors by approximately 80.83% relative to the classical ETKF under sparse or dynamic monitoring conditions. When combined with DFAS, the proposed framework further reduces key-region integrated concentration errors by about 61.33% compared with variance-based deployment. Overall, ICRETKF + DFAS achieves error reductions of approximately 92.59% relative to the ETKF with variance-based deployment and 81.17% relative to fixed-network assimilation, while advancing effective early-warning capability by approximately 60-80 h. These results highlight the effectiveness of sensitivity-informed dynamic monitoring deployment for large-scale radiological emergency response and risk mitigation. SYNOPSIS: This work proposes a sensitivity-informed dynamic monitoring framework that enhances early warning and concentration prediction for radiological emergencies, providing improved decision support for hazard mitigation.