Statistical optimization of a Kriging interpolation-based dose distribution estimation algorithm

Emergency workers responding to nuclear power plant accidents may be exposed to high radiation dose rates. Therefore, when deploying emergency workers, it is necessary to estimate dose distribution within the work area and establish work plans based on this information. This study optimized a Kriging interpolation-based dose distribution estimation algorithm to address limitations in handling extreme values, reflecting spatial trends, optimizing parameters, and improving computational efficiency. Five methodological improvements were implemented: drift-based estimation for coordinate-dependent trends, the robust Cressie-Hawkins estimator, Weighted Least Squares (WLS) fitting for theoretical variogram derivation, Restricted Maximum Likelihood (REML) for parameter optimization, and the Moving Window technique for computational efficiency. The optimized algorithm was validated against Monte Carlo N-Particle (MCNP) code results for Waste Evaporator Room, Holdup Tank Room and Waste Drum Leakage scenarios. Compared to the conventional algorithm, the developed algorithm achieved average error reductions of 82%, 85% and 97% in the respective scenarios, with computational time reduced by approximately 80%. Even using only 10% of measurement points, error rates remained below 17%, demonstrating effective dose distribution estimation with limited measured values. These results are expected to contribute to dose map generation during nuclear emergencies and worker exposure management according to the ALARA principle.