A spatiotemporally corrected radiological contamination mapping platform for multi-scenario nuclear and radiological emergencies.

Abstract

Rapid, quantitative assessment of the environmental consequences of radioactive contamination (and protection of responders operating within it) requires reconstructing continuous contamination fields from heterogeneous reconnaissance measurements. We present an automated platform that converts geo-referenced dose-rate data from ground and airborne assets into hazard-consequence maps. This study presents the platform architecture and quantitatively validates its spatial-reconstruction and decision-support modules; the altitude-correction and temporal-normalization modules are formalized as architecture components whose empirical validation is identified as future work. Airborne measurements are first converted to ground-equivalent dose rates through an altitude-dependent air-to-ground correction factor (AGCF) modeled as a segmented broken power-law (SBPL), then normalized to a common reference time using radionuclide half-life decay for radiological accidents or the Kaufmann power law for nuclear detonations. Normalized data are reconstructed into continuous 2-D fields by interpolation. Five methods, namely inverse distance weighting (IDW), ordinary kriging (OK), universal kriging (UK), anisotropic OK, and thin-plate spline (TPS), were evaluated in two scenarios: a 2000 kT nuclear surface burst (HotSpot 3.1.2) and the Fukushima Daiichi nuclear power plant accident (JAEA airborne surveys, 2011-2025). Performance was strongly scenario-dependent: for the detonation, TPS and anisotropic OK far outperformed isotropic OK (R²=0.866 and 0.849 vs. 0.256), a gap robust across resampling (95% CIs non-overlapping), whereas all methods converged in the Fukushima field (mean R²=0.938-0.957) under random hold-out; under spatially-independent block cross-validation this convergence is weaker and accuracy is lower (R²≈0.70-0.75), yet the anisotropy-aware advantage in the detonation field is reinforced. No single method is universally optimal; scenario-adaptive selection is essential for consequence estimation, exclusion-zone delineation, and minimum-dose routing in operational platforms.