KURAMA vs. Safecast: Radiation data comparison in Fukushima following whole-area decontamination.

Abstract

Following the Fukushima Daiichi Nuclear Power Plant (FDNPP) accident in 2011, environmental radiation monitoring is crucial for supporting reconstruction programs and ensuring the safety of returning residents. Although government-led monitoring systems yield relatively accurate data, limited coverage exists in some areas. Consequently, citizen-science initiatives, such as Safecast, emerged to address these gaps. We compare two major radiation monitoring systems in Fukushima Prefecture over the period 2019-2023: the Kyoto University Radiation Mapping (KURAMA) system and Safecast. As measurements are not co-located, all readings were aggregated onto a uniform 100 × 100 m grid for spatial comparison. At the current low ambient-dose levels, a nearly constant contribution from secondary cosmic radiation can measurably inflate readings from Geiger-Müller (GM) detectors. The analysis examines spatial coverage, radiation dose rate distribution, and data fitting through linear regression, error analysis, and Bland-Altman analysis. The results show that KURAMA provides extensive area coverage (1068-1419 km² per year) with sharp radiation dose transitions, particularly in high-exposure areas. In contrast, Safecast encompasses approximately 10 % of the area surveyed annually by KURAMA. We found that some Safecast areas cover residential areas and public facilities not monitored by KURAMA. Regression analysis indicates a strong linear correlation (R² = 0.8034). It also reveals a systematic bias in uncorrected Safecast data, yielding higher doses in low-exposure areas (<0.5 μSv/h) and lower doses in high-exposure areas (>1.0 μSv/h) compared to KURAMA. A key driver of the low-dose overestimation is the 31 nSv/h contribution of secondary cosmic radiation inherently counted by Safecast's GM detector. Subtracting this constant background (yielding Safecast_CR) lowers the Safecast median from 0.127 to 0.096 μSv/h and improves low-dose agreement with KURAMA while not changing the R² value. This claim of systematic bias is further strengthened by additional statistical analysis showing a consistent pattern of measurement discrepancies. These differences are influenced by detector characteristics, data collection methods, and operational variability, including the cosmic-ray component. Our findings indicate that integrating both systems, with appropriate calibration of Safecast data, can enhance the accuracy of radiation exposure assessments, improve post-decontamination monitoring, and contribute to more representative radiation maps for public safety and policy-making.