Ambient dose rate variation in the Fukushima region visualized using explainable AI techniques

Following the Fukushima Daiichi Nuclear Power Station accident in 2011, ambient dose rates in the surrounding region initially increased due to the deposition of radioactive materials on the ground, and have subsequently continued to decline owing to radioactive decay and decontamination efforts to the present. However, spatial variations in dose rate reduction remain insufficiently understood, particularly in forested areas where contamination persists. In this study, long-term trends in ambient dose rate changes were investigated using explainable AI techniques. An integrated dose rate map comprising fixed-point, walk, carborne, and airborne survey data collected over 12 years was used to analyze temporal and spatial patterns. We developed a predictive model using the Light Gradient Boosting Machine framework to estimate dose rate reduction ratios based on geographic and environmental features. SHapley Additive exPlanations were applied to quantify the contribution of each variable and enhance model interpretability. Our findings revealed that land use significantly influences dose rate reduction, with urban and agricultural areas showing faster reduction because of infrastructure and human activity, including decontamination work, whereas forests exhibit slower reduction. Notably, topographical features, such as elevation and slope, affect dose rate trends in undisturbed forests, with valleys and depressions showing stagnation. This paper provides the first visual validation of area-wide decontamination effects and demonstrates the utility of explainable AI in environmental radiation analysis. The proposed approach offers a robust framework for geospatial interpretation and, with further verification, is expected to support informed policymaking for regional recovery and forest utilization.