High-resolution spatiotemporal mapping of soil radon exhalation in China combining remote sensing and machine...

Radon (Rn), primarily generated via radium (Ra) decay in soil and exhaled across the soil-air interface, is a key source for natural ionizing radiation exposure. Accurate quantification of its exhalation is essential for health risk assessment, yet diverse geography, climate, and rapid urbanization processes of China induce spatiotemporal variability in Rn exhalation potential, limiting the representativeness and applicability of traditional measurements. Herein, we report the design of a two-stage model addressing Ra emanation and Rn exhalation processes: (1) Ra content prediction through an extremely randomized tree algorithm optimized with data augmentation, adaptive weighting, and cross validation; (2) Rn exhalation quantification by integrating a physical mechanistic model with soil remote sensing technology. Our models generated monthly 1-km Rn flux maps (validated at annual scale), achieved R² values of 0.72 and 0.71 with mean absolute percentage errors (MAPE) of 11.15% and 22.43% for Ra and Rn, respectively. However, expanding field surveys, particularly in remote high‑radon regions, remains essential to further improve model's precision (e.g., MAPE) and generalization. This study enables continuous, high-resolution radon exhalation mapping, and in turn, better understanding in its spatiotemporal variation, which serves as a preliminary screening tool to prioritize regions for targeted radon field surveys in areas heavily affected by radon exhalation.